CRWR Online Report 98-8
Use of Digital Soil Maps in a Rainfall-Runoff Model
by
Seann M. Reed, B.S., M.S.
Graduate Research Assistant
and
David Maidment, Ph.D.
Principal Investigator
December 1998
CENTER FOR RESEARCH IN WATER RESOURCES
Bureau of Engineering Research ? The University of Texas at Austin
J.J. Pickle Research Campus ? Austin, TX 78712-4497
This document is available online via World Wide Web at
http://www.ce.utexas.edu/centers/crwr/reports/online.html
ii
Acknowledgements
For helping me through all my years as a student, I would first like to thank my
parents, John and Kathie, and my sister, Monise.
During my graduate studies at The University of Texas, I have had the good
fortune to interact with many wonderful people.  Most importantly, I met my wife,
Cindy, who has brought great joy to my life.
I wish to thank my faculty advisor, David Maidment, not only for his scholarly
advice, but also for his never-ending optimism and kindness.  I would also like to
thank my other dissertation committee members, especially Jay Famiglietti, who has
made me feel welcome to participate in activities with the Geological Sciences
department.
A special thank you also goes to Francisco Olivera who has been both friend
and mentor.
I greatly appreciate the financial support provided by the Hydrologic
Engineering Center (HEC) and the technical guidance provided by John Peters and
Arlen Feldman of HEC.
Other colleagues who I would like to thank for providing help with technical
issues throughout my graduate studies include Zichuan Ye, Tom Evans, Dean
Djokic, Pawel Mizgalewicz, John Burgin, Laurie Wood, Tim Plover, and Ty
Lehman.
iii
I will always remember the warmth of the students, faculty, and staff in EWRE.
Thank you to the ultimate frisbee players for all the fun times.
I would like to acknowledge the support of the following organizations for
providing data used in this research free of charge:
  Agricultural Research Service, Grazinglands Research Laboratory, El Reno,
OK.
  Atmospheric Radiation Measurement (ARM) Program sponsored by the
U.S. Department of Energy, Office of Energy Research, Office of Health
and Environmental Research, Environmental Sciences Division.
  U.S. Geological Survey EROS Data Center, Sioux Falls, SD.
  U.S. Geological Survey, Oklahoma District, Oklahoma City, OK.
  U.S. Department of Agriculture Natural Resources Conservation Service.
  U.S. Environmental Protection Agency.
  National Weather Service Arkansas-Red Basin River Forecast Center, Tulsa,
OK.
  EOS Distributed Active Archive Center (DAAC), NASA Goddard Space
Flight Center, Greenbelt, MD.
iv
Use of Digital Soil Maps in a Rainfall-Runoff Model
Publication No. _______
Seann Mischa Reed, Ph.D.
The University of Texas at Austin, 1998
Supervisor: David R. Maidment
Computer programs, collectively named the Soil-water Balance Modeling
System (SWBMS), are developed to evaluate alternative methods for using digital
soil maps in rainfall-runoff modeling.  The two soils databases of interest are the
State Soil Geographic Database (STATSGO ? 1:250,000 scale) and the Soil
Survey Geographic Database (SSURGO ? 1:24,000 scale), both maintained by the
U.S. Department of Agriculture.  In SWBMS, simple conceptual models are used to
describe infiltration, percolation, and evaporation processes.  Use of simplified
process models is justified because there are large uncertainties involved with the
specification of soil hydraulic properties and in quantifying rainfall.  Representative
soil hydraulic properties are assigned based on USDA texture class.  Distributed soil
properties are used in conjunction with distributed NEXRAD Stage III rainfall data
to make to make rainfall excess calculations.
SWBMS is applied using hydrologic data from the Little Washita watershed in
southwestern Oklahoma.  The performance of SWBMS in predicting runoff is
evaluated using different levels of spatial complexity to describe soil properties.
Spatial complexity is varied in three different ways.  (1) Differences between a one
and a two-layer conceptual soil model are considered.  (2) Simulations in which a
single surface soil texture is assigned to the entire watershed, a single surface soil
v
texture is assigned to each NEXRAD rainfall cell, and multiple surface soil textures
are considered within each NEXRAD rainfall cell are compared.  (3) Soil texture
properties are assigned using soil maps created from sources at two different scales
(1:250,000 and 1:24,000).
Calibration and validation of the SWBMS model has yielded interesting
insights.  In the two-layer conceptual soil model, the top layer controls direct runoff,
but the bottom layer serves as an important control on how much water evaporates
and how much water percolates to the groundwater reservoir.  When the spatial
variability of surface texture is reduced through resampling of soil properties, model
performance decreases.  Although the 1:250,000 scale soil map contains
considerably less spatial detail than the 1:24,000 scale map, use of each data set as
input in model validation runs produced similar results.  Further simulations should
be made to verify these conclusions.
vi
TABLE OF CONTENTS
LIST OF TABLES....................................................................................................x
LIST OF FIGURES.................................................................................................xii
NOMENCLATURE..............................................................................................xvi
CHAPTER 1. INTRODUCTION...............................................................................1
1.1. Motivation......................................................................................................1
1.2. Objectives and scope......................................................................................3
1.3. Study area.......................................................................................................7
1.4. Organizational statement................................................................................9
CHAPTER 2. LITERATURE REVIEW..................................................................10
2.1. Surface and subsurface runoff: definition of terms???????.10
2.2. Lumped versus distributed modeling...........................................................11
2.2.1. NEXRAD Data.....................................................................................11
2.2.2. Partial area effects................................................................................13
2.3. Soil-water balance modeling........................................................................16
2.4. Infiltration....................................................................................................17
2.5. Evaporation..................................................................................................18
2.6. Percolation...................................................................................................22
2.7. Existing streamflow simulation modeling programs...................................22
2.8. GIS and Hydrologic Modeling.....................................................................23
2.8.1. Terrain analysis....................................................................................23
2.8.2. GIS preprocessors for hydrology models.............................................24
2.8.3. Use of soils data....................................................................................25
2.9. Summary......................................................................................................26
CHAPTER 3. METHODOLOGY............................................................................27
3.1. Overview......................................................................................................27
vii
3.2. Data description ...........................................................................................29
3.2.1. Topographic data..................................................................................32
3.2.2. Soils Data..............................................................................................35
3.2.2.1. STATSGO vs. SSURGO ..............................................................35
3.2.2.2. Estimation of soil hydraulic properties .........................................42
3.2.2.2.1. Soil attribute tables..................................................................42
3.2.2.2.2. The lookup table approach......................................................45
3.2.2.2.3. Selecting representative parameter values for texture classes 51
3.2.2.3. One-layer model versus a two-layer model ..................................61
3.2.3. Land use data........................................................................................68
3.2.4. NEXRAD data and spatial variability ..................................................70
3.2.5. Data used to estimate evaporation........................................................74
3.2.6. Data used for calibration and validation...............................................74
3.3. Process Representation ................................................................................77
3.3.1. Direct runoff computation ....................................................................77
3.3.1.1. Green-Ampt model .......................................................................79
3.3.1.2. Estimating wetting front suction head ..........................................85
3.3.1.3. Computational scheme for Green-Ampt infiltration under variable
rainfall ........................................................................................................87
3.3.2. Evaporation...........................................................................................90
3.3.2.1. Surface Resistance ........................................................................90
3.3.2.2. Data analysis: evaporation ............................................................92
3.3.2.3. Plotting surface resistance against soil piezometric head ...........104
3.3.2.4. Evaporation method used in this study: practical considerations111
3.3.3. Percolation..........................................................................................112
3.3.4. Soil layer mass balance.......................................................................113
3.3.4.1. Non-event mass balance for an individual time step ..................113
3.3.4.2. Event mass balance .....................................................................114
3.3.5. Runoff transform method : ?modClark?.............................................118
3.3.6. Baseflow .............................................................................................122
3.3.6.1. Single linear reservoir model ......................................................122
3.3.6.2. Baseflow separation ....................................................................123
3.3.7. Special considerations for the Little Washita.....................................126
3.4. Summary ....................................................................................................127
viii
CHAPTER 4. COMPUTER IMPLEMENTATION...............................................130
4.1. Collecting and reformatting input data ......................................................130
4.2. The ArcView Pre-Processor (AV-SWBMS) .............................................131
4.2.1. Pre-processing for a one-layer model.................................................133
4.2.2. Pre-processing for a two-layer model.................................................138
4.3. The Visual Basic Hydrology Program (VB-SWBMS)..............................141
4.4. Summary ....................................................................................................144
CHAPTER 5. ANALYSIS AND RESULTS..........................................................145
5.1. Overview....................................................................................................145
5.2. Calibration and validation procedures .......................................................145
5.2.1. Calibration ..........................................................................................147
5.2.2. Validation ...........................................................................................147
5.3. Comparison of NEXRAD and Micronet rainfall .......................................152
5.4. One-layer vs. two-layer model...................................................................163
5.4.1. Direct runoff .......................................................................................163
5.4.2. Baseflow .............................................................................................166
5.5. Horizontal soil variations...........................................................................169
5.6. Sensitivity analysis.....................................................................................173
CHAPTER 6. CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE
WORK ....................................................................................................................177
6.1. Conclusions................................................................................................177
6.2. Recommendations for future work ............................................................182
APPENDICES ........................................................................................................184
A. Description of the AV-SWBMS and VB-SWBMS programs .....................185
B. Time series data processing programs..........................................................205
ix
C. Description of files on the CD-ROM that was produced to archive the
research presented in this dissertation...............................................................217
D. Table used to reclassify 719 STATSGO texture names into 12 basic USDA
classes and the "other" category........................................................................222
REFERENCES........................................................................................................238
VITA .......................................................................................................................247
x
LIST OF TABLES
Table 3.1  List of data sources. .................................................................................30
Table 3.2  Projection parameters...............................................................................32
Table 3.3  Mean values and coefficients of variation for hydraulic parameters in the
12 basic USDA texture classes.  A comparison of values reported by Rawls,
Brakensiek, and Saxton (1982) and Carsel and Parrish (1988). .......................52
Table 3.4  Estimates for Green-Ampt wetting front suction head (h
f
) for an initially
dry soil...............................................................................................................61
Table 3.5 Description of soil drainage classes specified in the STATSGO comp
table...................................................................................................................66
Table 3.6 Impervious percentages associated with selected land use classes...........66
Table 3.7 Natural saturation values corresponding to the 12 USDA texture classes.
...........................................................................................................................82
Table 4.1 Spatial data required to run AV-SWBMS ..............................................132
Table 4.2  Description ArcView pre-processing programs under the menu
?SWBMS.?......................................................................................................135
Table 4.3 Description of menu items used to create input files for a two-layer model.
.........................................................................................................................138
Table 4.4  Input files for VB-SWBMS ...................................................................142
Table 4.5  Output files for VB-SWBMS.................................................................143
Table 5.1  Model parameters not specified from soils databases........................... 151
Table 5.2  Comparison of storm rainfall totals between Micronet and NEXRAD
data. ................................................................................................................ 153
Table 5.3  Calibrated parameter values for NEXRAD and adjusted NEXRAD
simulations. .................................................................................................... 157
Table 5.4  Statistics describing NEXRAD and adjusted NEXRAD calibration and
validation runs................................................................................................ 158
Table 5.5  Calibrated parameter values for the one-layer and two-layer models. . 163
Table 5.6  Statistics describing one-layer and two-layer calibration and validation
runs................................................................................................................. 164
xi
Table 5.7  Calibrated parameter values and validation statistics for the one-layer
model with different levels of model resampling and different soil map inputs.
........................................................................................................................ 172
Table 5.8  Statistics for level 2, STATSGO and level 0, 30 m grid simulations on
5/7/97 - 5/24/97.............................................................................................. 173
Table A1  Suggested AV-SWBMS menu sequences for different modeling options..
........................................................................................................................ 190
Table A2. VB-SWBMS File Names...................................................................... 191
Table B1:  Procedure for processing NEXRAD StageIII data............................... 215
xii
LIST OF FIGURES
Figure 1.1  Conceptual illustration of three resampling options.................................5
Figure 1.2  Location reference for the Little Washita River watershed......................9
Figure 2.1  Typical soil moisture extraction function (adapted from Shuttleworth,
1993) .................................................................................................................20
Figure 3.1 Conceptual storage zone model schematic..............................................28
Figure 3.2  Watersheds delineated from 100-m and 30-m digital elevation models
respectively. ......................................................................................................34
Figure 3.3  Flow length to the watershed outlet for the 30 m digital elevation model.
...........................................................................................................................35
Figure 3.4  STATSGO and 30 m soils grid...............................................................39
Figure 3.5 Mapunit OK103 polygons. ......................................................................40
Figure 3.6  Comparison of surface texture composition for the STATSGO soils map
and the 30-m soils grid with (a) soils grouped into the 12 USDA texture classes
and (b) loamy sand and sandy loam grouped together along with loam and silt
loam...................................................................................................................41
Figure 3.7  Relationships among the mapunit, comp, and layer table. .....................43
Figure 3.8. Example wetting and drying water retention curves for a loam soil.with
h
b
 = 27.8 and ? = 0.56.......................................................................................45
Figure 3.9  Mean porosity values for USDA texture classes with 1 standard
deviation error bars. ..........................................................................................55
Figure 3.10  Mean residual saturation values for USDA texture classes with 1
standard deviation error bars.............................................................................56
Figure 3.11  Mean ? values for USDA texture classes with 1 standard deviation
error bars. ..........................................................................................................57
Figure 3.12  Mean h
b
 values for USDA texture classes with 1 standard deviation
error bars. ..........................................................................................................58
Figure 3.13  Rawls, Brakensiek, and Saxton h
b
 values on the USDA textural
triangle. .............................................................................................................59
Figure 3.14  Carsel and Parrish h
b
 values on the USDA textural triangle. ...............60
xiii
Figure 3.15  Percentage of soil components with 1, 2, 3, 4, 5, or 6 layers (a) with
multiple distinguishing characteristics and (b) with the 12 basic USDA soil
classes as the only distinguishing characteristic. ..............................................63
Figure 3.16  Total depth of the soil zone defined for the two-layer model. .............67
Figure 3.17  Land use in the Little Washita watershed.............................................68
Figure 3.18 Intersecting NEXRAD cells with watershed boundary.........................71
Figure 3.19  Locations for gravimetric soil moisture sampling during SGP97. .......76
Figure 3.20  Locations of Micronet stations. ............................................................77
Figure 3.21  Average percent imperviousness in each NEXRAD cell. ....................79
Figure 3.22  Curved wetting front formed during infiltration into a homogeneous,
uniformly wetted soil. .......................................................................................80
Figure 3.23  Idealized infiltration into a homogeneous, uniformly wetted soil with a
sharp wetting front. ...........................................................................................80
Figure 3.24 Relative permeability curve for wetting.(? = 0.56, h
b
 = 27.8) ..............86
Figure 3.25  Pseudo-code for computation of Green-Ampt infiltration under variable
rainfall. ..............................................................................................................89
Figure 3.26  Locations of 22 ARM Measurement Sites ...........................................95
Figure 3.27  Surface radiation budget components at ARM SGP Site EF7 on July
29, 1997.............................................................................................................97
Figure 3.28  30 minute Bowen ratio values at the ARM SGP Site EF7 on July 29,
1997...................................................................................................................98
Figure 3.29  Wind speed (m/s) at site EF7 on July 29, 1997....................................99
Figure 3.30  Temperature (?C) and relative humidity (RH) at site EF7 on July 29,
1997.................................................................................................................100
Figure 3.31  Vapor pressure deficit and vapor pressure at site EF7 on July 29, 1997.
.........................................................................................................................100
Figure 3.32  Surface resistance (r
s
) vs. ln(|h+z|) for EF7 at a depth of 5 cm (a) with
regression line for all data points (b) with regression lines for individual
months.............................................................................................................103
Figure 3.33. Surface resistance (r
s
) vs. ln(|h+z||) for EF7 at a depth of 15 cm (a) with
regression line for all data points (b) with regression lines for individual
months.............................................................................................................106
Figure 3.34 Surface resistance (r
s
) vs. ln(|h+z|) for EF7 at a depth of (a) 25 cm, (b)
35 cm, (c) 60 cm, (d) 85 cm, (e) 125 cm, and (f) 175 cm...............................108
xiv
Figure 3.35  A comparison of surface resistance (r
s
) vs. ln(|?+z|) at two sites: EF7
and EF4. ..........................................................................................................110
Figure 3.36  Post-infiltration mass balance flow chart for the two-layer model. ...117
Figure 3.37  Post-infiltration mass balance flow chart for the one-layer model.....118
Figure 3.38  Average flow length from each model polygon to the watershed outlet.
.........................................................................................................................120
Figure 3.39  Baseflow separation method. (a) estimating baseflow at t = 2, (b)
estimating baseflow at t = 3, and (c) estimating baseflow at t = 4..................125
Figure 3.40  Baseflow separation using n = 15, 30, and 60 hours. .........................126
Figure 4.1 Screen capture of menus that are part of AV-SWBMS.........................133
Figure 4.2  Example input files for the one-layer model. .......................................137
Figure 4.3 Example input files for the two-layer model. ........................................140
Figure 4.4  Pseudo-code describing the overall computational sequence for VB-
SWBMS. .........................................................................................................142
Figure 5.1  A comparison of Micronet and NEXRAD Stage III rainfall estimates for
8 hours of the 7/10/97 storm. ......................................................................... 156
Figure 5.2  Calibrated flows using (a) NEXRAD and (b) adjusted NEXRAD
rainfall. ........................................................................................................... 159
Figure 5.3  Validation runs using parameters from Table 5.3.  (a) NEXRAD and (b)
adjusted NEXRAD rainfall. ........................................................................... 160
Figure 5.4  Comparison of modeled and observed average daily soil moisture for (a)
a loamy sand in cell (557, 309) and observed data averaged from sites 3, 4, and
5; (b) a sandy loam in cell (559, 305) and observed data from site 17; (c) a
loam in cell (560, 309) and observed data averaged from sites 11, 12, and 13;
and (d) a silt loam in cell (555, 307) and observed data averaged from sites 8
and 9............................................................................................................... 162
Figure 5.5  Flow during the validation period for the one-layer and two-layer
models. ........................................................................................................... 164
Figure 5.6  Attempts to simulate baseflow using the one-layer soil model. (a)
recession constant ? = 240 hours  (b) recession constant ? = 17520 hours... 168
Figure 5.7  Hydrographs from two-layer simulation with ? = 240 hours and p
2
 =
0.0015 cm/hour. ............................................................................................. 169
xv
Figure 5.8  Comparison of direct runoff from watershed level resampling,
STATSGO simulation with direct runoff from a no resampling, 30-m grid
simulation for 5/7/97 - 5/24/97. ..................................................................... 172
Figure 5.9 Tornado diagram showing deviations from the base case based on high
and low parameter values............................................................................... 175
Figure A1 Menus for AV-SWBMS. .......................................................................185
Figure A2 SWBMS user interface for a one-layer model with no baseflow
simulation........................................................................................................194
Figure B1 Graphical User Interface for several data processing programs. ...........206
xvi
NOMENCLATURE
(Symbols are listed in alphabetical order with lower case letters first, followed by
upper case letters, lower case Greek letters, and upper case Greek letters.)
Symbol Description Units
c
p
specific heat of moist air
J kg
-1
 ?C
-1
d layer depth cm
d
r
rooting depth cm
e vapor pressure kPa
e
s
saturation vapor pressure kPa
f(?) soil moisture extraction function
h suction head cm
h
b
air entry head (a.k.a. bubbling
pressure head)
cm
h
c
estimated crop height m
h
we
air exit head or water entry head cm
h
f
wetting front suction head cm
h
i
initial suction head cm
h
p
ponding depth cm
i infiltration rate cm hour
-1
k
f
hydraulic conductivity calibration
factor
k
rw
relative hydraulic conductivity
l latent heat of vaporization J kg
-1
n baseflow separation parameter
p percolation rate cm hour
-1
q Darcy flux cm hour
-1
r rainfall rate cm hour
-1
r
a
aerodynamic resistance s m
-1
r
s
surface resistance s m
-1
r
smin
minimum surface resistance s m
-1
t
p
time to ponding hours
u
z
wind speed at height z m s
-1
z elevation cm
z
e
height at which humidity
measurements are made
m
z
om
roughness height for momentum
transfer
m
z
ov
roughness height for turbulent
transfer of water vapor
m
z
u
height at which wind measurements m
xvii
are made
CV coefficient of variation
D vapor pressure deficit kPa
E actual evaporation mm day
-1
E
m
mass flux of water vapor kg m
-2
 s
-1
E
r
reference crop evaporation mm day
-1
FL
avg
average flowlength from a rainfall
cell to the watershed outlet
m
FL
max
maximum flowlength from a
rainfall cell to the watershed outlet
m
G ground heat flux W m
-2
GSM gravimetric soil moisture
H sensible heat flux W m
-2
I cumulative infiltration cm
I
p
cumulative infiltration at ponding
time
cm
K
ns
saturated hydraulic conductivity at
natural saturation
cm hour
-1
K
s
saturated hydraulic conductivity cm hour
-1
LAI leaf area index
L
f
wetting front depth cm
N van Genuchten water retention
parameter
P total percolation from a soil layer
during a given time step
cm
P
a
atmospheric pressure kPa
Qflow m
3
 s
-1
Q
o
observed flow m
3
 s
-1
Q
s
simulated flow m
3
 s
-1
Q
d
direct runoff m
3
 s
-1
Q
din
direct runoff into the linear storage
reservoir
m
3
 s
-1
Q
dout
direct runoff out of the linear
storage reservoir
m
3
 s
-1
R modClark storage coefficient hours
R
n
net radiation W m
-2
RH relative humidity
S groundwater storage m
3
T air temperature ?C
T
c
modClark time of concentration hours
T
m
travel time from a rainfall cell to
the watershed outlet
hours
xviii
?
van Genuchten water retention
parameter
cm
-1
?
Bowen ratio
?
groundwater storage constant hours
?
layer root fraction
? 
porosity
?
psychrometric constant
kPa ?C
-1
?
c
crop coefficient
?
co
potential crop coefficient
?
Brooks and Corey water retention
parameter
?
mean
?
volumetric water content
?
fc
water content at field capacity
?
i
initial water content
?
ns
water content at natural saturation
?
r
residual water content
?
wp
water content at plant wilting point
?
a
density of moist air kg m
-3
?
b
bulk density g cm
-3
?
w
density of water g cm
-3
?
gradient of the saturation vapor
pressure curve
kPa ?C
-1
?
effective saturation
?
d
drying phase effective saturation
?
i
initial effective saturation
?
ns
natural saturation
?
w
wetting phase effective saturation
1
CHAPTER 1. INTRODUCTION
1.1. Motivation
Traditionally, runoff simulation models have used lumped (watershed average)
rainfall and infiltration parameter values.  This approach has been practical because
rain-gage networks are typically sparse and use of data from these gages does not
justify a more detailed spatial model.  Development of next generation radar
(NEXRAD) products by the National Weather Service has made spatially
distributed hydrologic modeling a more practical alternative.  With gridded rainfall
products, distributed or quasi-distributed methods for watershed modeling merit
further investigation.
Researchers at the Hydrologic Engineering Center (HEC) of the U.S. Army
Corps of Engineers are interested in enhancing the capabilities of their Hydrologic
Modeling System (HMS) software to include capabilities for considering distributed
soil properties.  Geographic Information System (GIS) pre-processing software
programs have recently been developed to implement a quasi-distributed runoff
transform method called modClark in HMS (Peters and Easton, 1996; Reed and
Maidment, 1995; HEC, 1995; HEC, 1996).  This software is designed to work with
NEXRAD data.  Although this software makes it practical to transform rainfall
excess from gridded rainfall cells to the watershed outlet, it has only limited
capabilities for computing distributed loss rates.  GIS programs are available to
estimate the average Soil Conservation Service (SCS) curve number for each
NEXRAD grid cell.  Although knowing the spatial distribution of Curve Numbers is
an improvement over using watershed average loss rate parameters, this is a rather
unsatisfactory approach for dealing with losses when considering the many
weaknesses of the SCS Curve Number method, particularly when it is used in an
attempt to reproduce historical events ? not the purpose for which it was originally
developed (Goldman, 1989; Pilgrim and Cordery, 1993).  In addition to problems
2
with the SCS Curve Number method noted by Goldman (1989) and Pilgrim and
Cordery (1993), this method is ill suited for incorporation into a continuous soil-
moisture accounting model which can be used to predict antecedent moisture
conditions.  Methods to continuously model soil-water content through time are
important because rainfall excess is sensitive to antecedent moisture conditions
(Pilgrim and Cordery, 1993; Coles et al., 1997).
The focus of this manuscript is on the computation of rainfall excess using
distributed soil parameters and distributed rainfall estimates. Use of distributed
rainfall and soils data may help to alleviate some limitations of lumped continuous
simulation modeling methods currently being tested at HEC.  In simulations made
using the Continuous Soil Moisture Accounting Model being developed for HEC-
HMS, engineers have noticed that a model calibrated to large storms may predict no
response for small storms (Peters, pers. com., 1998).  Missing the small storms
could perhaps be due to the inability of a lumped model to capture partial area
effects or due to inaccurate estimation of antecedent soil moisture conditions.  These
issues are investigated in this research.
The ability of the soil to transmit and retain water influences the rate of runoff
during a storm and the rate at which the soil dries out between storms.  To model the
water fluxes into and out of the soil on a continuous basis, physically based process
representations are attractive ? as opposed to strictly empirical methods like the
SCS method.  For an appropriately formulated soil-water balance model, parameters
with a physical basis can be estimated using digital soils databases maintained by
the U.S. Department of Agriculture (USDA).  To date, the information in USDA
soils databases like STATSGO (State Soil Geographic Database) and SSURGO
(Soil Survey Geographic Database) has not been fully exploited by practicing
hydrologists.  This is due in part to the fact that practical schemes for using these
data in soil-water balance and rainfall-runoff modeling are lacking.  It would be
3
useful to be able to estimate parameter values that control infiltration, percolation,
and evaporation using information from these databases.
Distributed rainfall data, digital soils data, and the computer software to easily
manage these data sets are now available.  The challenging task is to develop new
methods to use these data and to determine the effectiveness of these new methods.
To evaluate new methods, a common approach is to compare predicted and
observed streamflows.  Although the total runoff generated in a watershed can be
compared with observed streamflow, it is often difficult to understand what is going
on within the watershed.  Just knowing the fraction of rainfall that does not become
streamflow does not provide information about how much of this water is stored in
the soil and groundwater and how much evaporates.  New data sets from intensive
hydrology field campaigns that are geared towards improving land surface and
cloud parameterizations in General Circulation Models offer unprecedented soil
moisture and meteorological measurements that may help to validate and
parameterize soil-water balance models.  Some of these data are used in this study.
The current investigation focuses on the soil-water balance component of
streamflow simulation models because this is an area where currently available soils
data and experimental observations have not been fully utilized.
1.2. Objectives and scope
The objectives of this study are:
? Develop a practical and automated method for using U.S. Department of
Agriculture digital soils data in a rainfall-runoff model.
? Compare the performance of modeling alternatives that use different
levels of spatial complexity.
? Quantify the uncertainties involved with the specification of soil
hydraulic properties.
4
? Make recommendations to HEC regarding the use of spatial soils data in
HEC-HMS.
To more specifically define the research scope, a summary of the modeling issues to
be addressed is provided here.
Spatial complexity:
Because the rainfall intensity is assumed to be uniform within a NEXRAD cell,
NEXRAD rainfall cell boundaries define the major units in which vertical flux
calculations are made.  A key issue addressed in this research is how to describe
spatial soil variability.  In the analysis of spatial data, resampling procedures are
often used to simplify complex data sets.  Some alternatives for resampling spatially
complex soils data are considered here.
? No resampling:  Estimate the percentage of each soil type or the percentage
of each distinct soil component within each NEXRAD cell, and make
vertical flux calculations separately for each of these units.
? NEXRAD cell level resampling:  Assign a representative soil type to each
NEXRAD cell.
? Watershed level resampling:  Assign a representative soil type to the entire
watershed (a lumped approach).
These three resampling options are illustrated in Figure 1.1.  For each of these three
resampling levels, modeling results may vary depending on the map scale of the
soils data.  The differences in simulation results obtained using 1:250,000 scale soil
maps (STATSGO) and 1:24,000 scale soil maps (SSURGO) are considered.  The
map scale issue is of particular interest because 1:250,000 scale data are readily
available for all of the United States, but 1:24,000 scale data are more difficult to
obtain.
5
No resampling.
NEXRAD cell level
resampling.
Watershed level
resampling.
 
S
LS
SL
L
SIL
CL
SICL
Other
Figure 1.1 Conceptual illustration of three resampling options.
In addition to considering horizontal variations in soil properties, tradeoffs
between parameterizing only near surface soils (one-layer model) and
parameterizing both surface soils and a second soil layer (two-layer model) that
restricts percolation to groundwater are investigated.  Given the results of this
investigation, a goal of this research is to assess whether further vertical partitioning
is desirable and could be supported by the available data.
6
Parameterization:
A simple scheme for estimating soil hydraulic properties based on the texture
attribute listed in standard soils data tables is proposed and implemented.  In the
two-layer model, other attributes in the soils data tables are considered to define the
lower boundary for the soil zone.
Process representation:
The soil-water balance is computed by estimating infiltration, percolation,
evaporation, and change in soil-water storage.  A concerted effort is made to
maintain an approximate physical basis for process representation whenever
possible.  Simple, straightforward equations (i.e. the Green-Ampt infiltration
equation) for computing infiltration and percolation as a function of measurable soil
properties and the current soil-water content are described in the literature.  Relating
evaporation rates to soil properties and soil-water content is more complex because
evaporation is dependent on many factors in addition to soil characteristics.
Methods used to estimate evaporation from soils range from very simple empirical
methods (bucket models) to very complex soil-vapor-atmosphere transfer schemes
(SVATS).  An effort is made in this research to take advantage of newly available
soil moisture and surface energy budget data in an effort to gain a greater
understanding of soil drying rates due to evaporation.  Using this data, the
possibility of deriving empirical relationships that define effective daily values of
evaporative surface resistance as a function of soil-water potential is investigated.
Computer implementation:
Using ArcView GIS and standard spatial data (i.e. STATSGO and 3? digital
elevation models) model pre-processing routines are automated, making it easy to
set up the model at any location in the United States.  The ramifications of model
7
complexity on computer implementation are discussed, particularly with regard to
the transfer of information from the GIS system to a hydrologic model like HEC-
HMS.
Evaluation:
The accuracy of alternatives for describing spatial complexity is evaluated by
comparing simulated runoff with observed runoff and simulated soil moisture with
observed soil moisture.
1.3. Study area
The Little Washita River Watershed in southwestern Oklahoma was selected as
the study area for this project for the following reasons:
? With a 600 km
2
 drainage area, the Little Washita River basin is at the lower
end of the size range for modeling units typical used by the U.S. Army
Corps of Engineers and the National Weather Service for river basin
forecasting.
? Archived Stage III NEXRAD Data, the most advanced National Weather
Service weather radar product, is publicly available for this area from the
Arkansas-Red Basin River Forecast Center.
? The state of Oklahoma is rich in meteorological and hydrological
observations.  Three research observation networks overlap in Oklahoma,
including the Department of Energy's Atmospheric Radiation Measurement
Program's Southern Great Plains Cloud and Radiation Testbed (DOE ARM
SGP CART) (www.arm.gov), the Oklahoma Mesonet operated jointly by the
University of Oklahoma and Oklahoma State University
(okmesonet.ocs.ou.edu), and the U.S. Department of Agriculture
Agricultural Research Service (USDA ARS) Micronet operated in the Little
8
Washita Watershed (grl.ars.usda.gov/sgp97-1.html).  In addition to the high
density of automated meteorological observations, a unique feature of these
three observation programs is the large number of automated soil moisture
observations that are being taken.  More than 90 locations within a 10,000
km
2
 area are now instrumented to produce profiles of soil-water and
temperature (Schneider and Fisher, 1997).  This network of 90 soil moisture
measurement stations has been given the name MOISTNET (Schneider and
Fisher, 1997).  Not all of the MOISTNET data are available free of charge
and some of the new equipment being used is still being calibrated.
Therefore, only a small sampling of the data from the MOISTNET sites is
analyzed in this study.  Rain gage data from the Little Washita Micronet are
compared with NEXRAD Stage III radar rainfall estimates in this study.
Gravimetric soil moisture measurements taken in the Little Washita
watershed during the Southern Great Plains 1997 Hydrology Experiment
(daac.gsfc.nasa.gov/CAMPAIGN_DOCS/SGP97/sgp97.html) are also used
for model calibration.
? County level (1:24,000 scale) soils are not generally available, but are
available for the Little Washita watershed.
The location of the Little Washita River, a tributary of the Chickasha River is
shown in Figure 1.2.
9
Little Washita 
River
Washita River
Watershed
Oklahoma
Figure 1.2 Location reference for the Little Washita River watershed
1.4. Organizational statement
The remainder of this dissertation includes:
  Chapter 2: A literature review.
  Chapter 3: A methodology chapter that includes a description of all input
data, methods used to estimate model parameters using GIS software, and
mathematical equations used to describe hydrologic processes.
  Chapter 4: A description of how the proposed model is implemented on a
computer.
  Chapter 5: An analysis and results chapter that describes calibration and
validation simulation runs and insights gained from these simulations related
to the study objectives.
  Chapter 6: Conclusions and recommendations for future work.
The methodology chapter is the longest because a large amount of the research
effort in this project has gone towards model development.
10
CHAPTER 2. LITERATURE REVIEW
This section includes a review of literature relevant to the stated objectives.
Topics covered include a summary of runoff terminology, NEXRAD data, partial
area effects, soil-water balance modeling, infiltration, evaporation, percolation,
existing streamflow simulation programs, and GIS applications for hydrologic
modeling.  Additional references related to the infiltration, evaporation, and
percolation models that have been selected for use in this study are provided in the
Chapter 3.  Chapter 3 also includes more information about the input data used in
this research.
2.1. Surface and subsurface runoff: definition of terms
Streamflow prediction is a fundamental problem in hydrology that has been
studied for many years.  Streamflow produced from flow over the land surface is
typically computed in a different manner than the flow that passes through the
subsurface because different mechanisms control the volume and timing of surface
and subsurface flows.  Subsurface flow is produced when soil-water or groundwater
discharges directly to a stream.  Subsurface flow with a short-term response is often
referred to as subsurface stormflow while flow that persists for long times after a
rainfall event is often referred to as baseflow.  The term direct runoff is used to refer
to streamflow generated from water that does not infiltrate and flows over that land
surface.  Direct runoff may be generated if the rainfall rate exceeds the infiltration
capacity of the soil (infiltration excess) or if the soil surface becomes saturated from
below (saturation excess).  The saturation excess mechanism for direct runoff
generation tends to be more important in watersheds with humid climates, concave
footslopes, and wide valley bottoms (Dunne, 1978).  In arid and sub-humid climates
and watersheds disturbed by human influence (i.e. agricultural watersheds), the
infiltration excess runoff mechanism tends to dominate.
11
2.2. Lumped versus distributed modeling
2.2.1. NEXRAD Data
A hydrologic model commonly used by the National Weather Service, the
Sacramento Soil Moisture Accounting Model (Burnash, Ferral, and McGuire,
1973), and a hydrologic model commonly used by the U.S. Army Corps of
Engineers, HEC-1, are both lumped models in which rainfall is assumed to be
uniform over an entire subwatershed.  Lumped models are currently used by the 13
National Weather Service River Forecast Centers (RFCs) to provide daily discharge
and stage forecasts at over 4,000 locations nationwide (Smith et al., 1996b).  The
installation of a national network of radars (Weather Surveillance Radar 1988-
Doppler (WSR88-D)) by the National Weather Service as part of the Next
Generation Weather Radar (NEXRAD) program makes the move from lumped
models to distributed models a real possibility.  NEXRAD products provide
precipitation estimates at higher spatial and temporal resolutions than have
traditionally been available for hydrologic modeling.  Better precipitation estimates
offer the potential to improve streamflow forecasts, increase the number of river
forecast points, and improve reservoir operations (Shedd and Fulton, 1993).  Other
applications for NEXRAD data include flash flood forecasting, long term water
supply forecasting, climatology, and engineering hydrometeorology (Smith et al.,
1996b).
Several levels of NEXRAD processing have been developed for different uses.
The lowest level NEXRAD product, Stage I, uses radar information alone to
generate 1-hour rainfall estimates.  In Stage II processing, rain gage data, satellite
information, and surface temperature information are used to improve the radar
estimates from Stage I.  In Stage III, rainfall estimates from several radars are
mosaicked into a common grid system at National Weather Service River Forecast
Centers (RFCs) so that basin-wide stream flow forecasts can be made.  Stage III
also incorporates interactive quality control by the forecaster.  Stage IV is a national
12
mosaic of hourly rainfall that uses both Stage II and Stage III data as input (Shedd
and Fulton, 1993).
The accuracy of NEXRAD data products is still a matter of study.  Smith et al.
(1996a) provide a good discussion of errors associated with measurement of rainfall
using WSR-88D radars based on a study of radar data from the Southern Great
Plains of the United States.  The conclusions of Smith et al. (1996a) include the
following:  (1) radar rainfall estimates are range-dependent with a tendency to
produce low biased estimates at close (0-40 km) and far (beyond 150 km) ranges
and higher estimates at intermediate ranges, (2) systematic differences in comparing
signals from overlapping radars indicate that calibrations may be inconsistent
among different radar sites, resulting in a 30% difference in average predicted
rainfall in the example provided, (3) improvements in NEXRAD algorithms have
eliminated a number of problems with anomalous propagation (high rainfall
estimates due to high radar reflectivity from areas not receiving rainfall) but
anomalous propagation is still problematic if embedded in actual precipitation, (4) a
comparison between radar and gage estimates indicates that radar systematically
underestimates rainfall, particularly at close and far ranges, and (5) despite the noted
problems, WSR-88D radars are far superior to rain gage networks in monitoring
heavy rainfall because it was found that a number of storm systems producing heavy
rain were completely missed by gage networks.  The NEXRAD Stage III estimates
used in this study incorporate information from several overlapping radars and have
been adjusted to reflect rain gage information; therefore, it is impossible to predict
whether an underestimate or overestimate of rainfall would be expected based on
the observations of Smith et al. (1996a).  In a study of nine watersheds near the
Oklahoma-Arkansas-Missouri state boundaries, Finnerty and Johnson (1997) found
that 7-month mean areal rainfall accumulations computed from NEXRAD StageIII
data were 10-25% lower than 7-month mean areal rainfall accumulations computed
from rain gage data alone.
13
Both the National Weather Service Hydrologic Research Lab (HRL) and the
U.S. Army Corps of Engineers Hydrologic Engineering Center (HEC) are currently
researching methods to use NEXRAD data in hydrologic forecasting.  As an
intermediate step between existing lumped basin models and grid-cell level
modeling, a ?semi-distributed? modeling approach has been adopted by the
Hydrologic Research Lab in which the basins in existing models are sub-divided
into smaller sub-basins and mean areal precipitation estimates are computed for
these basins using NEXRAD data (Smith et al., 1996b).  It is apparent from the
research of Smith et al. (1996b) that the Sacramento Soil Moisture Accounting
Model parameters that have been used in the past will require re-calibration if the
semi-distributed modeling approach using NEXRAD data is used.  As one might
expect, Smith et al. (1996b) show that spatially averaging rainfall data over different
size sub-basins produces notable differences (as high as a 600% difference in
surface runoff in the example given) if model parameters remain unchanged.  HEC
has implemented a ?quasi-distributed? cell-based model for using NEXRAD data in
the HEC-Hydrologic Modeling System (HEC-HMS) software.  Rainfall excess is
computed for each NEXRAD cell and routed to the watershed outlet using an
adaptation of the Clark unit hydrograph method called modClark (HEC, 1995;
Peters and Easton, 1996).  ModClark is ?quasi-distributed? rather than distributed
because there is no cell-to-cell flow routing.
2.2.2. Partial area effects
It is a well documented phenomenon that not all portions of a watershed
contribute equally to direct runoff during a storm.  This is due to both rainfall
variability and land surface heterogeneity.  Betson (1964) used the term partial
areas to refer to this phenomenon.  Betson (1964) did not observe large changes in
the partial areas contributing to direct runoff among different storms on the same
watershed, except during extreme conditions.  A similar term, variable source areas
14
(Hewlett and Hibbert, 1965), refers more specifically to the expansion and
contraction of saturated soil zones under different rainfall conditions, causing
variations in saturation excess runoff generation.
A number of hydrologic models have been formulated so that the soil
heterogeneities and/or spatially variable infiltration properties of a watershed are
described using statistical distributions.  In the Stanford watershed model and
subsequent variations thereof, the infiltration capacity in a watershed at a given time
is characterized by a simple probability distribution function (Hydrocomp, 1970).  It
is assumed that the infiltration capacity in a watershed can be described using a
uniform distribution ranging from a minimum to a maximum infiltration capacity.
This assumption has no physical basis but has proven to be effective in matching
observed runoff (Hydrocomp, 1970; Dawdy, Lichty, and Bergman, 1972).  This
same scheme is used in the U.S. Geological Survey Precipitation Runoff Modeling
System (PRMS) (Leavesley et al., 1983).  In a parameterization for a General
Circulation Model, Entekhabi and Eagleson (1989) described the spatial variability
of surface soil moisture using a probability distribution function that fits "well" with
field observations of other researchers.  Sivapalan and Wood (1986) and Wood et
al. (1988) used a lognormal distribution to describe the spatial variability of
saturated hydraulic conductivity for numerical experiments.  Assumed distributions
for soil properties or soil moisture are used to calculate the amount of runoff
generated by infiltration excess and the amount of evaporation.
Researchers have also developed simplified ways to explicitly model direct
runoff generated by the saturation excess mechanism.  This is done by using
topographic maps in conjunction with soil maps.  Localized areas of soil saturation
are predicted by effectively reducing the 3-dimensional watershed to two
dimensions using a topographic wetness index.  A topographic wetness index is
derived by assuming that surface saturation will occur where the accumulated
drainage flux exceeds the local transmissivity of the soil.  A common assumption is
15
that a perched water table forms at some depth due to the presence of a low
permeability layer and that surface saturation will occur when this water table rises
to the surface.  Beven and Kirkby (1979) and O'Loughlin (1981) were the first to
describe the use of topographic wetness indices.  A number of modelers have used
variations on Beven and Kirkby?s TOPMODEL (Beven, 1986; Sivapalan et al.,
1987; Famiglietti and Wood, 1994) and O?Loughlin?s wetness index (O'Loughlin,
1986; Moore et al., 1986; Moore et al., 1988).  The models of Beven (1986),
Sivapalan et al. (1987), and Famiglietti and Wood (1994) explicitly account for both
the infiltration excess and saturation excess mechanisms of runoff generation.
To a certain extent, field observations have validated the use of topographic
index methods as an indicator of lateral soil moisture distributions (Anderson and
Burt, 1977; Dunne, Moore, and Taylor, 1975; Zaslavsky and Sinai, 1981; and Burt
and Butcher, 1985).  Zaslavsky and Sinai (1981) observed higher soil-water
contents in hollows when compared with spurs even in situations where no perched
water table or impermeable layer existed.  Zaslavsky and Sinai (1981)
mathematically relate observed soil moisture to plan curvature.  Burt and Butcher
(1985) compare several different topographic indices with observed soil moisture
(a/s, ln(a/s), plan curvature, etc.) and find that perhaps the best correlation with
observed soil moisture results from the product plan curvature multiplied by (a/s).
In these expressions, a is the upslope drainage area and s is the local slope in
percent.  The analysis of Burt and Butcher (1985) was done on a 1.4 hectare
hillslope.  Based on their analysis alone, it is uncertain how well these topographic
indexes would perform if computed using digital elevation models with relatively
large cell sizes.  In a 100-m digital elevation model, for example, each cell covers
0.01 km
2
 so there would only be 1.4 cells in a 1.4 hectare area.
16
2.3. Soil-water balance modeling
A soil-water balance model is used to keep track of water content changes in
the soil zone.  Soil-water content is an important state variable that influences fluxes
into and out of the soil zone (infiltration, evaporation, and percolation) and the
energy balance at the land surface.  Soil-water balance models treat the soil zone as
a control volume.  The control volume approach provides a consistent means for
applying physical laws to hydrologic systems (Chow, Maidment, and Mays, 1988).
Using this approach, direct runoff, evaporation, and percolation are treated as losses
from the hydrologic system.
Thorough validation of soil-water balance models is typically only possible on
small, well-instrumented plots of land where the fluxes across the soil zone
boundaries are measured.  On small plots, evaporation is often estimated by
measuring rainfall or irrigation, changes in soil-water content, and sometimes
percolation.  Weighing lysimeters or neutron probes are typically used to measure
changes in soil-water content (ASCE, 1990).  Percolation may be measured with
some types of lysimeters or assumed negligible if measurements are taken several
days after irrigation (ASCE, 1990).  Evaporation may also be measured using an
energy balance approach like the Bowen ratio method or a mass transfer approach
like eddy correlation.  Descriptions of these methods are given by Brutsaert (1982)
and Kanemasu et al. (1992).  The Bowen ratio and eddy flux measurement
techniques require extensive instrumentation so data of this type has typically been
available only for short research studies (ASCE, 1990).  Data from the Department
of Energy?s Atmospheric Radiation Measurement Program (ARM) (ARM, 1998;
Peppler, Sisterson, and Lamb, 1997) provide unique data sets that might be helpful
in validating water balance models.  At some ARM measurement sites, Bowen ratio
measurement systems are co-located with soil-water measurement systems.
Schneider and Fisher (1997) note that the devices that are used to measure soil
moisture at these sites are still (as of the publication date) undergoing calibration but
17
some interesting insights related to evaporation can be still be drawn from the
available data sets as discussed in Section 3.3.2.
In the soil-water balance model described in this dissertation, changes in soil-
water content are estimated by making an initial guess at soil-water content and then
estimating infiltration, evaporation, and percolation fluxes using physically-based
mathematical equations and climatic forcing data.  To the extent possible,
gravimetric measurements of soil-water content from the 1997 Southern Great
Plains Hydrology Experiment are used to help calibrate the model (SGP, 1997).
2.4. Infiltration
Infiltration is the downward flow of water into the soil.  Infiltration rates are
controlled by soil properties and by the supply of moisture at the surface.
Infiltration is one mechanism for ?losses,? defined as the quantity of rainfall that
does not contribute to direct runoff.  Other loss mechanisms include interception,
depression storage, and streambed losses.
It is often adequate to use simple, non-physical models such as an initial
abstraction and constant loss rate model when calibrating an event simulation model
with gaged data (Goldman, 1989); however, it is desirable to relate loss rates to
physical characteristics of the watershed in continuous simulation so that loss rates
may be computed as a function of soil moisture conditions at the beginning of a
storm.  In addition, a model with a physical basis may be suitable for application in
ungaged areas.  Approximate theory-based solutions to Richard?s equation like the
Green-Ampt and Philip models and empirical loss models like Holtan?s method and
the SCS Curve Number method have parameters that relate to measurable watershed
characteristics.  Reviews of these methods are provided by Rawls et al. (1993) and
Goldman (1989).  The Green-Ampt and Philip methods are more practical to apply
than the Holtan method because simple relationships are available for deriving
Green-Ampt and Philip parameters from water retention and relative hydraulic
18
conductivity parameters of a given soil (Neuman, 1976; Clapp and Hornberger,
1978).  The SCS method is not considered as an option for the research described
herein because it is not well-suited for incorporation into a continuous soil moisture
simulation scheme and because it has a poor theoretical basis for time-varying
calculations.  In this research, the Green-Ampt model is selected over the Philip
model because the Green-Ampt model has received more attention in the
engineering literature, and the physical interpretation of Green-Ampt parameters is
more straightforward.
The Green-Ampt model has performed well in laboratory studies (Mein and
Larson, 1973; Rawls and Brakensiek, 1985), but uncertainty remains as to its
performance under field conditions with variable rainfall rates and heterogeneous
soil conditions.  A scheme for using the Green-Ampt model with variable rainfall
rates and heterogeneous soil conditions is proposed here.  Use of a simplified model
like the Green-Ampt model is justified when field data describing soil hydraulic
properties show large spatial variability, which is often the case (Govindaraju et al.,
1992).
2.5. Evaporation
Following the convention used by Shuttleworth (1993) the term evaporation in
this document refers to the ?rate of liquid water transformation to vapor from open
water, bare soil, or vegetation with soil beneath.?  Major factors that govern the rate
of evaporation from open water include the energy available at the surface and the
vapor pressure deficit.  Evaporation from bare soil is more complex because it also
depends on soil-water content and soil type.  Evaporation from plants is even more
complex because plant stomata open and close to control the release of water vapor.
Thus, the evaporation rate from plants is a function of the stomatal density and the
degree to which stomata are opened.  Guard cells on plant leaves may vary the size
of stomatal openings as a function of the amount of sunlight, the CO
2
 concentration,
19
the leaf-air vapor pressure gradient, the leaf temperature, and the leaf water content
(soil-water content can be used as a surrogate for leaf water content) (Dingman,
1994).  Modeling evaporation is further complicated because many of the factors
that influence evaporation vary significantly throughout the day and it is often the
case that only daily observations of meteorological variables are available.
Evaporation research has been conducted by investigators in many different
scientific and engineering disciplines with different purposes.  Agricultural and civil
engineers are interested in estimating irrigation requirements and evaporation rates
from surface water impoundments while climatologists and atmospheric scientists
are interested in the relationships between evaporation and regional and/or global
climate.  Perhaps the most rigorous evaporation models have been developed for use
with General Circulation Models (GCM) ? biophysically ?realistic? models
developed by Dickinson (1984) and Sellers et al. (1986) consider diurnal variations
in heat and vapor fluxes and complex feedbacks between plant growth, momentum
transfer, and the surface energy balance.  In these models, truly interdependent
parameters such as albedo, roughness length, and soil moisture are not prescribed
independently as is often the case (Sellers, 1992).  In engineering practice, such
biophysically realistic models are generally considered too complex to be practical
? parameters can often be determined only through numerical experiments and the
detailed meteorological inputs required are not generally available (outside of a
GCM).
Shuttleworth (1993) and ASCE (1990) provide thorough reviews of common
methods for estimating evaporation in engineering applications.  A widely accepted
approach for estimating evaporation from irrigated field crops is to first estimate the
so-called reference crop evaporation (E
r
) [mm day
-1
] and then multiply this by an
empirical crop coefficient (?
c
) which depends on the crop type, stage of
development, and to some degree on the average climate where the empirical
calibration was carried out.  The crop coefficient also depends on the dryness of the
20
soil when the calibration was carried out.  A slightly different crop coefficient is
defined for conditions when the water supply is adequate so that plant growth is not
limited (?
co
).  Under conditions when soil moisture becomes a significant limiting
factor on evaporation, empirical studies have been made to determine the
relationship between the ratio of actual evaporation (E) to reference crop
evaporation (E
r
) and soil-water content.  Dyck (1983), Shuttleworth (1993), and
Brutsaert (1982) summarize the work of several researchers who have plotted
empirical relationships between soil-water content (?) and the evaporation ratio
(E/E
r
).  Although the precise relationship between ? and E/E
r
 varies under different
conditions, Shuttleworth (1993) notes that the results of many investigators follow
the general trend shown in Figure 2.1.
moi
s
ture
 extra
c
tion
f
unct
i
on, f(
?
)
?
wp
?
c
?
fc
1
0
soil moisture fraction
Figure 2.1  Typical soil moisture extraction function (adapted from
Shuttleworth, 1993)
In Figure 2.1, ?
wp
 is the wilting point and ?
fc
 is the field capacity.  According to the
relationship in Figure 2.1, drying proceeds unrestricted by soil-water content until a
critical level (?
c
) is reached.  At this point, the evaporation rate decreases until the
wilting point is reached and the evaporation becomes zero.  ?
c
 is typically 50 to 80
21
% of ?
fc
.  Using this type of relationship, the evaporation from a vegetated surface is
estimated as:
Equation 2.1
rco
EfE ?? )(= [mm day
-1
] 
The amount of water accessible to plants for evaporation depends on the rooting
depth, which varies with location and season.  When a soil root zone is divided into
multiple layers, one approach to estimating evaporation from each layer is to take
layer evaporation as the product of E
r
, the fraction of roots in that layer, and f(?)
(Shuttleworth, 1993).  This approach is discussed further in Section 3.3.2.
Part of the reason for the variability found in the literature for plots like that
shown in Figure 2.1 may be that the true relationship is dependent on soil type,
which makes sense because, as noted by Shuttleworth (1993), plants are more
sensitive to suction head (h) than soil-water content (?) and the relationship between
? and h varies for different soils.  Zahner (1967) used different f(?) relationships for
sand, loam, and clay while making soil-water balance calculations.  The
relationships proposed by Zahner (1967) are also used in the USGS PRMS model
(Leavesley et al., 1983).
The method of Equation 2.1 is attractive because of its simplicity.  For practical
application with agricultural crops, empirical ?
co
 factors for different stages in the
plant growth cycle are tabulated for different crop types and different climates
(Doorenbos and Pruitt, 1977; Shuttleworth, 1993).  However, Shuttleworth (1993)
points out that the crop coefficient ?
c
 is poorly defined and has implicit
meteorological dependence (the same point would apply to ?
co
).  Shuttleworth
(1993) also recommends that future research projects into plant control on
evaporation should attempt to estimate and document a purer measure of stomatal
control, surface resistance (r
s
), rather than the crop coefficient (?
c
).  An empirical
study that attempts to relate a daily average surface resistance (r
s
) to soil-water
22
matric potential is presented in Section 3.3.2.  This approach is investigated as a
possible alternative to using Equation 2.1.
2.6. Percolation
Kinematic models for the redistribution and percolation of infiltrated water are
discussed by Jury, Gardner, and Gardner (1991) and Charbeneau (1989).  A simple
kinematic model for percolation is used in this study, as described in Section 3.3.3.
2.7. Existing streamflow simulation modeling programs
This section is not intended to be an exhaustive review of existing streamflow
simulation programs.  Reviews of well-known hydrologic simulation programs are
provided by DeVries and Hromadka (1993), Francini and Pacciani (1991) and
McFadden (1994).  None of the programs described in these reviews includes the
capability to use digital soil maps to automatically estimate soil hydraulic properties
or the capability to use distributed radar rainfall estimates.  Francini and Pacciani
(1991) compare the predictive capabilities of seven conceptual storage models
(Stanford, Sacramento, TANK, APIC, SSARR, Xinanjiang, and Arno).  Francini
and Pacciani conclude that all of these models (with the exception of APIC) produce
?similar and equally valid results.?  Therefore, the most attractive models are those
that are easy to calibrate, require few parameters, and consider the physics of the
problem (Francini and Pacciani, 1991).  The model described in this dissertation
meets all of these criteria.
Streamflow simulation models are often calibrated and validated using
observed streamflow data.  Models with a large number of parameters are often
difficult to calibrate.  In models with many parameters, more than one parameter
can have a similar effect on the predicted hydrograph.  In these cases, one approach
to calibration is to isolate times in the hydrologic record when a certain parameter
23
has the most effect (USACE, 1994).  It takes a minimum of 5 to 8 years of
hydrologic data to properly calibrate the Sacramento model used by the National
Weather Service (Smith et al., 1996b).  Estimating some model parameters directly
from physical watershed characteristics simplifies calibration procedures.
2.8. GIS and Hydrologic Modeling
Geographic Information System (GIS) software programs provide tools for
storing, analyzing, and visualizing geographical and related tabular data.  Different
data structures are used in a GIS to store different types of data.  Three of the most
common data structures are (1) the vector data structure, which uses points, lines,
and polygons to represent features in two dimensions, (2) the raster data structure
which uses a regular grid of cells to represent continuous surfaces and categorical
data, and (3) triangulated irregular networks (TINs) which are used to model 3-
dimensional surfaces.  The GIS programs described in this dissertation use the
ArcView 3.0a GIS with its Spatial Analyst 1.1 extension to manipulate vector and
raster data describing topography, soils, land use, and rainfall cells.  The Avenue
script language is used to manipulate objects in ArcView and automatically prepare
an input file for hydrologic calculations.  Previous work on preparing input data for
streamflow models is discussed briefly in this section.  Many other hydrology
related GIS applications are not discussed.  A review of some other hydrology
related applications is provided on the GISHydro98 CD-ROM (CRWR-ESRI,
1998).
2.8.1. Terrain analysis
Digital terrain models are widely used in hydrology.  A digital terrain model is
an ordered array of elevation values, commonly stored in one of three data
structures: contours, grids, or triangulated irregular networks (TINs).  Moore,
Grayson, and Ladson (1991) review the advantages and disadvantages of each of
24
these three data structures.  Grid terrain models are attractive because they are
widely available, terrain analysis methods with the grid model are simple, and the
grid structure is compatible with remote sensing techniques.  A grid terrain model
consists of regularly spaced elevation values in two dimensions.  The term adopted
by the USGS to describe digital terrain data in a grid format is digital elevation
model (DEM).
A basic task that can be accomplished easily with GIS software is the
delineation of drainage paths and watershed boundaries from a digital elevation
model.  Other hydrologic parameters that can be estimated from a digital elevation
model include drainage area, slope, aspect, and flow path length.  A more
exhaustive list of topographically derived attributes is provided by Moore, Grayson,
and Ladson (1991).  A simple and widely used method for approximating cell-to-
cell flow paths and delineating drainage areas from a digital elevation model is
described by O?Callaghan and Mark (1984) and Jensen and Domingue (1988).  Both
the Arc/Info and the ArcView 3.0 with the Spatial Analyst software packages
contain built in functions for terrain analysis and watershed delineation based on the
algorithms described by Jensen and Domingue (1988).  Maidment and Mizgalewicz
(1993) and Reed and Maidment (1995) describe Arc/Info procedures for watershed
delineation.  The ArcView GUI offers a more convenient user interface to do the
same types of analysis.  Customized ArcView programs for watershed and stream
network delineation are briefly summarized in the introduction section of Olivera,
Reed, and Maidment (1998).
2.8.2. GIS preprocessors for hydrology models
Customized tools for hydrologic modeling have been developed to pass
information from GIS databases to widely used engineering hydrology models.
Programs described by Hellweger (1997) with modifications described by Olivera,
Reed, and Maidment (1998) are capable of preparing a basin file for the Hydrologic
25
Modeling System (HMS).  This basin file contains enough information for HMS to
automatically create a schematic network for lumped hydrologic modeling.
Reaches and watersheds in this model are automatically attributed with parameters
derived from the topography, soil, and land use data layers.  Arc/Info GIS software
programs that can be used to prepare a cell parameter file for HEC?s quasi-
distributed modClark routing program (a component of HMS) are described by
Reed and Maidment (1995) and HEC (1996).  The Watershed Modeling System
developed at the Brigham Young University Engineering Computer Graphics
Laboratory has the capability to use GIS data in preparing an input file for the HEC-
1 hydrology model and other models used for hydrologic and hydraulic design
(WMS, 1998).
2.8.3. Use of soils data
In this dissertation, a systematic approach is proposed to estimate hydrologic
soil properties from standard soils data distributed and maintained by the U.S.
Department of Agriculture.  In previous work, Miller and White (1998) have
developed a 1 km grid with multi-layer hydrologic soil characteristics for the
conterminous United States by interpreting and processing information in the USDA
STATSGO database.  In attempting to develop a generic gridded database intended
for applications such as soil-vegetation atmosphere transfer schemes (SVAT),
climatology, and other environmental models, Miller and White (1998) were forced
to make numerous assumptions in the interpretation of STATSGO data.  Although
different assumptions are preferred for the current work, the work of Miller and
White (1998) is similar to the present study in the sense that both are geared towards
making soil information more accessible to hydrologic modelers.
Additional references describing spatial data sources (topography, soils, and
land use) and procedures for processing and interpreting these data are provided in
Chapter 3.
26
2.9. Summary
Distributed rainfall estimates from NEXRAD products offer the potential to
improve streamflow prediction models.  Traditional, lumped approaches to
hydrologic modeling need to be adapted in order to fully utilize the spatial
information provided by NEXRAD products.  Efforts to use distributed rainfall
estimates in operational models have been initiated by the U.S. Army Corps of
Engineers Hydrologic Engineering Center (HEC) and the National Weather Service
Hydrologic Research Lab (HRL).  Geographic Information Systems provide useful
tools for implementing distributed modeling schemes.
Existing distributed modeling schemes lack the ability to describe spatially
variable loss rates.  Digital soil maps (STATSGO and SSURGO) maintained by the
U.S. Department of Agriculture contain spatial information that may be used along
with distributed rainfall to estimate distributed loss rates.  Research is needed to
develop and evaluate practical methods for using these soil maps in operational
hydrology models.  The remaining chapters of this dissertation address this issue.
Loss rates depend on the wetness of the soil at the beginning of a storm.  A
continuous soil-water balance model can be used to model the changes in soil-water
content due to infiltration, evaporation and percolation.  Because field data
describing soil properties show large spatial variability, it is believed that the use of
simple process models, like the Green-Ampt infiltration equation, is justified.  The
parameters for simple flow models can be estimated using information in the
STATSGO and SSURGO databases.  A method to parameterize simple infiltration,
evaporation, and percolation equations is proposed in Chapter 3.
27
CHAPTER 3. METHODOLOGY
3.1. Overview
This section describes a conceptual model, input data, data interpretation
methods, and model equations.  The conceptual model and model equations are
specifically chosen to maintain a physical basis when practical, use methods that are
well supported by previous studies, require a small number of calibration
parameters, and be practical to apply using spatial rainfall and soils data.  In terms
of input data and data processing, the most emphasis is placed on the estimation of
soil hydraulic properties; however, all of the input data sets are described in some
detail in Section 3.2.
A diagram illustrating the proposed conceptual storage zone model is shown in
Figure 3.1.  Well-known streamflow simulation models like the USGS Precipitation
Runoff Modeling System (Dawdy, Lichty, and Bergmann, 1972; Leavesley et al.,
1983), the Sacramento model (Burnash, Ferral, and McGuire, 1973), and the
Streamflow Synthesis and Reservoir Regulation Model (USACE, 1994) all contain
components similar to those shown in Figure 3.1.  All of these models are more
complex, dealing with additional factors such as snowmelt, interception and
depression storage, and explicitly estimating interflow, but the fundamental
approach is the same.  The surface and subsurface are divided into zones in which
the storage and movement of water are controlled by different factors.  Equations
are formulated to describe the movement of water into and out of these zones and
mass balance equations are used to keep track of the water stored in each zone.
Forcing data, including rainfall and other meteorological variables used to estimate
evaporation, are used to drive the model.
28
transmission
zone
percolation
infiltration evaporation
direct 
runoff
groundwater storage
zones(s)
subsurface 
runoff
root zone
Figure 3.1 Conceptual storage zone model schematic
The simplified conceptual model in Figure 3.1 is used in this research because
the emphasis here is on developing and testing parameterizations for the soil zone.
Infiltration, evaporation, and percolation estimates are point calculations.  Different
soil properties in different parts of a watershed will result in different amounts of
excess runoff when subjected to the same amount of rainfall.  The routing of excess
runoff from different points in the watershed to the watershed outlet depends on
overland flow velocity and storage effects.  These effects are approximated using
the modClark method (see Section 3.3.5).  Although the spatial variabiltiy of soil
properties within a watershed are considered in this model, the subsurface response
is modeled as a function of a single lumped parameter.  Two alternatives for
determining baseflow are used in this research, simulating baseflow with a simple
linear reservoir model and applying a baseflow separation technique to observed
flow data.  Reasons for using these two baseflow approaches are discussed in
Section 3.3.6.
29
In this research, all spatial analysis tasks are automated using ArcView GIS and
all hydrologic calculations are done using Visual Basic.  Computer implementation
and data management are discussed further in Chapter 4.
3.2. Data description
A summary of data used in this research is provided in Table 3.1.  Not all of the
data sources listed in Table 3.1 are required to run a simulation.  Required inputs
include a digital elevation model, a soil map (STATSGO or SSURGO), a land use
map that uses the Anderson Level II classification scheme, rainfall data and an
associated map of rainfall cells, and enough meteorological data to make an
estimate of reference crop evaporation.  The meteorological variables required to
estimate reference crop evaporation depend on the estimation method used.  In this
research, the Penman-Monteith reference crop equation is used, requiring estimates
of available heating, temperature, relative humidity, atmospheric pressure, and wind
speed; however, a program user may compute reference crop evaporation using any
method and store these values in a model input file.  The model operates using a
nominal time step of one hour because this is the time interval for which NEXRAD
rainfall data are available.  All of the required inputs for this model are relatively
easy to obtain.  Other data sets listed in Table 3.1 that are used for calibration and
validation are unique to the Southern Great Plains region and/or the Little Washita
River watershed.
Tab
l
e 3.1  List of
 
d
a
ta sou
rces.
Category
D
escription
M
ap
Sca
l
e
Source of Infor
m
ation
M
e
t
hod of
Acquis
i
t
i
o
n
Dates
 for Wh
ich
Data w
a
s
 
Ob
tain
ed
T
o
p
ograp
h
y
30 m
 di
g
i
t
a
l
 el
ev
a
t
i
o
n
 
m
odel
~
 1:
24,0
0
0
A
R
S
-
G
RL
, El
 R
e
n
o
, OK
g
r
l
.
ars
.
us
da.g
ov
Pe
r
s
on
a
l
requ
es
t
3
?
 (~100
 m
)
 di
gi
t
a
l
 el
ev
a
t
i
o
n
 m
odel
1:
250,
000
US
GS
 EROS
 Dat
a
 
C
e
n
t
er.
edc
w
ww
.cr.us
gs
.g
o
v
/w
ebg
lis
In
tern
et
do
w
n
l
o
ad
So
ils
S
T
A
T
S
GO f
o
r Ok
l
a
h
o
m
a
1:
250,
000
US
D
A
 N
R
C
S
,
w
w
w
.f
tw
.n
r
c
s
.
u
s
d
a
.
g
ov
/
s
o
i
ls_
d
a
ta.ht
m
l.
In
tern
et
do
w
n
l
o
ad
30 m
 s
o
i
l
s
 g
r
i
d
 ? S
S
U
R
GO l
e
vel
 of
sp
atial d
e
tail; attrib
ute tab
l
es are
struct
ured
 si
m
ilarl
y
 to
 ST
A
T
SGO.
~ 1:24,0
0
0
A
R
S
-
G
RL
, El R
e
n
o
, OK
g
r
l
.
ars
.
us
da.g
ov
Pe
r
s
on
a
l
requ
es
t
La
nd Use
/
La
n
d
Cover
G
I
R
A
S La
nd
use
/
La
nd
c
o
ve
r
 
fil
e
s w/
And
e
rso
n
 
L
e
ve
l II land
 use cla
ssi
f
i
catio
n
sche
me.
1:
250,
000
EPA
.
www
.epa.
g
ov
/n
gi
s
pgm
3/n
s
di/pro
j
ects
/
g
i
ras
.
ht
m
In
tern
et
do
w
n
l
o
ad
M
e
teorol
o
g
i
c
al
Data
NWS
 NEXRA
D
 S
t
ag
e III data: h
o
u
r
l
y
esti
m
a
tes o
f
 p
r
ecip
itatio
n o
n
 grid
 cells
approx
i
m
a
t
ely
 4 km
 on
 a si
de.
NW
S A
r
k
a
nsas
-
R
ed B
a
si
n
R
i
v
e
r Forecast 
C
e
nter.
i
n
fo
.
a
b
r
fc
.
n
o
a
a
.
go
v
In
tern
et
do
w
n
l
o
ad
T
h
e arch
iv
ed
produ
ct
 i
s
 av
ai
l
a
bl
e
Ju
n
e
 24, 199
4 ?
presen
t.  Data 
w
a
s
obt
ai
n
e
d f
o
r Oct
ober
1995 ? 
S
e
pt
em
ber
1997
Micro
n
et:  5
-
m
i
n
u
te c
u
m
u
lat
i
ve rai
n
f
a
ll
e
s
ti
ma
te
s 
fr
o
m
 4
2
 ga
ge
s in a
n
d
 a
r
o
und
the 
L
ittle W
a
s
h
ita 
w
a
ters
hed
.
  Used
 to
ch
eck
 t
h
e NEX
R
A
D
 data.
AR
S
-
G
R
L
,
 El R
e
n
o
, OK.
g
r
l
.
ars
.
us
da.g
ov
/s
g
p97-
1.h
t
m
l
In
tern
et
do
w
n
l
o
ad
J
u
n
e
, J
u
l
y
, a
n
d
Aug
u
s
t
, 1997 *
m
a
ny
 of
 th
e dat
a
 f
o
r
earl
y
 an
d 
m
i
d-
J
u
n
e
are m
i
ss
ing
30
Table 3.1 (cont.) Category
D
escription
M
ap
Sca
l
e
Source of Infor
m
ation
M
e
t
hod 
of
Acquis
i
t
i
o
n
Dates
 for Wh
ich
Data w
a
s
 
Ob
tain
ed
M
e
teorol
o
g
i
c
al
Data (con
t
.
)
DOE ARM SGP
-
C
A
RT
: 30 m
i
n
u
t
e
En
ergy
 B
u
dg
et Bo
w
e
n
 
R
a
tio (EBBR
)
d
a
ta strea
m
; d
a
ta f
r
o
m
 
site EF2
6
 used
 to
esti
m
a
te p
o
t
ential e
v
ap
o
r
atio
n f
o
r
si
m
u
lat
i
o
n
s i
n
 t
h
e Little W
a
s
h
ita, d
a
ta
f
r
o
m
 s
i
t
e
s
 EF
4 an
d EF
7 us
ed f
o
r
ev
aporation
 stu
d
y
 (see Sectio
n
 3.3.2).
DOE-ARM. w
w
w
.
a
r
c
hi
ve
.
a
r
m
.
g
o
v
/d
a
t
a
/o
r
d
e
r
ing.
ht
ml
In
tern
et
do
w
n
l
o
ad
EF
26:
 Oct
ober 1
995
? S
e
pt
em
ber 1997;
EF4: Ma
y
,
 J
u
n
e
,
Ju
ly
, 1997;
 EF
7:
J
une
,
 J
u
l
y
,
 
A
u
gu
st,
S
e
pt
e
m
ber, 1997.
DOE ARM SGP
-
C
A
RT
: 30 m
i
n
u
t
e
Su
r
f
ace Meteorolog
i
cal Observ
atio
n
S
y
s
t
e
m
 (SM
O
S) d
a
ta strea
m
; d
a
ta f
r
o
m
s
i
t
e
s
 EF
4 an
d EF
7 us
ed f
o
r ev
aporat
i
o
n
s
t
u
d
y
 (
s
ee S
ect
i
o
n
 3.3.2).
DOE-ARM. w
w
w
.
a
r
c
hi
ve
.
a
r
m
.
g
o
v
/d
a
t
a
/o
r
d
e
r
ing.
ht
ml
In
tern
et
do
w
n
l
o
ad
EF4: Ma
y
,
 J
u
n
e
,
Ju
ly
, 1997;
 EF
7:
J
une
,
 J
u
l
y
,
 
A
u
gu
st,
S
e
pt
e
m
ber, 1997.
Strea
m
f
low
1
5
 m
i
nu
te
 flo
w
 d
a
ta
 fo
r
 U
S
G
S
 ga
gi
ng
s
t
ation
 on
 
L
i
t
tle Was
h
i
t
a R
.
 Eas
t
 of
 Ea
s
t
Ni
nn
ek
a
h
, OK, 073275
50.
Ok
la
h
o
m
a
 U
S
GS
P
e
rson
al
requ
es
t
Oct
ober 19
95 
?
Oct
ober 19
97.
Observed Soil Moi
s
tu
r
e
DOE ARM SGP
-
C
A
RT
: 30 m
i
n
u
t
e 
Soil
W
a
ter an
d 
T
e
m
p
eratu
r
e S
y
s
t
em
s
(SW
A
T
S
) d
a
ta strea
m
; 
m
a
tric p
o
t
ential
dat
a
 f
r
o
m
 s
i
t
e
s
 EF
4 an
d EF
7 u
s
ed 
f
o
r
ev
aporation
 stu
d
y
 (see Sectio
n
 3.3.2).
DOE-ARM. w
w
w
.
a
r
c
hi
ve
.
a
r
m
.
g
o
v
/d
a
t
a
/o
r
d
e
r
ing.
ht
ml
In
tern
et
do
w
n
l
o
ad
EF4: Ma
y
,
 J
u
n
e
,
Ju
ly
, 1997;
 EF
7:
J
une
,
 J
u
l
y
,
 
A
u
gu
st,
S
e
pt
e
m
ber, 1997.
S
o
u
t
h
e
rn
 Great
 Pl
ai
n
s
 H
y
drol
ogy
Ex
perim
e
n
t
 1997, S
G
P ?
97.
  
T
w
i
ce dai
l
y
g
r
av
i
m
etric 
soil m
o
ist
u
re m
eas
u
r
e
m
e
n
ts.
USD
A
/N
AS
A
,
daac.g
s
f
c
.
n
asa.
g
o
v
/
C
A
MP
A
I
GN_DO
CS/
S
GP
97/g
r
av
i_
m
o
ist.
ht
m
l
In
tern
et
do
w
n
l
o
ad
Ju
n
e
 18 ? Ju
l
y
 16,
1997.
31
32
In order to use spatial data from different sources in conjunction with one
another, each data layer must be transformed into a common map projection.  The
map projection used in this study is the same map projection used in USGS Open-
File Report 95-727 for watershed and elevation maps of Oklahoma.  The parameters
for this projection are given in Table 3.2.
Table 3.2  Projection parameters
Projection Albers Equal Area Conic
Datum NAD83
Units Meters
Spheroid GRS1980
Xshift 0
Yshift 0
1st standard parallel 34.0
2nd standard parallel 36.5
Central meridian -98.0
Latitude of projection's origin 33.0
False easting (meters) 0
False northing (meters) 0
3.2.1. Topographic data
Digital elevation models derived from map sources at two different scales are
used in this study.  A 3-arc-second digital elevation model was obtained from the
U.S. Geological Survey.  The horizontal and vertical spacing between elevation
values in this data set is 3-arc-seconds, hence the name.  3 arc-second spacing is
equivalent to 92 meters at the Equator.  The earth distance between elevation values
along parallels of latitude decreases with increasing latitude, but the earth distance
between elevation values along meridians of longitude changes very little.  The 3-
arc-second data are projected and resampled to a 100-m grid for use in this study.
33
The map scale associated with these 3-arc-second data is 1:250,000 because the data
were derived from topographic information on maps of this scale.
A 30-meter digital elevation model was obtained from the Agricultural
Research Service Grazinglands Research Laboratory in El Reno, Oklahoma.  The
map scale associated with the 30-m digital elevation model is 1:24,000.
The area draining to the streamflow gage of interest, USGS 07327550 East of
Ninnekah, Oklahoma was delineated using each of the two digital elevation models.
Procedures for watershed delineation using Arc/Info and ArcView are reviewed in
the Watershed Characterization Module of GISHydro ?98 (CRWR-ESRI, 1998).
The watershed delineation method used in these programs requires that an integer
flow direction value be assigned to each grid cell, indicating to which of its eight
neighboring cells water will flow.  Flow from a given cell must go to one and only
one of its neighboring cells.  This flow direction model is sometimes referred to as
the D8 or eight-direction pour-point model.  All cells draining to a selected outlet
cell can be determined using the flow direction grid, thus defining the watershed for
that outlet.  Also using the grid of flow directions, the path from each cell to the
watershed outlet cell can be traced, allowing the approximate flow length to the
watershed outlet to be determined for each grid cell.  Diagrams illustrating grid-
based flow direction and flow length calculations are provided in the online
documentation for Arc/Info and ArcView software.
Watersheds delineated from the 100-m and 30-m elevation grids are illustrated
in Figure 3.2.  The areas delineated using the two different terrain models are 597.0
km
2
 (59,700 cells) for the 100-m grid and 600.1 km
2
 (666,778 cells) for the 30-m
grid.  The flow path traced from a selected cell (A) in the watershed to the outlet (B)
is also shown in Figure 3.2.  The average flow length from this point to the
watershed outlet is 31,155 m using the 100-m grid and 31,444 m using the 30-m
grid.  A map showing the flow length from each cell in the 30-m grid to the
34
watershed outlet is shown in Figure 3.3.  The way in which flow length values are
used in runoff transform calculations is discussed in Section 3.3.5.
Elevation (m)
320 - 340
340 - 360
360 - 380
380 - 400
400 - 420
420 - 440
440 - 500
  
 
USGS Gaging Station 
07327550
100 m DEM
30 m DEM
A
B
A
B
 
AB length = 31,444 m
AB length  = 31,154 m
Figure 3.2  Watersheds delineated from 100-m and 30-m digital elevation
models respectively.
35
Figure 3.3  Flow length to the watershed outlet for the 30 m digital elevation model.
3.2.2. Soils Data
3.2.2.1. STATSGO vs. SSURGO
Two soils databases maintained by the United States Department of Agriculture
are of interest in this study, the Soil Survey Geographic (SSURGO) database and
the State Soil Geographic (STATSGO) database.  The SSURGO database contains a
greater level of spatial detail than STATSGO.  SSURGO maps are created using
field methods at scales ranging from 1:12,000 to 1:63,360, and are designed
primarily for landowner, township, and county level natural resources planning and
management (USDA, 1995).  STATSGO maps are created either by generalizing
more detailed soil maps such as SSURGO or by assimilating data on geology,
topography, vegetation, and climate and deducing the probable classification and
extent of various soil types.  With the exception of Alaska, the STATSGO data are
mapped at 1:250,000 scale.  STATSGO data are designed primarily for regional,
river basin, state, and multi-state level planning (USDA, 1994).
Both STATSGO and SSURGO maps are divided into map units.  ?A map unit
is a collection of areas defined and named the same in terms of their soil
components or miscellaneous areas or both.? (USDA, 1993)  A soil component is
typically associated with a single soil series phase (USDA, 1994).  A soil series
36
consists of soils with similar properties.  Miscellaneous areas are areas that have
little or no soil such as a rock outcrop.  Rock outcrop areas are listed as separate
components in the STATSGO attribute tables.
The attribute database associated with STATSGO and SSURGO includes a
description of the physical and chemical properties of 18,000 soil series in the
United States.  Soil series are differentiated primarily by the kind and character of
soil properties and the arrangement of horizons within the profile.  Soil series are
the most homogeneous classification category used for soils in the United States
(USDA, 1993).  Since map components are commonly associated with a single soil
series phase, map components are also relatively homogeneous.  However, even
map units that only contain a single component are likely to contain some
inclusions.  Map units are typically designed so that inclusions for which use and
management might differ constitute less than 15% of the map unit area.
There is a distinct difference between the amount of spatial detail provided in
the STATSGO and SSURGO databases.  STATSGO map units may consist of
anywhere from 1 to 21 components for which the spatial delineations are not shown,
while SSURGO map units most commonly contain only 1 component but may
contain up to 3 components (for which the spatial delineations are not shown).
STATSGO map units may consist of one or more non-contiguous polygons while
map units in SSURGO consist of a single polygon.  The location of individual
components within map unit polygons is not shown on the map.
Currently, STATSGO data files for any of the 50 United States are readily
available from the USDA Natural Resource Conservation Service (NRCS).  The
availability of SSURGO data is much more limited.  Information about the current
status of SSURGO data is provided on the NRCS World Wide Web page for soils
(www.ftw.nrcs.usda.gov/soils_data.html).  As of July 1, 1998, certified SSURGO
data files were only available for 2 counties in Oklahoma and 17 counties in Texas.
In some areas, the availability of SSURGO data is increasing.  In the past year, the
37
number of counties for which data is available in Texas increased from 5 to 17.  The
USDA Natural Resources Conservation Service (NRCS) is proposing to digitize 150
county soil surveys in Texas and share costs through the Strategic Mapping
Initiative (www.tnris.state.tx.us/stratmap.html).  One goal of this research is to
assess the value of using the more detailed SSURGO data for hydrologic modeling.
If the SSURGO level of detail is shown to be much more valuable then availability
becomes an issue.  Although certified SSURGO maps are not available for the three
counties intersecting the Little Washita watershed (Caddo, Grady, and Comanche),
a 30-m soils grid for the watershed has been obtained from the Agricultural
Research Service Grazinglands Research Laboratory.  The spatial detail in this grid
is equivalent to that in SSURGO maps, and the data tables associated with this grid
are very similar to those used for STATSGO/SSURGO, although the differences are
significant enough to require separate GIS pre-processing routines.
Figure 3.4 illustrates the differences in spatial detail provided by the
STATSGO and the 30-m soils grid.  The soil types listed in the legend correspond to
the predominant soil surface texture.  The shading scheme in Figure 3.4 is selected
so that darker shades represent soils with a tendency to produce more runoff.  The
location of different soil components within a STATSGO mapunit is not known.
For example, there are 14 components associated with Mapunit OK103.  The
percentage of each component in OK103 is known and the component surface
texture is also known.  Surface texture names reported in the database tables are
reclassified into the 12 basic USDA soil texture classes as explained in Section
3.2.2.2.  Using this information, it was determined that the surface texture at any
point in OK103 has a 56% likelihood of being a Loam, a 30% likelihood of being a
Silt Loam, and a 14% likelihood of being a Sandy Loam.  The portion of
STATSGO Mapunit OK103 shown in Figure 3.4 is shaded as if it were a Loam
because this is the predominant soil type.  These percentages do not guarantee,
however, that any soil with a Silt Loam (for example) surface texture lies within the
38
Little Washita portion of Mapunit OK103 or that the surface texture in the Little
Washita portion of Mapunit OK103 is not entirely Silt Loam.  The Little Washita
portion of Mapunit OK103, in fact, constitutes only 10% of the total area for
Mapunit OK103.  Figure 3.5 shows the location of all polygons in Mapunit OK103
polygons relative to the Little Washita watershed.
The percentage of the Little Washita watershed with surface textures specified
as each of the 12 basic USDA soil textures was computed using both STATSGO
and the 30-m soils grid.  The results of this calculation are shown in Figure 3.6a.
For calculations with the 30-m soils grid, some grid cells may have two possible
surface textures.  The predominant texture was used in calculations for Figure 3.6a.
Nearly all of the surface soils in the Little Washita fall into one of four categories ?
loamy sand, sandy loam, loam, or silt loam.  At first glance, there appear to be
significant differences in the surface texture compositions estimated from the two
different map scales.  However, the hydraulic properties of loamy sand and sandy
loam are similar and the hydraulic properties of loam and silt loam are similar as
discussed in Section 3.2.2.2.3.  If these two sets of soil types are grouped together,
then the surface texture compositions predicted by the two soil maps are very
similar (Figure 3.6b).  This observation may explain the small differences in
simulation results obtained when the two different data sets are used (See Chapter
5).
So
i
l
 T
y
p
e
s
S LS SL L
SI
L
CL SI
C
L
Othe
r
(
wa
ter
,
 roc
k, e
t
c
.
)
Ma
punit OK103 
(
1
4
c
o
mpone
nts)
    56%
 Loa
m
    30% S
ilt L
o
am
    14%
 S
a
ndy Loa
m
STATSGO
3
0
 m s
o
i
l
s
 grid
F
i
gure 3.4  STATSG
O and 30-m soils grid.
39
40
Mapunit OK103
Watershed Boundary
Figure 3.5 Mapunit OK103 polygons.
41
0
10
20
30
40
50
60
S LS SL SCL L SIL CL SI C SC SICL SIC O
USDA Soil Types
Percentag
e
STATSGO 30 m Grid
(a)
0
10
20
30
40
50
60
S LS/SL SCL L/SIL CL SI C SC SICL SIC O
USDA Soil Types
Percentage
STATSGO 30 m Grid
(b)
Figure 3.6  Comparison of surface texture composition for the STATSGO soil
map and the 30-m soil grid with (a) soils grouped into the 12 USDA texture
classes and (b) loamy sand and sandy loam grouped together along with loam
and silt loam.
42
3.2.2.2. Estimation of soil hydraulic properties
3.2.2.2.1. Soil attribute tables
Many soil attributes are stored in tables associated with the STATSGO and
SSURGO soil maps.  STATSGO tables are described here.  The tables associated
with the 30-m soils grid contain most of the same information.  The attribute tables
of interest here are the mapunit table, the comp (component) table, and the layer
table.  The mapunit table holds the key to relating soil attributes to polygons on the
map.  One record exists in the mapunit table for each mapunit.  A single record in
the mapunit table relates to one or more components in the comp table.  In
STATSGO there can be up to 21 components in a mapunit while SSURGO
mapunits typically contain only 1 component but may contain up to 3.  A single
record in the comp table relates to 1 to 6 records in the layer table.  The
relationships among these three tables is illustrated in Figure 3.7 for mapunit
OK103.  The total number of attributes available in the comp table is 53 and the
total number of attributes available in the layer table is 55.  Several attributes in the
comp and layer tables are relevant to hydrology but interpretation and use of these
values for modeling is not straightforward.
43
14 components associated
with mapunit OK103
5 layers associated with
component 1 of mapunit
OK103
mapunit OK103
Figure 3.7  Relationships among the mapunit, comp, and layer table.
The following soil properties must be estimated in order to use the infiltration,
evaporation, and percolation equations that are described in Section 3.3:  saturated
hydraulic conductivity (K
s
), porosity (?), residual water content (?
r
), the Brooks and
Corey water retention parameters (? and h
b
) defined by Equation 3.1, field capacity
(?
fc
), and wilting point (?
wp
).  ?
fc
 and ?
wp
 do not need to be prescribed independently
because they can be computed from ? and h
b
.  In Equation 3.1, ?
d
 is the effective
saturation associated with soil drying and h is the absolute value of suction head
[cm].  Equation 3.2 shows a similar equation that is used to describe the water
44
retention function in the wetting phase.  Differences in soil behavior during the
wetting and drying phases is referred to as hysteresis (Jury, Gardner, and Gardner,
1991).  The distinction between wetting and drying phases becomes important in the
derivation of the equations to estimate wetting front suction head given in Section
3.3.1.2.  In Equation 3.2, h
we
 [cm] is the water entry head.  Although h
b
 in Equation
3.1 is sometimes referred to as the air entry head, use of the symbol h
b
 (the subscript
b stands for bubbling pressure head) is commonly seen in the literature and will be
used here.  Based on the experimental work of Bouwer (1966), h
we
 is commonly
taken as half of h
b
.  An illustration of the wetting and drying water retention curves
for a loam soil is given in Figure 3.8.
Equation 3.1
?
??
??
?
?
?
?
?
?
=
?
?
=?
h
h
b
r
r
d
Equation 3.2
?
??
??
?
?
?
?
?
?
=
?
?
=?
h
h
we
r
r
w
45
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1 1.2
?
cm of wate
r
Drying
Wetting
h
we
h
b
Figure 3.8. Example wetting and drying water retention curves for a loam soil
with h
b
 = 27.8 and ? = 0.56.
The model equations described in Section 3.3 have been selected in part
because a rational method for estimating the required parameters is available.
3.2.2.2.2. The lookup table approach
A good deal of attention has been paid in the literature to the estimation of
hydraulic properties from the information in soils databases.  McCuen, Rawls, and
Brakensiek (1981) concluded that Brooks and Corey water retention parameters and
the Green-Ampt infiltration parameters vary systematically across the USDA soil
texture classes, thus suggesting that representative hydraulic properties for each of
the 12 USDA soil texture classes might be used in hydrologic models when only the
46
soil texture is known.  To arrive at this conclusion, McCuen, Rawls, and Brakensiek
(1981) analyzed 1085 water retention data sets and used statistical analysis to show
that the Brooks and Corey and Green-Ampt infiltration parameters are statistically
different among soil classes.  Rawls, Brakensiek, and Saxton (1982) performed a
similar analysis on a larger data set data and presented a table with means and
standard deviations for ?, ?
r
, ?, h
b
, ?
fc
, and ?
wp
 within 11 of the 12 basic USDA
textural classes.  Clapp and Hornberger (1978) had also performed a similar analysis
on a subset of the data used by Rawls, Brakensiek, and Saxton (1982).  Rawls,
Brakensiek, and Saxton (1982) also ?visually? estimated representative values of
saturated hydraulic conductivity (K
s
) for each of the 11 texture classes from plots of
experimental data but emphasize that other characteristics besides texture will
influence K
s
.
Rawls and Brakensiek (1985) developed regression equations that can be used
to estimate ?
r
, h
b
, ?, and K
s
 as a function of percent clay, percent sand, and porosity.
Using a large soils database with information on bulk density, sand, and clay
contents for thousands of soils, Carsel and Parrish (1988) used the regression
equations of Rawls and Brakensiek (1985) to produce empirical distributions for K
s
,
?
r
, and the van Genuchten water retention parameters (? and N) defined by
Equation 3.3.
Equation 3.3
)/11(
])(1[
)(
NN
r
r
h
?
+
?
+=
?
??
??
Subsequently, Carsel and Parrish (1988) used statistical techniques to determine the
means, variances, and covariances for each of these parameters (K
s
, ?
r
, ?, ?) within
each of the 12 USDA texture classes.  Carsel and Parrish (1988) used the
relationships presented in Equation 3.4 and Equation 3.5 in order to use the Rawls
and Brakensiek (1985) regression equations for ? and h
b
 to predict ? and N.  Thus,
47
the representative values presented by Carsel and Parrish for ? and N can precisely
be converted back to ? and h
b 
using these same equations.
Equation 3.4 1?= N?
Equation 3.5
?
1
=
b
h
The Brooks and Corey (? and h
b
) and van Genuchten (? and N) relationships
become identical at large capillary heads when Equation 3.4 and Equation 3.5 hold
true.
Considering this previous work, two general approaches may be considered for
estimating  the parameters K
s
, ?, ?
r
, ?, h
b
, ?
fc
, and ?
wp
 from the information in
STATSGO or SSURGO data tables, a regression approach or a lookup table
approach.  A third alternative might be considered for estimating K
s
 and ? alone.
The first, more complex, approach would be to try to use layer table attributes
describing the percent of soil passing through American Society for Testing
Materials sieves (Attribute names are No10l, No10h, No200l, and No200h, where
No10l is the low estimate for the fraction of soil passing the number 10 sieve and
No10h is the high estimate and so forth.) in conjunction with information about
percent clay (clayl and clayh) to estimate the percent sand and percent clay.  Using
this information along with bulk density (bdh and bdl) to estimate porosity (?), the
regression equations of Rawls and Brakensiek (1985) could be applied to estimate
K
s
, ?
r
, ?, and h
b
.  There are at least two practical difficulties in using this approach.
In Oklahoma, some inconsistencies are found between the sieve values and the
percent clay estimates if high and low attribute values are averaged.  For example,
in some layers the average percent material passing sieve No. 200 is greater than the
average percent clay, a condition that should not be possible.  Miller and White
(1998) note similar difficulties in estimating percent, sand, silt, and clay in
Pennsylvania.  A second difficulty with using the regression approach is that some
48
non-water, non-bedrock layers in the Oklahoma layer table have zero values for
bulk density, preventing direct estimates of porosity for these layers.  Miller and
White (1998) also note this problem in Pennsylvania.
A simpler parameter estimation approach adopted here is to estimate K
s
, ?
r
, ?,
and h
b
 as a function of USDA texture class by using a lookup table.  Field capacity
(?
fc
) and wilting point (?
wp
) values are estimated using the lookup table values for ?
and h
b
 and Equation 3.1.  The matric potential corresponding to field capacity is
taken as -10 kPa for coarse textured soils (S, LS, SL) and ?20 kPa for medium to
fine textured soils (SCL, L, SIL, CL, SI, C, SC, SICL, SIC) as recommended in
ASCE (1990).
Although all of the required soil parameters can be derived using the texture
lookup table approach advocated here, one might argue that layer attributes other
than texture should be considered to estimate K
s
 and ? more directly.  The reasons
for not doing this are discussed briefly here.
The bulk density (bdh, bdl) attribute might be used to estimate porosity more
directly; however, as stated previously, non-water, non-bedrock layers with zero
values for bulk density cause problems with this approach.
The ?permeability? attribute (with units of inches/hour) might be used to
estimate K
s
, but a study of the physical meaning and derivation of permeability
attribute in STATSGO and SSURGO has led the author to conclude that this
approach would be no better than the lookup table approach and more complex to
implement.  The USDA Soil Survey Manual (1951) describes two methods for
determining soil permeability.  One method is a laboratory test to determine the
drainage through a saturated soil core sample subject to a constant ponded head and
the other is to assign a permeability class to a soil based on structure, texture,
porosity, cracking and other horizon characteristics.  The USDA Soil Survey
Manual (1993) describes a similar system for assigning a saturated hydraulic
conductivity class to soils based on texture and bulk density.  The term saturated
49
hydraulic conductivity in the 1993 manual is used in lieu of permeability that was
used in the 1951 manual.  Measured values of saturated hydraulic conductivity can
easily vary by an order of magnitude within a soil series or even between soil
samples taken within centimeters of one another (USDA, 1993).  Different
measurement techniques may produce drastically different results; however, field
measurement techniques are generally preferred over laboratory techniques (USDA,
1993).  Of the model parameters (?, ?
r
, K
s
, ?, h
b
) used in this study, K
s
 has by far
the largest coefficient of variation based on the analysis of Carsel and Parrish
(1988).  This is a major reason why a hydraulic conductivity factor is chosen for
model calibration (see Chapter 5).
Upon examination of the STATSGO layer data for Oklahoma, it seems likely
that reported saturated hydraulic conductivity values are based on a general
classification system like that described in the USDA Soil Survey Manual (1951).
Typically only ranges of values like 0-0.06 inches/hour, 0.2-0.6 inches/hour, 0.6?2.0
inches/hour, 2-6 inches/hour, 6-20 inches/hour etc., are given in STATSGO.  The
ranges prescribed by the 1951 manual are 0-0.05 inches/hour, 0.05-0.2 inches/hour,
0.2-0.8 inches/hour, 0.8-2.5 inches/hour, 2.5-5 inches/hour, 5-10 inches/hour, and >
10 inches/hour.  2927 out of 7597 (39%) of the soil layers in Oklahoma have the
permeability range 0.6-2.0 inches/hour.  At least 6 different USDA texture classes
have permeability specified in this range.  Thus, other factors besides texture are
influencing the permeability class, perhaps bulk density or layering.  Using both
bulk density and texture to estimate saturated hydraulic conductivity would be better
than using texture alone; however, the additional complications introduced by using
this approach seem to outweigh the benefits at this time, particularly since hydraulic
conductivity will be used in model calibration and there are missing data values in
STATSGO for bulk density.  Carsel and Parrish (1988) showed that there is a
significant statistical correlation between samples of K
s
 and the water retention
50
parameters within each texture class.  Therefore, an additional advantage of using a
lookup table alone is that all parameter values are mutually consistent.
One last argument in favor of using texture as the only distinguishing factor
among components is that this significantly simplifies model data management, at
least for the one-layer model discussed in the next section.  When only a single
distinguishing factor is used, components with a common texture can be lumped
together and the spatial soils information passed from the GIS to the hydrology
model is set at 13 different specifications (12 USDA texture classifications and an
?other? category).  If multiple distinguishing factors are used, then components with
a common surface texture may differ in other characteristics so that there is a need
to pass more information about individual components to the hydrology model.
In application, the texture names specified in the STATSGO database cannot be
passed directly to a texture-parameter lookup table because the letter codes used to
specify soil texture may include qualifiers or letters used in lieu of texture.  BY-SL
(bouldery sandy loam) and CB-SL (cobbly sandy loam) are examples of texture
codes with qualifiers.  PEAT and UWB (unweathered bedrock) are examples of
codes used in lieu of texture.  Miller and White (1998) compiled a lookup table that
reclassifies all 719 possible texture name combinations listed in STATSGO into one
of the 12 basic USDA classes or the other category.  A slightly modified version of
the Miller and White reclassification table is used here.  With two exceptions, the
reclassifications specified in the Miller and White table make sense.  In the original
table, LFS (loamy fine sand) and LCOS (loamy coarse sand) were reclassified as S
(sand).  These reclassifications were changed to LS (loamy sand) for the current
application because this seemed more intuitive.  Due to its length, presentation of
this lookup table is left to the Appendix D.
51
3.2.2.2.3. Selecting representative parameter values for texture classes
Both Rawls, Brakensiek, and Saxton (1982) and Carsel and Parrish (1988)
present tables with representative values for K
s
, ?, ?
r
, ?, and h
b
 in the basic USDA
texture classes.  Rawls, Brakensiek, and Saxton and Carsel and Parrish used
information from thousands of soil samples to derive their respective tabulated
parameter values; however, their approaches used to derive parameter values
differed and so do their tabulated results.
Rawls, Brakensiek, and Saxton visually estimated values of K
s
 for each texture
class by plotting unsaturated hydraulic conductivity data from the literature, but
noted that other factors besides texture will influence K
s
.  Both Rawls, Brakensiek,
and Saxton and Carsel and Parrish report means and standard deviations for porosity
(?) based on sample values obtained directly from soil survey reports.  To estimate
values of ?
r
, ?, and h
b
, Rawls, Brakensiek, and Saxton plotted water retention-matric
potential data from the literature and used optimization to determine the parameter
values (?
r
, ?, h
b
) that best fit each data set.  Rawls, Brakensiek, and Saxton report
the mean values for these fitted parameters and values corresponding to plus and
minus one standard deviation.  In contrast, Carsel and Parrish generated a sampling
of values for K
s
, ?
r
, ?, and ? in each texture class by using regression equations
developed by Rawls and Brakensiek (1985).  Carsel and Parrish used Equations 3.3
and 3.4 to compute ? and N from regression equations for ? and h
b
.  The inputs to
these regression equations are percent clay, percent sand, and porosity (?) (values
obtained from Soil Conservation Service Soil Survey Information Reports).  Carsel
and Parrish statistically analyzed the K
s
, ?
r
, ?, and N samples in each texture class
and reported means and standard deviations as well as the correlations among the
parameters in each texture class.
A tabulation of the parameter means (?) and coefficients of variation (CV)
from Rawls, Brakensiek, and Saxton and Carsel and Parrish is presented in Table
3.3.  Rawls, Brakensiek, and Saxton do not report the coefficients of variation but
52
do report values corresponding to one standard deviation above and below the mean
(x
+
, x
-
).  Approximate coefficients of variation (CV*) from Rawls, Brakensiek, and
Saxton are computed using the following equation.
Equation 3.6 100*
2/)(
*
?
?+
?
=
xx
CV
Table 3.3  Mean values and coefficients of variation for hydraulic parameters
in the 12 basic USDA texture classes.  A comparison of values reported by
Rawls, Brakensiek, and Saxton (1982) and Carsel and Parrish (1988).
Saturated Hydraulic Conductivity                         Porosity
R,B, & S C & P R, B, & S C & P
K
s
K
s ? ?
??
CV(%)
?
CV*(%)
?
CV(%)
S 23.6 29.7 52 0.44 14 0.43 15
LS 5.98 14.6 78 0.44 16 0.41 22
SL 2.18 4.42 127 0.45 23 0.41 21
SCL 0.3 1.31 209 0.40 17 0.39 18
L 1.32 1.04 175 0.46 19 0.43 22
SIL 0.68 0.45 275 0.50 16 0.45 19
CL 0.2 0.26 267 0.46 12 0.41 22
SI 0.25 130 0.46 17
C 0.06 0.2 210 0.48 10 0.38 24
SC 0.12 0.12 234 0.43 14 0.38 14
SICL 0.2 0.07 289 0.47 11 0.43 17
SIC 0.1 0.02 453 0.48 11 0.36 20
53
Residual Water Content                                     Lambda
R,B, & S C & P R, B, & S C & P
?
r
CV*(%) ?
r
? ?
??
CV(%)
?
CV*(%)
?
CV(%)
S 0.02 95 0.045 22 0.69 57 1.68 20
LS 0.035 91 0.057 26 0.55 58 1.28 12
SL 0.041 129 0.065 27 0.38 63 0.89 9
SCL 0.068 101 0.1 6 0.32 75 0.48 9
L 0.027 137 0.078 17 0.25 66 0.56 7
SIL 0.015 193 0.067 22 0.23 55 0.41 9
CL 0.075 116 0.095 10 0.24 71 0.31 7
SI 0.015 0.034 30 0.37 3
C 0.09 108 0.068 50 0.17 78 0.09 8
SC 0.109 88 0.1 13 0.22 78 0.23 8
SICL 0.04 148 0.089 11 0.18 78 0.23 5
SIC 0.056 121 0.07 34 0.15 73 0.09 5
                                    h
b
R, B, & S C&P
h
b
h
b
?
CV*(%)
?
CV(%)
S 7.3 257 6.9 20
LS 8.7 230 8.1 35
SL 14.7 200 13.3 49
SCL 28.1 242 17.0 65
L 11.2 335 27.8 57
SIL 20.8 281 50.0 65
CL 25.9 212 52.6 78
SI 62.5 45
C 37.3 241 125.0 160
SC 29.2 286 37.0 62
SICL 32.6 233 100.0 62
SIC 34.2 233 200.0 114
* Computed using Equation 3.6.
With the exception of porosity (?), the Rawls, Brakensiek, and Saxton
parameter estimates tend to have much higher coefficients of variation than those of
Carsel and Parrish.  K
s
 has by far the highest variability among the parameter values
reported by Carsel and Parrish.  Rawls, Brakensiek, and Saxton do not report
54
uncertainty for K
s
.  The h
b
 values have the highest coefficients of variation among
the parameter estimates tabulated by Rawls, Brakensiek, and Saxton.  The most
striking differences in the mean values from the two data sets occurs in the ? values
for coarse textured soils and the h
b
 values for fine soils.
Plots of the mean values of ?, ?
r
, ?, and h
b
 with one standard deviation error
bars have been prepared to provide a visual basis for comparing the two sets of
values.  These plots are shown in Figures 3.9 - 3.12.  In these figures, textures are
listed in order of decreasing saturated hydraulic conductivity from top to bottom.
Preparation of the Rawls, Brakensiek, and Saxton data series on these plots is
straightforward because their paper reports parameter values corresponding to one
standard deviation above and below the mean.  From the Carsel and Parrish paper,
the parameter values corresponding to these points are not explicitly reported and
must be computed for the skewed distributions.  Statistical parameter values
reported in Table 6 of Carsel and Parrish for the transformed ?
r
, ?, and N values
were untransformed to create the error bars in these figures.
The Carsel and Parrish estimates for ? tend to be lower than those of Rawls,
Brakensiek, and Saxton but the differences between mean ? values tend to be no
more than one standard deviation, with the exception of silty clay (SIC).  Mean ?
r
values from Carsel and Parrish tend to be higher than mean values from Rawls,
Brakensiek, and Saxton.  For ?, there seems to be an increasing discrepancy in the
mean parameter values from the two data sets when moving from fine to coarse
textured soils.  The error bars do not even overlap in Figure 3.11 for loam, sandy
loam, loamy sand, and sand soils.  For h
b
, discrepancies in the mean values increase
when moving from fine textured soils to coarse textured soils.  The error bars for h
b
still overlap for coarse textured soils because of the large uncertainty in the h
b
values reported by Rawls, Brakensiek, and Saxton.
Another way of visualizing the variations in parameter values across the
different texture classes is to use the USDA textural triangle.  Figures 3.13 and 3.14
55
show h
b
 values printed on the USDA textural triangle.  Looking at the textural
triangle, one would intuitively expect bubbling pressure head to increase with
decreasing sand content (along the line from point 1 to point 2) and increase with
increasing clay content (along the line from point 1 to point 3).  This trend is seen in
both Figures 3.13 and 3.14 with one notable exception, the Rawls, Brakensiek, and
Saxton h
b
 value for the loam soil appears to be too low.  If figures similar to 3.13
and 3.14 are created for the ? values, all parameter values are consistent with the
expected trends for both data sets.
0.1 0.2 0.3 0.4 0.5 0.6 0.7
? 
SICL
SIC
C
SC
S
LS
SL
L
SIL
CL
SCL
SI
Values from Carsel and Parrish (1988)
Values from Rawls, Brakensiek, and
Saxton (1982)
Error bars denote +/- 1 standard deviation.
Figure 3.9  Mean porosity values for USDA texture classes with 1 standard
deviation error bars.
56
0 0.1 0.2 0.3
?r 
SICL
SIC
C
SC
S
LS
SL
L
SIL
CL
SCL
SI
Carsel and Parrish (1988)
Error bar end points are determined by 
back-transforming values corresponding 
to +/- 1 standard deviation above and below
the mean in a transformed space.
Rawls, Brakensiek, and Saxton (1982)
Error bars correspond to +/- 1 standard
deviation values reported by Rawls, 
Brakensiek, and Saxton (1982).
Figure 3.10  Mean residual saturation values for USDA texture classes with 1
standard deviation error bars.
57
00.511.52
? 
SICL
SIC
C
SC
S
LS
SL
L
SIL
CL
SCL
SI
Carsel and Parrish (1988)
Error bar end points are determined by 
back-transforming values corresponding 
to +/- 1 standard deviation above and below
the mean in a transformed space.
Rawls, Brakensiek, and Saxton (1982)
Error bars correspond to +/- 1 standard
deviation values reported by Rawls, 
Brakensiek, and Saxton (1982).
Figure 3.11  Mean ? values for USDA texture classes with 1 standard deviation
error bars.
58
0 50 100 150 200
h
b
 (cm)
SICL
SIC
C
SC
S
LS
SL
L
SIL
CL
SCL
SI
Carsel and Parrish (1988)
Error bar end points are determined by 
back-transforming values corresponding 
to +/- 1 standard deviation above and below
the mean in a transformed space.
Rawls, Brakensiek, and Saxton (1982)
Error bars correspond to +/- 1 standard
deviation values reported by Rawls, 
Brakensiek, and Saxton (1982).
Figure 3.12  Mean h
b
 values for USDA texture classes with 1 standard
deviation error bars.
59
% sand
% s
i
lt
%
 clay
S
LS
SL
L
SIL
SI
SICLCL
SIC
C
SC
SCL
7.3
8.7
14.7
20.8
11.2
28.1
29.2
25.9
32.6
34.2
37.3
1
2
3
20406080
20
40
60
80
20
40
60
80
Figure 3.13  Rawls, Brakensiek, and Saxton h
b
 values on the USDA textural
triangle.
60
% sand
% s
i
lt
%
 clay
S
LS
SL
L
SIL
SI
SICLCL
SIC
C
SC
SCL
7.3
8.7
14.7
20.8
11.2
28.1
29.2
25.9
32.6
34.2
37.3
1
2
3
20406080
20
40
60
80
20
40
60
80
Figure 3.14  Carsel and Parrish h
b
 values on the USDA textural triangle.
A method for computing Green-Ampt wetting front suction head (h
f
) values
from the Brooks and Corey water retention parameters (?, h
b
) is discussed in detail
in Section 3.3.1.2.  For discussion here, the Green-Ampt wetting front suction head
parameter values computed from the Rawls, Brakensiek, and Saxton mean ? and h
b
values and the Carsel and Parrish mean ? and h
b
 values for an initially dry soil are
given in Table 3.4.  The wetting front suction head values derived from the Carsel
and Parrish ? and h
b
 parameter values are much more consistent with values of
wetting front suction head shown graphically in Figure 5.5.4 of Rawls et al. (1993).
For this reason, and the fact that there is an internal inconsistency in the h
b
 value for
loam soil shown in Figure 3.13, the Carsel and Parrish mean parameter values are
used for simulations described in Chapter 5.  To get a sense of the impact of
61
parameter uncertainty on simulation results, a simple sensitivity analysis is included
in Section 5.6.
Table 3.4  Estimates for Green-Ampt wetting front suction head (h
f
) for an
initially dry soil.
Texture h
f
 from Rawls,
Brakensiek, and
Saxton (cm)
h
f
 from Carsel
and Parrish (cm)
S 4.8 4.0
LS 6.0 4.9
SL 11 8.5
SCL 22 12
L 8.7 19
SIL 16 36
CL 20 40
SI 46
C 31 112
SC 23 29
SICL 27 78
SIC 29 179
3.2.2.3. One-layer model versus a two-layer model
One of the questions to be addressed in this research is whether increased
vertical detail in representing soil water processes is beneficial, in addition to the
increased horizontal detail associated with moving from lumped to distributed
models.  Therefore, the tradeoffs between using a one-layer and a two-layer soil
model are considered.  It is hypothesized that partitioning beyond two layers would
introduce unnecessary complication into the model.  Computer codes have been
written in this research to use a one-layer or a two-layer approach.
The interpretation of soil component data for use in a one-layer model is
straightforward.  The texture for the one layer is simply taken from the surface
texture attribute (surftex) of the comp table and translated into soil hydraulic
parameters using the lookup tables described in the Section 3.2.2.2.  The top layer is
assumed to be homogeneous in terms of its hydraulic properties.  The top layer is
62
loosely defined as the depth of soil that reacts quickly to atmospheric forcing.  The
depth of the top layer is important conceptually because infiltrated water is averaged
over the depth of the top layer at the end of a rainstorm.  If the depth of the top layer
is too large, then a realistic degree of soil moisture variability is not simulated.  In
the model, each soil component is assigned the same top layer depth, and top layer
depth essentially becomes a model parameter.
Each component in either the STATSGO or SSURGO database corresponds to
anywhere from 1 to 6 layers in the layer table.  The percentage of soil components
with 1, 2, 3, 4, 5, or 6 layers in STATSGO for Oklahoma is illustrated in Figure
3.15a.  92% of the components have 4 or fewer layers and 71% of the components
have 3 or fewer layers.  When information about multiple layers exist, it is often the
case that adjacent layers correspond to the same basic USDA texture class,
effectively reducing the amount of vertical heterogeneity given the proposed
modeling scheme.  Differences in soil characteristics other than texture distinguish
these adjacent layers.  Considering adjacent layers of the same basic USDA texture
as one layer, the percentage of components with 1, 2, 3, 4, or 5 distinct layers is
illustrated in Figure 3.15b.  In this case, 93% of the Oklahoma STATSGO
components contain 3 or fewer layers and 64% of the components contain 2 or
fewer layers.  In light of this information, it does not seem that any vertical
discretization (based on texture) beyond 3 layers could be supported by the available
data and only a two-layer model could be supported in the majority of components.
63
2 (23%)
3 (46%)
4 (21%)
5 (7%)
6 (1%)
1 (2%)
% of components with
x layers or fewer
100 
92 
71 
25 
1 (22%)
2 (42%)
3(29%)
4(6%)
100 
93 
64
22
Basic STATSGO
layers
STATSGO Layers with distinct
soil texture
x layers (% of
components
with x layers)
(a) (b)
Figure 3.15  Percentage of soil components with 1, 2, 3, 4, 5, or 6 layers (a) with
multiple distinguishing characteristics and (b) with the 12 basic USDA soil
classes as the only distinguishing characteristic.
The idea in this study is to use both a one-layer and a two-layer model and
evaluate whether adding the complexity of a second layer is valuable.  The coding,
data management and calibration for a two-layer model is significantly more
complex than for a one-layer model, as discussed in Chapters 4 and 5; therefore,
adding a third layer was not considered, a decision which appears to be justified
based on modeling results.
The two-layer model developed in this study is only an approximation of the
layering that exists in the soils databases.  For implementation, the soil hydraulic
properties of the second model layer are taken as the properties associated with the
layer below the top layer with the lowest saturated hydraulic conductivity (K
s
).  This
64
approach is used because it is common that the most restrictive layer controls
percolation out of the soil zone (USDA, 1951).  As in the one-layer model, the depth
of the top layer in the two-layer model is treated as a model parameter.  The total
soil zone depth is specified based on additional information in the soil layer table as
described in the next paragraph, and the depth of the second layer is taken as the
difference between the total soil zone depth and the depth of the top layer.  The
second layer serves as a reservoir from which water may evaporate or percolate to
the groundwater reservoir.  This approach is sensible in terms of restricting the
percolation out of the soil zone based on the most impermeable layer, but does not
reflect the true thickness of the soil whose properties are assigned to the second
layer and may not be the most appropriate assumption in terms of estimating
evaporation from the second layer.  The emphasis here is on simplicity.
The total depth for the soil zone determines the depth of the second layer in
each soil component.  The depth of the second layer influences the water content of
the second layer which in turn influences the rate of percolation out the bottom of
the second layer and thus the recharge rate to groundwater.  Information in the soils
layer tables is used to estimate the total depth of the soil zone.  In the proposed
model, there are three pieces of information that may define the bottom of the soil
zone.  The total soil zone depth is taken as the minimum of (1) the deepest layer
depth reported, (2) the depth to a non-soil, low permeability layer if one exists, and
(3) the depth to an assumed water table if it exists.  The definitions of non-soil, low
permeability layers and the rationale for assuming a water table depth are discussed
in the next paragraph.  One motivating factor for estimating the total soil zone depth
is to determine whether shallow soil layers exist that may become saturated during
heavy rainfall.
The layer texture codes are used to identify any non-soil, low permeability
layers at depth.  In this study, the top depth (laydepl) for a layer with any of the
following codes is taken as the bottom of the soil zone: cemented (CEM), indurated
65
(IND), unweathered bedrock (UWB), and weathered bedrock (WB).  If limited by
such a layer, the water storage capacity and the evaporation parameters for the
second layer are determined by the properties of the soil above such a layer, but this
low permeability layer now limits the percolation.  These layers are not assumed to
be entirely impermeable because, if they are, simulations show that there will not be
enough recharge to the groundwater.  The model user specifies a constant
percolation rate for these layers.
Information about the depth to water table in STATSGO is implicit in the
?drainage? attribute of the comp table.  Drainage classes for soils defined by the
USDA are listed in Table 3.5.  In excessively (E), somewhat excessively (SE), and
well (W) drained soils the depth to the water table is greater than 6 feet and the
presence of a water table is not considered in the soil-water balance model.  For
other drainage classes, water table depths are assumed based on information in
Table 3.5.  Clearly the water table depths will vary with season, and the values in
Table 3.5 are only a first approximation.  In the Little Washita soils data obtained
for this study, the STATSGO comp table contains the drainage class attribute but the
comp table associated with the 30-m soils grid does not.  Based on the STATSGO
database, the most limiting drainage class that actually exists in the Little Washita is
SP (somewhat poorly) with an assumed water table depth of 2 ft.  In the model runs
described in Chapter 5, the SP drainage class does not influence direct runoff
because soils with this drainage class never become fully saturated.
66
Table 3.5  Description of soil drainage classes specified in the STATSGO comp
table.
Drainage Class Symbol Description* Assumed water
table depth.
Very poorly VP ?Soils are wet to the surface most of the
time.  Depth to water table is less than 1
foot, or is ponded.?
15.2 cm (0.5 ft)
Poorly P ?Soils may have a saturated zone, a layer
of low hydraulic conductivity or seepage.
Depth to water table is less than 1 foot.?
15.2 cm (0.5 ft)
Somewhat poorly SP ?Soils commonly have a layer with low
hydraulic conductivity, wet state high in
profile, etc.  Depth to water table is 1 to 3
feet.?
61 cm (2 ft)
Moderately well MW ?Soils have a layer of low hydraulic
conductivity, wet state high in the profile.
Depth to water table is 3 to 6 feet.?
137 cm (4.5 ft)
Well W ?Soils have intermediate water holding
capacity.  Depth to water table is more
than 6 feet.?
--
Somewhat
excessively
SE ?Soils have high hydraulic conductivity
and low water holding capacity.  Depth to
water table is more than 6 feet.?
--
Excessively E ?Soils have very high and high hydraulic
conductivity and low water holding
capacity.  Depth to water table is more
than 6 feet.?
--
* Descriptions are from USDA (1994).
A map showing the total depth of the soil zone (used for the two-layer model
only) derived from the 30-m soils grid is shown in Figure 3.16.  The total soil zone
depth in 47% of the watershed is defined by a non-soil, low permeability layer.
67
Figure 3.16  Total depth of the soil zone defined for the two-layer model.
Assessing the value of using the more complex two-layer approach is one goal
of this research.  There are two reasons why the added complexity of a two-layer
model is of interest.  By defining a lower limit on the soil zone, a limit is now
placed on the cumulative infiltration that may occur.  There is no limit on the
cumulative infiltration in the one-layer model.  If the infiltration depth predicted
using the Green-Ampt model exceeds the capacity of the top layer in the one-layer
model, then any excess infiltration is passed to the groundwater reservoir or treated
as a sink.  In contrast, if the predicted infiltration volume exceeds the capacity of the
soil zone in the two-layer model, any infiltration that is not accommodated adds to
68
the direct runoff.  The second reason that adding a second layer is of interest is that
it provides an additional storage zone for water before it is passed to the
groundwater reservoir.
3.2.3. Land use data
In this research, land use data are used to estimate impervious areas in a
watershed.  1:250,000 scale GIRAS Land Use/Land Cover (LULC) files covering
the Little Washita watershed were obtained from the EPA Internet site listed in
Table 3.1. Data in these files were originally collected by the U.S. Geological
Survey and converted to Arc/Info format by the U.S. Environmental Protection
Agency.  Land use characteristics in these files were specified in the mid 1970?s and
early 1980?s.  In some parts of the United States, significant land use changes have
occurred since this time; however, this is not as big issue in the Little Washita
watershed.  A map showing broad land use classes in the Little Washita is provided
in Figure 3.17.  It is interesting to note overall similarities between land use and
soils (Figure 3.4).  Areas with sandy soils correspond to rangeland while silty and
loamy soils correspond to agricultural land.
Figure 3.17  Land use in the Little Washita watershed
69
Each polygon in an LULC map is assigned a code indicating land use type
based on the Anderson Level II land use classification scheme (Anderson et al.,
1976).  Anderson Level II codes consist of two integers, the first integer specifies a
general category (i.e. 1 = urban or built-up land) and the second integer is more
specific (i.e. 11 = residential).  In this study, Table 3.6 has been developed to
associate an impervious cover percent with several of the Anderson level II land use
codes.  Impervious percent estimates for codes 11, 12, and 13 are made based on
information in Chow, Maidment, and Mays (1988) Table 5.5.2.  Other estimates are
considered reasonable guesses.  Zero impervious percent is assumed for land use
categories not listed (i.e. agriculture, rangeland, and forest land).  Subcategories of
the Tundra (8) and Perennial Snow or Ice (9) Level 1 categories are not considered
here.
Table 3.6 Impervious percentages associated with selected land use classes.
Land use
code
Description Assigned
% Imperv.
11 Residential 36
12 Commercial and services 85
13 Industrial 72
14 Transportation, communication,
utilities.
100
16 Mixed urban or built-up land 50
17 Other urban or built-up land 50
51 Streams and canals 100
52 Lakes 100
53 Reservoirs 100
54 Bays and estuaries 100
61 Forested wetland 100
62 Non-forested wetland 100
74 Bare exposed rock 100
70
3.2.4. NEXRAD data and spatial variability
The primary rainfall data used for this study are hourly NEXRAD Stage III data
from the National Weather Service Arkansas-Red Basin River Forecast Center
(ABRFC).  NEXRAD Stage III data are mapped using the Hydrologic Rainfall
Analysis Project (HRAP) grid (Greene and Hudlow, 1982) with cells that are
approximately 4 km on a side in Oklahoma.  A detailed discussion of mapping
issues associated with using HRAP grid cells in GIS-based hydrologic modeling is
provided by Reed and Maidment (1998).  The NEXRAD rainfall cells define the
modeling units for this study.  Programs developed by Reed and Maidment (1995)
were used to create a polygon coverage of rainfall cells covering the Little Washita
watershed.  These cells are intersected with the watershed boundary as shown in
Figure 3.18.  Portions of 55 different NEXRAD cells intersect the watershed
boundary.  Many of the NEXRAD cells that intersect near the watershed boundary
form small polygons that are only a fraction of the NEXRAD cell size.  The number
of whole NEXRAD cells within the Little Washita watershed can be estimated by
dividing the watershed area (600 km
2
) by the approximate area of NEXRAD cells
(16 km
2
) giving 38 cells.  Thus, it would take about 40 gages uniformly spread
around the watershed to get the same spatial resolution.
(
552, 303)
 
(
563, 311)
(hr
a
px, hr
apy)
M
o
d
e
l p
o
l
ygon
s
F
i
gure 3.18 Int
ersecting NEXRAD 
cells w
i
th w
a
tershed bo
undary.
71
72
Three levels of resampling are considered for simplifying the description of soil
properties in the model.
  No resampling:  Estimate the percentage of each soil type or the percentage
of each distinguishable soil component profile within each cell, and make
vertical flux calculations separately for each of these units.
  NEXRAD cell level resampling.  Assign a representative soil type to each
NEXRAD cell.
  Watershed level resampling:  Assign a representative soil type to the entire
watershed.
The same rainfall data are used at all three levels; therefore, watershed level
resampling is lumped in terms of soil properties but not lumped in terms of rainfall.
For NEXRAD cell and watershed level resampling, the representative property is
assigned based on the predominant soil type in the watershed or NEXRAD cell
respectively.  This approach is used rather than a weighted average because
complex issues arise when trying to average soil properties such as water retention
parameters that exhibit and control non-linear behavior.  For each of the three
resampling levels, soil properties may be estimated using either the STATSGO data
or the 30-m soils grid.
Both the one-layer and two-layer models are considered for the case of no
resampling.  For the one-layer model this means that the percentage of each surface
texture class (and the ?other? class) falling within each NEXRAD cell must be
computed.  For the two-layer model, a list of components and the percentage area
covered by each component are computed.  In addition, a list of soil properties
associated with each component is also computed for the two-layer model.  A
description of how this information is transferred from Arcview GIS to the
hydrologic model written in Visual Basic is provided in Chapter 4.  For all
resampling levels, several non-soil parameters are computed for each NEXRAD cell
73
including area, average flow length to the watershed outlet, and percent
imperviousness.
Although NEXRAD cells are used exclusively in this study, the spatial analysis
programs used are general enough so that any polygons could be used to define
uniform rainfall units (i.e. Thiessen polygons).
The acquisition of rainfall time series data is a somewhat tedious process.
Archived, hourly Stage III data files are available from the Arkansas-Red Basin
River Forecast Center (ABRFC) from July 24, 1994, to present.  The Arkansas-Red
Basin River Forecast Center is responsible for flood forecasting over a 538,487 km
2
area that covers seven states.  Hourly Stage III files produced at ABRFC contain
53,265 data values.  The archived files are in a special binary format called netCDF
(Network Common Data Format) developed at the Unidata Program Center in
Boulder, Colorado.  Semi-automated procedures were developed to download and
process netCDF rainfall files into a more tractable format, consisting of a comma
delimited text file containing one column of time series data for each of the 55
NEXRAD cells intersecting the Little Washita watershed.
NetCDF files are obtained from the ABRFC through their Internet server
(address given in Table 3.1).  After downloading the data, a UNIX shell script is
executed which runs a series of processing routines including a freeware program
that converts netCDF files into ASCII files, and C and FORTRAN routines that are
used to reformat the files and extract only the data for cells in the area of interest.
The specific cells of interest in the Little Washita watershed are hardwired into the
code.  These programs are described further in Appendix B.  Two months of
netCDF data files (1440 files) can be processed in about 30 minutes running this
procedure on a Sun Ultrasparc1.  Comma delimited rainfall files containing data for
the area of interest were prepared for October, 1995 ? September, 1997, although
not all of these data are used for analysis.
74
3.2.5. Data used to estimate evaporation
A conventional approach to estimating evaporation was discussed in Chapter 2.
The development of an alternative method is considered in Section 3.3.2.  Data from
Department of Energy Atmospheric Radiation Measurement program Southern
Great Plains Cloud and Radiation Testbed (DOE ARM SGP CART) sites are used
in the development of the proposed method and a description of these data is left to
Section 3.3.2.  Although the data used for analysis in Section 3.3.2 are for locations
outside of the Little Washita watershed, data from a different SGP CART site that
lies within the Little Washita is used for simulation runs.  The use of these data is
also discussed in Section 3.3.2.
3.2.6. Data used for calibration and validation
15 minute runoff data for USGS gaging station 07327550 near Ninnekah,
Oklahoma, for October, 1995, - October, 1997, were obtained from the Oklahoma
USGS office.  A Visual Basic program was written to compute hourly totals, flag
missing data, and put the data into a more usable format.  Further description of this
data processing program is provided in Appendix B.
Gravimetric soil moisture measurements from Southern Great Plains 1997
Hydrology (SGP97) experiment are used in this study.  SGP97 was an intensive
hydrology field experiment organized jointly by the USDA and NASA that took
place from June 18 - July 17, 1997.  There were several objectives in SGP97 related
to the use of in-situ and remote sensing observations to understand soil moisture and
energy balance dynamics at different spatial scales.  Numerous field and remote
sensing measurements were taken during SGP97.  A subset of the gravimetric soil
moisture observations taken in SGP97 are used for model calibration in this study.
Gravimetric soil samples were taken at 23 different sites in the Little Washita
watershed from June 18 ? July 16, 1997.  The locations of these sampling sites is
shown in Figure 3.19.  9 to 14 samples were taken from the top 5 cm of soil twice
75
daily at each of the sites.  Daily averages and standard deviations for gravimetric
soil moisture (GSM = mass water(g)/mass of soil(g)) and bulk density  (?
b
 = mass
of soil(g)/total volume(cm
3
)) were obtained from the SGP Internet site listed in
Table 3.1.  Using this information, daily average water content (? ) values for the
top 5 cm of soil are computed using Equation 3.7.
Equation 3.7
w
b
GSM
?
?
? *=
In Equation 3.7, ?
w
 is the density of water taken as 0.997 g/cm
3
.
The soil textures at each of the 23 sampling sites are provided in the SGP
documentation.  Among the 23 sites, four different soil textures are represented:
sandy loam (SL), loamy sand (LS), loam (L), and silt loam (SIL).  Sites 3, 4, and 5
are all loamy sand sites and fall within the same NEXRAD cell (557, 309).  The
daily values from these three sites are averaged so that a comparison can be made
with simulated results for loamy sand in cell (557, 309).  Other sites chosen for
analysis include sites 8 and 9 with silt loam soils that fall within cell (555, 307),
sites 11, 12, and 13 with loam soils that fall within cell (560, 309), and site 17 with
sandy loam soil that falls within cell (559, 305).  Use of observed soil moisture for
model calibration is discussed in Chapter 5.
76
Sites 5,4,3 in
cell 557, 309
Sites 8, 9 in cell
555, 307
Site 17 in cell 
559, 305
Sites 11,12,13 in
cell 560, 309
Figure 3.19  Locations for gravimetric soil moisture sampling during SGP97.
A dense meteorological observation network operated by the Agricultural
Research Service Grazinglands Research Laboratory includes 42 stations covering
areas in and around the Little Washita.  This network is called the Micronet.  The
Micronet, which contains about 1 rain gage per NEXRAD cell, has an unusually
high network density that cannot be found at many locations in the world.  The
locations of the Micronet rain gages are shown in Figure 3.20.  Rainfall data from
this network are available for the months of June, July, and August, 1997, from the
SGP Internet site listed in Table 3.1, although many stations were not reporting until
mid-June.  Hourly average values for the Little Washita as a whole were computed
from 5 minute cumulative totals available for each gage.  As discussed in Chapter 5,
these hourly averages are compared with spatially averaged NEXRAD Stage III
radar estimates to try to assess the quality of the NEXRAD data being used for
simulations.
77
Figure 3.20  Locations of Micronet stations.
3.3. Process Representation
3.3.1. Direct runoff computation
Direct runoff forms the rapidly varying portions of watershed hydrographs.  In
event simulation, direct runoff can be defined as the fraction of rainfall that is not
lost to interception, depression storage, or infiltration.  In this study, infiltration is
the only loss mechanism that is explicitly modeled, in part, because infiltration
parameters can be estimated from available data, but no satisfactory scheme has
been developed to parameterize other mechanisms such as interception, depression
storage, and channel storage.  Direct runoff could evaporate while flowing to the
watershed outlet, but this additional loss mechanism is not considered in this study
due to the relatively short time of concentration (12 hours).
Interception is an important factor in forested areas, accounting for as much as
40% of the gross precipitation in some forests (Dingman, 1994) but is not as
important in the agricultural and rangeland areas, which predominate in the Little
Washita watershed.  Previous researchers have used the leaf area index (LAI),
78
defined as the leaf area per unit of ground area to estimate interception.  It is
difficult to support seasonally varying estimates of leaf area index with the data
used in this research.  Although beyond the scope of this research, the use of remote
sensing data to estimate leaf area index values is discussed briefly in Section 3.3.2
on evaporation.  The model described in this manuscript has proven useful and
insightful without an explicit parameterization for interception.
Since the Little Washita watershed is an agricultural watershed in a sub-humid
climate with moderately rolling topography (Allen and Naney, 1991), the dominant
control on direct runoff generation is assumed to be soil texture.  Although
topography influences soil moisture locally, it appears from the remote sensing
images of Jackson et al. (1996) that soil texture plays a bigger role than vegetation
or topography in controlling large scale soil moisture distributions in the Little
Washita watershed.
Impervious areas will significantly influence direct runoff production in a
watershed.  The percent impervious area in each model polygon is computed using
the land use coverage (Figure 3.15) and Table 3.6.  Impervious areas generate direct
runoff even during small storms when the infiltration capacity of even the least
permeable soils is never exceeded.  The percent imperviousness for each model
polygon in the Little Washita is shown in Figure 3.21.
79
P e rce nt Impe rvio usness
0 - 1
1 - 5
5 - 10
10 - 14
Figure 3.21  Average percent imperviousness in each NEXRAD cell.
3.3.1.1. Green-Ampt model
For reasons discussed in Section 2.4, the Green-Ampt infiltration model is used
in this study.  The Green-Ampt model approximates vertical infiltration into a semi-
infinite homogeneous soil under ponded surface conditions.  Experimental evidence
indicates that a short period of time after surface ponding, a plot of soil moisture
versus depth reveals a curved wetting front, which maintains a constant shape as it
advances into the soil (Jury, Gardner, and Gardner, 1991).  An analogy is often
made between this wetting front and the movement of a piston, which advances but
does not change shape (Figure 3.22).  To derive a simple expression for the
infiltration rate as a function of cumulative infiltration, an abrupt wetting front is
assumed in the Green-Ampt model as shown in Figure 3.23.  Two parameters that
control the rate of infiltration are the hydraulic conductivity behind the wetting front
and the average wetting front suction head.
80
?
r
?
i
h
p
?
ns
h=0
h=h
e
h = h
i
Figure 3.22  Curved wetting front formed during infiltration into a
homogeneous, uniformly wetted soil.
h
p
?
r
L
f
z
wetted zone
(conductivity = K
ns
)
?
i
?
ns
?
ns
 - ?
i
Figure 3.23  Idealized infiltration into a homogeneous, uniformly wetted soil
with a sharp wetting front.
81
The rate of infiltration and the wetting front suction head (h
f
) itself are both
functions of the initial water content of the soil (?
i
).  The physical meaning of the
wetting front suction head (h
f
) has been the subject of debate in the literature.  Mein
and Larson (1973) suggest that h
f
 can be estimated using Equation 3.8.
Equation 3.8
?
=
i
h
rwf
dhhkh
0
)(
In Equation 3.8, k
rw
 is the relative permeability (K(h)/K
s
), h is suction head, h
i
 is the
suction head at the initial saturation of the soil.  Evaluation of the integral in
Equation 3.8 is discussed in the next section.  Neuman (1976) shows how the
expression in Equation 3.8 can be derived theoretically by presenting a formal
derivation of the Green-Ampt infiltration equation.  The derivation of Neuman is
reproduced here.
Darcy?s law quantifies the vertical flux at any point in the soil column.  With
the z-coordinate positive downward as in Figure 3.23, Darcy?s can be written as:
Equation 3.9 )1)((
)(
)( +=
?
?
?
?
?
? +
=
dz
dh
hK
dz
zhd
hKq
Considering Figure 3.22, the total flux into the soil column can be estimated by
rearranging Equation 3.9 and integrating  from the ponded surface to the bottom
edge of the wetting front.
Equation 3.10
?? ?
+=
fi
p
f
Lh
h
L
dzhKdhhKqdz
00
)()(
or
Equation 3.11
????
++=
0
000
)()()(
p
iff
h
hLL
dzhKdhhKdhhKqdz
It is assumed that the hydraulic conductivity is constant between ponding depth (h
p
)
and elevation zero.  Because an abrupt wetting front is assumed, then q is constant
and the hydraulic conductivity is constant from the ground surface to the abrupt
front.  Neuman (1976) takes the constant hydraulic conductivity behind the wetting
82
front as K
s
, the saturated hydraulic conductivity associated with air entry that is
normally reported in the literature.  Based on these assumptions, the integral on the
left-hand-side of Equation 3.11 and the first and third integrals on the right-hand-
side can be evaluated as follows:
Equation 3.12
?
++=
i
h
fspsf
LKdhhKhKqL
0
)(
Factoring K
s
 out of the remaining integral and dividing through by L
f
 yields an
equation for infiltration rate (i).
Equation 3.13
f
h
o
prwfs
L
hdhkLK
qi
i
?
++
==
)(
The remaining integral is the same as Equation 3.8 and defines the wetting front
suction head term (h
f
).  A method for evaluating this integral (h
f
) is presented in the
next section.
Bouwer (1969) notes that the actual hydraulic conductivity of the wetted zone
in the Green-Ampt model is less than the saturated hydraulic conductivity (K
s
) due
to entrapped air.  Bouwer (1969) suggests estimating the hydraulic conductivity of
the wetted zone as half of the saturated hydraulic conductivity (K
s
) based on his
review of hydraulic conductivity-suction head relationships presented in the
literature.  This assumption is cited by Rawls, Brakensiek, and Miller (1983) and
Charbeneau and Asgian (1991).  Following Charbeneau and Asgian (1991), the term
natural saturation (K
ns
) will be used here to refer to the hydraulic conductivity
behind the wetting front that is taken as half of K
s
.
Equation 3.14
2
s
ns
K
K =
Replacing K
s
 by K
ns
 in Equation 3.13 and assuming surface ponding (h
p
) is
negligible yields:
83
Equation 3.15 
f
h
o
rwfns
L
dhkLK
i
i
)(
?
+
=
Strictly speaking, the term K
ns
 should be substituted for K
s
 in Equation 3.12 rather
than in Equation 3.13.  Doing this would produce a factor of 1/2 in front of the
integral expression for h
f
, giving the value of h
f
 as half of the value computed using
Equation 3.8.  In evaluating the integral expression for h
f
 (described in the next
section), this factor of 1/2 is not included so that the results are consistent with the
expression for h
f
 derived by Brakensiek (1977) that appears to have gained
acceptance in the literature.
Given the assumption that there is trapped air in the wetted soil, the wetted
zone does not have a saturation equal to 1.  During an event, the water content
jumps from its initial value (?
i
) to the water content (?
ns
) that corresponds natural
saturation (?
ns
) (Charbeneau and Asgian, 1991).  ?
ns
 can be computed by
substituting K
ns
 = ? K
s
 into the Brooks and Corey relative permeability relationship
(Equation 3.16) and then solving for ?
ns
 (Equation 3.17).
Equation 3.16 ()
2
1)(
/23
/23
=
?
?
?
?
?
?
?
?
?
?
=?=
+
+
?
?
??
???
r
rns
ns
s
ns
K
K
Equation 3.17
rrns
????
?
+
?
?
?
?
?
?
?=
?
?
?
?
?
?
?
?
+
2
3
1
2
1
)(
Using the Carsel and Parrish values for the Brooks and Corey parameters given in
Table 3.3, natural saturation values for each of the 12 USDA soil texture classes are
computed and given in Table 3.7.  In the proposed model, the top layer water
content is not allowed to exceed ?
ns
 to ensure that the water deficit (?
ns
 ? ?
i
) is
always positive.  In natural soils, ?
ns
 may be exceeded if water is not drained and
trapped air has time to escape from the soil.
84
Table 3.7 Natural saturation values corresponding to the 12 USDA texture
classes.
Class Name Symbol ?
ns
sand S 0.85
loamy_sand LS 0.86
sandy_loam SL 0.88
sandy_clay_loam SCL 0.91
loam L 0.90
silt loam SIL 0.92
clay_loam CL 0.93
silt SI 0.92
clay C 0.97
sandy_clay SC 0.94
silty_clay_loam SICL 0.94
silty_clay SIC 0.97
The jump in water content from ?
i
 to ?
ns
 defines the relationship between
cumulative infiltration (I) and the depth of the wetting front (L
f
)
Equation 3.18
fins
LI )( ?? ?=
Solving Equation 3.18 for L
f
 and substituting into Equation 3.13 yields an equation
where the infiltration rate (i) is a function of cumulative infiltration (I).
Equation 3.19
I
hIK
dt
dI
i
insfns
))(( ?? ?+
==
An expression for the cumulative infiltration as a function of time can be derived by
separating and integrating Equation 3.19 from 0 to t.
Equation 3.20 tK
h
I
hI
ns
insf
insf
=
?
?
?
?
?
?
?
?
+
?
?? 1
)(
ln)(
??
??
This equation can be solved for I using iterative substitution or by using the
Newton-Raphson root finding method.  The Newton-Raphson method is used in the
computer codes developed for this study.  Equations 3.19 and 3.20 only apply if the
rainfall has ponded to a small depth at the soil surface, meaning that the rainfall rate
85
has exceeded the potential infiltration rate.  Therefore, time to ponding must be
determined in order to make direct runoff computations.
Mein and Larson (1973) describe a method for determining ponding time under
a constant rainfall rate (r).  The assumption is that all rainfall infiltrates before
ponding occurs and ponding begins when the potential infiltration rate (computed
with Equation 3.19) drops below the rainfall rate.  Therefore, time to ponding (t
p 
)
occurs when the potential infiltration rate (i) equals the rainfall rate (r).  Since all
rainfall that occurs before ponding infiltrates, the cumulative infiltration at ponding
time (I
p
) equals rt
p 
when the rainfall rate is constant.  Substituting this expression for
I
p
 into Equation 3.19 and setting i = r yields an expression for the cumulative
infiltration at ponding under a constant rainfall rate.
Equation 3.21
)(
)(
ns
insfns
pp
Kr
hK
rtI
?
?
==
??
Computing infiltration under variable rainfall conditions is discussed in Section
3.3.1.3.
3.3.1.2. Estimating wetting front suction head
The derivation method presented here follows that of Brakensiek (1977).  To
estimate the wetting front suction head at the onset of each storm, the Brooks and
Corey relative permeability relationship is substituted into the integral defining the
wetting front suction head in Equation 3.8.  Graphically, h
f
 is the area under the
relative permeability curve for wetting shown in Figure 3.24 (for a loam soil) and
described by Equation 3.22.
Equation 3.22 ()
23
2
3
+
+
?
?
?
?
?
?
=?=
?
?
h
h
k
we
wrw
86
0
50
100
150
200
0 0.2 0.4 0.6 0.8 1
suction head 
(
cm)
h
we
k
rw
Figure 3.24 Example relative permeability curve for wetting (? = 0.56, h
b
 =
27.8).
Following Brakensiek (1977), the wetting front suction head can be calculated
by breaking Equation 3.8 into two parts as shown in Equation 3.23.
Equation 3.23
???
+==
i
we
iwe
h
h
rw
hh
rwf
dhhkdhdhhkh )()(
00
During sorption, k
rw
(h) is 1 between h = 0 and h = h
we
.  Above h
we
, k
rw
(h) begins to
decrease.  Given the relationship shown in Equation 3.22, the second integral on the
right hand side of Equation 3.23 can be written in terms of effective saturation and
solved.  To do this, a variable substitution is made by solving Equation 3.22 for
suction (h) as a function of effective saturation (?) and taking the derivative of h
with respect to ?.
87
Equation 3.24
()1
1
??
??=
?
?
?
w
we
w
h
d
dh
Equation 3.23 then becomes
Equation 3.25
()
????=
??
?
+
?
d
h
hh
w
we
wwef
i 11
1
)
2
3(
??
?
which evaluates to
Equation 3.26
()
we
wewwe
f
h
hh
h +
+
+
+
?
?=
+
1313
1
3
??
?
This result is consistent with the result presented by Brakensiek (1977) for wetting
front suction head into a completely dry soil if the effective saturation is set equal to
zero and the substitution h
we
 = h
b
/2 is made.  That result is:
Equation 3.27
?
?
?
?
?
?
+
+
=
13
32
2 ?
?
b
f
h
h
3.3.1.3. Computational scheme for Green-Ampt infiltration under variable rainfall
To make infiltration calculations under variable rainfall using the Green-Ampt
model, the computational scheme described in Figure 5.4.1 of Chow, Maidment,
and Mays (1988) is used.  This solution scheme uses the integral form of the
infiltration equation rather than a time stepping method so there is no error
associated with the numerical algorithm.  A key idea in implementing this method is
that the cumulative infiltration at the end of a time interval ?t can be written as a
function of the cumulative infiltration at the beginning of that interval as follows.
Equation 3.28 tK
hI
hI
hII
ns
insft
insftt
insfttt
?=
?
?
?
?
?
?
?
?
?+
?+
???
?+
?+
)(
)(
ln)(
??
??
??
Equation 3.28 is derived by writing Equation 3.20 for both time t and time t+?t and
then subtracting the two equations.
The pseudo-code in Figure 3.25 summarizes the scheme for computing
infiltration.
88
In both the one-layer and two-layer soil models, the top layer properties are
assumed to control infiltration.
89
'------------------------------------------------------------------------------------
' variables:
' I final cumulative infiltration
' i potential infiltration rate
' r rainfall rate
' subscripts t and t+dt refer to current and next time step respectively
' subscript p refers to the ponding time
' first_time:  boolean variable that is true if t = 1
'------------------------------------------------------------------------------------
Initialize I, I
t
, and I
t+dt
, to 0
MAIN LOOP: For all time steps t with positive precipitation
If this is the first time step
     Set I
t
 = 0
     Set first_time = True
Endif
If this is not the first time step then
     Compute infiltration rate at time t (i
t
) with Equation __
Endif
If (ponding has not occurred (i
t
 > r
t
)) or (if this is the first time step)
     Calculate tentative values I?
t+dt 
and i?
t+dt
 assuming that all rainfall infiltrates
     If (i?
t+dt
 > r
t
) then ponding does not occur in the interval
          I
t+dt
 = I?
t+dt
     Else ponding occurs at some point in the interval
          Calculate the cumulative infiltration at ponding I
p
 based on Equation __
          with r = r
t
          Determine the ponding time t+?t? where ?t? = (I
p
 - I
t
) / r
t
          Solve Equation 3.28 iteratively for I
t+dt
 with I
t
 = I
p
 and ?t = ?t - ?t? using
          the Newton-Raphson method
     End If
     first_time = False
Else ponding occurs throughout interval
     Solve Equation 3.28 for I
t+dt 
using the Newton-Raphson method
End If
' reinitialize cumulative infiltration before the next time step
I
t
 = I
t+dt
END MAIN LOOP through time steps
Figure 3.25  Pseudo-code for computation of Green-Ampt infiltration under
variable rainfall.
90
3.3.2. Evaporation
In Section 2.5, a conventional method for estimating evaporation was described
that uses an empirical crop coefficient.  Shuttleworth (1993) notes that the crop
coefficient has a poor theoretical basis and implicitly has meteorological
dependence.  He recommends that future research be directed towards estimating
surface resistance values rather than crop coefficients.  An analysis of data from
SGP CART sites presented here lends insight into the relationship between surface
resistance and soil moisture, and an alternative to the conventional method of
estimating evaporation under moisture limiting conditions (Figure 2.1) is
investigated.
3.3.2.1. Surface Resistance
The Penman-Monteith method is a widely accepted approach for estimating
reference crop evaporation.  The Penman-Monteith method is attractive because it
has a sound physical basis and because its accuracy is proven when compared with
lysimeter measurements.  In fact, this method was rated first in accuracy among
twenty methods for estimating reference crop evaporation in ASCE (1990).  The
Penman-Monteith equation can be written as follows:
Equation 3.29
)1(
)(
a
s
a
pa
n
m
r
r
r
Dc
GR
lE
++?
++?
=
?
?
In Equation 3.29, lE
m
 is the energy flux used for evaporation [W m
-2
] (l is the
latent heat of vaporization in [J kg
-1
] and E
m
 is the mass flux of water vapor [kg m
-2
s
-1
]), R
n
 is the net radiation absorbed at the land surface [W m
-2
], G is the ground
heat flux [W m
-2
], ? is the gradient of the saturation vapor pressure curve [kPa ?C
-1
],
?
a
 is the density of moist air [kg m
-3
], c
p
 is the specific heat of moist air [J kg
-1
 ?C
-1
],
91
D is the vapor pressure deficit [kPa], ? is the so-called psychrometric constant [kPa
?C
-1
], r
a
 is the aerodynamic resistance to diffusive mass transfer of water vapor
[s m
-1
], and r
s
 is the surface resistance to mass transfer due to the vegetation canopy
[s m
-1
].  The variables and parameters R
n
, G, ?,  ?
a
, c
p
, and ? in Equation 3.29 can
either be measured directly or estimated in a fairly straightforward manner from
temperature and pressure.  Critical variables that cannot be measured directly and
are more difficult to estimate are the aerodynamic (r
a
) and surface (r
s
) resistances.
The units for r
a
 and r
s
 are [s m
-1
] because these resistances are equal to a
concentration difference [kg m
-3
] divided by a mass flux [kg m
-2
 s
-1
].
The aerodynamic resistance (r
a
) is a function of both the crop height and the
wind profile.  r
a
 is often estimated by assuming a logarithmic wind profile and
making an estimate of the surface roughnesses for heat and vapor transfer as a
function of the crop height.
The surface resistance (r
s
) refers to the total resistance of all leaves in a canopy.
When estimates of reference crop evaporation are made, a specific crop type (grass
or alfalfa) and height are assumed.  With this information, the minimum surface
resistance (r
smin
) may be estimated using an approximation for surface resistance as
a function of leaf area index (LAI) (Shuttleworth, 1993).  The variable r
smin
 here is
equivalent to the parameter "r
s
" used by Shuttleworth (1993).  Surface resistance is
inversely proportional to leaf area index.  The minimum surface resistance (r
smin
) is
the value of surface resistance when there is no water stress (when potential
evaporation is occurring).
Use of the Penman-Monteith equation to estimate reference crop evaporation is
feasible and is done commonly in practice.  It is a much more difficult problem to
predict actual evaporation when soil-water is limiting by using Equation 3.29.
Monteith (1980) notes that ?it is impossible to predict with confidence how r
c
[equivalent to r
s
 used here] will change during a growing season as soil water is
depleted by root extraction and replenished by rain or irrigation.?  The two key
92
factors cited by Monteith (1980) that control surface resistance are changes in soil-
water content which affects the leaf water content and therefore the stomatal
resistance of individual leaves, and changes in the leaf area index which affect the
net resistance for the entire canopy.
Using an analogy to electrical resistance for parallel resistors, the net surface
resistance (r
s
) for a canopy is approximated as the individual leaf resistance (r
sleaf
)
divided by the leaf area index.  Monteith (1980) suggests that empirical evidence
may be used to determine a value for r
s
 as a function of soil-water potential.
Monteith also suggests that ?progress [in calculating the actual evaporation rate as a
function of soil-water content] in the immediate future calls for the shrewd
interpretation of evidence from carefully designed field experiments rather than the
development of more comprehensive simulation models.?  The work presented in
the following sections represents an attempt to pursue this line of thinking, although
the field experiments were not carefully designed by the author ? rather, selected
data sets from sites being extensively monitored for other purposes are used to
illustrate an idea.  The idea is to see if useful empirical relationships between
surface resistance (r
s
) and soil-water potential can be derived.
3.3.2.2. Data analysis: evaporation
This section describes the analysis of field data that were collected as part of
the Department of Energy?s Atmospheric Radiation Measurement (ARM) program.
The ARM data collection program is an ongoing effort to improve scientific
understanding of radiative energy fluxes in the Earth's atmosphere.  One goal of the
program is to improve radiative transfer parameterizations in General Circulation
Models.  Three field measurement sites called Cloud and Radiation Testbeds
(CART) are part of the ARM program.  These sites include the Southern Great
Plains site in Kansas and Oklahoma, USA, a Tropical Western Pacific Site, and a
93
North Slope of Alaska site.  More information about this program can be found on
the Internet at www.arm.gov.
Three data streams from the Southern Great Plains site are used in the analysis
presented here.  The word ?streams? is used by in the ARM program to refer to the
stream of measurements that are produced by selected group of instruments.  These
data streams include Energy Balance and Bowen Ratio (EBBR) data, Surface
Meteorological Observation System (SMOS) data, and Soil Water and Temperature
Sensors (SWATS) data.  EBBR, SMOS, and SWATS data for certain facility
locations and time periods are easily accessible from the ARM Archive Internet site
(www.archive.arm.gov/data/ordering.html).  The EBBR and SMOS data streams are
classified as "production" data streams because these data streams have routine and
well-developed procedures for instrument operations, data processing, and data
review.  The SWATS data streams are classified as "developmental" data streams as
of June 1, 1998, indicating that data quality control procedures are relatively
immature (www.arm.gov/docs/instruments/static/swats.html).  Calibrations to
convert SWATS matric potential estimates into volumetric water content estimates
are still being refined.
The idea proposed here is to examine functional relationships between surface
resistance and soil-water potential using EBBR, SMOS, and SWATS data.  In
theory, this type of relationship could be used in a continuous soil-water balance
model to estimate r
s
 at any given time.  Values of r
s
 are estimated by solving
Equation 3.29 for r
s
 and determining values for all other variables in Equation 3.29
using the EBBR and SMOS data.  Before describing these calculations in more
detail, further description of the ARM data sets is provided.
There are a number of instrumented sites in the ARM SGP study area covering
central Oklahoma and south-central Kansas.  To give an idea of the areal extent of
the SGP study area, the locations for 22 of the instrumented sites are shown in
94
Figure 3.26.  These points were plotted based on latitude and longitude values for
extended facilities provided by Peppler, Sisterson, and Lamb (1997).
Many different instruments are used throughout the ARM SGP study area, but
the instrumentation varies from site to site.  To facilitate the proposed analysis, the
EBBR, SMOS, and SWATS instrument sets must coexist at selected sites and data
from these instruments must be available for coincident time periods.  According to
Peppler, Sisterson, and Lamb (1998), 8 of 26 extended facilities should be equipped
with all three required instrument sets by the end of 1997.  Unfortunately, one or
more of the desired data streams at some of these sites were either not available
from the ARM data archive or contained measurements of suspect quality.  Data
from extended facilities EF-4 (Plevna, KS) and EF-7 (Elk Falls, KS) were used for
analysis because the necessary data were available for these sites at the time of this
study.  Data from these two sites are adequate to illustrate the proposed idea.  Large
volumes of data that could support further study are continuously being created at
the ARM sites.
95
Kansas
Oklahoma
E4
E7
E26
Figure 3.26  Locations of 22 ARM Measurement Sites
Equation 3.29 is rearranged here so that r
s
 is written as a function of variables
that can be determined from instrument measurements.
Equation 3.30
a
a
pa
n
s
r
El
r
Dc
GR
r
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
++?
= 1
)(
??
?
Bars over the variables indicate daily average values.  Although the Penman-
Monteith equation is only theoretically valid at a single instant in time and is more
accurate when used with hourly data than with daily data, the Penman-Monteith
equation can provide reliable estimates of evaporation when daily mean climatic
input values are used (ASCE, 1990).  Although the ARM data would support
estimating hourly values of r
s
, meteorological observations available in most areas
will not support hourly computations for evaporation.  For these reasons,
representative daily surface resistance values are computed here rather than hourly r
s
96
values.   Monteith, Szeicz, and Waggoner (1965) showed that surface resistance can
vary considerably throughout the day.  A large part of this variation is attributable to
the amount of sunlight received by the plants, which influences the degree to which
stomata are opened.  Making daily calculations averages out diurnal variations in
surface resistance.  When r
s
 is computed using Equation 3.30, the result is really the
combined resistance due to both bare soil and plants; however, when vegetation is
dense or when the soil surface is dry, evaporation from the bare soil is negligible
(Monteith, Szeicz, and Waggoner, 1965).
In Equation 3.30, R
n
, G, and lE are estimated using EBBR data; ?, ?
a
, D, ?, and
r
a
 are estimated using SMOS data; and c
p
, the specific heat of moist air, is taken as
1013 J kg
-1
 C
-1
.
The Energy Balance Bowen Ratio (EBBR) systems are primarily designed to
estimate vertical fluxes of sensible and latent heat at the land surface.  Net radiation
(R
n
) absorbed at the Earth's surface is consumed by either heating the air (sensible
heat flux, H), converting liquid water into water vapor (latent heat flux, lE), or
heating the soil (ground heat flux, G).  The sum of the energy fluxes at the Earth's
surface is equal to zero.
Equation 3.31 0=+++ GlEHR
n
In Equation 3.31, the units of each term are [J m
-2
 s
-1
] or [W m
-2
].  At each EBBR
location, R
n
 and G are measured.  The sum of R
n
 and G is sometimes referred to as
the available heating.
The Bowen ratio (?), the ratio of sensible heat flux to latent heat flux (H/lE), is
used to partition the available heating between latent and sensible heat fluxes.  The
Bowen ratio is estimated using measurements of temperature and relative humidity
(converted to vapor pressure) at two heights.  Derivations of equations that can be
used to estimate the Bowen ratio from such measurements are presented by Chow,
Maidment, and Mays (1988).  Given 30 minute average estimates of R
n
, G, and ?,
30 minute average estimates of lE and H are estimated using
97
Equation 3.32
)1(
)(
?+
+?
=
GR
lE
n
and
Equation 3.33 lEH ?=
For use in Equation 3.30, daily average values of R
n
, G, and lE are computed from
the 30 minute values.  The Bowen ratio method becomes unstable if ? is close to ?1,
which may occur in the evening or morning hours when the sensible heat flux
changes direction.  To avoid this type of problem, the lE estimates for time periods
in which ?1.5 < ? < -0.65 are not considered in the daily averaging.
-600
-400
-200
0
200
400
600
800
0 5 10 15 20
Time (hours)
Energy Flux (W/m
2
)
Rn + G Latent Sensible
Figure 3.27  Surface radiation budget components at ARM SGP Site EF7 on
July 29, 1997.
98
-5
-3
-1
1
3
5
0 5 10 15 20
Time (hours)
Bowen Rati
o
Figure 3.28  30 minute Bowen ratio values at the ARM SGP Site EF7 on
July 29, 1997.
Typical surface radiation flux values and Bowen ratio values are shown in
Figures 3.27 and 3.28 for ARM SGP site EF7 on July 29, 1997.  Negative radiation
fluxes are directed away from the earth's surface and positive fluxes are directed
towards the earth's surface.  Net radiation is positive during the day but becomes
slightly negative during the night.  A downward spike occurs in the 27th half-hour
period for the day, 1:00 - 1:30 p.m.  A possible explanation for this spike is that a
cloud covered the measurement site during this time.  The measured Bowen ratio
often remains relatively constant during daytime hours but can become unstable at
night.
The gradient of the saturation vapor pressure curve (?), the air density ?
a
, the
vapor pressure deficit (D), the psychrometric constant (?), and the aerodynamic
resistance (r
a
) are estimated using SMOS data.  The SMOS measurements that are
99
used for these calculations include wind speed at 10 m, air temperature and relative
humidity (RH) at 2 m, and barometric pressure at 1 m.  Although 24 hour averages
of these variables are used for calculations, it is interesting to see plots showing how
some of these meteorological variables vary throughout the day.  Plots of wind
speed, temperature and relative humidity, and vapor pressure deficit are shown in
Figures 3.28, 3.29, and 3.30 for ARM SGP site EF7 on July 29, 1997.
0
1
2
3
4
5
6
7
0 5 10 15 20
Time (hours)
Wind Speed (m/s
)
Figure 3.29  Wind speed (m/s) at site EF7 on July 29, 1997.
100
10
15
20
25
30
35
0 5 10 15 20
Time (hours)
Temperature (C
)
0
20
40
60
80
100
Relative Humidity (
%
Temp RH
Figure 3.30  Temperature (?C) and relative humidity (RH) at site EF7 on July
29, 1997.
0
1
2
3
4
5
0 5 10 15 20
Time (hours)
Vapor Pressure (kPa
)
Deficit(es-e) es
Figure 3.31  Vapor pressure deficit and vapor pressure at site EF7 on July 29,
1997.
101
? is computed using the average daily air temperature (T) at 2 m and the
following two equations:
Equation 3.34
?
?
?
?
?
?
+
=
T
T
e
s
3.237
27.17
exp6108.0
Equation 3.35
2
)3.237(
4098
T
e
s
+
=?
T is air temperature [?C] and e
s
 is the saturation vapor pressure [kPa] corresponding
to the temperature T.  The density of moist air [kg m
-3
]is estimated as
Equation 3.36
T
P
a
+
=
275
486.3?
following Shuttleworth (1993).  P is atmospheric pressure [kPa] and T is
temperature [?C].  Daily average values for P and T were used to make calculations.
The vapor pressure deficit (D) [kPa] is the difference between the saturation vapor
pressure and the actual vapor pressure.
Equation 3.37 )1( RHeeeD
ss
?=?=
Equation 3.38
s
e
e
RH =
This deficit appears in Equation 3.30 because the gradient between saturated air
inside the leaf stomata and ambient air is one of the driving forces for evaporation.
The average daily deficit is estimated as the average of 24 hourly deficit values
When hourly values are not available, approximate methods for estimating daily
average deficits are used ? these different methods may yield significantly
different results due to the non-linear nature of Equation 3.34.  Alternative methods
for estimating a representative daily vapor pressure deficit are described by
Shuttleworth (1993) and ASCE (1990).  Shuttleworth (1993) recommends specific
methods depending on the types of meteorological measurements available.
102
? [kPa ?C
-1
], the so-called psychrometric "constant" is a function of pressure (P)
and the latent heat of vaporization (l) [MJ kg
-1
](l is a function of temperature).  The
following two equations are used to estimate ?:
Equation 3.39
s
Tl 002361.0501.2 ?=
Equation 3.40
l
P
0016286.0=?
T
s
 should be temperature at the water surface [?C] but the ambient air temperature is
used here as a surrogate for water temperature.  Daily average pressure and
temperature values are computed from 30 minute observations and substituted into
Equations 3.41 and 3.42.
The aerodynamic resistance term, r
a
, is computed using Equation 4.2.25 from
Shuttleworth (1993).
Equation 3.41
z
ovceomcu
a
u
zhzzhz
r
2
)41.0(
]/67.0ln[]/)67.0ln[( ??
=
Equation 3.42
com
hz 123.0=
Equation 3.43
cov
hz 0123.0=
In this equation, z
u
 is the height at which wind measurements are made (10 m), z
e
 is
the height at which relative humidity measurements are made (2 m), u
z
 is the
observed wind speed, h
c
 is an estimated crop height, z
om
 is the roughness height for
momentum transfer, and z
ov
 is the roughness height for turbulent transfer of water
vapor.  Average crop heights (h
c
) were estimated for the EF 4 and EF 7 sites used in
this study based on limited information available from the EBBR data files ? notes
in the EBBR files describe the vegetation found during instrument installation (both
instruments were installed during September).  An estimated grass height of 1.25 m
was used for EF 4 and 0.25 m was used for EF 7.  Substituting values for z
u
, z
e
, and
h
c
 into Equations 3.41-3.43 for EF 4 and EF 7 yields the relationships.
103
Equation 3.44
z
a
u
r
105
=  for EF4
Equation 3.45
z
a
u
r
219
=  for EF7
Daily average values for wind speed (u
z
) are used to calculate r
a 
in this study.
In the ARM SWATS developmental data streams for sites EF4 and EF7,
estimates of soil temperature, matric potential, and volumetric water content are
provided at eight different depths.  The SWATS systems use heat dissipation
sensors to measure matric potential in the soil (Reece, 1996).  The theory behind
heat dissipation sensors is that the thermal conductivity of a ceramic cylinder in
contact with the soil can be computed knowing the rate of heat dissipation from a
line heat source at the center of the cylinder.  The thermal conductivity of the
ceramic cylinder changes with the water content in the cylinder.  In the laboratory,
temperature changes observed by the sensor are calibrated against matric potential,
rather than water content.  For ARM SWATS field installations, the soil-water
retention curve of the soil surrounding the sensors is determined in order to translate
matric potential into volumetric water content (Schneider and Fisher, 1997).
Because the accuracy of calibrations that are used to convert matric potential
estimates to volumetric water content are still being assessed
(www.arm.gov/docs/instruments/static/swats.html), only matric potential
measurements are used in this study.  This may be a logical approach regardless of
the data quality because plants are more sensitive to matric potential than water
content and a derived empirical relationship between r
s
 and matric potential is
independent of soil type.
All of the ARM data sets used in this study were downloaded from the ARM
Archive accessible from the Internet (www.arm.gov/docs/data.html).  Data obtained
from the ARM Archive were distributed in network Common Data Format
(netCDF).  All netCDF files were converted to ASCII files using a freeware utility,
104
and Visual Basic programs were written to read and manipulate data from these
files.  A description of the programs used to manipulate ARM data is provided in
Appendix B.
3.3.2.3. Plotting surface resistance against soil piezometric head
The intent here is to investigate how daily surface resistance (r
s
) to evaporation
changes as the soil dries out.  r
s
 values are computed using Equation 3.30 and
plotted against soil piezometric head values (h + z).  Piezometric head values are
used rather than matric potential so graphs showing measurements at different
depths can be compared without bias.  An interesting question that can be explored
is how surface resistance relates to piezometric head measured at different depths.
At site EF7, concurrent EBBR, SMOS, and SWATS data were available for
June, July, August, and September of 1997.  Computed r
s
 and (h + z) values at a 5
cm depth are plotted in Figure 3.32.  Not all days in June, July, August, and
September are represented by points on Figure 3.32 because (1) during some days
there were obvious errors or unusual behavior in the input data (i.e. values were out
of a reasonable range or missing), and (2) days during which rain occurred were not
considered.  The abscissas in Figure 3.32 show the natural logarithms of the
absolute values of the piezometric heads rather than their absolute values, because it
was found that using a logarithmic scale makes the data easier to interpret.  If data
for all four months are looked at collectively (Figure 3.32 a), the correlation
between increasing ln (h + z) and increasing r
s
 is very weak.  It is interesting to note,
however, that correlations for individual months are better, as shown in Figure 3.32
b, indicating that a seasonal factor such as vegetation is influencing surface
resistance.  A similar observation can be made for plots of r
s
 versus (h + z) at a 15
cm depth as shown in Figure 3.33a and b.
Similar plots of r
s
 versus (h + z) at greater depths are shown in Figure 3.34 (a-
f).  In these plots the correlations for data points for individual months begin to
105
deteriorate as the depth increases.  Data points for individual months become
aligned vertically, indicating that there is little variation in soil matric potential
among the days plotted.  Because many data points are aligned vertically, no
regression lines have been included in Figure 3.34 (a-f).   At the ARM EF7 site,
there are actually two instrumented profiles separated by 1 meter, a west profile and
an east profile.  The data shown in Figures 3.31, 3.32, and 3.33 are for the west
profile.
106
y = 19.21x - 17.93
R
2
 = 0.09
-100
0
100
200
300
246810
ln (|h+z||) (ln cm)
Surface Resistance (s/m)
Jun Jul Aug Sep All Months Linear (All Months)
(a)
y = 20.11x - 66.95
R
2
 = 0.62
June
y = 29.39x - 99.92
R
2
 = 0.2061
July
y = 42.441x - 139.88
R
2
 = 0.4996
August
y = 18.602x + 41.769
R
2
 = 0.0527
Sep
-100
0
100
200
300
246810
ln (|h+z|) (ln cm)
Surface Resistance (s/m)
Jun Jul Aug Sep
Linear (Jun) Linear (Jul) Linear (Aug) Linear (Sep)
(b)
Figure 3.32  Surface resistance (r
s
) vs. ln(|h+z|) for EF7 at a depth of 5 cm (a)
with regression line for all data points (b) with regression lines for individual
months.
107
y = 15.25x + 14.41
R
2
 = 0.045
-100
0
100
200
300
246810
ln (|h+z||) (ln cm)
Surface Resistance (s/m)
Jun Jul Aug Sep All Data Linear (All Data)
(a)
y = 23.29x - 77.0
R
2
 = 0.52
June
y = 37.66x - 134.6
R
2
 = 0.36
July
y = 69.98x - 254.3
R
2
 = 0.54
August
y = 32.48x - 16.59
R
2
 = 0.10
September
-100
0
100
200
300
246810
ln (|h+z||) (ln cm)
Surface Resistance (s/
m
Jun Jul Aug Sep
Linear (Jun) Linear (Jul) Linear (Aug) Linear (Sep)
(b)
Figure 3.33. Surface resistance (r
s
) vs. ln(|h+z|) for EF7 at a depth of 15 cm (a)
with regression line for all data points (b) with regression lines for individual
months.
108
-100
0
100
200
300
246810
-100
0
100
200
300
246810
S
u
r
f
ac
e R
e
sist
ance (s/m)
(a) (b)
ln(|piezometric head|)(ln cm)
-100
0
100
200
300
246810
-100
0
100
200
300
246810
S
u
rface R
e
s
i
stan
ce 
(s
/
m
)
ln(|piezometric head|)(ln cm)
(d)
(e)
-100
0
100
200
300
246810
(e)
-100
0
100
200
300
246810
(f)
Surfa
c
e
 Re
s
i
st
a
n
c
e
 
(s/
m
)
ln(|piezometric head|)(ln cm)
Figure 3.34 Surface resistance (r
s
) vs. ln(|h+z|) for EF7 at a depth of (a) 25 cm,
(b) 35 cm, (c) 60 cm, (d) 85 cm, (e) 125 cm, and (f) 175 cm.
109
It appears from Figures 3.32 and 3.33 that factors other than soil piezometric
head influence the surface resistance and perhaps a key factor is the month of the
year.  Monteith, Szeicz, and Waggoner (1965) showed that values of surface
resistance computed in a manner similar to that used here are sensitive to leaf area
index.  Although additional factors that influence stomatal openings might come
into play (i.e. solar radiation, air temperature, humidity deficit, and CO
2
concentration), Fukami et al. (1995) indicate that soil moisture and leaf area index
(LAI) are the most important factors controlling surface resistance based on
modeling "experience."
If leaf area index data were available (LAI data were not available for this
study.), then graphs like those shown in Figures 3.32 and 3.33 could be normalized
to surface resistance for a single leaf by multiplying the surface resistance by LAI.
Perhaps this would bring results for different months closer together.  With this
being considered, the results in Figures 3.32b and 3.33b are consistent with models
for surface resistance used by Dickinson (1984) and Noilhan and Planton (1989) in
which surface resistance is inversely proportional to water content and leaf area
index.  An inverse relationship with water content is similar to a linear relationship
with the logarithm of piezometric head shown here if one considers the popular
water retention relationship of Brooks and Corey in which water content is
exponentially related to the inverse of matric potential.  Knowledge of LAI would
not only be valuable for estimating surface resistance but can also be related to
vegetation height, which would help in computing the aerodynamic resistance (r
a
).
Although beyond the scope of this study, the use of remote sensing Advanced
Very High Resolution Radiometer Normalized Difference Vegetation Index
(AVHRR NDVI) data to infer leaf area index is a promising science.  Senay and
Elliot (1997) use AVHRR NDVI data to characterize seasonal vegetation changes in
the state of Oklahoma.
110
Surface resistance for selected days was also computed at site EF4.  The
concurrent availability and quality of required data (EBBR, SMOS, and SWATS)
was more limited at this site.  Therefore, only values for selected days in June, 1997,
are plotted.  Figure 3.35 shows a comparison of June 1997 values at the two sites
(EF7 and EF4) at a 5 cm depth.  Surface resistance values at site EF4 far exceed
those at site EF7.  Not enough information is available to determine the reason for
large discrepancies between sites EF4 and EF7, but clearly factors other than matric
potential are influencing the surface resistance.  Based on the discussion of leaf area
index in the last paragraph, one might infer that the leaf area index at site EF4 is
lower than that at EF7.  Unfortunately, leaf area index data were not available for
this study.
y = 16.06x + 183.5
R
2
 = 0.17
Site EF4
y = 19.21x - 17.93
R
2
 = 0.09
Site EF7
-100
0
100
200
300
400
500
246810
ln (|h+z|) (ln cm)
Surface Resistance (s/m)
Site EF7
Site EF4
Linear (Site EF4)
Linear (Site EF7)
Figure 3.35  A comparison of surface resistance (r
s
) vs. ln(|?+z||) at two sites:
EF7 and EF4.
111
3.3.2.4. Evaporation method used in this study: practical considerations
In the preceding sections, an attempt was made to develop empirical equations
for estimating surface resistance as a function of piezometric head.  The graphs
developed indicate that other factors besides soil piezometric head significantly
influence surface resistance.  There appears to be a seasonal influence on surface
resistance.  It is suspected that this seasonality is related to vegetation growth and
could perhaps be quantified using the leaf area index; however, leaf area index data
were not available for this study.  Further investigation into the use of the leaf area
index subject is suggested as future work.
Based on a review of the literature and the empirical analysis presented here,
the conventional approach to estimating evaporation under conditions of limiting
soil moisture specified by Equation 2.1 in Section 2.5 is the most practical approach,
and this approach is used for simulations described in Chapter 5.
A difficult question to answer when applying Equation 2.1 (or any function that
might be used to estimate evaporation) is to what soil depth is water being extracted
by plants.  Maximum rooting depth estimates may be obtained in the literature for
agricultural crops (Doorenbos and Pruitt, 1977) or for general vegetation types
(Canadell et al., 1996), but actual rooting depth is a more difficult parameter to
specify.  In this research, rooting depth (d
r
) is used as a model calibration parameter.
Physically-based pecification of rooting depth (i.e. as a function of vegetation type
and season) is left as a subject for future research.
The amount of water accessible to plants for evaporation depends on the
rooting depth, which varies with location and season.  When a soil root zone is
divided into multiple layers, one approach to estimating evaporation from each layer
is to take layer evaporation as the product of E
r
, the fraction of roots in that layer (?),
and f(?) (Shuttleworth, 1993).  Thus, in application to individual soil layers,
Equation 2.1 is modified to include the layer root fraction (?):
Equation 3.46
rco
EfE ??? )(= [mm day
-1
]
112
Root fraction is assumed to be proportional to soil layer depth in the proposed
model.  If the model user specifies a total rooting depth (d
r
), then the root fractions
for the top layer (layer 1) and bottom layer (layer 2) are given by Equation 3.47 and
3.48 respectively.
Equation 3.47
r
d
d
1
1
=?
Equation 3.48
r
r
d
dd
1
2
?
=? )(
21
ddd
r
+<
In the one-layer model, only the evaporation from layer 1 is computed.  For the two-
layer model, rooting depth must be less than the total soil zone depth.  The rooting
depth parameter (d
r
) is used to help calibrate the proposed model against observed
soil water content as discussed in Chapter 5.  Rooting depth significantly effects the
drying rate from layer 1 and consequently influences direct runoff.
For the simulation runs described in Chapter 5, the input data used to estimate
reference crop evaporation (E
r
) are taken from the EBBR data stream at site EF26
shown in Figure 3.19.  Equation 3.29 is used to estimate reference crop evaporation.
Using the definition prescribed by Shuttleworth (1993) for a reference crop, the
surface resistance (r
s
) is taken as 69 s m
-1
 and the crop height is taken as 0.12 m.
The wind speed at site EF26 is measured at 3.4 m and the humidity measurement
used is taken at 1.84 m.  Substituting these values into Equations 3.41 - 3.43 yields
the following relationship for estimating the aerodynamic resistance:
Equation 3.49
4.3
228
u
r
a
=  for EF26
3.3.3. Percolation
In both the one-layer and two-layer models, the water content of layer 1
controls the rate of infiltration into layer 1 and the rate of percolation out of layer 1.
The percolation out of layer 1 is a function of the water content in layer 1 only, and
113
the percolation out of layer 2 (if it exists) is a function of the water content in layer 2
only.  In the one-layer model, percolation out of layer 1 is passed directly to the
groundwater reservoir.  In the two-layer model, percolation from layer 1 is passed to
layer 2 and percolation from layer 2 is passed to the groundwater reservoir.
Percolation rates are specified using a kinematic model, which assumes that matric
potential gradients are negligible relative to gravity.  Based on this assumption, the
percolation rates for each layer are simply specified as the hydraulic conductivity
corresponding to the average effective saturation in the respective soil layers.  The
Brooks and Corey relationship between hydraulic conductivity and effective
saturation is used to define percolation (p)
Equation 3.50
)
2
3(
)(
j
jjsj
Kp
?
+
?=   j = 1, 2
The subscript j indicates that a separate computation is made for layer 1 and layer 2.
The bar over the effective saturation (?) indicates layer average effective saturation.
When percolation is controlled by a non-soil, low permeability layer such as
weathered bedrock (see Section 3.2.2.3), p
2
 is assigned a constant value specified by
the user.
3.3.4. Soil layer mass balance
Using the equations describing infiltration, evaporation, and percolation, the
mass balance on each soil layer is computed.  Computations during an event are
different than computations between events and will be discussed separately in the
next two sections.
3.3.4.1. Non-event mass balance for an individual time step
When there is no infiltration occurring, the mass balance equation on layer 1
can be written as follows
114
Equation 3.51 0)(
11
1
11
=??
?
? pe
dt
d
d
r
??
where the first term represents the change in storage, e
1
 represents evaporation, and
p
1
 represents percolation.  d
1
 is the depth of layer 1.  In Equation 3.51 both e
1
 and p
1
are non-linear functions of ?
1
 which means that Equation 3.51 is a non-linear, non-
separable, ordinary differential equation for which it is difficult if not impossible to
determine an analytical solution.  For this reason, a time stepping method is used to
solve this equation in which estimates of e
1
 and p
1
 are functions of the effective
saturation (?) at the beginning of the time step.  In the two-layer model, a time
stepping method is also used to compute the mass balance on layer 2.
The discrete mass balance equations for the case of no infiltration are:
Equation 3.52
t
r
dtt
t
d
pe
1
11
11
1
)(
)(
?+?
?
+
?=?
+
??
   layer 1
Equation 3.53
t
r
dtt
t
d
epp
2
22
221
2
)(
)(
?+?
?
??
=?
+
??
   layer 2 
Because of the highly non-linear nature of the percolation function, a limit (0.1)
is placed on the amount of change in saturation that can occur during a single time
step.  If this is not done, then a saturated layer with a coarse textured soil might
completely dry out during a one hour period if the percolation rate for the entire
hour is assumed to be governed by the effective saturation at the beginning of the
hour.  If the 0.1 limit is exceeded, the computation time step is cut in half until
either this limit is satisfied or a maximum number of time step subdivisions (128) is
reached.  Additional precautions are taken to make sure that effective saturation in
each layer never drops below 0 or exceeds 1.
3.3.4.2. Event mass balance
During a rainfall event, infiltration is modeled as piston flow, and the rate of
advancement of this piston is a function of initial water content of soil layer 1 and
115
the cumulative infiltration for that event.  To keep things simple, infiltrated water is
not redistributed throughout the soil layer(s) until the rainfall event has ended.
Evaporation is assumed to be negligible during a rainfall event, but percolation is
not neglected during an event.  Initial estimates of event percolation from both
layers are made based on the initial saturation levels.  By neglecting evaporation and
not accounting for infiltrated water until the end of the storm, an analytical solution
for the percolation from layer 1 during an arbitrary rainfall period can be
determined.  The mass balance equation for layer 1 is:
Equation 3.54 0)()(
)/23(
11
1
111
1
=?+
?
?
+ ?
??
sr
K
dt
d
d
Equation 3.54 can be separated and integrated from the known initial effective
saturation (?
1i
) to the unknown final effective saturation (?
1f
) over the time period t
i
to t
f
.
Equation 3.55 dt
d
K
d
f
i
f
i
t
t
r
s
??
?
?
+?
?
?=??
1
1
1
)(
11`1
1
1
)/23(
1
??
?
Solving for ?
1f
 yields:
Equation 3.56
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?? +
?
?
?
?
?
?
?++?
?
=?
11
1
1
2
1
2
2
1
1
11
1
1
)
1
1)((
)(
2
?
?
?
?
??
iif
r
s
f
tt
d
K
To simplify the mathematics for layer 2, the initial estimate for layer 2 percolation is
calculated independently of the added percolation from layer 1 during the event
(making the math identical to that for layer 1).  The exception occurs when the
percolation rate out the bottom of layer 2 is assumed to be a constant because of a
non-soil, low permeability layer.  In this case, the equation for the final saturation in
layer 2 is as follows:
Equation 3.57
i
r
f
t
d
p
2
222
2
2
)(
?+?
?
?=?
??
116
The initial estimates of total percolation (P
j
) (note that lower-case p stands for
percolation rate) for each layer during a storm are given by Equation 3.58.  The
units for total percolation are [cm] and the subscript j indicates layer number.
Equation 3.58 ( )()
jjrjfjij
dP ?? ????= j = 1, 2
Although the assumptions that allow these analytical solutions are imperfect from a
physical point of view, percolation computations are simplified greatly.  As
described in the next paragraph, if necessary, adjustments are made to P
1
 and P
2
 to
ensure that the physical limits on effective saturation are not violated.
At the end of a rainfall event, the water contents in each layer are adjusted to
account for the event percolation (P
j
) defined by Equation 3.58 and event infiltration
(I).  As much of the infiltrated water as possible remains in the top layer, without
allowing the effective saturation to exceed natural saturation.  Any remaining
infiltrated water is passed to the second layer.  The second layer may also become
full and if it does, the excess infiltration is added to the storm direct runoff and
distributed evenly over the storm time periods.  In the one-layer model, the total
depth of the soil zone is unspecified so any infiltration that exceeds that capacity of
layer 1 is treated as percolation.  In both models, layer 1 is allowed to fill only to ?
ns
because any higher saturation might cause an error in the infiltration calculations
(which only depend on the saturation in the top layer).  Layer 2 is allowed to fill to
? = 1.
Because percolation and infiltration are occurring simultaneously, but the
magnitudes for these flux computations are estimated independently from one
another, a series of checks are made to ensure that the physical limits on effective
saturation are not violated at the end of a storm.  A flow chart that illustrates the
sequence of calculations ensuring post-infiltration mass balance in the two layer
model is shown in Figure 3.37.
If
 [I 
<
 
(d
1
(
?
1n
s
 - 
?
1i
)+
P
1
)]
(
c
heck f
o
r
 available sp
a
c
e
i
n
 layer
 1)
 
If
 
?
1f
 
> 
?
1n
s
If
 
?
2f
 
>1
If
 I
r
 < 
(
d
2
(
?
2
 - 
?
2i
) +
 P
2
-P
1
)
(
c
heck f
o
r
 available sp
a
c
e
i
n
 layer
 2)
C
o
m
pute pos
t-
s
t
or
m
?
1f
 a
n
d 
?
2f
 
ba
s
e
d on
I, P
1
,
 a
n
d P
2
-
 C
o
mpute exces
s in 
layer
 2:
E
2
 = 
(
?
2f
 
- 1
`
)
(
?
2
-
?
2r
)d
2
- R
e
duce
 
P
1
 
by E
2
- Add E
2
 
ba
ck i
n
to layer
 1
- Se
t 
?
2f
 
= 1
-
 C
o
mpute exces
s in 
layer
 1:
E
1
 = 
(
?
1f
 
- 
?
ns
)(
?
1
-
?
1r
)d
1
- Add E
1
 
to dir
ect runof
f
- Se
t 
?
1f
 
= 
?
1n
s
Y
e
s
 (
u
nlikely, but
 c
o
uld 
oc
c
u
r
 i
f
 P
1
 >> 
P
2
)
ye
s
- Se
t 
?
1f
 = 
?
ns
- C
o
mpute remaining infiltration (
I
r
):
I
r
 
=
 (I
 -(
 d
1
(
?
1n
s
 - 
?
i
) +
 P
1
))
no
Ad
j
u
s
t
 
?
2f
 
t
o
 re
fl
e
c
t
 
I
r
.
F
i
nis
h
ed pos
t-
s
t
or
m
calculat
i
ons
- Se
t 
?
2f
 = 
1
- C
o
mpute remaining
i
n
f
ilt
r
a
tion (
I
rr
)
I
rr
 = 
I
r
 - (
d
2
(
?
2
-
?
2i
) +
 P
2
 - 
P
1
)
-Add I
rr
 
to dir
ect runof
f
ye
s
no
*
 s
ubscr
ip
t i indicates the
v
a
l
u
e a
t
 the beginning
 of
t
h
e s
t
orm and s
ubs
cr
ipt f
i
ndicates t
h
e
 
value at the
end
 of
 the s
t
orm
No
No
Ye
s
F
i
g
u
re 3.36  Post-infiltration mass balance flow 
chart fo
r the two-la
y
e
r model.
117
118
If [I < (d
1
(?
1ns
 - ?
1i
)+P
1
)]
(check for available space
in layer 1)
Compute post-storm
?
1f
 based on I and P
1
yes
- Set ?
1f
 = ?
ns
- Add remaining infiltration,
I
r
 = (I -( d
1
(?
1ns
 - ?
i
) + P
1
)) to P
1
.
no * subscript i indicates the value at the
beginning of the storm
and subscript f indicates the value at
the end of the storm
Figure 3.37  Post-infiltration mass balance flow chart for the one-layer model.
3.3.5. Runoff transform method : ?modClark?
Although watershed runoff routing is not the focus of this research, a runoff
transform method is included so that predicted rainfall excess can be compared with
observed runoff.  Direct runoff from each model cell is transformed into a
streamflow response at the watershed outlet using the modClark transform method.
The modClark method was developed at the U.S. Army Corps of Engineers
Hydrologic Engineering Center (HEC) for use in quasi-distributed modeling with
NEXRAD rainfall data (HEC, 1995; Peters and Easton, 1996).  The modClark
method is also applied using NEXRAD data here; however, the same methodology
can be applied to any arbitrarily defined precipitation polygons.  ModClark is an
adaptation of the Clark unit hydrograph method.  In modClark, direct runoff pulses
from each cell are first translated to the outlet based on a travel time that is linearly
proportional to travel length.  At the outlet, arrays of lagged direct runoff pulses are
summed and then passed through a linear reservoir to generate the direct runoff
hydrograph.  As in the original Clark method, the linear reservoir is intended to
represent stream channel storage effects.
119
In addition to the time series of rainfall excess determined from the soil-water
balance, implementation of the modClark method requires estimates of (1) the
average flow length from each model cell to the watershed outlet, (2) cell areas, (3)
the time of concentration (T
c
) for the watershed as a whole, and (4) a linear reservoir
storage coefficient (R).  Area is an attribute of the model polygons shown in Figure
3.18.  The average flow length for each NEXRAD polygon is determined by
averaging the flow lengths of all digital elevation model cells (Figure 3.3) within
that polygon.  There are approximately 1,600 of the 100-m DEM cells  and 17,778
of the 30-m DEM cells in a NEXRAD cell.  A map showing the average flow length
for each model polygon is shown in Figure 3.38.  T
c
 and R may be estimated from
previous studies, regionalization, or calibration (Peters and Easton, 1996).  The
travel time (T
m
) from a model cell to the watershed outlet is estimated using the
following equation:
Equation 3.59
max
FL
FL
TT
avg
cm
=
In Equation 3.59, FL
avg
 stands for the average flow length from a given NEXRAD
cell and FL
max
 stands for the maximum of the cell average flow lengths.
Equation 3.60 is used to route water through a linear reservoir at the outlet
(HEC, 1995).
Equation 3.60
dout
tt
din
dout
t
Q
tR
t
Q
tR
t
Q
???
?
?
?
?
?
?+
?
?+
?
?
?
?
?
?
?+
?
=
)5.0(
1
5.0
din
Q  is the average flow rate into the linear reservoir during the current time step.
dout
tt
Q
??
 is the flow rate out of the linear reservoir at time t-?t, and 
dout
t
Q (the unknown
quantity) is the flow rate out of the linear reservoir at time t.  The ?d? in the
superscript is intended to distinguish direct runoff from baseflow.  To be consistent
with the original references, this mathematical equation describing a linear reservoir
differs slightly from that developed in the next section for the baseflow component,
120
however, the numbers resulting from calculation with the two different forms are
very similar.
Figure 3.38  Average flow length from each model polygon to the watershed
outlet.
3.3.6. Baseflow
Two methods are used to generate baseflow in this study, baseflow simulation
with a single linear reservoir model and baseflow separation using observed flow
data.  Both methods require a single parameter that cannot be estimated from the
physical characteristics of the watershed.  The baseflow separation approach is
simpler to apply and is adequate to address the objectives of this research dealing
with the effects of soil heterogeneity on direct runoff generation.  When the
121
baseflow separation method is used, any percolation out the bottom of the soil zone
is treated as a sink in the model.  With the single linear reservoir baseflow model,
percolation out the bottom of the soil zone is passed to the linear reservoir.
Although not absolutely necessary to meet the objectives of this research, a
baseflow simulation scheme is necessary to apply the proposed model in areas
where no observed flow data is available.
It will be shown in Chapter 5 that use of the two-layer soil model in
conjunction with the single linear reservoir baseflow model may provide reasonable
baseflow results, but when the one-layer soil model is used, the single linear
reservoir model is not adequate to simulate baseflow response.  This is because too
much water gets passed to the linear reservoir in the one layer model.  Existing
models like PRMS (Leavesley et al., 1983) and the HEC Continuous Soil Moisture
Accounting Model (unpublished by Bennett, 1997) allow for multiple groundwater
storage reservoirs.  One of these reservoirs might serve the same role as the second
soil layer in the model used here.  Multiple groundwater zones may also be used to
distinguish between shorter and longer term recession responses for a basin
(sometimes referred to as ?subsurface? and ?groundwater? flow respectively).  The
drawback to introducing multiple groundwater zones is the addition of more
parameters that cannot be estimated from watershed characteristics.  In the HEC
Continuous Soil Moisture Accounting model, the user must supply at least two
parameters for each groundwater zone ? the maximum storage in that zone and a
linear reservoir storage coefficient for modeling the release of water.  A third
parameter, maximum percolation rate, is required for a given zone if another zone
exists below it.
More detailed descriptions of the simulation and separation baseflow
approaches are provided in the next two sections.
122
3.3.6.1. Single linear reservoir model
Using a linear reservoir to model baseflow is a popular approach because the
linear reservoir model generates exponentially decaying baseflow that is consistent
with observed hydrographs and theoretically consistent with the equation for 1-
dimensional flow from a phreatic aquifer (USACE, 1994).  Although the
mathematics of linear reservoirs are widely documented, an explanation is included
here for the sake of completeness.  By definition, the storage (S [L
3
]) in a linear
reservoir is proportional to the outflow (Q
b
 
[L
3
 T
-1
]) with ? [T] being the constant of
proportionality (reservoir constant) as shown in Equation 3.61.
Equation 3.61 S = ? Q
b
  
Combining Equation 3.61 with the continuity equation (Equation 3.62) and then
integrating from time t to time t+?t yields Equation 3.63, which can be used to
calculate the discharge from the reservoir during any time interval.
Equation 3.62
b
QInflow
dt
dS
?=    
Equation 3.63
?
?
?
?
?
?
?+=
????
?+
??
t
t
t
b
t
b
tt
eInfloweQQ 1
Inflow
t
 is the sum of the percolation out of the bottom soil layer for all soil
components in all cells.  
b
t
Q  is the baseflow at time t and 
b
tt
Q
?+
 is the baseflow at
time t+?t.
Using only one groundwater reservoir, there is only one groundwater
parameter, the reservoir storage coefficient (?).  This is a lumped parameter for the
entire watershed and is determined by calibration against observed flow data.
123
3.3.6.2. Baseflow separation
In this study, the purpose of the baseflow separation is to extract a direct runoff
hydrograph from the observed flow record for comparison with the direct runoff
hydrograph predicted by the soil-water balance model.  Historically, a number
different methods have been used for baseflow separation.  Chow, Maidment, and
Mays (1988) describe three possible baseflow separation techniques.  A baseflow
separation technique described by Olivera, Maidment, and Charbeneau (1996) is
used here.  The baseflow separation program described by Olivera, Maidment, and
Charbeneau (1996) was written in FORTRAN but has been rewritten in Visual
Basic with slight modifications for this research.  The main algorithm used by the
baseflow separation program is unchanged.  This baseflow separation algorithm is
illustrated in Figure 3.39.  In the example of Figure 3.39, baseflow at the first time
step (t(1)) is set equal to the total flow Q
b
(1) = Q(1) at time 1.  The baseflow at the
next time step Q
b
(t+1)
 
depends on the separation parameter n, which specifies the
number of future hours that are considered in estimating Q
b
(t+1).  The parameter n
is taken as 3 hours in Figure 3.39.  n lines are drawn from the known point Q
b
(1),
one to each of the n subsequent observed flow values (Q(t)), and the slopes of these
lines are calculated.  The baseflow hydrograph is assumed to follow the minimum
slope defined by this set of lines.  Thus Q
b
(t+1) is computed using the following
equation:
Equation 3.64
?
?
?
?
?
?
?
? ???
+=
+++
+
n
QQQQQQ
QQ
b
tnt
b
tt
b
ttb
t
b
t
,...
2
,
1
min
21
1
Choice of the parameter n can significantly influence the baseflow estimates
derived using this method.  An example application using streamflows from July 7 ?
July 17 at the USGS flow station East of Ninnekah on the Little Washita River is
shown in Figure 3.40 to illustrate this point.  For the observed direct runoff
hydrographs shown in Chapter 5, an n value of 60 hours was used.  If a different n
value was chosen, then some of the calibrated parameter values presented in
124
Chapter 5 would change.  The absolute accuracy of direct runoff predicted in the
validation runs might also change, but the relative performance of the different
modeling scenarios would likely remain the same.
Although there is no physical basis for selecting an n value, there is no physical
basis for the specification of the storage constant (?) used in the baseflow
simulation scheme either.  The baseflow separation method described here is
attractive because it is easily automated and can be applied consistently to different
storms in a long flow record.
125
0
20
40
60
80
024681012
Slope j = 1:  (12 -10)/1 = 2
Slope j = 2:  (25 - 10)/2 = 7.5
Slope j = 3:  (60 - 10)/3 = 16.66
Q
b
(1) = 10
Q
b
(2) = 10 + 2 = 12
j = 1
j = 2
j = 3
Fl
ow
 (
c
fs
)
(a)
0
20
40
60
80
024681012
Slope j = 1:  (25 -12)/1 = 13
Slope j = 2:  (60 - 12)/2 = 24
Slope j = 3:  (40 - 12)/3 = 9.3
Q
b
(3) = 12 + 9.3 = 21.3
j = 1
j = 2
j = 3
Fl
ow
 (
c
fs
)
Flow
Baseflow
(b)
0
20
40
60
80
024681012
Time (hours)
Fl
ow
 (
c
fs
)
Slope j = 1:  (60 -21.3)/1 = 38.7
Slope j = 2:  (40 - 21.3)/2 = 9.3
Slope j = 3:  (30 - 21.3)/3 = 2.9
Q
b
(4) = 21.3 + 2.9 = 24.2
(c)
Figure 3.39  Baseflow separation method. (a) estimating baseflow at t = 2, (b)
estimating baseflow at t = 3, and (c) estimating baseflow at t = 4.
126
-500
0
500
1000
1500
2000
2500
0 50 100 150 200 250
Time (hours)
Flow (cfs)
Observed Flow n = 15 n = 30 n = 60
Figure 3.40  Baseflow separation using n = 15, 30, and 60 hours.
3.3.7. Special considerations for the Little Washita
Although the Little Washita River Watershed is an attractive study area due to
the large amount and variety of hydrologic data available, one drawback in studying
this watershed is that it contains 45 floodwater detention structures (as of 1985)
installed by the Soil Conservation Service (Allen and Naney, 1991) which alter the
natural hydrology of the watershed.  These detention structures influence 46% of the
watershed area, and appear to have a significant influence on the hydrographs.  It is
127
hypothesized that the recession limbs of storm hydrographs do not exhibit obvious
logarithmic recession because of these detention structures.  For the purposes of
predicting direct runoff volume, the detention structures are ignored because it is
assumed that these structures primarily affect runoff timing and not total storm
volume.  The storage effects of these reservoirs are roughly accounted for in the
modClark parameters that are used to transform rainfall excess to the watershed
outlet.
In addition to these man-made detention structures, Allen and Naney (1991)
report that 19.5 percent of the Little Washita watershed was controlled by farm
ponds as of 1983.  According to Allen and Naney, these farm ponds drain areas that
do not contribute to runoff.  In application, 19.5 percent of the direct runoff
predicted by the model is assumed to flow to farm ponds where it is lost to
evaporation.  An option is available in the Visual Basic user interface to specify the
fraction of contributing area.  If this estimate of non-contributing areas is not
accurate, then the calibrated model parameter values given in Chapter 5 would
change slightly, but this is not likely to alter the conclusions reached in Chapter 5
because the same assumption is made for all modeling scenarios being compared.
3.4. Summary
A study of the available data and relevant literature has led to the development
a soil-water balance and runoff model for which many of the required parameters
can be estimated from available spatial databases.  Before continuing on to a
discussion of computer implementation (Chapter 4) and modeling results (Chapter
5), a summary of the key model features is provided here.
To implement the model, a watershed is divided into cells of uniform
precipitation (NEXRAD rainfall cells in this case) and vertical flux calculations are
made independently for each rainfall cell.  Rainfall excess in each cell is
128
transformed to the watershed outlet using the modClark method developed at the
Hydrologic Engineering Center.
Vertical flux calculations are dependent on soil hydraulic properties that vary in
space.  The model is designed so that different levels of spatial resampling in
describing soil properties can be tested.  Figure 3.4 shows that there are significant
differences in the spatial detail provided by the 1:250,000 STATSGO soils data and
the 1:24,000 30-m soils grid; however, the overall surface textural composition for
the watershed predicted from the two data sets is similar (Figure 3.6b).
A straightforward method to estimate soil hydraulic properties (K
s
, ?, ?
r
, ?, h
b
,
?
fc
, and ?
wp
) from soil texture is proposed.  This method involves two lookup tables.
The first table translates texture names in the STATSGO database into one of the 12
basic USDA texture classes or the "other" category, and the second table assigns
mean hydraulic properties to each texture class.  Two key papers by Rawls,
Brakensiek, and Saxton (1982) and Carsel and Parrish (1988) tabulate mean values
for hydraulic parameters in each texture class.  The water retention parameter values
(?, h
b
) reported in these two papers are considerably different.  The reason(s) for
these differences cannot be discerned from the literature but the Carsel and Parrish
values have been selected for use here because they are more consistent with Figure
5.5.4 of Rawls et al. (1993).  The sensitivity of model results to specifying a range
of parameter values is discussed in Section 5.6.
Relatively simple infiltration, evaporation, and percolation models have been
described.  Simple infiltration (Green-Ampt) and percolation (kinematic) models
described in the literature are easily amenable to a continuous soil-water balance
scheme.  Large uncertainties in parameter estimates and input data justify the use of
simple flow models.  Models available in the literature for predicting evaporation
under soil moisture limited conditions are either theoretically weak or intractably
complex.  The possibility of developing a model of intermediate complexity was
explored.  An attempt was made to develop an empirical relationship between
129
evaporative surface resistance and soil suction using detailed meteorological data
from DOE ARM CART measurement sites.  Looking at this data, it is clear that this
approach will not have general applicability because other factors besides soil
moisture are strongly influencing surface resistance.  This is a subject for further
study.  A conventional method for estimating evaporation discussed in Section 2.5
has been implemented in the proposed model.
It is a relatively simple task to partition the surface soils in a watershed into
different texture classes, but considering increased vertical subdivision significantly
complicates the problem.  Based on the type of layer information available and the
desire to keep the model as simple as possible, use of only a one-layer and a two-
layer soil model are considered.  The two-layer model uses more information
provided in the soils databases.  In the two-layer model, the top layer controls direct
runoff while the bottom layer primarily serves as a storage reservoir for infiltrated
water.  Specific differences in model complexity, in terms of computer
implementation, are described in Chapter 4.
130
CHAPTER 4. COMPUTER IMPLEMENTATION
To implement the model described in Chapter 3, there are three major steps: (1)
collect spatial and time series data and convert these data into the required formats
as necessary, (2) analyze spatial data to define modeling units and estimate model
parameters, and (3) make hydrologic computations.  Acquisition and formatting of
input data is discussed briefly in Section 4.1.
Spatial analysis and parameter estimation are automated using the ArcView
GIS and its Avenue script language.  ArcView procedures are referred to as pre-
processing steps.  Hydrologic computations are done using Visual Basic.  The
Avenue scripting language is not used for the hydrologic calculations because the
Avenue language is interpreted, and tends to be much slower than compiled
languages like Visual Basic.  Collectively, the programs developed in this research
are given the name Soil Water Balance Modeling System (SWBMS).  The ArcView
pre-processor is referred to as AV-SWBMS and the Visual Basic programs are
referred to as VB-SWBMS.
In addition to speed issues, an important reason for separating the spatial
analysis computations from the hydrologic computations is to make it possible to
pass parameter information derived in the GIS environment to existing hydrologic
models developed elsewhere, like HEC-HMS.  Working with researchers at HEC
over the past several years, a simple strategy of passing parameter information from
GIS to HEC-HMS through ASCII files has proven effective.  The same scheme is
used here, but hydrologic calculations are done using VB-SWBMS.
Sections 4.2 and 4.3 are devoted to a discussion of the two principal model
components, AV-SWBMS and VB-SWBMS.
4.1. Collecting and reformatting input data
In this study, collecting and reformatting the spatial and time series data was
the most tedious of the three major processing steps because it is the least
131
automated, but this step is typically only done once.  The sources for acquiring
spatial and time series data are given in Table 3.1.  The preparation of time series
data was more time consuming than the preparation of spatial data because the time
series data that were obtained from different sources are distributed in different
formats.  Therefore, customized data reformatting programs are required for each
time series data source so that the data can be read by the AV-SWBMS code.  Since
the programs for data conversion are not critical to the modeling concepts, a
description of these programs is left to Appendix B.
The spatial data, including soils data, digital elevation models, and land use
data are distributed in standard formats.  Procedures for converting spatial data
sources from their original formats to formats readable by the ArcView software are
standard GIS operations.  In addition to the information found in this chapter,
information about converting these standard spatial data sources may be found at the
web sites listed in Table 3.1, in the Arc/Info or ArcView software documentation, or
in one of the references cited in Section 3.2.  As discussed in Section 3.2, all spatial
data must be converted to a common map projection.
4.2. The ArcView Pre-Processor (AV-SWBMS)
AV-SWBMS prepares parameter files for hydrologic calculations in VB-
SWBMS.  AV-SWBMS is packaged as an ArcView extension, which contains a set
of scripts and associated menus that are added to the View GUI when the extension
is loaded.  The View GUI includes menus, buttons, tools that are seen when an
ArcView user has a View window active.  In ArcView, View windows are used to
display and manipulate map layers called Themes.  A Theme may contain vector or
raster (gridded) data.  A summary of the Themes and tabular data required to run the
AV-SWBMS programs is provided in Table 4.1.  All of the input Themes must
provide complete coverage for the watershed of interest and be in the same map
projection.
132
Table 4.1 Spatial data required to run AV-SWBMS
Theme Suggested method of acquisition Comments
A polygon Theme
defining cells of
uniform rainfall. (i.e.
NEXRAD cells)
Programs for generating a polygon
coverage of NEXRAD rainfall cells
are described at
www.ce.utexas.edu/prof/maidment/gis
hydro/seann/nexrad.htm and in Reed
and Maidment (1995).
Although not needed for AV-
SWBMS, a time series file
containing a column for each
rainfall cell must be related to
cells in this Theme to run a
hydrologic simulation.  The
rainfall cells do not have to
be uniform in shape.
A watershed
polygon
A watershed polygon can be derived
from a digital elevation model using
the Watershed Delineator, CRWR-
Prepro, or the Hydrologic Modeling
extension (Olivera, Reed, and
Maidment, 1998).  Information on
obtaining digital elevation models
through the Internet may be found at
www.ce.utexas.edu/prof/maidment/gis
hyd97/library/websites/webtoc.htm.
At this time, the SWBMS
system can only model single
watersheds because there is
no river routing component.
A grid Theme with
values of flow
length from each
cell to the watershed
outlet
The flow length grid can also be
derived from a digital elevation model
using either the Watershed Delineator,
CRWR-Prepro, or the Hydrologic
Modeling extension (Olivera, Reed,
and Maidment, 1998).
A STATSGO soil
polygon Theme for
the area of interest
with the associated
component
(comp.dbf) and layer
(layer.dbf) tables or
a soil grid Theme
and associated tables
www.ftw.nrcs.usda.gov/soils_data.ht
ml
The AV-SWBMS programs
should also work with
SSURGO vector data as is or
with minor modifications.
Use of SSURGO vector data
has not been tested.
A land use polygon
coverage for which
the attribute table
contains a field with
Anderson Level II
land use code.
www.ce.utexas.edu/prof/maidment/gis
hyd97/library/websites/webtoc.htm or
www.epa.gov/ngispgm3/nsdi/projects/
giras.htm
133
The AV-SWBMS extension contains two menus, one called "SWBMS" and
one called "SWBMS-Twolay," that become visible when a View is active.  A screen
capture of these two menus is shown in Figure 4.1.
?SWBMS? ?SWBMS-Twolay?
Figure 4.1 Screen capture of menus that are part of AV-SWBMS.
The ?SWBMS? menu contains a sequence of programs that are run to create an
input file for a one-layer model run, and the ?SWBMS-Twolay? menu contains
additional programs that are required to create input files for a two-layer model.  A
single ASCII file is created as input for a one-layer VB-SWBMS run, and two
ASCII files are created for a two-layer VB-SWBMS run.  Important differences in
the data transfer between the one-layer and two-layer models are discussed in
Sections 4.2.1 and 4.2.2.
For both the one-layer and two-layer models, separate algorithms are required
to deal with the STATSGO soils data in vector format and the 30-m gridded soils.
In addition, separate algorithms are required to prepare an input file for the different
resampling levels.
4.2.1. Pre-processing for a one-layer model
ArcView pre-processing steps used to create a one-layer VB-SWBMS input file
are summarized in Table 4.2.  Table 4.2 lists each item under the menu "SWBMS"
and explains its purpose.  A more detailed description of each of these programs is
provided in Appendix A.1.  The first three menu items are generic tools that are
134
used both for the one-layer and the two-layer model while the last three menu items
are specific tools that generate input files for VB-SWBMS.
The format of the VB-SWBMS input files is the same regardless of whether
vector or gridded soils data are used as input.  The transfer file format for NEXRAD
cell level resampling is slightly different from that for no resampling as shown in
Figure 4.2.  Note that only one line of information is required for each NEXRAD
cell in the sample input files shown in Figure 4.2.  The actual input files are comma
delimited but are presented here in columnar format to be more readable.
In the descriptions of the last three menu items in Table 4.2 there is no mention
of creating an input file for watershed level resampling.  To create input files for the
watershed level resampling option in this study, it was easiest just to edit the input
files for the no resampling option using Microsoft Excel.  To do this, the
percentages for the predominant watershed soil type (i.e. silt loam) were set to 100
for all cells, while the percentages for all other soil textures in all cells were set to 0.
Although all the pre-processing procedures described in Table 4.2 are
automated, special situations may arise when it is necessary to manually edit an
entry in one of the ASCII transfer files.  A situation that requires manual editing
occurs when only a tiny portion of the watershed being modeled intersects a
NEXRAD cell.  If this tiny intersected polygon is small enough so that it contains
no underlying elevation grid cells, then the flow length from that tiny polygon to the
watershed outlet is recorded as a Null Number.  A problem will also occur if the
program attempts to determine the soil type(s) within this polygon if gridded soils
data are being used.  This problem occurs where the Little Washita watershed
intersects NEXRAD cell (553,306) forming a thin polygon with an area less than
0.01 km
2
.  Grid cell resolution is almost always an issue when dealing with both
grid and vector data layers.  In the author?s experience, polygon/polygon operations
are generally more robust than polygon/grid operations.
135
Table 4.2  Description ArcView pre-processing programs under the menu
?SWBMS.?*
Step # Must be
preceded
by steps
Menu item Description
1 Intersect Intersect watershed polygon with rainfall cells to define
modeling units.
2 1 Percent
Impervious
Compute percent imperviousness in each model
polygon.
3 1 Polygon/Grid
Avg. Property
Determine the average value of all grid cells in each
model polygon or the most prevalent value of the grid
cells in each model polygon.  This program is used to
compute the average flow length to the watershed
outlet in each model polygon and is also used to
compute the predominant soil type grid code in each
cell when gridded soils data and NEXRAD cell level
resampling are used.
4 1, 2, 3 SWBMS Input
File
(STATSGO)
Creates an input file for VB-SWBMS using STATSGO
soils data.  Tallies the percentage of each soil type that
falls within that polygon.  The user can choose between
writing a file with NEXRAD cell level resampling or
no resampling.  Example input files for NEXRAD cell
level resampling and no resampling are shown in
Figure 4.1.
5 1, 2, 3 SWBMS Input
File (gridded
soils)
Creates an input file for VB-SWBMS using 30 m soils
data.  Tallies the percentage of each soil type that falls
within each model polygon.  Creates an input file with
no resampling.
6 1, 2, 3 SWBMS Input
File (grid,
most likely
value)
Creates an input file for VB-SWBMS using 30 m soils
data and NEXRAD cell level resampling.  Program
assumes that predominant soil type in each model
polygon has already been computed from the soils grid
using the menu item Polygon/Grid Avg. Property.
*  A more detailed description of each of these steps is provided in Appendix A.1.
F
i
gur
e
 4.2  Exa
m
ple
 input file
s for
 the
 one
-
laye
r
 m
o
de
l.
Ce
ll prop
ert
ie
s
 fi
le
 
for
 
NE
XRAD ce
ll le
ve
l resa
mp
lin
g:
Ce
ll prop
ert
ie
s
 fi
le
 
for
 
no
 resa
mp
lin
g:
No
te
s:
  
 Uni
t
s
 f
o
r
 cel
l
a
r
ea 
ar
e m
2
.
  
 U
n
i
t
s fo
r flo
w
le
n
g
th
 
a
r
e
 m.
No re
sa
m
p
lin
g fil
e
 c
o
n
tin
ue
d.
hr
a
p
x
h
r
a
p
y
c
e
l
l
a
r
e
a
f
lo
w
l
e
ngt
h
p
c
t
_
i
m
p
so
i
l
t
ype
560
311
1570369
7329
14
6
561
311
356931
5836
0
6
556
310
3402604
27999
0
3
557
310
10245727
25071
5
3
..
.
.
..
..
.
.
..
h
r
apx
h
r
a
py
c
e
l
l
ar
e
a
f
l
ow
len
g
t
h
p
c
t
_im
p
pc
t_S
p
c
t
_L
S
p
c
t
_SL
p
c
t
_SC
L
. . .
560
311
1570369
7329
14
0
0
6.364
0
.
 . .
561
311
356931
5836
0
0
0
0
.557
0
.
 . .
556
310
3402604
27999
0
0
48.616
51.384
0
 . . .
557
310
10245727
25071
5
0
10.409
89.591
0
 . . .
.
.
.
.
..
..
.
.
.
.
.
..
..
.
pct
_
L
p
ct
_SI
L
pct
_
C
L
pct
_
SI
pct
_
C
p
ct
_SC
p
ct
_SI
C
L
pct
_
SI
C
p
ct
_O
37.457
56.179
0
0
0
0
0
0
0
23.354
76.089
0
0
0
0
0
0
0
0
0
0
000
00
0
0
0
0
000
00
0
.
.
.
...
..
.
.
.
.
...
..
.
136
137
4.2.2. Pre-processing for a two-layer model
Data processing for the two-layer model is much more complicated than that
for the one-layer model.  This is because more than one factor is used to distinguish
mapped soil components in the two-layer model.  In the one-layer model, surface
texture is the only distinguishing factor while in the two-layer model there are three
distinguishing factors for each soil component: top layer texture, bottom layer
texture (which is the texture of the most restrictive layer in the layer table), and total
soil zone depth.  When the two-layer model is applied, only the no resampling
option is used because it does not make sense to compute the predominant soil type
in two different layers and the predominant soil zone depth independently of one
another.
Another difference in the way the one-layer and two-layer models are
implemented in this research is that hydraulic properties associated with each soil
texture are assigned within AV-SWBMS for the two-layer model but within VB-
SWBMS for the one-layer model.  This difference reflects a change in philosophy
that occurred during the course of this research.  It now seems most logical keep the
amount of information that is passed from AV-SWBMS to VB-SWBMS to a
minimum.  Passing only texture information rather than each parameter value
associated with that texture is one way to reduce this amount of information.  The
one-layer model was developed more recently and therefore adheres to this newer
philosophy.
The programs under the pull-down menu ?SWBMS-Twolay? that are used to
create input files for a two-layer SWBMS run are described in Table 4.3.  Items 1, 2,
and 3 from the ?SWBMS? menu (Table 4.2) are prerequisites to run items 3, 4, and
5 in Table 4.3.  Only a general description of the programs is provided in Table 4.3
with more details provided in Appendix A.1.
Two ASCII files get passed from AV-SWBMS to VB-SWBMS for the two-
layer model, a cell properties file and a component properties file.  The format for
138
both of these files is shown in Figure 4.3.  Each line in the cell file contains a
number indicating the number of components in that cell.  The component file is
ordered such that all of the components for the first cell are listed first, components
for the second cell are listed second, and so forth.  The actual input files are comma
delimited but are presented here in columnar format to be more readable.
Table 4.3 Description of menu items used to create input files for a two-layer
model.
# Must be
preceded by
steps
Menu item Description
1 STATSGO
Comp
Properties
This program processes the comp and layer soil
attribute tables.  The program loops through all
components in the comp table and for each component
loops through the associated layers.  The program
creates a new data table, which lists each component
and the associated soil hydraulic properties for the top
layer and the most restrictive layer in that component
and the total depth for that component computed using
the criteria described in Section 3.1.1.3.
2 30 m Grid
Comp
Properties
Performs exact same functions as above, but there are
differences in the code to account for differences in the
attribute tables.
31 from
Table 4.1
and 1 from
this table.
Comp Props
File
(STATSGO)
Creates one of two required input files for VB-
SWBMS (the component properties file) using the
STATSGO Theme as input.  One line gets written to an
ASCII file for each distinguishable component in each
model cell. The format for this file is shown in Figure
4.3.
41 from
Table 4.1
and 2 from
this table
Comp Props
File (Gridded
soils)
Same as above except the 30-m soils grid is used as
input.  Algorithm is slightly different due to the use of
gridded data but the result is identical.
5 1, 2, and 3
from Table
4.1 and
either 3 or 4
from this
table
Cell Properties
File
Creates the second of two required input files for VB-
SWBMS, the cell properties file.  See Figure 4.3.
F
i
g
u
re 4.3 Ex
ample input files for the two-l
a
y
e
r model.
hrap
x
h
rap
y
ce
l
l
area
num
co
m
p
s
f
lo
w
l
engt
h
p
ct_im
p
560
311
287
20
67
12
756
1.
4
6
13
561
311
522
94
1
4
616
3.
6
6
3
556
310
267
87
11
11
278
12
.
3
0
..
.
.
.
.
..
.
.
.
.
Co
mponent properties file:
Cell p
roperties 
file
:
Th
e
 first 1
2
co
mpon
en
t
s
li
s
t
e
d
 i
n
 the
co
mpon
en
t
 pro
p
e
r
t
i
es
t
a
b
l
e
 ar
e 
f
o
r
ce
ll
 (5
6
0
,3
1
1
).
m
u
i
d
gi
scl
a
ss
co
m
p
p
c
t
t
ot
de
pt
h
k
l
u
p
o
rl
u
t
h
e
t
a
_
r
l
a
m
b
da
hb
f
c
w
p
51
002
71
6.
675
16
2.
6
0
.
4
5
0
.
4
5
0
.
067
0.
41
50
0.
239
0.
103
7
51
014
82
1.
159
5
1
1
6
.
8
0.
45
0.
45
0.
067
0.
41
50
0.
239
0.
103
7
51
015
83
0.
689
4
1
1
4
.
3
0.
45
0.
45
0.
067
0.
41
50
0.
239
0.
103
7
51
017
84
23
.
2
2
1
6
1
8
2
.
9
0.
45
0.
45
0.
067
0.
41
50
0.
239
0.
103
7
..
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
r_kl
u
r
_p
o
r
l
r
_t
het
a
_
r_lam
b
d
a
r_hb
r_f
c
r_w
p
0.
0
2
0.
3
6
0.
0
7
0.
0
9
20
0
0
.
3
36
7
0
.
2
66
1
0
0
.
4
5
0
.
0
67
0.
4
1
50
0.
2
3
9
0
.
1
03
7
0
0
.
4
5
0
.
0
67
0.
4
1
50
0.
2
3
9
0
.
1
03
7
0
0
.
3
8
0
.
0
68
0.
0
9
12
5
0
.
3
46
6
0
.
2
70
3
..
.
.
.
.
.
..
.
.
.
.
.
..
.
.
.
.
.
139
140
4.3. The Visual Basic Hydrology Program (VB-SWBMS)
Several different versions of the VB-SWBMS program have been created to
investigate the different modeling alternatives discussed in Chapter 3 and Chapter 5.
All equations used to make calculations in VB-SWBMS are described in Chapter 3.
The overall sequence of computations is the same for different program versions,
although the details differ.  The base time step for all model versions is one hour.
The pseudo-code in Figure 4.4 summarizes the overall computational scheme for the
main program.  Several subprograms are called from the main program to perform
tasks listed in Figure 4.4.  A description of all the subprograms is provided in
Appendix A.2.
G89 initialization and setup
? open output files
? initialize variables
? read input files
G83 cell properties file
G83 component properties file (only for two-layer model)
G83 texture to soil properties lookup table
G83 precipitation file
G83 climate data used to estimate potential evaporation
G83 initial soil moisture levels (optional)
G83 observed flow data (optional)
G83 file specifying cells and components for which daily soil water content should be
exported (optional)
G89 loop through all computational cells
G83 initialize precipitation and direct runoff arrays for the current cell
G83 loop through soil components (or 12 possible soil types)
GBE initialize soil hydraulic properties for current component
GBE loop through simulation time steps
G99 if rainstorm is reached
G39 increment time steps and store storm precip array
G39 compute direct runoff due to infiltration excess (GA
algorithm - see Figure 3.25) and impervious cover and
increment totals for the current cell for all storm time steps
G39 make post-storm soil moisture adjustments (see Figures 3.36
and 3.37)
G39 increment subsurface recharge for the current cell in all
storm time steps
G99 else if no rain
G39 estimate evaporation and percolation based on initial water
content in each layer
141
G39 compute mass balance on each soil layer
G39 if change in effective saturation is greater than 0.1 then cut
time step in half and repeat mass balance calculations
G39 after mass balance, make sure physical limits on effective
saturation are not violated -- adjust results if necessary
G39 increment subsurface recharge for the current cell
G39 increment time step
G99 end if rain
GBE end of loop through time periods
GBE check mass balance on soil zone layers
G83 end of loop through soil components
G83 write final soil moisture values for each component to a file
G83 convert units on direct runoff and subsurface recharge for each cell and for all times
G83 compute modClark translated runoff array in each cell
G89 end of loop through cells
G89 route translated runoff array for each cell through modClark linear reservoir
G89 compute baseflow (optional)
G89 check mass balance on the entire watershed
G89 write output files
Figure 4.4  Pseudo-code describing the overall computational sequence for VB-
SWBMS.
A number of checks were made during code development to eliminate errors.
Two specific checks that are described here apply to two of the more complicated
sections of code, infiltration calculations and soil layer mass balance calculations.
The Green-Ampt infiltration code was checked against Example 5.4.1 in Chow,
Maidment, and Mays (1988).  There are two places in the VB-SWBMS where mass
balance checks are written into the code.  A mass balance check ensures that the
inflow minus the outflow is equal to the change in storage within a specified
tolerance.  A mass balance check for the soil zone is included at the end of the loop
through all time steps in each component.  This check is applied both to the top
layer and to the soil zone as a whole (in the two-layer model).  A mass balance
check for the watershed as a whole is also included at the end of the program.  Error
messages are given if tolerances get violated during these mass balance checks.  The
mass balance checks are not critical to the operation of the program, but are left in
as a debugging tool to help with any future algorithm changes.
A list of input files required by VB-SWBMS is provided in Table 4.4 and a list
of output files is provided in Table 4.5.  The formats for the files that are used to
142
transfer spatial information from ArcView are described in Figures 4.1 and 4.2.  The
required formats for other input files and the formats for output files are described in
Appendix A.  A description of the VB-SWBMS user interface and brief operating
instructions are is also discussed in Appendix A.
Table 4.4  Input files for VB-SWBMS
File description Notes
control file Stores all input parameters and input filenames so
that user does not have to redefine all this
information each time the model is used
cell properties file See Figure 4.2 or 4.3
component properties file Only used in two-layer model versions (see Figure
4.3)
soil parameters lookup table Only used in one-layer model versions; actual
parameter values are passed in the two-layer model.
rainfall file
potential evaporation file This file may contain daily meteorological data in
the format specified in Appendix A.2 or previously
computed daily potential evaporation estimates
using any method.
observed runoff data optional
soil moisture initialization
file
optional, If soil moisture is not initialized with a
file then the user can specify a uniform initial
effective saturation or a uniform suction head for
the entire watershed being modeled.
list of components for which
predicted soil water content
data will be exported
optional
143
Table 4.5  Output files for VB-SWBMS
File description Notes
flow file Contains direct runoff, baseflow, total
runoff, and observed flow (if available)
summary file Summary of statistics and input
parameters for simulation run
final storage values file Could be used to initialize the soil
moisture values in each layer and
component and the subsurface storage
level for a subsequent simulation
daily average soil moisture for selected
components
optional, Used to compare predicted
soil water contents with observed soil
water contents in the Little Washita
4.4. Summary
The Soil Water Balance Modeling System code is divided into two parts, an
ArcView pre-processor called AV-SWBMS and a Visual Basic component for
hydrologic computations called VB-SWBMS.
A key aspect of this two part model is that spatial information is transferred
from AV-SWBMS to VB-SWBMS using ASCII files.  This information transfer
becomes significantly more complicated for a two-layer model than for a one-layer
model because there is more than one characteristic that differentiates soil
components in the two-layer model.  With no resampling and a one-layer model, the
description of soil properties in each cell can always be accomplished with 13
values (percentage of soil in each texture class and the ?other? class), no more and
no less.  With no resampling and the two-layer model, both the number of
components in each cell and the properties associated with each component must be
passed to the VB-SWBMS and the number of components will vary from cell to
cell.  At a minimum, the percent of the cell area covered, the total soil zone depth,
the surface soil texture, and the percolation restricting texture must be passed for
each component.  In the current version, soil textures are converted to hydraulic
parameters before this information is passed in the two-layer model (see Figure 4.3).
144
In addition to data transfer, the mass balance calculations for the two-layer
model are significantly more complicated.  Although code complexity does not
preclude the use of a two-layer model, it is interesting to note that going from a one-
layer to a two-layer model requires 3-4 times as many lines of code in AV-SWBMS
and about 2 times as many lines of code in VB-SWBMS.  These figures are
estimated based only on code sections that significantly change between the two
cases.
In previous research, Reed and Maidment (1995) developed programs to create
a parameter file for the quasi-distributed modClark transform program that is now
part of HEC-HMS.  The modClark parameter file has a simple structure that is
similar to the cell properties files shown in Figure 4.2.  The modClark file includes
HRAP cell coordinates, area, and average flow length for each rainfall cell.  If it
desired to implement a similar soil-water balance routine to that described here in
HEC-HMS, it would be a straightforward task to add the additional columns shown
in Figure 4.2 to the modClark parameter file.  In a simple case, the one-layer model
with NEXRAD cell level resampling, only the percent imperviousness and soil
texture would need to be added to the modClark file.  Other cases could be
accommodated with the addition of more fields ? 13 fields for the one-layer, no
resampling case.  Even the two-layer model could be accommodated by this data
transfer structure (a two file data transfer scheme was used in this study), but the
number of additional fields required would vary from cell to cell.
Different version of SWBMS are run in Chapter 5 to evaluate different
modeling scenarios.  The SWBMS programs are designed to answer specific
research questions, not to provide a fool proof user interface (described in Appendix
A.2).  Based on the evaluation of different modeling scenarios in Chapter 5, future
work can be geared towards cleaning up and integrating the most useful versions of
SWBMS and making them more user friendly.
145
CHAPTER 5. ANALYSIS AND RESULTS
5.1. Overview
The purpose of this chapter is to examine the influence of the following factors
on simulated direct runoff:  (1) different levels of spatial detail in describing soil
properties, (2) uncertainty in rainfall estimates, and (3) uncertainty in model
parameter values associated with each of the basic USDA texture classes.  This
overview section lays out the strategy for making simulation runs, including a
description of the alternatives being studied and the calibration and validation
procedures.
As described in Section 3.2.4, three levels of resampling are considered for
representing soil variability.
  No resampling: Estimate the percentage of each soil type or the percentage
of all distinguishable soil components within each cell and make vertical
flux calculations separately for each soil type.
  NEXRAD cell level resampling: Assign a representative soil type to each
NEXRAD cell.
  Watershed level resampling: Assign a representative soil type to the entire
watershed.
The rainfall excess in each cell is routed to the watershed outlet the same way in all
three cases.
Several model variations in addition to spatial resampling are also of interest,
including a study of the differences caused by the use of input data at two different
map scales, the implications of choosing between a one-layer and a two-layer soil
model, and the advantages and disadvantages of simulating baseflow versus using
baseflow separation.  Considering all of these factors, there are many combinations (3
resampling levels x 2 map scales x 2 layer options x 2 baseflow methods = 24 total
146
combinations) that may be of interest.  A strategy is proposed here so that all of these
combinations do not need to be analyzed.  Insights about the one-layer versus two-
layer approach and using baseflow simulation versus baseflow separation are derived
for the most complex soil representation (no resampling and using the 30-m soils grid
as input), and it is assumed that these insights hold for the simpler representations.
The insights gained from these initial simulations lead the author to believe that using
a one-layer soil model and baseflow separation is adequate to compare the remaining
six scenarios which involve the three resampling levels and two map scales.
Another important issue to consider in these runoff modeling scenarios is the
accuracy of the rainfall data.  In examining the NEXRAD Stage III rainfall data and
comparing it with data from the Little Washita Micronet, significant differences in
rainfall magnitudes are noted for the brief period when Micronet data are available,
although the spatial distributions of rainfall in the two data sets is similar.  A data set
in which the NEXRAD rainfall field has been adjusted to be more consistent with
rainfall magnitudes reported by the Micronet is used for the majority of simulation
runs.  A comparison of NEXRAD and Micronet values and the preparation of this
adjusted rainfall data set is described in Section 5.3.
The remainder of this chapter includes four sections.
? Section 5.2:  A description of a general procedure for calibration and
validation.
? Section 5.3:  A comparison of NEXRAD rainfall estimates with Micronet
rainfall estimates and creation of an adjusted NEXRAD rainfall field for
subsequent simulation runs.
? Section 5.4:  A comparison of the one-layer and two-layer simulation models
and the implications for the use of a baseflow simulation or baseflow
separation approach.
? Section 5.5:  A study of simulations using the different horizontal soil
representations (using three different resampling levels and using source data
from two different map scales).
147
? Section 5.6:  A sensitivity analysis aimed at understanding the impacts of
model input uncertainties on runoff.  Input uncertainties include rainfall error,
soil moisture initialization, and the uncertainties inherent to specifying soil
hydraulic properties based on USDA texture class.
5.2. Calibration and validation procedures
For each selected set of modeling options (i.e., a one-layer model with no
resampling, 30-m soils data, and baseflow separation), both a calibration and a
validation run are made.  The calibration and validation runs are made for different
time periods.  In the calibration time period, model parameters are adjusted in an
effort to provide the best match between simulated and observed flow data (and
simulated and observed water content data in one case).  For simulations during the
validation time period, the optimal parameter values determined during the
corresponding calibration time period are used.  Performance during the validation
phase is used to compare modeling scenarios.
5.2.1. Calibration
Calibration is done for a time period during the Southern Great Plains 1997
hydrology experiment so that both observed runoff data and observed soil moisture
data can be used.  The dates used for calibration are 6/18/97 ? 7/22/97.  Although a
greater premium is placed on matching runoff data, it is interesting to see how well
the variability of soil moisture is described by the model.
Runoff data recorded at 15 minute intervals were obtained from the Oklahoma
U.S. Geological Survey and resampled to hourly average values for comparison with
hourly modeled flows.  As described in Section 3.2.6, gravimetric soil moisture
samples were taken at 23 different sites in the Little Washita watershed from 6/18/97
? 7/16/97.  The average of twice daily soil moisture measurements from selected
sandy loam, loamy sand, loam, and silt loam sites are compared with simulated values
in selected NEXRAD cells to assist with calibration.
A general calibration procedure is summarized here.
148
1. Initialize top layer soil moisture values using representative gravimetric soil
moisture measurements from SGP97 for all sandy loams (SL), loamy sands
(LS), loams (L), and silt loams (SIL) in the model.  Other textures found in
the watershed cover only a small fraction of the area but their water contents
also must be initialized.  To do this, the suction head of one of the texture
values for which water content is known (SL, LS, L, or SIL) is used to
estimate the initial water content in the other textures with the most similar
properties.  For example, the initial water content of sandy loam is known
but the initial water content for sand is not known.  The initial suction head
for sandy loam is computed and then it is assumed that the initial suction
head for sand is the same.  Subsequently, the initial water content for sand is
estimated using Equation 3.1.  In the two-layer model runs, there are no data
that can be used to assign initial water content for the bottom layer.  Initial
water content in the bottom layer may significantly influence baseflow.  The
top layer depth is set to 5 cm to compare results with observed soil moisture
values that are from the top 5 cm of soil.
2. A user defined hydraulic conductivity calibration factor (k
f
) is multiplied by
the saturated hydraulic conductivity for each soil unit being modeled.  This
calibration factor is modified and calculations are repeated until either the
total volume of simulated direct runoff matches the total observed direct
runoff (baseflow separation case) or the total volume of modeled runoff and
baseflow matches the total volume of observed streamflow (baseflow
simulation case).  Total observed direct runoff (ODR) and total simulated
direct runoff (SDR) are computed using Equations 5.1 and 5.2.
Equation 5.1 ODR = 
?
?tQ
d
o
Equation 5.2 SDR = 
?
?tQ
d
s
149
In application, reasonable results are obtained through manual calibration and
by carrying k
f
 to one decimal place for most cases.  Calibrated k
f
 values range
from 0.5 to 5.4 for different modeling options.
3. ModClark transform parameters (T
c
, R) are modified to approximate the peak
and spread in observed hydrographs.  Changing these parameters will not
influence the calibration of k
f
 unless a significant amount of surface water is
still stored in the watershed at the end of the simulation time period.  T
c
 and R
parameters were adjusted manually in an attempt to minimize the HEC-1
objective function (HECOBJ) defined by Equations 5.3 and 5.4 (HEC, 1997).
For all simulation results presented in this section, the values R = 9 hours and
T
c
 = 12 hours are used.  The HEC-1 objective function is designed to give
greater weight to flow deviations associated with higher flows.
Equation 5.3 HECOBJ = 
n
Q
QtQ
tQtQ
A
Ao
s
n
t
o
2
))((
))()((
2
1
+
?
?
=
Equation 5.4
?
=
=
n
i
OA
Q
n
Q
1
1
4. The rooting depth parameter (d
r
) is modified to get the closest fit to daily
average soil moisture observations for the four available textures.  The sum of
absolute residuals (SAR) for 26 days and four different soil textures is used as
a measure of the fit.  The mathematical definition of SAR is given in Equation
5.5.  Rooting depth values are calibrated manually to the nearest 5 cm.
Calibration against soil moisture is only implemented in runs with no
resampling because all of the correct textures do not exist in the correct
rainfall cells with the other resampling levels.  The rooting depth parameter
values determined using the no resampling option are used for simulations
with NEXRAD cell level and watershed level resampling.
150
Equation 5.5 SAR = 
??
==
?
4,126,1jk
so ??
5. After adjusting the rooting depth parameter (d
r
), check to make sure that the
optimal k
f
 value has not changed.  If it has changed, repeat steps 2 and 4.
A summary of the model parameters that cannot be specified directly from the
soils databases is provided in Table 5.1.  In the third column of this table there is an
explanation of how values for each of these parameters is determined.
151
Table 5.1  Model parameters not specified from soils databases.
Symbol Description Method of specification
k
f
hydraulic
conductivity
calibration factor
Determined through calibration against
direct runoff.  In this study, values of k
f
range from 0.5 to 5.4, depending on the
modeling options.
d
1
top layer depth Taken as 5 cm for calibration and
validation runs because observed soil
moisture samples are from the top 5 cm of
soil.
d
r
root zone depth Determined through calibration against
observed soil moisture.  Values
determined in different calibration runs
were 35, 40, and 50 cm.
p
2
percolation out
of layer 2 if it is
underlain by a
non-soil, low
permeability
layer
Used in two-layer model only.  Varied in
attempt to match observed baseflow (see
Section 5.4).  A value of 0.0015 cm hour
-1
seems reasonable.
?
baseflow storage
constant
Only required when baseflow is
simulated.  Estimated based on a visual
comparison of simulated flow and the
baseflow recession curve.  A value of 240
hours seems reasonable.
n baseflow
separation
parameter
Only required when baseflow separation
is used.  Taken as 60 hours.
T
c
modClark time
of concentration
Initial estimates of T
c
 and R were
determined by scaling estimates from a
previous modClark study in another part
of Oklahoma.  Values of T
c
 and R were
adjusted manually in an effort to minimize
HECOBJ function during the calibration
period.  The resulting values used in all
simulation runs were R = 9 hours and T
c
 =
12 hours.
R modClark
storage constant
see above
152
5.2.2. Validation
Validation runs are made during a time period preceding the calibration period,
6/01/97-6/18/97.  A big storm occurs just prior to this validation period on 5/30/97 so
the top layer soils are assumed to be initially saturated.  The k
f
 and d
r
 values
determined from the corresponding calibration run are used in validation runs.
Modeled runoff from storms that occur on 6/10/97, 6/13/97, and 6/17/97 is compared
with observed runoff.
For selected cases, validation runs are also made for the time period 5/7/97-
5/24/97.
5.3. Comparison of NEXRAD and Micronet rainfall
Although the NEXRAD Stage III data are derived using both radar and rain gage
information, the density of rain gage information used to develop the Stage III
product is much less than that of the Micronet rain gage network.  In this section it is
assumed that when all Micronet stations are reporting, the Micronet data provide a
better indicator of rainfall depths, and are used here to check the reliability of the
NEXRAD data.  Micronet data from the Agricultural Research Service are available
for June, July, and August of 1997.  As discussed in Section 3.2.6, there are 42
Micronet sites where rainfall measurements are taken; however, not all 42 stations
reported valid data during a significant portion of June, 1997.  Hourly rainfall totals
have been computed from this data set by accumulating 5 minute data for each
station.  A summary of storm average rainfall totals computed based on Micronet data
and NEXRAD data is provided in Table 5.2.  The Micronet storm totals are computed
by first determining the arithmetic average of the station values in a given hour and
then summing the hourly averages for all hours in a storm.  NEXRAD storm totals are
computed similarly, first hourly arithmetic averages are computed from the values in
all cells and then the hourly averages are summed over the storm duration.  No area-
weighting was used to determine the storm totals in Table 5.2.
153
The ratios of Micronet storm totals to NEXRAD storm totals (M/N) and the
differences (M ? N) between these quantities are tabulated in the last two columns of
Table 5.2.  For the ratios, there is no clear trend as to whether the NEXRAD totals
tend to be greater than or less than the Micronet totals.  The mean of the M/N ratio for
all storms is 0.88 with a standard deviation of 0.31.  The absolute differences between
the two estimates are not exceptionally large in most cases except the storm of
7/10/97 where the NEXRAD data yield 2.72 cm more rain than the Micronet data.
Since this is a critical storm within the selected calibration period, a further study of
rainfall differences for this storm has been made.
Table 5.2  Comparison of storm rainfall totals between Micronet and NEXRAD
data.
Storm begins Storm
duration
(hrs)
Micronet Storm
Totals, M (cm)
Nexrad
Storm
Totals, N
(cm)
# of
Micronet
Stations
Reporting
M/N M - N
6/13/97 6:00 5 1.85 2.08 11 0.89 -0.24
6/15/97 4:00 3 0.10 0.17 16 0.57 -0.07
6/17/97 2:00 14 0.85 1.26 19 0.68 -0.40
6/22/97 22:00 22 1.24 0.83 23 1.50 0.41
6/26/97 12:00 12 0.19 0.17 32 1.10 0.02
6/27/97 18:00 28 0.42 0.91 36 0.46 -0.49
7/4/97 9:00 5 0.48 0.54 41 0.88 -0.06
7/9/97 19:00 6 0.17 0.16 41 1.09 0.01
7/10/97 19:00 16 4.78 7.51 42 0.64
-2.72
7/15/97 18:00 6 0.95 0.94 42 1.01 0.01
Maps in Figure 5.1 show the spatial distribution of rainfall estimated from the
two data sources for 8 hours during the July 10, 1997 storm.  The Micronet data are
mapped using thiessen polygons.  There are actually fewer than 42 thiessen polygons
in these maps because some stations were missing data during part of the storm and
no polygons were created for these stations.  The fact that the number of rainfall
stations reporting valid data changes with time is a clear drawback of using thiessen
polygons for spatially distributed modeling.  Each time there is a missing station or an
added station, the spatial network changes.
154
It is encouraging to see that the spatial distributions of rainfall from the two data
sources are comparable.  Differences in the areas covered by intense rainfall are
evident at 7/11/97 00:00 GMT.
155
Micronet NEXRAD
156
Micronet NEXRAD
Figure 5.1  A comparison of Micronet and NEXRAD Stage III rainfall estimates
for 8 hours of the 7/10/97 storm.
157
It is interesting to see how differences in rainfall magnitudes influence the values
of parameters determined through calibration.  A comparison is made here between a
calibration run made using the NEXRAD rainfall field and a run using an adjusted
NEXRAD rainfall field in which all rainfall values for the July 10, 1997 storm are
multiplied by the M/N ratio for that storm (0.64).  The one-layer model with 1:24,000
scale soils data, baseflow separation, and no resampling is used for this illustration.
The calibrated parameter values (k
f
 and d
r
) for the two cases are given in Table 5.3.
Statistics for both the calibration and validation runs are given in Table 5.4.
Figure 5.2 shows calibrated direct runoff (6/18/97-7/22/97) generated using the
two rainfall input data sets and Figure 5.3 shows plots of direct runoff for the
validation runs (6/1/97 ? 6/18/97).  As one might expect, there is little difference
between the observed and predicted direct runoff during the calibration phase for both
rainfall data sets because the value of k
f
 is modified to achieve the best fit in both
cases.  There is a big difference in the calibrated k
f
 values for the two cases, 5.4 for
non-adjusted and 1.9 for adjusted rainfall.  The validation results for the adjusted
rainfall are significantly better than the validation results for non-adjusted rainfall as
indicated by the SDR/ODR ratios in Table 5.4 and the direct runoff hydrographs
shown in Figure 5.3.
Table 5.3  Calibrated parameter values for NEXRAD and adjusted NEXRAD
simulations.
Parameter NEXRAD Adjusted NEXRAD
k
f
5.4 1.9
d
r
50 cm 40 cm
158
Table 5.4  Statistics describing NEXRAD and adjusted NEXRAD calibration
and validation runs.
Statistic NEXRAD Adjusted NEXRAD
Calibration:
   ODR (m
3
) 2,989,816 2,989,816
   SDR (m
3
) 2,965,103 2,935,168
   SDR/ODR 0.99 0.98
   HECOBJ (m
3
/s) 2.39 2.44
   SAR for soil
water content
4.91 4.45
Validation:
   ODR (m
3
) 1,658,031 1,658,031
   SDR (m
3
) 741,265 1,225,445
   SDR/ODR 0.45 0.74
   HECOBJ (m
3
/s) 3.57 2.79
159
0
10
20
30
40
50
60
0 150 300 450 600 750 900
Time (hours)
Flow (cms
)
Observed direct runoff (cms) Simulated direct runoff (cms)
(a)
7/21/97 23:00
6/18/97 0:00
0
10
20
30
40
50
60
0 150 300 450 600 750 900
Time (hours)
Flow (cms
)
Observed direct runoff (cms) Simulated direct runoff (cms)
(b)
7/21/97 23:00
6/18/97 0:00
Figure 5.2  Calibrated flows using (a) NEXRAD and (b) adjusted NEXRAD
rainfall.
160
0
5
10
15
20
25
30
0 150 300 450
Time (hours)
Flow (cms
)
Observed direct runoff (cms) Simulated direct runoff (cms)
6/1/97 0:00
6/18/97 23:00
(a)
0
5
10
15
20
25
30
0 150 300 450
Time (hours)
Flow (cms
)
Observed direct runoff (cms) Simulated direct runoff (cms)
(b)
6/1/97 0:00
6/18/97 23:00
Figure 5.3  Validation runs using parameters from Table 5.3.  (a) NEXRAD and
(b) adjusted NEXRAD rainfall.
161
As seen in Table 5.4, the SAR statistic between modeled and observed soil water
content is lower for the case with adjusted rainfall, another indication that the model
performs better using the adjusted rainfall.  Figure 5.4 is included to give an idea of
how well the model predicts average daily soil water content values when the
adjusted rainfall is used.  The graphs in Figure 5.4 represent the lowest overall SAR
value that could be achieved by varying d
r
 by increments of 5 cm with k
f
 equal to 1.9.
The best match for an individual soil type occurs with the loam soil (SARl = 0.72).
Although the trends make sense, there are considerable differences between observed
and modeled soil water content values.
Given the comparisons between NEXRAD Stage III data and Micronet data and
between simulations made using the original NEXRAD data and adjusted NEXRAD
data, the adjusted NEXRAD data are used for simulations described in Sections 5.4,
5.5, and 5.6.
F
i
g
u
re 5.4  Comparison of mod
e
led and obse
r
ved a
v
er
a
g
e dail
y
 soil moisture for (
a
) a lo
am
y
 s
a
nd in cell (557, 309) 
and
observed dat
a
 ave
r
a
g
ed 
from sites 3, 4, and 5; (b) 
a sand
y
 loam in c
e
ll (559, 305) and obs
erved d
a
ta f
r
om site 17; (c) a
loam in cell (560, 309) and observed d
a
ta av
eraged from sites 11, 12, and 13; and 
(d) a silt loam in cell (555, 307) 
and
observed dat
a
 ave
r
a
g
ed 
from sites 8 and 9.
162
0
0.
1
0.
2
0.
3
0.
4
6/
16/
97
6/
26/
97
7/
6/
97
7/
16/
97
Water conten t
0
0.
1
0.
2
0.
3
0.
4
6/
16/
97
6/
26/
97
7/
6/
97
7/
16/
97
Water conten t
0
0.
1
0.
2
0.
3
0.
4
6
/
16/
97
6
/
26/
97
7
/
6/
97
7
/
16/
97
Water conten t
0
0.
1
0.
2
0.
3
0.
4
6/
16
/
9
7
6
/
2
6
/
97
7/
6/
97
7/
16
/
9
7
Water conten t
(d) s
ilt loam
SAR(
SI
L
)
 =
 1
.
3
7
(b) s
a
ndy lo
am
(a) loam
y sand
S
A
R
(
LS
) = 1
.
2
0
SAR(
SL
)
=
 1
.
1
6
SAR(
L
)
 =
 0
.
7
2
(c) loam
Mod
e
le
d
Ob
ser
ve
d
163
5.4. One-layer vs. two-layer model
In this section, the differences in predicted direct runoff from a one-layer and
two-layer model are compared.  In addition, simulations are presented showing the
advantage of using a two-layer model when it is desired to simulate baseflow.
5.4.1. Direct runoff
To compare direct runoff, calibration and validation runs were made using the
two-layer model and the one-layer model, each using the 1:24,000 scale spatial data
and the baseflow separation option.  Calibrated parameters for the one and two-layer
models are given in Table 5.5 and statistics for these simulation runs are given in
Table 5.6.  Figure 5.5 shows the direct runoff generated by the one-layer and two-
layer models during the validation period.  Note that the one-layer results are identical
to those from Section 5.3 (for the case of the adjusted rainfall field which is also used
in this section).
Table 5.5  Calibrated parameter values for the one-layer and two-layer models.
Parameter One-layer Two-layer
k
f
1.9 1.8
d
r
40 cm 35 cm
164
Table 5.6  Statistics describing one-layer and two-layer calibration and
validation runs.
Statistic One-layer Two-layer
Calibration:
   ODR (m
3
) 2,989,816 2,989,816
   SDR (m
3
) 2,935,168 3,043,277
   SDR/ODR 0.98 1.02
   HECOBJ (m
3
/s) 2.44 2.52
   SAR for soil water content 4.45 4.41
Validation:
   ODR (m
3
) 1,658,031 1,658,031
   SDR (m
3
) 1,225,445 1,352,298
   SDR/ODR 0.74 0.82
   HECOBJ (m
3
/s) 2.79 2.65
0
5
10
15
20
25
30
0 150 300 450
Time (hours)
Flow (cms
)
Observed direct runoff One-layer direct runoff
Two-layer direct runoff 
6/1/97 0:00
6/18/97 23:00
Figure 5.5  Flow during the validation period for the one-layer and two-layer
models.
165
The calibrated parameter values for the one and two-layer models are very
similar.  The difference in the calibrated parameters is caused primarily by a
difference in the assumption that is made to deal with non-soil cells in the 30-m soils
grid (i.e. water, sewage lagoon, and limestone).  Note that these non-soil cells are not
assumed to be impermeable because this might be redundant with the land use data,
which also identifies water bodies.  The simplicity of data structure in the one-layer
model made it easy to divide the percentage of each NEXRAD cell taken up by non-
soil grid cells proportionally among the existing texture classes.  Because of the
greater complexity of the two-layer data structure, all non-soil cells are assigned the
properties of a single soil type in this model.  The differences that cause the calibrated
k
f
 value to be 1.9 for the one-layer model and 1.8 for the two layer model are very
subtle.  The two-layer simulation with a k
f
 value of 1.8 is only slightly better than a
two-layer simulation with a k
f
 value of 1.9.  In fact, the best k
f
 value for both the one
and two-layer models is close to 1.85.  Thus, one reason why the validation run for
the two-layer model is slightly superior to that for the one-layer model is due to
rounding in the calibration phase.
For the two-layer model, a small amount of saturation excess runoff (23,575 m
3
)
is generated during the validation phase when shallow soil components underlain by
bedrock become saturated.  This amount of saturation excess runoff constitutes only
1.7% of the simulated direct runoff during the validation phase.  The amount of
saturation excess that is generated is a function of the user specified percolation rate
for non-soil, low permeability layers (p
2
).  The value assigned to p
2
 for the
simulations described in Table 5.6 and Figure 5.5 is 0.0015 cm/hour, which seems to
be a reasonable value as discussed later in this section.  The other major factor that
influences the amount of saturation excess runoff is the depth of the soil zone.  The
small amount of saturation excess runoff in this example was generated from soils
with less than 20 cm of total depth.
166
Based on the evidence in Tables 5.5 and 5.6, and Figure 5.5, the two-layer model
is only slightly superior to the one-layer model in predicting direct runoff during the
validation period; therefore, given the additional complications associated with using
the two-layer model, the one-layer model is used for further analysis in Sections 5.5
and 5.6.
5.4.2. Baseflow
Although the simplicity of the one-layer model is preferred for simulating direct
runoff, the second soil layer provides an advantage when simulating baseflow.
Although the focus of this study is on direct runoff and not baseflow, a simple
example is included here to illustrate this point.
If percolation out the bottom of a one-layer soil model is passed directly to a
single linear reservoir for simulating baseflow, then this subsurface reservoir receives
so much flow that it is not possible to match hydrograph recession.  To show this, a
calibration run is initialized with a recession constant of ? equal to 240 hours and a
subsurface storage (S) of 2,510,000 m
3
 so that the initial streamflow is equal to the
initial observed streamflow of 2.9 m
3
/s, (See Figure 5.6a).  Provided there is not a
large amount of initial percolation, a value of 240 hours for the recession constant is
reasonable to approximate the recession rate for the first 120 hours in Figure 5.6a.
However, when a storm occurs, the percolation influx to the subsurface creates
unreasonably high subsurface storage and baseflow.  This problem cannot be
alleviated by simply increasing the subsurface storage constant.  An extreme example
of attempting to increase the subsurface storage constant is shown in Figure 5.6b.  For
Figure 5.6b, the subsurface storage constant is two years (? = 17520 hours) and the
initial subsurface storage is set to 183,957,998 m
3
 in order to match the observed
baseflow of 2.9 m
3
/s.  With such a large storage constant, the ridiculously high
baseflows of Figure 5.6a are eliminated but the baseflow curve becomes flat and
cannot reflect the natural recession.
167
Using the two-layer soil model, the second soil layer provides a reservoir for
infiltrated water which can subsequently evaporate or percolate to the subsurface
reservoir.  Simulation runs indicate that a more reasonable approximation for
baseflow can be achieved using the two-layer model than using the one-layer model.
Use of the two-layer model requires that the user specify a percolation rate (p
2
) for
soil components underlain by non-soil material such as weathered bedrock.  As noted
in Section 3.2.2.3, this percolation rate influences 47% of the Little Washita area
being modeled.  Using a recession constant of ? equal to 240 hours, an initial
subsurface storage of 2,510,000 m
3
, and a non-soil percolation rate equal to 0.0015
cm/hour, a more reasonable baseflow hydrograph is simulated, as shown Figure 5.7.
The simulation results shown in Figure 5.7 do not represent an optimal solution and
are only presented with the intent to contrast the results of Figure 5.6.
The use of a two-layer model with baseflow simulation has not been fully
explored in this research.  Different combinations of p
2
, ?, initial subsurface storage,
and initial water content in the second layer may provide equally valid or superior
results to those shown in Figure 5.7.  The second layer water content influences both
the total amount of evaporation from the soil profile and the amount of percolation to
the subsurface reservoir.  Questions remain unanswered as to how the lower layer
water content should be initialized and how realistic the final water content of the
lower layer is relative to the initial water content.  Perhaps, these are appropriate
subjects for further research.
One important conclusion from the comparison of the one-layer and two-layer
models is that for the validation period selected, the two-layer model does not
significantly outperform the one-layer model in terms of predicting direct runoff.  The
one-layer model is much more practical to implement and is adequate to assess the
effects of horizontal soil variability on direct runoff as described in the next section.
168
0
20
40
60
80
100
0 150 300 450 600 750 900
Time (hours)
Flow (cms
)
Observed flow Base flow Simulated flow 
7/21/97 23:00 GMT
6/18/97 0:00 GMT
(a)
0
10
20
30
40
50
60
0 150 300 450 600 750 900
Time (hours)
Flow (cms
)
Observed flow Base flow Simulated flow 
7/21/97 23:00 GMT6/18/97 0:00 GMT
(b)
Figure 5.6  Attempts to simulate baseflow using the one-layer soil model. (a)
recession constant ? = 240 hours  (b) recession constant ? = 17520 hours.
169
0
10
20
30
40
50
60
0 150 300 450 600 750 900
Time (hours)
Flow (cms
)
Observed flow Base flow Simulated flow 
7/21/97 23:00 GMT
6/18/97 0:00 GMT
Figure 5.7  Hydrographs from two-layer simulation with ? = 240 hours and p
2
 =
0.0015 cm/hour.
5.5. Horizontal soil variations (spatial resampling and map scale issues)
An important objective of this research is to explore the implications of spatial
resampling and map scale on runoff prediction.  Six sets of calibration and validation
runs using three resampling levels and two map scales are summarized in Table 5.7.
Values of the ratio SDR/ODR closest to 1 are best while low values of HECOBJ are
best.  With one exception (watershed level resampling with STATSGO data), as
spatial resampling decreases (meaning that spatial variability increases) the results
improve.  Looking at validation runs using the 30-m soils grid, the SDR/ODR ratio is
the highest with no resampling (0.74) and decreases by 32% to 0.5 with NEXRAD
cell level resampling and by another 8% to 0.46 with watershed level resampling.
Still looking at validation runs using the 30-m soils grid, the HEC-1 objective
function is the best (lowest) for no resampling and gets worse (increases) by 18%
when NEXRAD cell level resampling is used.  The HEC-1 Objective function
170
actually decreases slightly (3%) moving from NEXRAD cell level to watershed level
resampling.  Based on the numbers in Table 5.7, it appears that there is a significant
drop-off in performance between no resampling an NEXRAD cell level resampling
and a much smaller drop-off between NEXRAD cell level and watershed level
resampling.
The differences between STATSGO and 30-m grid simulations are quite
noticeable for watershed level resampling but not as marked for NEXRAD cell level
resampling and no resampling.  With NEXRAD cell level resampling, the ratio
SDR/ODR indicates slightly better performance using STATSGO data, but the
HECOBJ statistics indicate slightly better performance with the 30-m grid.  With no
resampling, the SDR/ODR ratio indicates slightly better performance using the 30-m
grid data while the HECOBJ function values are identical.  A graph of the validation
flows for the no resampling, 30-m grid case was shown in Figure 5.3b.
Since the good performance of the watershed level resampling, STATSGO
combination is inconsistent with the trend of improved performance with decreased
model resampling, additional validation runs have been made to compare the
watershed level resampling, STATSGO combination with the no resampling, 30-m
grid combination.  These validation runs were made for the time period 5/7/97 to
5/24/97 and the same parameter values determined from the original calibration runs
(Table 5.7) were used.  In theory, the same k
f
 should be applicable for this time
period but perhaps not the same d
r
 value.  Figure 5.8 shows graphical results from this
second validation period and validation statistics are given in Table 5.8.  Since no
observed soil moisture data are available for this time period, all soils were initialized
to the same effective saturation and this initial value was adjusted to get the best fit
with observed data.  Since the model overpredicts runoff for this period, a low
effective saturation of 0.1 was used to initialize the model for the results shown here.
The results in Figure 5.8 and Table 5.8 show that the no resampling, 30-m grid
171
combination significantly outperforms the watershed level resampling, STATSGO
combination in this period.
172
Table 5.7  Calibrated parameter values and validation statistics for the one-layer
model with different levels of model resampling and different soil map inputs.
Calibrated
parameter
values
Statistics for validation runs
Resampling Soils Data
Source
k
f
d
r
*** SDR/ODR
HECOBJ (m
3
)
watershed STATSGO** 0.49 40 0.66 2.5
watershed 30-m grid* 3.0 40 0.46 3.2
NEXRAD cell STATSGO 2.2 40 0.55 3.4
NEXRAD cell 30-m grid 2.4 40 0.50 3.3
none STATSGO 1.7 40 0.69 2.8
none 30-m grid 1.9 40 0.74 2.8
*  Entire watershed assumed to be silt loam.
**  Entire watershed assumed to be sandy loam.
***  d
r
 is only calibrated for the no resampling option.
0
10
20
30
40
50
60
0 150 300 450
Time (hours)
Flow (cms
)
Observed direct runoff
No resampling  (30-m grid)
Watershed level resampling  (STATSGO)
5/7/97 0:00 GMT
5/24/97 0:00 GMT
Figure 5.8  Comparison of direct runoff from watershed level resampling,
STATSGO simulation with direct runoff from a no resampling, 30-m grid
simulation for 5/7/97 - 5/24/97.
173
Table 5.8  Statistics for the watershed level resampling, STATSGO combination
and the no resampling, 30-m grid combination for validation during 5/7/97 -
5/24/97.
Resampling Soils Data
Source
SDR/ODR
HECOBJ
watershed STATSGO* 1.42 9.3
none 30 m grid 1.13 5.4
*  Entire watershed assumed to be silt loam.
5.6. Sensitivity analysis
From the discussion in Section 3.2.2.2, it is clear that there is considerable
uncertainty involved with specifying the soil parameter values (?, ?
r
, h
b
, and ?) based
on texture class.  There is also considerable uncertainty in specifying K
s
 but this
parameter is not considered here because the calibration factor k
f
 is used to adjust this
parameter.  To get a feel for how the uncertainty in input parameters influence model
results, the nominal range method of sensitivity analysis is used.  In this method, a
model is run using high and low values for a given parameter while keeping the other
parameters at their nominal values (Morgan and Henrion, 1990).
The mean soil parameter values reported by Carsel and Parrish (1988) are
considered the nominal values.  Low values and for each parameter are taken as the
minimum of the Carsel and Parrish (1988) and Rawls, Brakensiek, and Saxton (1982)
values corresponding to 1 standard deviation below the mean.  High values are taken
as the maximum of the Carsel and Parrish (1988) and Rawls, Brakensiek, and Saxton
(1982) values for 1 standard deviation above the mean.  This is equivalent to taking
the minimum and maximum parameter values within each texture class from Figures
3.8 - 3.11.  For some textures, the high values for ?
r
 exceeded the nominal wilting
point values computed from the nominal h
b
 and ? values.  In these cases, the high ?
r
values were set equal to the wilting point to prevent an error in the model mass
balance.
174
Sensitivity runs were made for the 6/18/97-7/22/97 using the model scenario that
showed the best overall performance in Section 5.5, a one-layer model with no
resampling that uses the 30-m soils grid.  The results of the sensitivity analysis are
presented in the form of a tornado diagram shown in Figure 5.9.  The tornado
diagram shows the deviations in the ratio SDR/ODR from the nominal case caused by
using high and low parameter values.  Parameters are ranked from top to bottom by
the degree to which their uncertainty influences model results.  In addition to
uncertainty in soil parameter values, model sensitivity to uncertainty in rainfall and
initial soil water content is also shown in Figure 5.9.  The results of simulation runs
with rainfall depths varied by +/-20% and +/-30% from the adjusted NEXRAD field
are shown.
Model sensitivity to initial soil moisture is tested in two ways.  In one set of
sensitivity runs, the initial effective saturation (?
i
) is set to high and low values of 1
and 0 for all soils on 6/18/97.  In this case, the modeled direct runoff is insensitive to
initial soil moisture because the major storm does not occur until 7/10/97.  For the run
with nominal values, soil moisture on 6/18/97 is initialized using observed data as
described in Section 5.2.1.  In a second set of runs, simulations are begun on 7/10/97,
the morning of the big storm, and initial effective saturation values for all soils are set
to high and low values of 1 and 0.  In this case, the initial soil moisture does have a
significant impact on the modeled runoff.  The results for this case are denoted by the
symbol ?
i
* in Figure 5.9.  Note that the nominal value of the ratio SDR/ODR
changes for this case and this is reflected in the results shown.
175
% deviation in SDR/ODR from
the nominal case
-100 -50 0 50 100
- 30% rain
High h
b
-20% rain
?
i
* = 0
High ?
High ?
Low ?
r
?
i
 = 0
+ 30% rain
Low h
b
+20% rain
?
i
* = 1
Low ?
Low ?
High ?
r
?
i
 = 1
Figure 5.9 Tornado diagram showing deviations from the base case based on
high and low parameter values.
The results shown in Figure 5.9 indicate that the uncertainty in specifying h
b
causes more uncertainty in model output than the uncertainty in specifying ?, ?, or ?
r
.
Understanding the reasons for differences in the Carsel and Parrish and Rawls,
Brakensiek, and Saxton parameter values would certainly reduce the uncertainty
associated with specifying h
b 
because the ranges for h
b
 specified by these two studies
are drastically different as indicated by Figures 3.9-3.12.  Among the parameters ?, ?,
?
r
, and h
b
, part of the reason why the specification of h
b
 introduces the largest amount
of uncertainty in the model output is because the coefficients of variation for this
parameter tend to be higher (see Table 3.3).  If averages of the coefficients of
variation are computed for h
b
 and ? using the Rawls, Brakensiek, and Saxton data set
(which exhibits a wider range of variability than the Carsel and Parrish data set), the
176
average coefficient of variation for h
b
 is 3.7 times higher than that for ?.  Another
indication of why the modeled direct runoff is less sensitive to ? than it is to h
b
 is that
in Equation 3.27 for the Green-Ampt wetting front suction head, ? appears both in the
numerator and the denominator.
177
CHAPTER 6. CONCLUSIONS AND RECOMMENDATIONS FOR
FUTURE WORK
6.1. Conclusions
General objectives that have been met in this research include:
? A practical and automated method for using U.S. Department of
Agriculture digital soil maps in a hydrology model has been developed
and implemented.
? Model performance has been tested using different levels of spatial
complexity to describe soil properties.  This includes testing of a one and
a two-layer conceptual soil model, three different levels of horizontally
resampling soil properties, and the use of soils data from two different
map scales (1:250,000 and 1:24,000).
? Uncertainties associated with soil parameter specification and rainfall
inputs have been quantified.
The computer programs and the underlying conceptual model that have been
developed for this dissertation are referred to collectively as the Soil Water Balance
Modeling System (SWBMS).  Summaries of important points and conclusions are
grouped here according to specific modeling issues: spatial complexity, the use of
NEXRAD Stage III rainfall data, parameterization, process representation, and
computer implementation.  It is important to point out that the conclusions below
are based on a limited set of validation runs in a single watershed.  Further work is
suggested in Section 6.2 to verify these conclusions.
Spatial complexity
? Only minor differences were observed between the direct runoff volume
predicted by a one-layer and a two-layer model conceptual model.  Given the
SWBMS conceptual model and the soil characteristics of the Little Washita
watershed, only a small fraction of the direct runoff (1.7% of the total simulated
178
direct runoff) was generated due to the characteristics of the second layer in a
validation run using the two-layer model.
? In SWBMS, a second soil layer serves as a storage reservoir from which water
may evaporate or percolate to the groundwater reservoir.  As discussed in
Section 5.4, the combination of a single linear reservoir with a one-layer soil
model is inadequate to simulate baseflow, but the inclusion of the second soil
layer produces more reasonable baseflow results.
? Compared with a one-layer model, running a two-layer model requires
significantly more complex pre-processing and simulation codes.  In addition,
the two-layer model requires the somewhat arbitrary specification of an
additional parameter defining the non-soil (i.e. weathered bedrock) percolation
rate and the specification of initial water content for the second layer.
? Resampling of soil properties is considered as an option to spatially simplify the
water balance model.  Three levels of resampling are considered: no resampling,
resampling at the NEXRAD cell level, and resampling at the watershed level.
When the spatial variability in surface soil texture is reduced through
resampling, model performance decreases.  In validation runs, when surface soil
texture was resampled to the NEXRAD cell level and the 30-m soils grid was
used, the ratio of total simulated to total observed direct runoff (SDR/ODR)
decreased from 0.74 to 0.5 (a 32% decrease) and the HEC-1 objective function
(lower is better) increased from 2.8 to 3.3 (18%).  Based on these results,
resampling is not recommended.
? When the no resampling or the NEXRAD cell resampling option is used, the
differences in simulation results between using STATSGO soils data (1:250,000
179
map scale) and a 30-m soils grid (1:24,000 map scale) as input are relatively
small when compared with the differences in simulation results among
resampling levels.  This result is consistent with the observation that the overall
textural composition of the Little Washita watershed computed from the two
data sources is similar (Figure 3.6b).  It would be interesting to see if this result
holds true in a watershed containing a broader range of soil textures.
NEXRAD data
? A comparison of Micronet gage rainfall estimates and NEXRAD StageIII
rainfall estimates for a short period in June and July 1997 reveals substantial
discrepancies in rainfall depth estimates; however, the spatial distribution of
rainfall from these two data sources is similar for a storm on July 10, 1997.  In a
study of 10 storms, the mean ratio of average Micronet rainfall estimates to
average NEXRAD StageIII estimates was 0.88 and these ratios ranged from 0.5
to 1.5.
Parameterization
? A practical scheme for estimating soil hydraulic properties from soils databases
is proposed and implemented.  The method involves the use of two lookup
tables.  The first table translates texture names in the STATSGO database into
one of the 12 basic USDA texture classes or the "other" category, and the second
table assigns mean hydraulic properties to each texture class.
? As discussed in Section 3.2.2.2.2, there are several reasons why the texture
lookup table approach is used rather than trying to estimate individual parameter
values from soil data table attributes other than texture.  The Brooks and Corey
parameters required to estimate the wetting front suction head and calculate
180
percolation cannot be estimated directly from the soil attribute tables and
therefore must be estimated using either the regression equations of Rawls and
Brakensiek (1985) or as a function of texture alone.  In order to automatically
apply the regression equations of Rawls and Brakensiek (1985), a complicated
set of assumptions would be required to deal with layer table attributes that have
missing values or internal inconsistencies.  This level of complication does not
seem justified when considering all the approximations in the modeling process.
Therefore, the lookup table approach has been adopted.
? Several research papers have been published that deal with the specification of
soil hydraulic properties as a function of texture class.  Unfortunately, there are
large discrepancies between some of the hydraulic parameter estimates reported
in the two papers that have used the largest soils databases to develop parameter
estimates ? Rawls, Brakensiek, and Saxton (1982) and Carsel and Parrish
(1988).  The Carsel and Parrish values are used here, primarily because wetting
front suction head values computed using Carsel and Parrish ? and h
b
 values are
more consistent with Figure 5.5.4 in Rawls et al. (1993) than wetting front
suction head values computed using Rawls, Brakensiek, and Saxton (1982)
values.
? Among the parameters of interest in this study, saturated hydraulic conductivity
(K
s
) exhibits the most natural variability within a given texture class.  For this
reason, a hydraulic conductivity factor (k
f
) is used for calibration.  k
f
 is
multiplied by the K
s
 value for all soil types at all points in space.  In this study,
values of k
f
 ranged from 0.5 to 5.4 depending on the modeling scenario being
considered.
181
? Among the soil parameters not used for calibration (?, h
b
, ?, and ?
r
),
specification of the Brooks and Corey parameter (h
b
) involves the most
uncertainty and has the biggest impact on simulated runoff.
? Gravimetric estimates of soil-water content from the Little Washita watershed
are used to calibrate a rooting depth parameter (d
r
).  Calibrated d
r
 values were 35
cm, 40 cm, and 50 cm for different modeling scenarios.  Developing a
physically based method to estimate the rooting depth (d
r
) would require further
research.
Process representation
? Simplified flow models for infiltration and percolation are used because the
uncertainties in model inputs do not justify the use of more complex
computational algorithms.  The Green-Ampt model is used for infiltration and a
simple kinematic model is used for percolation.
? Existing methods used to estimate evaporation from unsaturated soils range
from simple empirical methods (bucket models) to complex soil-vapor-
atmosphere transfer schemes (SVATS).  The possibility of developing a model
of intermediate complexity was explored using new data sets being generated at
DOE ARM CART sties.  Plots relating evaporative surface resistance to
piezometric head were generated.  Looking at this data, it is clear that this
approach will not have general applicability because other factors besides soil-
water potential are significantly influencing surface resistance.
? A conventional bucket model for estimating evaporation is used in this study.
Initial results indicate that uncertainties in soil parameter specification and
182
rainfall have a greater impact on direct runoff than initial soil moisture.  In light
of this, perhaps no further complication in evaporation estimation is justified.
Computer implementation:
? Automated pre-processing routines are used to estimate hydrologic model
parameters using ArcView GIS.  The required spatial data describing soil,
topography, and land use are readily available on the Internet for any location in
the United States.
? Transfer of the required spatial information from the GIS (AV-SWBMS) to an
external hydrologic model (VB-SWBMS) is easily achieved using an ASCII
file.  The proposed format of using a single line to describe the properties of
each NEXRAD cell is compatible with the format of the modClark parameter
file that is already used in HEC-HMS.
? The most tedious procedures in using the SWBMS model developed here are
obtaining and reformatting the precipitation, climate, and runoff data.  The GIS
analysis of spatial data is more straightforward because the data formats for the
original data sources are standardized.
6.2. Recommendations for future work
A key impetus for this research is the prospect of the eventual implementation
of a GIS-based soil parameterization scheme within HEC-HMS.  Based on limited
trial simulations, use of the no resampling option showed better results than using
either NEXRAD cell or watershed level resampling.  Using a one-layer soil model,
it would be a simple task to include a list of the percentages for 12 different soil
textures (no resampling) on the line for each cell in an HEC-HMS parameter file
(see Figure 4.2).  Passing 12 values per cell versus passing one value per cell is not
183
overly cumbersome when procedures are computer automated.  The number of
vertical flux computations does increase when moving from NEXRAD level
resampling to no resampling, although not necessarily by a factor of 12, if all 12
textures are not represented in each NEXRAD cell.
The observations in the preceding paragraph are immediately relevant to event
simulation modeling in HEC-HMS.  For continuous simulation, it would be
interesting to try to combine the best features of the SWBMS model and the HEC
Continuous Soil Moisture Accounting Model that is now under development.
Baseflow simulation is an important component of continuous simulation.  Issues
involved with using the two-layer SWBMS to simulate baseflow have not been fully
explored.
The SWBMS programs have only been applied in one watershed and for
limited time periods.  It would be prudent to apply SWBMS for different time
periods and in different locations to verify the conclusions of this study.  It would be
interesting to test SWBMS in locations with a broader range of soil textures,
particularly with regard to comparing the use of the STATSGO and SSURGO soil
maps.
It would be interesting to explore the impacts of changing the top layer depth
on model results.  All simulations in this study were made with a top layer depth of
5 cm.  A small top layer depth is more likely to be fully saturated after a storm but
will also drain and return to mid-range saturation levels more quickly than a thicker
top layer initialized at the same water content.
It is questionable whether the calibrated rooting depth values derived in this
study would be appropriate to estimate evaporation during different seasons of the
year.  Use of remote sensing to estimate leaf area index values seems to be a
promising avenue for improving the physical basis for evaporation estimates over
large areas.  This is a large research area that could be the subject of many future
dissertations.
184
APPENDICES
The Appendices are divided into 5 parts.
A. Description of the AV-SWBMS and VB-SWBMS programs.
B.  Time series data processing programs
C.  Description of files on the CD-ROM that was produced to archive the
research presented in this dissertation.
D.  Table used to reclassify 719 STATSGO texture names into 12 basic USDA
classes and the "other" category.
185
APPENDIX A:  Description of the AV-SWBMS and VB-SWBMS
programs.
A.1  AV-SWBMS
A general description of the subprograms in AV-SWBMS is provided in
Chapter 4.  A more detailed description of each of these programs and the actual
filenames for these programs is provided here.
The only computer file required to access all of the capabilities of AV-SWBMS
is called swbms.avx.  The .avx filename extension indicates that this file is an
ArcView ?extension? file.  To make this extension accessible to ArcView users, it
should be placed in the directory ~/arcview/ext32 that gets created when ArcView is
installed on PCs.  When the AV-SWBMS extension is added to ArcView, two
menus get added to the standard View GUI as shown in Figure A1.
?SWBMS? ?SWBMS-Twolay?
Figure A1. Menus for AV-SWBMS.
When each of the menu items is selected, an Avenue script gets executed.  A list of
menu items, script names, and a detailed description of the script functions is
provided here.  Inputs and outputs for each program are also described.  The spatial
data required to run AV-SWBMS are summarized in Table 4.1.
?SWBMS? Menu Items
Menu item:  Intersect
Script name:  swbp.intersect
Must be preceded by menu items:  none
Description: Intersect watershed polygon with rainfall cells to define modeling
units.
Input(s):  watershed polygon, NEXRAD cell polygons
Output(s):  intersected shapefile of model polygons (see Figure 3.17)
186
Menu item:  Percent Impervious
Script name:  swbp.percentimp
Must be preceded by menu items:  Intersect
Description:  Computes percent imperviousness in each model polygon.  Adds a
field called "pct_imp" to the model polygon attribute table.  Impervious percent
values associated with specific land use codes that are shown in Table 3.5 are
hardwired into this program.  The program loops through all model polygons and
for each model polygon it loops through all intersecting land use polygons.  For
each land use polygon the percent imperviousness is determined based on the values
in Table 3.5.  An area-weighted average is used to determine the percent
imperviousness for the entire model polygon.
Input(s):  model polygons, land use polygons with a field containing the Anderson
Level II classification code
Output(s):  A new field called "pct_imp" is added to the model polygon attribute
table (Figure 3.20 displays values for the pct_imp attribute.)
Menu item:  Polygon/Grid Avg. Property
Script name:  swbp.gridavg
Must be preceded by menu items:  Intersect
Description:  Given any grid and any polygon coverage, this program computes
either the average value of all grid cells in each polygon or the most prevalent value
of the grid cells in each polygon.  The input polygon Theme must contain a field
with a unique ID for each polygon.  If the input polygon theme does not have a field
with a unique ID for each polygon, then a field containing record numbers can
easily be added to the polygon Theme using a script called
View.AddrecnumfieldtoThemes that is freely available at www.esri.com.  A copy of
this script is saved in the scripts directory on the CD-ROM produced with this
dissertation with the filename adrecno.ave.  The attribute (either an average or a
most prevalent value) computed by this program is written to a new field in the
model polygon Theme.  The user is prompted for the name of this new field.  For the
most prevalent value option, the input grid must be an integer grid.  In AV-
SWBMS, this program is used to compute the average flow length in each model
polygon and is also used to compute the predominant soil type code in each cell
when NEXRAD cell level resampling and gridded soils data are used.
IMPORTANT:  In order for the subsequent scripts to work, the flow length field
MUST be named ?flength.?
Input(s):  flow length grid or integer soil texture classification grid (For the soil
texture grid, the grid textural codes should be specified as follows: S = 1, LS = 2, SL
= 3, SCL = 4, L = 5, SIL = 6, CL = 7, SI = 8, C = 9, SC = 10, SICL = 11, and SIC =
12), model polygons
187
Output(s):  A field added to the model polygon Theme with a name specified by the
user.  The flow length field MUST be named "flength." (Average flow length values
for each cell are shown in Figure 3.35.)
Menu item:  SWBMS Input File (STATSGO)
Script name:  swbp.writeinputfile
Must be preceded by menu items:  Intersect, Percent Impervious, Polygon/Grid
Avg. Property
Description:  Creates an input file for VB-SWBMS using STATSGO soils data.
Loops through each model polygon and tallies the percentage of each soil type that
falls within that polygon.  The user is given a choice as to whether a file is written
with the predominant soil type only (NEXRAD cell level resampling) or with the
percentage of each soil type (No resampling) in each model polygon.
Input(s):  model polygons, soil comp table, STATSGO Theme, lookup table that
translates STATSGO texture class into one of the 12 basic USDA classes or the
"other" category (MUST be named statusda.txt)
Output(s):  VB-SWBMS input file (see Figure 4.1)
Menu item:  SWBMS Input File (gridded soils)
Script name:  swbp.writeinputfilegrid
Must be preceded by menu items:  Intersect, Percent Impervious, Polygon/Grid
Avg. Property
Description:  Creates an input file for VB-SWBMS using gridded soils data.  Loops
through each model polygon and tallies the percentage of each soil type that falls
within that polygon.  Writes a file with the percentage of each soil type in each
model polygon (No resampling).
Input(s):  model polygons, soil comp table, soil grid Theme, lookup table that
translates STATSGO texture into one of the 12 basic USDA classes or the "other"
category (MUST be named statusda.txt)
Output(s):  VB-SWBMS input file (see Figure 4.1)
Menu item:  SWBMS Input File (grid, most likely value)
Script name:  swbp.writeinputfilegrid_b
Must be preceded by menu items:  Intersect, Percent Impervious, Polygon/Grid
Avg. Property
Description:  Creates an input file for VB-SWBMS using NEXRAD cell level
resampling and assuming that a predominant soil type for each model polygon has
already been computed from a soils grid using Polygon/Grid Avg. Property.  The
188
user is prompted for the name of the field in the model polygon Theme containing
soil codes.
Input(s):  model polygons ? must contain field with predominant soil type
computed with Polygon/Grid Avg. Property
Output(s):  VB-SWBMS input file (see Figure 4.1)
?SWBMS-Twolay? Menu Items
Menu item:  STATSGO Comp Properties
Script name:  swbp.comp3
Must be preceded by menu items:  none
Description:  This program processes the comp and layer soil attribute tables and
creates a new table which lists the key hydraulic properties for each component that
are required for a two-layer VB-SWBMS model.  The program loops through all
components in the comp table and for each component loops through the associated
layers to determine the required parameter values.  This program uses both the
"statusda.txt" lookup table to translate STATSGO texture names into the 12 basic
USDA classes and the "usdapar.txt" lookup table to define hydraulic properties for
each texture class.  The lookup tables must have these names and be present in the
project when the program is run.  The program can process the comp and layer
tables for the entire state of Oklahoma in less than 30 minutes on a PC with a 200
MHz processor.
Input(s):  Lookup tables: statusda.txt, usdapar.txt (note that these two file names are
hardwired into the codes), comp table, layer table; all of these tables must exist in
the ArcView project.
Output(s):  A table of component properties called compjoin.dbf (this filename is
hardwired into the code)  Attributes in this compjoin.dbf include the texture code,
K
s
, ?, ?
r
, ?, h
b
, ?
fc
, and ?
wp
 for both the top layer and the soil layer with the lowest
K
s
 value (most restrictive layer) and the total soil zone depth.  Field names for the
restrictive layer attributes start with "r_".
Menu item:  30 m Grid Comp Properties
Script name:  swbp.compssurg2
Must be preceded by menu items:  none
Description:  Performs exacts same functions as the menu item STATSGO Comp
Properties but there are differences in the code to account for differences in the table
structures.
Input(s):  Same as for menu item STATSGO Comp Properties
Output(s):  Same as for menu item STATSGO Comp Properties except the filename
for the table that gets created is not hardwired into the code.
189
Menu item:  Comp Props File (STATSGO)
Script name:  swbp.writecompprops
Must be preceded by menu items:  Intersect, STATSGO Comp Properties
Description:  Creates one of two files input files for a two-layer VB-SWBMS
simulation ? the component properties file ? using the STATSGO soil map.  One
line gets written to an ASCII file for each distinct component in each model cell.
Information on all components in the first cell is written, followed by all
components for the second cell, etc.  Each line contains information about the
percent area taken by that component (comppct), the total soil zone depth (totdepth)
and K
s
 (k_lu), ? (por_lu), ?
r
 (theta_r), ? (lambda), h
b
 (hb), ?
fc
 (fc), ?
wp
 (wp) for both
the top layer and the most restrictive layer (r_klu, r_porlu, r_theta, r_lambda, r_hb,
r_fc, and r_wp).  IMPORTANT:  the comppct here is not the same as the comppct
attribute in the STATSGO comp table.  The comppct here is weighted by the model
cell fraction covered by the STATSGO component's mapunit.  Adding all of the
comppct values for a given cell will yield 100%.  Components with identical
properties are eliminated from the lists written to the component properties file.  The
routine for eliminating redundant components is time consuming but saves time
during the VB-SWBMS computations.  It is expected that VB-SWBMS
computations will be repeated many more times than the pre-processing routines.
Input(s):  model polygons, compjoin.dbf table created using menu item STATSGO
Comp Properties, STATSGO soils Theme
Output(s):  Component properties file (see Figure 4.2 for an example).  Adds an
attribute to the model polygon Theme indicating the number of distinct soil
components in each polygon.  The field name for this attribute is ?numcomps.?
Menu item:  Comp Props File (Gridded soils)
Script name:  swbp.writecomppropsgrid
Must be preceded by menu items:  Intersect, 30 m Grid Comp Properties
Description:  Creates the component properties file for VB-SWBMS using the 30 m
soils grid.  Procedures are identical to those above although the coding for spatial
analysis differs because of the differences between grid and vector data.
Input(s):  model polygons, component properties table created by menu item 30 m
Grid Comp Properties, soil grid Theme
Output(s):  Same as for menu item Comp Props File (STATSGO).
Menu item:  Cell Properties File
Script name:  swbp.writecellprops
Must be preceded by menu items:  Intersect, Percent Impervious, Polygon/Grid
Avg. Property
190
Description:  Creates a cell properties file for VB-SWBMS.  Reads attributes from
the model polygon Theme that were created using menus Intersect, Percent
Impervious, Polygon/Grid Avg. Property and either Comp Props File (STATSGO)
or Comp Props File (Gridded soils).  This program works independently of whether
grid or vector soils data are being used.
Input(s):  model polygons
Output(s):  Cell properties file (see Figure 4.2).
Table A1 describes menu item sequences that were used to make input files for
the modeling scenarios discussed in Chapter 5.
Table A1.  Suggested AV-SWBMS menu sequences for different modeling
options.
Model options Menu sequence
One-layer, STATSGO, No
resampling
Intersect, Percent Impervious, Polygon/Grid
Avg. Property [create ?flength? field],
SWBMS Input File (STATSGO)
One-layer, STATSGO, NEXRAD
cell level resampling
same as above
One-layer, gridded soils, No
resampling
Intersect, Percent Impervious, Polygon/Grid
Avg. Property [create ?flength? field],
SWBMS Input File (gridded soils)
One-layer, gridded soils, NEXRAD
cell level resampling
Intersect, Percent Impervious, Polygon/Grid
Avg. Property [create ?flength? field],
Polygon/Grid Avg. Property (create
?soiltype? field), SWBMS Input File (grid,
most likely)
Two-layer, STATSGO, No
resampling
Intersect, Percent Impervious, Polygon/Grid
Avg. Property [create ?flength? field],
STATSGO Comp Properties, Comp Props
File (STATSGO), Cell Properties File
Two-layer, gridded soils, No
resampling
Intersect, Percent Impervious, Polygon/Grid
Avg. Property [create ?flength? field], 30 m
Grid Comp Properties, Comp Props File
(Gridded Soils), Cell Properties File
191
A.2  VB-SWBMS
Several different versions of VB-SWBMS have been created.  The source files
for all program versions are stored on the project CD-ROM.  Table A2 lists each
model version and gives the source file names and directory locations.  The two
main model types are the one-layer and two-layer models for which there are
significant differences in the code.  The code differences among the one-layer
versions and among the two-layer versions are not large.  With more time for
developing the user interface, it would be better to have a single interface to all
modeling options of interest.
Table A2. VB-SWBMS File Names
Model version Source file names Directory
location
one-layer, no baseflow
simulation, no resampling
swbms1lay3.vpb, .vbw,
.frm
single_layer
one-layer, baseflow simulation,
no resampling
swbms1base.vbp, .vbw,
.frm
single_layer_base
one-layer, no baseflow
simulation, NEXRAD cell level
resampling*
swbms1lay4.vbp,.vbw,.fr
m
single_layer_pred
two-layer, baseflow simulation,
no resampling
swbm2lay.vbp,.vbw,.frm,
.frx
double_layer
two-layer, no baseflow
simulation, no resampling
swbms2_24.vbp, .vbw,
.frx, .frm
double_nobase
* Watershed level resampling can be achieved in any of the one-layer model
versions by editing the cell parameter file.
A.2.1  Example user interface description
Since the user interface features are similar for all versions, only one example
is given here.  The user interface for a one-layer model without baseflow simulation
is shown in Figure A2.  In the upper left-hand corner of Figure A2, there is a File
Listbox.  Underneath this box is a Directory Listbox and a Drive Listbox.  By
clicking on these boxes with the mouse, a user can navigate the computer's directory
structure and highlight a file of interest in the File Listbox.  Rectangular shaped
objects that appear raised and have writing on them are command buttons.  A mouse
click on one of these buttons executes a subprogram.  The rectangular white boxes
are text boxes.  A user can add text to any of these boxes by inserting the cursor and
typing.  The small white squares are check boxes.  A check placed in one of these
boxes indicates that a given option is turned on.  The small white circles are option
192
buttons.  These buttons can be highlighted with a click of the mouse and are used to
choose between different modeling options (i.e. between different options for
specifying the initial soil-water content).  Related buttons and text boxes have been
grouped with the frames labeled ?Identify Input Files,? ?Simulation Times,?
?Detention and Farm Ponds,? ?Initial Conditions,? ?PE Method,? ?Reference Crop
Parameters,? ?User Defined and/or ?Calibration? Factors,? and ?Transform
Parameters.?
The numbers that are seen labeling the command buttons serve only as a
reference for this text description and do not appear on the actual program user
interface.  Buttons 1-4 and 6 are used to select the desired input files.  For example,
rather than typing the full path name of the input precipitation file, a user can simply
navigate the directory structure, highlight the appropriate input file and click on
command button 1, causing the full path name for the highlighted input file to be
written in the text box next to button 1.  Button 1 is used to specify the precipitation
file.  Button 2 is used to specify a file containing either the required data for
estimating potential evaporation or pre-computed potential evaporation estimates
depending on which option button is selected.  Button 3 is used to specify the cell
parameter file.  Button 4 is used to specify the name of an observed runoff file.  The
check box next to this button should be checked if an observed runoff file is
provided.  If this box is not checked, then the user will not be able to write to the
observed runoff text box.  Button 5 is not used to specify an input file but is used to
specify the name of an output file, the final soil moisture file.  The full path name is
not required for this file.  All output files are written to the output directory written
next to button 11.  Button 6 is used to specify the filename for the soil parameter
lookup table.  In the two-layer user interface, there is an additional button to select
the component properties file which is not shown in Figure A2.
The starting date and ending date for the simulation are specified by typing in
the text boxes next to the words ?Start Day? and ?End Day.?  Simulations always
start and end at midnight and the time step is always one hour.  The date format
(mm/dd/yy) shown in Figure A2 should be used.
In the ?Detention and Farm Ponds? frame, all items except the top text box can
be ignored.  The top text box is used to specify the percent contributing area or the
?pond factor.?  The other items were used to test detention storage models that are
not used in this dissertation.
In the ?Initial Conditions? frame, the user chooses from three different methods
for specifying initial soil-water levels.  In the example shown, the user has chosen to
specify initial soil-water content from a file.  The filename can be specified using
button 7 in the same way that buttons 1-6 operate.  Other options to specify the
initial soil-water state are to specify initial suction head in cm or an initial effective
saturation.  These initial values can be typed in the appropriate text boxes and will
apply to all soils in the watershed.
193
The ?PE Method? frame is used to specify whether potential evaporation
estimates will be made using raw meteorological measurements and the
Penman/Monteith reference crop method or whether potential evaporation estimates
are supplied in an input file.
If the ?Penman/Monteith Reference Crop? option is selected, then ?Reference
Crop Parameters? need to be specified.  ?ra numerator? is the numerator for the
aerodynamic resistance function like that of Equation 3.49.  ?rsmin? is the minimum
stomatal resistance corresponding to conditions when soil moisture is not limiting
evaporation.
The frame ?User Defined and/or ?Calibration? Factors? allows the user to type
preferred values for model parameters.
The frame ?Transform Parameters? is used to specify the modClark storage and
time of concentration parameters.
Command button 13 is used to select a file that specifies soil components for
which average daily soil moisture values are to be exported.  Exporting these values
only occurs if the box next to ?Output WC for Selected Components? is checked.
Button 8 operates in the same way as buttons 1-6, but is used to select a model
control file.  The control file is just an ASCII file that contains a list of input files
and model parameter values.  If a control file is selected, clicking on button 9 will
read all values from the control file into the user interface.  Clicking on button 10
will write all current settings to a new control file.  The file specified in the text box
next to button 8 is the file which gets read from and written to.  Clicking on button
10 will overwrite an existing control file or create a new file if the specified
filename does not exist.  Button 11 can be used to specify the model output
directory.  All output files are written to this directory.  To set the output directory,
highlight the desired directory in the Directory Listbox and then click on button 11.
The selected directory name will get written to the text box next to button 11.
Once all input file names and parameter values have been set, button 14 is used
to run a simulation.  Optional text boxes include the observed runoff box and the
output component list box.  Only one of the text boxes in the ?Initial Conditions?
frame must contain a value.  All other text boxes should contain text.  The progress
of a simulation will be indicated by a progress bar at the bottom of the form.
Clicking on button 12 will exit the program.
1 2 3
4
5
1
6
7
8 9
10
11
12
14
13
F
i
g
u
re A2.  SW
B
M
S user inter
f
a
ce fo
r a one
-la
y
e
r
 model with no baseflow simulation.
194
195
A.2.2  Formats for input and output files:
A list of input and output files is given in Tables 4.4 and 4.5.  Input files
include a control file, a soil parameter lookup table, a rainfall file, a potential
evaporation file, an observed runoff file, a soil-water content initialization file, and a
file listing components for which soil-water content will be exported.  Examples
that show the required formats for input files are given below.  An example of a cell
properties file and a component properties file (two-layer model only) are not given
here because these files are illustrated in Figures 4.1 and 4.2.  Procedures for
creating a rainfall file, a potential evaporation file, and an observed runoff file are
described in Appendix B.  Output files include a flow file, a summary statistics file,
a file with final storage values (soil-water content and subsurface reservoir storage),
and daily average soil-water content for selected components (optional).  For files
that may be exceptionally long and/or wide (i.e. rainfall file), only a few lines and
columns are shown to illustrate the format.
196
Example VB-SWBMS input files.
# Control file for Soil Water Balance Program
# full path names of input data files
C:\washita\precip\06to07971to24mod10.csv
c:\washita\potent_evap\0497to0997day1.txt
c:\washita\single_layer\swbms1to24cc.csv
c:\washita\streamflow_data\bfj18tojl20.txt
c:\washita\soils\usdapar.txt
C:\washita\single_layer\cal1to24adj
c:\washita\single_layer\sminit.csv
smfinal.txt
# start time and end time for simulation
6/18/97
7/22/97
# soil moisture initialization method
fromfile
# initial effective saturation in layer 1
# initial soil suction in layer 1
# modClark storage (hours)
9
# modClark time of concentration (hours)
12
# Potential evaporation method: PMRC = Penman-Monteith Reference Crop, File = input values from a file
pmrc
# ks calibration factor
1.9
# layer 1 depth (cm)
5
# rooting depth (cm)
40
# Brooks and Corey pore-size index factor
1
# potential crop coefficient
1
# detention constant
0
# fraction of basin influence by detention
0
# farm pond factor
0.805
# numerator for aerodynamic resistance function
228
# minimum stomatal resistance
69
# use observed runoff
 1 
Example control file for a one-layer model with no baseflow simulation.
197
class,abbrev,ks,poros,theta_r,fc,wp,lambda,hb
sand,S,29.7,0.43,0.045,0.0456,0.0450,1.68,6.90
loamy_sand,LS,14.59,0.41,0.057,0.0600,0.0570,1.28,8.06
sandy_loam,SL,4.42,0.41,0.065,0.0845,0.0657,0.89,13.33
sandy_clay_loam,SCL,1.31,0.39,0.1,0.1688,0.1111,0.48,16.95
loam,L,1.04,0.43,0.078,0.1645,0.0883,0.56,27.78
silt loam,SIL,0.45,0.45,0.067,0.2390,0.1037,0.41,50.00
clay_loam,CL,0.26,0.41,0.095,0.2687,0.1493,0.31,52.63
silt,SI,0.25,0.46,0.034,0.2567,0.0897,0.37,62.50
clay,C,0.2,0.38,0.068,0.3466,0.2703,0.09,125.00
sandy_clay,SC,0.12,0.38,0.1,0.2666,0.1701,0.23,37.04
silty_clay_loam,SICL,0.07,0.43,0.089,0.3374,0.1962,0.23,100.00
silty_clay,SIC,0.02,0.36,0.07,0.3367,0.2661,0.09,200.00
Example soil parameter lookup table.
time,p557_304,p558_304,p559_304,p554_305,p555_305,p556_305. .
.
5059702,0,0,0,0,0,0. . .
5059703,0,0,0,0,0,0. . .
5059704,0,0,0,0,0,0. . .
5059705,0,0,0,0,0,0. . .
5059706,48,0,0,0,0,0. . .
5059707,204,227,295,0,0,0. . .
5059708,0,0,0,0,0,0. . .
.    .    .
.    .    .
.    .    .
55 comma delimited columns (one for each cell)
(54 for the 1:24,000 scale simulation runs)
t
i
me (mmddyy
hh
)
Rainfall units are 1/100 mm.  These are the units of the data distributed by
ABRFC.  SWBMS divides values from these files by 1000 and works with
units of cm/hour.
Example of rainfall input file.
198
"day,avail_heating,tempavg,tempmax,tempmin,presavg,rhavg,windavg,deficitavg"
"04/01/97",38.13,14.31,18.96,9.7,96.9,.5926,5.1,.679
"04/02/97",28.43,16.88,19.77,14.37,97.19,.7722,3.92,.449
"04/03/97",40,16.4,20.89,13.25,96.74,.8038,3.74,.404
"04/04/97",-.59,16.11,17.6,15.46,95.69,.9894,3.96,.019
"04/05/97",156.26,16.46,23.35,9.4,95.23,.5253,5.75,1.02
"04/06/97",133.41,9.22,14.89,3.65,97.05,.4464,3.66,.689
"04/07/97",138.71,12.54,20.32,4.24,97.3,.4649,3.35,.943
"04/08/97",9.4,6.99,13.07,1.23,97.27,.8893,4.43,.149
"04/09/97",62.67,6.16,11.24,1.49,97.47,.8622,3.34,.151
"04/10/97",60.07,15.61,19.26,10.32,96.52,.7386,6.1,.487
.                 .                 . 
.                 .                 .
.                 .                 .
Units: avail_heating (W/m
2
), tempavg (C), tempmax (C), tempmin (C), 
presavg (kPa), rhavg(none), windavg(m/s), deficitavg (kPa)
Example input file for Penman-Monteith potential evaporation option.
date,observed_flow,base_flow
#1997-06-18#,103,103
#1997-06-18 01:00:00#,99.75,99.75
#1997-06-18 02:00:00#,97,97
#1997-06-18 03:00:00#,95.5,95.5
#1997-06-18 04:00:00#,93.5,93.5
#1997-06-18 05:00:00#,92.25,92.25
#1997-06-18 06:00:00#,90.5,90.5
#1997-06-18 07:00:00#,89.25,89.25
#1997-06-18 08:00:00#,88.5,88.5
#1997-06-18 09:00:00#,88,88
#1997-06-18 10:00:00#,87,87
.                .               .
.                .               .
.                .               .
Flow units: cfs. 
Example observed flow file.
199
cell,S_wc,LS_wc,SL_wc,SCL_wc,L_wc,SIL_wc,CL_wc,SI_wc,C_wc,SC_wc,SICL_wc,SIC_wc
1,0.1,0.18,0.22,0,0.3,0.24,0.23,0,0.33,0,0.3,0
2,0.1,0.18,0.22,0,0.3,0.24,0.23,0,0.33,0,0.3,0
3,0.1,0.18,0.22,0,0.3,0.24,0.23,0,0.33,0,0.3,0
4,0.1,0.18,0.22,0,0.3,0.24,0.23,0,0.33,0,0.3,0
5,0.1,0.18,0.22,0,0.3,0.24,0.23,0,0.33,0,0.3,0
6,0.1,0.18,0.22,0,0.3,0.24,0.23,0,0.33,0,0.3,0
7,0.1,0.18,0.22,0,0.3,0.24,0.23,0,0.33,0,0.3,0
8,0.1,0.18,0.22,0,0.3,0.24,0.23,0,0.33,0,0.3,0
.           .             .
.           .             .
.           .             .
Example soil-water content initialization file.
lwid,hrapx,hrapy,soiltype
3_4_5,557,309,2
8_9,555,307,6
11_12_13,560,309,5
17,559,305,3
File listing soils for which soil-water content data
will be exported.  ?Lwid? numbers correspond to 
the numbers in Figure 3.18. 
?Soiltype? codes.
S = 1, LS = 2, SL = 3, SCL = 4, 
L = 5, SIL = 6, CL = 7, SI = 8,
 C = 9, SC = 10, SICL = 11, SIC = 12
200
Example VB-SWBMS output files.
date,direct_runoff,baseflow,total_runoff,observed_flow
#1997-06-18#,0,103,103,103
#1997-06-18 01:00:00#,0,99.75,99.75,99.75
#1997-06-18 02:00:00#,0,97,97,97
#1997-06-18 03:00:00#,0,95.5,95.5,95.5
#1997-06-18 04:00:00#,0,93.5,93.5,93.5
#1997-06-18 05:00:00#,0,92.25,92.25,92.25
#1997-06-18 06:00:00#,0,90.5,90.5,90.5
#1997-06-18 07:00:00#,0,89.25,89.25,89.25
#1997-06-18 08:00:00#,0,88.5,88.5,88.5
#1997-06-18 09:00:00#,0,88,88,88
#1997-06-18 10:00:00#,0,87,87,87
.        .          .
.        .          .
.        .          .
Example output flow file.  This file is automatically 
named ?flowfile.txt? and will overwrite any existing 
file with this name in the working directory.
Units: cfs
201
#1899-12-30 16:49:55#,#1899-12-30 16:56:10#
HEC Objective function:      146.740387455355 
Total precipitation (m3):   51370941.7
Total evaporation (m3):     13723076.1
Total mod Clark runoff (m3):              2931621.6
Change in mod Clark storage (m3):         325.
Total modeled direct runoff (m3):         2931946.6
Total observed direct runoff (m3):        2989815.9
Modeled - observed direct runoff (m3):    -57869.4
Total direct runoff due to impervious cover alone (m3): 903965.6
Total change in soil-water storage (m3):  -3572678.4
Total volume diverted to ponds (m3):      710223.1
Total volume sent to subsurface (m3):     37578116.5
Net mass unaccounted for (m3):            257.9
Input precip file:          C:\washita\precip\06to07971to24mod10.csv
Input climate file:         c:\washita\potent_evap\0497to0997day1.txt
Input cell file:            c:\washita\single_layer\swbms1to24cc.csv
Input observed runoff file: c:\washita\streamflow_data\bfj18tojl20.txt
Soil moisture initialization file:        c:\washita\single_layer\sminit.csv
Final soil moisture file:   smfinal.txt
Simulation start time:      6/18/97
Simulation end time:        7/22/97
Initial effective saturation in layer 1:  
modClark storage (hours):   9
modClark time of concentration (hours):   12
ks calibration factor:      1.9
Layer 1 depth (cm):         5
Rooting depth:              40
Brooks and Corey pore-size index factor:  1
Potential crop coefficient: 1
Farm pond factor:           0.805
Numerator for aerodynamic resistance function:          228
Minimum stomatal resistance 69
Example output summary file.  This file is automatically 
named ?summary.txt? and will overwrite any existing 
file with this name in the working directory.
202
cell,S_wc,LS_wc,SL_wc,SCL_wc,L_wc,SIL_wc,CL_wc,SI_wc,C_wc,SC_wc,SICL_wc,SIC_wc
1,0.,0.,0.,0.,0.1473,0.1894,0.,0.,0.,0.,0.,0.
2,0.,0.,0.,0.,0.1392,0.1814,0.,0.,0.,0.,0.,0.
3,0.07,0.0858,0.1206,0.,0.,0.23,0.,0.,0.,0.,0.,0.
4,0.045,0.057,0.0747,0.,0.,0.,0.,0.,0.,0.,0.,0.
5,0.,0.057,0.0657,0.,0.1141,0.1571,0.,0.,0.,0.,0.,0.
6,0.,0.057,0.0651,0.,0.0975,0.1401,0.,0.,0.,0.,0.,0.
.              .                .
.              .                .
.              .                .
The last line of this file will be the final subsurface reservoir storage (m3) if an SWBMS
version with baseflow simulation is being used.  
Final storage values example file.  This example is for a one-layer model.  The format
is slightly different for a two-layer model.  Reported values are water content.   
3_4_5,8_9,11_12_13,17
0.113,0.229,0.236,0.156
8.46E-02,0.210,0.196,0.123
6.76E-02,0.192,0.175,0.105
5.72E-02,0.175,0.158,8.83E-02
5.70E-02,0.159,0.141,7.20E-02
0.126,0.214,0.197,0.156
.        .         .
.        .         .
.        .         .
Example of average daily soil-water content file.  
The numbers in the header correspond to the numbers 
in Figure 3.18.
A.2.3  Description of subprograms.
There are many subprograms within VB-SWBMS.  The main program is called
cmdrun_click and is initiated from button 14 (?Run Simulation?).  This program
calls several subprograms and functions to make the program more readable.  A list
of the most important subprograms and their description is provided here.  Programs
that perform tasks not essential to the hydrologic calculations (i.e. the programs that
tell the File Listbox and the Directory Listbox how to communicate) are not
included.
203
  calcpevap:  Read average daily meteorological values from a climate file and
compute potential evaporation using the Penman-Monteith method.
  cmdctlread_click:  Read the selected control file and update text boxes,
check boxes, and options in the user interface.  Associated with the ?Read
Control File? button.
  cmdctlwrite_click:  Write a control file based on the current entries and
selections in the user interface.  Associated with the ?Write Control File?
button.
  cmdrun_click:  Main program.  Initiated with a mouse click on the ?Run
Simulation? button.  Tasks performed by this program are summarized in
Figure 4.3.  Calls many subprograms.
  cuminffunct:  A function that calculates the left-hand-side of Equation 3.28
minus the right-hand-side of Equation 3.28 ? used to solve for the root of
this equation.
  dcuminffunct:  A function that computes the derivative of the left-hand-side
of Equation 3.28 minus the right-hand-side for use in the Newton-Raphson
root finding subprogram.
  green_ampt:  Computes cumulative infiltration for an event using the
algorithm described in Section 3.3.1.
  hecobject:  Compute the HEC Objective Function given input arrays of
observed flow and simulated flow and an average observed flow value.
  hffunct:  A function that computes the wetting front suction head as a
function of the Brooks and Corey parameters and the initial effective
saturation.
  newt_raphs:  Solves for the cumulative infiltration at time step t+?t in
Equation 3.28 using the Newton-Raphson root finding method.
  outcompsub:  Assign values to the output array that stores average daily soil-
water content values for selected components.
  readprecip:  Read input precipitation file.
  readpevap:  Read potential evaporation estimates from a file.
204
  readcell:  Read cell parameter information contained in a file.
  readflowdata:  Read observed streamflow data from a file.
  readoutcomps:  Read the file containing a list of soil components for which
average daily soil-water content will be written to a file.
  readsminit:  Read soil-water content initialization file.
  readusdapar:  Read soil parameter lookup table from a file.
205
APPENDIX B:  Time series data processing programs
Many programs were written to process raw data files obtained from various
sources and convert them into formats useful in this research.  A summary of all
data used in this research is provided in Table 3.1.
Many of the data processing programs are written in Visual Basic and run from
the common GUI called "Data Processing Programs" shown in Figure B1.  The
Visual Basic files associated with the programs shown in Figure B1 are called
readdata.vbp, readdata.vbw, and readdata.frm.  Tasks performed by these programs
include processing ARM data sets, processing 15 minute USGS streamflow data,
baseflow separation, and processing Micronet rainfall data.  An additional utility is
provided to convert the flow units in the SWBMS output file (flowfile.txt) from cfs
to cms.  Processing programs not run from this interface include programs for
processing NEXRAD data and a program for converting files in netCDF binary
format into ASCII format.  These programs are also described in this section.
As with any program, a user of these programs should always check input and
output files to make sure the correct data transformations have been made.
206
Figure B1  Graphical User Interface for several data processing programs.
B.1 Processing ARM data
Instructions for obtaining ARM data sets are available at
www.archive.arm.gov/data/ordering.html.  ARM data are distributed in netCDF
(Network Common Data Format).  Although there are FORTRAN and C utilities
207
distributed on the Internet for working directly with binary netCDF files, the
approach taken in this research was to convert all netCDF files to ASCII files and
then extract the desired data from these files using Visual Basic programs.  A utility
called ncdump was used to convert netCDF files from binary to ASCII format.  The
FORTRAN source code for this program was obtained through
http://unidata.ucar.edu.  Ncdump was compiled on a Sun UltraSparc1 workstation.
Since the ncdump utility only operates on one file at a time, a semi-automated
procedure using UNIX batch files was used to process many files at once.  Here is
an example of how this procedure works:
? 79 EBBR data files (one for each day from 5/01/97 to 8/22/97) are downloaded
for site E4 to a single directory.
? A list of these files is written to an ASCII file called ?files.txt? using the UNIX
"ls" command.
? A C program called rdump.c is then run which creates a new file called
?macro.txt? with a list of UNIX shell commands appropriate for running the
ncdump program on each of EBBR files.
? In order to run the UNIX shell commands in ?macro.txt,? the chmod command
must be used to make ?macro.txt? an executable file.  The program rdump.c is
dependent on the file name length so it may need to be modified when applied to
different data streams (for example, SWATS file names are 1 character longer
than EBBR file names).
?Read Selected EBBR File(s)?:  Clicking on this button in Figure B1 executes a
Visual Basic sub-program called "cmdebbr_click()."  Two different tasks are
performed by this sub-program ?  (1) EBBR data used for the analysis described in
Section 3.3.2 are extracted from the ASCII versions of ARM data files and (2)
EBBR data used by the SWBMS program to estimate potential evaporation are
transformed to the correct format.  In both cases, data from many files may be
consolidated into a single file.  The following steps are needed to execute this
program:
? Using the Drive Listbox and the Directory Listbox shown in Figure B1, move
to the directory where the raw ASCII files (that have been converted from
netCDF format) are located.
? Highlight all files of interest in the ?Files in Working Directory? box using the
Windows standard combination of a mouse click and the shift key.
? Click the button "Select Files."  All highlighted files should now be listed in
the "Selected Files" list box.  If a mistake is made in this selection, just click
the ?Select Files? button again and the previously selected file list will be
overwritten.
208
? Click on "Read Selected EBBR File(s)."
? Enter an output file name including the desired file name extension.  This file
is written to the currently active directory in the Directory Listbox.
? As the program runs, the user has a choice of whether or not to create a file in
simulation format.  A file in simulation format is in a format that can be read
by another data reformatting program, the result of which can be used as input
to SWBMS if the Penman-Monteith option is selected for estimating potential
evaporation.  A file in simulation format is a comma delimited file with the
following entries on each line for each 30 minute period: time (this time has
been converted to Central Standard time from the Universal Standard Time
reported in the ARM data files); net radiation [W m
-2
]; ground heat flux [W m
-
2
]; temperature [?C]; atmospheric pressure [kPa]; relative humidity; and wind
speed [m s
-1
].  If simulation format is selected then the user will get no further
questions and a message saying when the computations are complete.  A
second program needs to be run to compute daily values for input to the
SWBMS model from the file just created.  To prepare a daily input file,
  Locate the 30 minute file just created from the "Read Selected EBBR
File(s)" button using the Directory Listbox.  Highlight the file name with
the mouse and click on "Select Files."
  Type the start date and the end date for the time period of interest in
the text boxes next to the words "Start" and "End."  The start date must
exist in the input file for the program to work.  An acceptable format for
the start and end dates is mm/dd/yy.
  Click on the button "Create Daily PE Input File."
  The user is prompted to name the output file.  The file will be written
to the active directory in the Directory Listbox.  A message will appear
when the program has completed.  The file created is a comma delimited
ASCII file with the following information:  date, available heating (R
n
 +
G) [W m
-2
], average temperature [?C], maximum temperature [?C],
minimum temperature [?C], average pressure [kPa], average relative
humidity, average wind speed [m s
-1
], and average vapor pressure deficit
[kPa].
? If the simulation format is not chosen, then a file in a different format is created.
This file contains 30 minute values:  time (CST), Bowen ratio, net radiation [W
m
-2
], latent heat flux [W m
-2
]), sensible heat flux [W m
-2
], soil heat flux [W m
-2
],
and soil-water content.  In order to come up with the estimate of soil-water
content, the user must enter a porosity value.  In some of the ARM data files, the
surface soil texture at a given site is noted.  This information may be used to
estimate porosity.  Suspect values for latent and sensible heat fluxes are flagged
with -1.
209
?Read Selected SWATS File(s):?  This button in Figure B1 runs a sub-program
called ?cmdswats_click().?  This program does not prepare input data for the
SWBMS program.  The purpose of this subprogram is only to reformat SWAT
ASCII files and put selected data in a format that is useful for making calculations
and graphs described in Section 3.3.2.  This program can be used to consolidate
information from many files into one file.  The following steps are required to run
this program:
? Using the Drive Listbox and the Directory Listbox shown in Figure B1, move
to the directory where the raw SWAT ASCII files (that have been converted
from netCDF format) are located.
? Highlight all files of interest in the ?Files in Working Directory? box using the
Windows standard combination of a mouse click and the shift key.
? Click the button "Select Files."  All highlighted files should now be listed in
the "Selected Files" list box.  If a mistake is made in this selection, just click
the ?Select Files? button again and the previously selected file list will be
overwritten.
? Click on "Read Selected SWATS File(s)."
? The user is prompted to enter an output file name including the desired file
name extension.  This file is written to the currently active directory in the
Directory Listbox.
? The program user is asked whether he/she wishes to use the ?West Profile.?
SWATS data files contain data for both an East profile and a West profile
separated by 1 meter.  Clicking yes will extract data for the West profile and
clicking no will extract data for the East profile.
? The output from the program is an ASCII text file in which each line contains
hourly observations of the following variables:  time (CST), average water
content from the top 3 measurement locations (watc_3), average suction head
from the top 3 measurement locations [cm] (wats_3), water content from the
top measurement location (watc_1), and piezometric head from all 8
measurement locations [cm] (watp_1, watp_2, watp_3, watp_4, watp_5,
watp_6, watp_7, watp_8).  Note that unit conversions have made from the
original SWAT files.  The SWAT data files used in this study contain
measurements from 8 depths: 5 cm, 15 cm, 25 cm, 35 cm, 60 cm, 85 cm, 125
cm, and 175 cm.
?Get Retention Data:?  This button in Figure B1 runs a sub-program called
?cmdretent_click().?  This program extracts water content and water potential data
for a selected layer so that a plot of the water retention curve for that layer can be
made.  The retention curves derived from the SWATS data files are not presented in
this dissertation because calibrations used to convert matric potential to water
210
content are still being refined (as of June 1, 1998,
www.arm.gov/docs/instruments/static/swats.html).  This program can be used to
consolidate information from many files into one file.  The following steps are
required to run this program:
? Using the Drive Listbox and the Directory Listbox shown in Figure B1, move
to the directory where the raw SWAT ASCII files (that have been converted
from netCDF format) are located.
? Highlight all files of interest in the ?Files in Working Directory? box using the
Windows standard combination of a mouse click and the shift key.
? Click the button "Select Files."  All highlighted files should now be listed in
the "Selected Files" list box.  If a mistake is made in this selection, just click
the ?Select Files? button again and the previously selected file list will be
overwritten.
? Type the layer number for which data will be extracted in the text box under
the words ?Layer #.?
? Click on "Get Retention Data."
? Enter an output file name including the desired file name extension.  This file
is written to the currently active directory in the Directory Listbox.
? The program automatically extracts data for the ?West Profile? and creates an
ASCII text file in which each line contains two values, water content
(wat_cont) and matric potential [cm] (wat_pot), for the selected layer.
?Read Selected SMOS File(s)?:  This button in Figure B1 runs a sub-program called
?cmdsmos_click().?  This sub-program does not prepare input data for the SWBMS
program.  The purpose of this subprogram is only to reformat SMOS ASCII files
and put selected data in a format that is useful for making calculations and graphs
described in Section 3.3.2.  This program can be used to consolidate information
from many files into one file.  The following steps are required to run this program:
? Using the Drive Listbox and the Directory Listbox shown in Figure B1, move
to the directory where the raw ASCII files (that have been converted from
netCDF format) are located.
? Highlight all files of interest in the ?Files in Working Directory? box using the
Windows standard combination of a mouse click and the shift key.
? Click the button "Select Files."  All highlighted files should now be listed in
the "Selected Files" list box.  If a mistake is made in this selection, just click
the ?Select Files? button again and the previously selected file list will be
overwritten.
? Click on "Read Selected SMOS File(s)."
211
? Enter an output file name including the desired file name extension.  This file
is written to the currently active directory in the Directory Listbox.
? The file created by this program contains 30 minute values for time (CST),
precipitation [mm], snow depth [mm], wind [m s
-1
], temp [?C], relative
humidity, and barometric pressure [kPa].
Select Files to Merge, Clear Selected, and Merge Selected Files:   These buttons in
Figure B1 are used to merge files containing EBBR, SWATS, and SMOS data.
Merging the data from these files into one file made the analysis for Section 3.3.2
easier.  The three input files containing EBBR, SWATS, and SMOS data must be in
the format created by the ?Read Selected EBBR File(s),? ?Read Selected SWATS
File(s),? and ?Read Selected SMOS File(s)? buttons.  These files must be named
with the file extensions ?.ebbr?, ?.swat?, and ?.smos?.  To create a merged file,
? Using the Drive Listbox and the Directory Listbox shown in Figure B1, move
to the directory where each input file is located and select each input file by
clicking the ?Select Files to Merge? button.  Selected files get written to the
?Files to merge? text box.  Selected files can be cleared using the ?Clear
selected? button.
? Type the start date and end date for the data that you wish to merge in the text
boxes next to the words ?Start? and ?End? (mm/dd/yy).  The data for these
dates must exist in the input files for the program to work.
? Click the ?Merge Selected Files? button.
? The user is prompted to enter an output file name including the desired file
name extension.  The output file is written to the currently active directory in
the Directory Listbox.
B.2  Processing Streamflow Data
?Read USGS Streamflow?:  This button in Figure B1 runs a sub-program called
?cmdusgs_click().?  This program was used to convert 15 minute flow data obtained
from the Oklahoma office of the U.S. Geological Survey into a more useful format.
The 15 minute data obtained from the USGS were in a complex, multi-column
format with several lines of text separating the values for individual months.
Cmdusgs_click() computes hourly average flows from the 15 minute data and
creates a much simpler text file with only two columns, time and flow (cfs).  No
time conversions are made by this program.  To convert the flow time series (which
are reported in local time) to Greenwich Mean Time (GMT), the ?Shift USGS to
GMT? button is used.  To run cmdusgs_click() ,
? Using the Drive Listbox and the Directory Listbox shown in Figure B1, move
to the directory where the raw ASCII file is located.
212
? Highlight the file of interest in the ?Files in Working Directory? box using the
Windows standard combination of a mouse click and the shift key.
? Click the button "Select Files."  The highlighted file should now be listed in
the "Selected Files" list box.  If a mistake is made in this selection, just click
the ?Select Files? button again and the previously selected file list will be
overwritten.
? Click on "Read USGS Streamflow."
? Enter an output file name including the desired file name extension.  The
output file is written to the currently active directory in the Directory Listbox.
To shift the output flow file to GMT,  repeat the above procedure for identifying the
input file but select the output file that was just created as input.  Click on the button
?Shift USGS to GMT? (runs a subprogram called ?cmdshift_click()?) to create a
new ASCII file with hourly flow values in GMT.
?Baseflow Separation:?  Clicking on this button in Figure B1 runs a sub-program
called ?cmdbflow_click().?  Cmdbflow_click() computes baseflow using the
algorithm described in Section 3.3.6.2.  The input file and output file format for this
program are shown below.  Flow can be in any units.  A single line header must
exist in the input file.  An input file could be created using the ?Read USGS
Streamflow? button.
date,observed_flow,base_flow
#1997-06-18#,103,103
#1997-06-18 01:00:00#,99.75,99.75
#1997-06-18 02:00:00#,97,97
#1997-06-18 03:00:00#,95.5,95.5
#1997-06-18 04:00:00#,93.5,93.5
#1997-06-18 05:00:00#,92.25,92.25
.        . 
.        .
.        .
Time,flow_cfs
#1997-06-18#,103
#1997-06-18 01:00:00#,99.75
#1997-06-18 02:00:00#,97
#1997-06-18 03:00:00#,95.5
#1997-06-18 04:00:00#,93.5
#1997-06-18 05:00:00#,92.25
.        . 
.        .
.        .
Input file format
Output file format
? The baseflow parameter must be specified by typing a value into the text box
next to the words ?Baseflow Parameter? before clicking on the ?Baseflow
Separation? button.  The baseflow parameter is in the same units as the time
step for the input file (hours in the above example).
? The dates of interest must be specified in the ?Start? and ?End? text boxes.
The program will not work if the start date does not exist in the input file.
213
? The input file is specified by using the Drive Listbox and the Directory
Listbox shown in Figure B1, moving to the directory where the input file is
located, highlighting the input file with the mouse, and clicking on the "Select
Files" button so that the appropriate file is listed in the ?Selected Files? box.
?Read Micronet File(s) for a Single Station? and ?Merge Micronet:?  These program
buttons in Figure B1 are specifically designed to reformat the Little Washita
Micronet 5 minute rainfall data.  In the Micronet data downloaded from
grl.ars.usda.gov/sgp97-1.html, there is a file for each day and each station.  The
format that was useful in this research was to have a single file that contains hourly
rainfall values for all stations.  Thus, the data processing procedure was divided into
two steps.  In the first step, data files for all days of interest from a single station are
read, hourly values are computed and a single file with hourly values for that station
is created.  In the second step, data files from each station are merged into a single
file.
? Clicking on the button ?Read Micronet File(s) for a Single Station?
(cmdmicronet_click()) will process all files listed in the ?Selected Files? box.
The only other input required by the user is an output file name.
? Clicking on the button ?Merge Micronet? (cmdmergemnet_click()) will merge
single station files listed in the ?Selected Files? box into one file.  Two output
files are created by this program.  One output file contains hourly data for all
stations and the other file contains the hourly arithmetic average rainfall for all
stations.  NOTE:  Even though this is a merge operation, the box labeled ?Files
to Merge? is not used.  No dates need to be specified for this program to run.
cfs_to_cms:   The button cfs_to_cms (cmdcfstocms_click()) in Figure B1 was
created to convert the units in the VB-SWBMS output flow file called flowfile.txt
from cfs to cms.  In future versions it will be more practical to just have VB-
SWBMS write the output file in cms or give the user a choice of units.  The
cfs_to_cms button operates on a single file listed in the ?Selected Files? box.
B.3  Processing NEXRAD StageIII Rainfall Data
Hourly NEXRAD StageIII data are archived by the Arkansas-Red Basin River
Forecast Center (ABRFC ? info.abrfc.noaa.gov) in network Common Data Format
(netCDF ? www.unidata.ucar.edu/packages/netcdf/).  A semi-automated procedure
was developed to facilitate the tedious task of downloading hourly netCDF files
from ABRFC and extracting the data that covers the Little Washita watershed.  This
procedure could certainly be improved with more work.
As discussed in Section 3.2.4, the Hydrologic Rainfall Analysis Project
(HRAP) grid is used to map radar rainfall estimates.  The first step in the process of
214
using the StageIII data is to determine the HRAP grid cells which are completely or
partially within the Little Washita watershed.  Programs described at
www.ce.utexas.edu/prof/maidment/gishydro/seann/nexrad.htm were used to create
an HRAP grid coverage and a standard GIS intersection procedure was used to
identify the cells of interest as shown in Figure 3.17.  Based on the watershed
delineation using the 3 arc-second digital elevation model, there are 55 HRAP cells
that intersect the Little Washita watershed.  These 55 cell HRAP ID?s are:
hrapx, hrapy values
553, 306-309
554, 305-309
555, 305-309
556, 305-310
557, 304-310
558, 304-310
559, 304-310
560, 305-311
561, 306-311
562, 309
A subtle point that was not discovered by the author until all precipitation
reformatting had been completed is that the watersheds delineated from the 3 arc-
second digital elevation model and the 1 arc-second differ significantly enough so
that tiny portions of the 1 arc-second watershed intersect HRAP cells that do not
intersect with the 3 arc-second watershed.  Tiny portions of HRAP cells (559, 311),
(562, 308), and (562, 307) intersect with the 1 arc-second watershed but not with the
3 arc-second watershed.  The total of the intersected areas for these cells amounts to
only 0.1 km
2
, so these cells were ignored in calculations made with the 1 arc-second
data.  Rainfall files used with the 1:24,000 data only contain data from 54 of the 55
cells because a tiny intersection with cell (553, 306) does not completely contain
any 30-m grid cells.
The procedure used to extract the data for these cells was implemented on a
Sun UltraSparc1 Workstation with the UNIX (cshell) operating system.  To
download StageIII files for June, 1997, the following sequence of commands was
used.
ftp -i info.abrfc.noaa.gov
login: anonymous
password: your_email_address
cd archive/1997/jun/1hr_netcdf
bin
mget *
215
This procedure will take several minutes depending on the rate of data transfer.  720
zipped files, each with 53,265 data values get transferred.
A shell script containing the following commands is run to process the
downloaded files for a month at a time and create a file in the correct format for
input into VB-SWBMS.  This procedure could certainly be streamlined with more
work but serves its purpose.
Table B1:  Procedure for processing NEXRAD StageIII data.
gunzip -v * Unzip all files; unzipped files will have the .nc
extension.
\rm files.txt Delete any old versions of files.txt.
ls .nc > files.txt Create a new file called files.txt which
contains a list of files with the .nc ending.
rdump Run rdump program. Rdump is a compiled
version of a C program called rdump.c.
Rdump creates a file called macro.txt which
contains a list of commands necessary to
convert all of the .nc files listed in files.txt into
ASCII files with a .tx1 ending.  The commands
include a call to the utility ncdump described
in the ?Processing ARM Data? section of
Appendix B.  Note: As written, rdump.c
requires a specific directory structure.
macro.txt This command runs an executable text file
called macro.txt (see above).  This file can be
made executable using the chmod command.
\rm *.nc Remove all raw netCDF files to save space.
\rm files1.txt Delete any old versions of files1.txt.
ls *.tx1 > files1.txt List all *.tx1 files created by macro.txt to a file
called files1.txt.
replace2 Run replace2 program.  Replace2 is a
compiled version of a C program called
replace2.c.  Replace2 eliminates all header
information and delimiting commas from all of
the *.tx1 files and creates a new set of files
with the *.tx2 ending.
\rm *.tx1 Remove all *.tx1 files to save space.
\rm files2.txt Delete any old versions of files2.txt.
ls *.tx2 > files2.txt List all *.tx2 files in a file called files2.txt.
readvals2 Run readvals2 program.  Readvals2 is a
compiled version of a FORTRAN program
216
called readvals2.f.  Readvals2 reads through all
the StageIII files (720 for a month) with the
*.tx2 ending and extracts the data for the
HRAP cells in the Little Washita listed above.
This program must be changed to use it in any
location other than the Little Washita.  A
comma delimited file called precip.txt with one
line for each hour and one column for each
rainfall cell is created.
mv precip.txt 0697prec.txt Rename the precipitation file to so that it will
not be overwritten during future processing.
The text header with a list of cell IDs required
by VB-SWBMS is not created by readvals2
and was created manually and copied to the
top of each precipitation file.  With the
addition of this header, file 0697prec.txt would
be in the input format required by VB-
SWBMS.
\rm *.tx2 Remove all
The time stamp for each hourly NEXRAD StageIII data file represents the time
at the end of the hour for which rainfall is recorded.  Because the flow files used in
this research have their time stamp at the beginning of the hour in which flow is
recorded, the time stamps for the rainfall data are shifted to the beginning of the
hour for consistency.  Since this fact was not discovered until after the rainfall
processing described above, a button was added to the Visual Basic ?Data
Processing Programs? shown in Figure B1 to shift the time stamp on the rainfall
data.
Shift NEXRAD Data: The ?Shift NEXRAD Data? button operates on a single file
listed in the ?Selected Files? box.  The file format for the input and output files are
the same as the rainfall input format for the VB-SWBMS program.  The time stamp
corresponding to each set of rainfall values is shifted back one hour.
217
Appendix C:  Description of files on the CD-ROM that was produced to
archive the research presented in this dissertation.
Type of
File*
File Name Description
* Filetypes: ascii (text), excel (Microsoft Excel), grid (Arc/Info
Grid), shape (ArcView shapefile), cover (Arc/Info coverage), word
(Microsoft Word), ppoint (Microsoft PowerPoint).  Directories
are denoted by Dir.  Top level directories are printed in bold.
Dir avfiles contains AV-SWBMS program files
ascii swbms.apr ArcView project file that contains scripts and menus
for AV-SWBMS (also contains many scripts not used
in AV-SWBMS)
ascii swbms.avx ArcView extension described in Appendix A.1
Dir arm contains EBBR, SMOS, and SWATS data files
obtained from the DOE ARM data archive.
Dir e4ebbr contains data files with 30 minute energy budget
bowen ratio (EBBR) measurements at site E4 in the
Southern Great Plains Cloud and Radiation Testbed
ascii sgp30ebbrE4.a1.970501.000
000.txt
EBBR raw data file for 5/01/97.  There are 79 data
files with similar names for each day from 5/01/97 to
8/22/97.
ascii 0597.ebbr, 0697.ebbr,
0797.ebbr, 0897.ebbr
comma delimited text files containing selected 30
minute surface energy budget components for each
month; these files were created using the "Read
Selected EBBR File(s)" button in the "readdata"
Visual Basic program.
Dir e4merge directory in which merged EBBR, SMOS, and
SWATS data files for site E4 are analyzed
excel e4julmerg_r.xls,
e4junmerg_r.xls,
e4maymerg_r.xls
spreadsheets used for analysis of data extracted from
EBBR, SMOS, and SWATS data streams as described
in Section 3.3.2.
Dir e4smos contains data files with 30 minute surface
meteorological observation system (SMOS)
measurements at site E4 in the Southern Great Plains
Cloud and Radiation Testbed
ascii sgp30smosE4.a1.970501.00
0000.txt
SMOS raw data file for 5/01/97.  There are 80 data
files with similar names for each day from 5/01/97 to
8/22/97.
ascii 0597.smos, 0697.smos,
0797.smos
comma delimited text files containing selected 30
minute SMOS data for each month; these files were
created using the "Read Selected SMOS File(s)"
button in the "readdata" Visual Basic program.
Dir e4swats contains data files with 30 minute soil water and
temperature sensors (SWATS) measurements at site
E4 in the Southern Great Plains Cloud and Radiation
Testbed
ascii DsgpswatsE4.a1.970521.00
0700.txt
SWATS raw data file for 5/21/97.  There are 186 data
files with similar names for each day from 5/21/97 to
218
12/31/97.  The upper case ?D? at the beginning of the
file name indicates that this data is from a
developmental data stream.
ascii 0597.swats, 0697.swats,
0797.swats
comma delimited text files containing selected 30
minute SWATS data for the specified month; these
files were created using the "Read Selected SWATS
File(s)" button in the "readdata" Visual Basic
program.
Dir e7ebbr contains data files with 30 minute energy budget
bowen ratio (EBBR) measurements at site E7 in the
Southern Great Plains Cloud and Radiation Testbed
ascii sgp30ebbrE7.a1.970601.000
000.txt
EBBR raw data file for 6/01/97.  There are 110 data
files with similar names for each day from 6/01/97 to
9/30/97.
ascii 0697.ebbr, 0797.ebbr,
0897.ebbr, 0997.ebbr
comma delimited text files containing selected 30
minute surface energy budget components for each
month; these files were created using the "Read
Selected EBBR File(s)" button in the "readdata"
Visual Basic program.
Dir e7merg directory in which merged EBBR, SMOS, and
SWATS data files for site E7 are analyzed
excel e7augmerg_r.xls,
e7julmerg_r.xls,
e7junmerg_r.xls,
e7sepmerg_r.xls
spreadsheets used for analysis of data extracted from
EBBR, SMOS, and SWATS data streams as described
in Section 3.3.2.
Dir e7smos contains data files with 30 minute surface
meteorological observation system (SMOS)
measurements at site E7 in the Southern Great Plains
Cloud and Radiation Testbed
ascii sgp30smosE7.a1.970609.00
0700.txt
SMOS raw data file for 6/09/97.  There are 233 data
files with similar names for each day from 6/09/97 to
3/07/98.
ascii 0697.smos, 0797.smos,
0897.smos, 0997.smos
comma delimited text files containing selected 30
minute SMOS data for each month; these files were
created using the "Read Selected SMOS File(s)"
button in the "readdata" Visual Basic program.
Dir e7swats contains data files with 30 minute soil water and
temperature sensors (SWATS) measurements at site
E7in the Southern Great Plains Cloud and Radiation
Testbed
ascii DsgpswatsE7.a1.970609.00
0700.txt
SWATS raw data file for 6/09/97.  There are 233 data
files with similar names for each day from 6/09/97 to
03/07/98.  The upper case ?D? at the beginning of the
file name indicates that this data is from a
developmental data stream.
ascii 0697all.swat, 0797all.swat,
0897all.swat, 0997all.swat
comma delimited text files containing selected 30
minute SWATS data for the specified month; these
files were created using the "Read Selected SWATS
File(s)" button in the "readdata" Visual Basic
program.
219
Dir delin1sec contains files related to watershed delineation and
spatial analysis using the 1-second (30 m) DEM
and the 30-m soils grid
grid dem1alb 30-m DEM, vertical units are ft
grid fdr_clip flow direction grid derived from 30-m DEM, clipped
to area draining to Ninnekah gage.
grid filclip filled DEM, vertical units are ft, clipped to area
draining to Ninnekah gage
grid filclip_m filled DEM, vertical units are m, clipped to area
draining to Ninnekah gage
grid flen1clip flow length grid derived from filclip_m
grid shd1mask mask of 30-m cells draining to Ninnekah gage, cells
inside the watershed have the value 1, cells outside
the watershed are assigned NODATA.
grid soil30clip 30-m soils grid clipped to the area draining to the
Ninnekah gage
grid soil30m 30-m soils grid, unclipped
grid soiltypes 30-m soils grid, values reclassified into basic USDA
texture classes; 1 = S, 2 = LS, 3 = SL, 5 = L, 6 = SIL,
11 = SICL
grayelev.avl ArcView legend file for elevation grayshades
ascii lwdemw.img raw 30-m elevation data obtained from El Reno ARS
ascii lwdemw.two raw 30-m elevation data reformatted from a 1-
dimensional to a 2-dimensional ASCII array.
ascii onetotwo.f FORTRAN program used to reformat lwdemw.img
into lwdemw.two
shape shd1secp.shp,.dbf,.shx watershed polygon boundary delineated from 30-m
DEM
shape shdnex1.shp,.dbf,.shx shd1secp.shp intersected with NEXRAD cells
shape shdnex_clip.shp,.dbf,.shx shdnex1.shp with 3 tiny polygons removed, these
removed polygons correspond to HRAP cells (559,
311), (562, 308), and (562, 307); these tiny polygons
do not appear in the equivalent coverage from the 3-
arc second (100-m) delineation; contains flowlength
and percent imperviousness attributes.
Dir delin3sec contains files related to watershed delineation and
spatial analysis using the 3-second DEM, the
1:250k soils data, and land use data
grid dem3al1 square portion of dem3alb clipped to the area of
interest
grid dem3alb 3-second DEM with data from USGS 1:250k
quadrangles Clinton (East), Oklahoma City (West),
Lawton (East), and Ardmore (West); grid has been
merged and projected to a 90 m grid.
grid demfclip filled 3-second DEM clipped to the watershed
boundary
grid demfdr2 flow direction for demfil1
grid demfil1 filled dem3al1
grid fdrclip flow direction grid clipped to the watershed boundary
grid flen1 flow length grid computed from demfil1
220
grid flenclip flow length grid clipped to the watershed boundary
grid shdmask mask of watershed boundary with 100-m cells
shape cliptheme.shp,.dbf,.shx rectangle used to clip other data sets
dBase compjoin.dbf dBase file containing computed component attributes
(used for two-layer model only)
shape landuse.shp,.dbf,.shx 1:250K GIRAS landuse polygons for Ardmore and
Lawton quadrangles
shape lwnexrad.shp,.dbf,.shx NEXRAD cells in the vicinity of the Little Washita
R.; a 12x9 mesh of HRAP cells; information about
creating a mesh of HRAP cells may be found at
www.ce.utexas.edu/prof/maidment/gishydro/seann/ne
xrad.htm
shape modelp.shp,.dbf,.shx intersection of NEXRAD cells and watershed
boundary; KEY COVERAGE - contains cell
attributes used for modeling
shape nexlu.shp,.dbf,.shx intersection of NEXRAD cells, land use, and
watershed boundary
shape nexshdsoil.shp,.dbf,.shx intersection of NEXRAD cells, watershed boundary,
and STATSGO soil polygons
shape ninnekahshd.shp,.dbf,.shx watershed boundary; area draining to the Ninnekah
gage
shape shdlu.shp,.dbf,.shx watershed boundary intersected with land use
polygons
shape soillu2.shp,.dbf,.shx STATSGO polygons intersected with land use
polygons for the area of interest
shape statnearlw.shp,.dbf,.shx STATSGO polygons near the Little Washita R.
shape statsclipfig.shp,.dbf,.shx STATSGO polygons clipped for large area in central
Oklahoma used to make Figure 3.5.
shape statsshd.shp,.dbf,.shx STATSGO polygons intersected with the watershed
boundary
Dir double_layer contains files used for simulations with two-layer
model described in Section 5.4.
Dir cal1to24adj contains files for calibration runs (6/18/97-7/22/97)
using 1:24K input data and adjusted rainfall
ascii 0807ctl.txt control file for VB-SWBMS
excel flow.xls plot of simulated flows in Excel
ascii flowfile.txt VB-SWBMS flow output file (cfs)
ascii flowm3.txt VB-SWBMS flow output file converted to cms
ascii smf.txt output file from VB-SWBMS with final soil moisture
values
ascii summary.txt summary output file created by VB-SWBMS
ascii 0810cellss.txt cell parameter file for the two-layer model with 1:24K
input data (?ss? stands for SSURGO)
ascii 0810compss.txt component properties file for the two-layer model
with 1:24K input data (?ss? stands for SSURGO)
ascii cellscompd24.txt file which lists HRAP cell coordinates and the soil
grid-code for locations at which modeled average
daily soil water content values are to be compared
with predicted values
ascii sminit_d24.csv file used to initialize soil moisture in the two-layer
221
model with 1:24K soils data (second layer is
initialized to field capacity in this file)
ascii sminit_d24_1.csv file used to initialize soil moisture in the two-layer
model with 1:24K soils data (second layer is
initialized to a water content of 0.1 in this file)
swbm2lay.frm, .vbp,.vbw Visual Basic files for two-layer model with baseflow
Dir double_nobase contains files used for simulations with two-layer
model with no baseflow described in Section 5.4.
ascii 0810cellss.txt cell parameter file for the two-layer model with 1:24K
input data (?ss? stands for SSURGO)
ascii 0810compss.txt component properties file for the two-layer model
with 1:24K input data (?ss? stands for SSURGO)
ascii cellscompd24.txt file which lists HRAP cell coordinates and the soil
grid-code for locations at which modeled average
daily soil water content values are to be compared
with predicted values
ascii sminit_d24no.csv soil-water content initialization file
swbms2_24.frm, .frx, .vbp,
.vbw
two-layer model modified to work specifically with
the 1:24K input data and input data for the baseflow
separation alternative
Dir cal1to24adj files associated with ?double_nobase? calibration
run using 1:24K input data
ascii 0807ctl.txt VB-SWBMS control file
excel flow.xls spreadsheet used to plot flow results
ascii flowfile.txt VB-SWBMS output flow file
ascii flowm3.txt flowfile.txt converted from cfs to cms
ascii smdailavg.txt VB-SWBMS output file containing soil-water content
values
ascii smf.txt VB-SWBMS output file of final soil-water content
values
excel soil_moist.xls spreadsheet used to plot soil-water content results
ascii summary.txt VB-SWBMS output summary file
Dir val1to24adj files associated with ?double_nobase? validation
run using 1:24K input data
ascii 0807ctl.txt VB-SWBMS control file
ascii flowfile.txt VB-SWBMS output flow file
ascii flowm3.txt flowfile.txt converted from cfs to cms
excel flowv2.xls spreadsheet used to plot flow results
ascii smdailavg.txt VB-SWBMS output file containing soil-water content
values
ascii smf.txt VB-SWBMS output file of final soil-water content
values
ascii summary.txt VB-SWBMS output summary file
Dir micronet contains unprocessed and processed Micronet
rainfall data files
Dir a110, a111, a121, etc. . . . These 42 directories contain the raw data files for
each of 42 Micronet stations for June, July, and
August, 1997 (from grl.ars.usda.gov/sgp97-1.html).
In each directory, there are raw 5 minute
meteorological data files for each day in each month.
222
Each of these directories also contains two summary
files for the months of June and July, 1997, containing
hourly rainfall totals.  These files are named after their
respective directories concatenated with the endings
?jul.txt? and ?jun.txt.?
Dir julsum contains files summarizing rainfall values for July,
1997
ascii jul10storm.txt file used to help create Figure 5.1
excel julyavg.xls comparison of areal average hourly rainfall values
from the Micronet and StageIII rainfall data sources
Dir junsum contains files summarizing rainfall values for June,
1997
excel junavg.xls comparison of areal average hourly rainfall values
from the Micronet and StageIII rainfall data sources
shape microdel.shp, .shx, .dbf point shapefile of Micronet stations with reported data
througout the July 10 storm (36 points)
shape microdth.shp, .shx, .dbf thiessen polygons from microdel.shp
shape micronet.shp, .shx, .dbf point shapefile of all 42 Micronet stations
ascii micronetloc.txt latitude and longitude coordinates of Micronet
stations
excel micronetloc.xls latitude and longitude coordinates of Micronet
stations
shape shdthiess.shp, .shx, .dbf intersection of watershed boundary with microdth.shp
(contains only 32 polygons)
Dir office_docs Microsoft Office documents
Dir Excel Microsoft Excel documents
excel baseillust.xls used to generate figures illustrating baseflow
separation (Figures 3.39 and 3.40)
excel radexample.xls example DOE ARM EBBR data for July, 1997 ? used
to create Figures 3.27, 3.28, 3.29, 3.30, and 3.31
excel soil_moist.xls example comparison between predicted and observed
soil-water content (see Figure 5.4)
excel stypes.xls used to create bar chart showing the distribution of
surface soil textures in the Little Washita (Figure 3.6)
excel tornado.xls used to create Figure 5.9
excel usdapar.xls used to create Figures 3.9-3.12
Dir Power Point related Power Point Presentations
ppoint defens_slides2.ppt slides used at dissertation defense (Oct. 12, 1998)
ppoint figures.ppt, figures0.ppt,
figures2.ppt, figures3.ppt,
figures4.ppt
figures used in dissertation
ppoint hecoct98_c.ppt slides for presentation at HEC (Oct. 9, 1998)
ppoint rainfigs.ppt used for Figure 5.1 of dissertation
ppoint lecture_43098.ppt GIS class lecture related to dissertation
Dir Word Microsoft Word Documents
word anal_results.doc Chapter 5: Analysis and results
word appendices2.doc Appendices
word comp_imp.doc Chapter 4: Computer implementation
word conclusions.doc Chapter 6: Conclusions
word fpages.doc Front Pages (copyright, title, etc.)
223
word fpages2.doc Fornt Pages (cont.?acknowledgements, TOC, etc.)
word intro.doc Chapter 1: Introductions
word litreview.doc Chapter 2: Literature Review
word method.doc Chapter 3: Methodology
word references.doc References
word VITA.doc Vita
Dir potent_evap contains data used to estimate potential
evaporation
ascii 0497to0997day1.txt daily average meteorological data from April to
September, 1997 (used for Penman-Montieth
calculations); this is an input file for VB-SWBMS
model runs
Dir precip processed NEXRAD StageIII precipitation data
and associated files
ascii 0197prec.csv, 0297prec.csv,
0397prec.csv, 0497prec.csv,
0597prec.csv, 0697prec.csv,
0797prec.csv, 0897prec.csv,
0997prec.csv, 1095prec.csv,
1196prec.csv, 1296prec.csv,
1to496prec.csv,
5to896prec.csv
StageIII hourly rainfall estimates in 55 cells for the
month specified in the filename
ascii 06to07971to24.csv StageIII hourly rainfall estimates in 54 cells for June
and July, 1997; ?1to24? indicates that this file was
used in simulation runs using 1:24 k scale input data
ascii 06to07971to24mod10.csv StageIII hourly rainfall estimates in 54 cells for June
and July, 1997; ?1to24? indicates that this file was
used in simulation runs using 1:24 k scale input data;
?mod10? indicates that the data for the July 10
th
 storm
has been adjusted by a factor of 0.64 to reflect
Micronet observations (see Section 5.3)
ascii 06to0797mod10.csv StageIII hourly rainfall estimates in 55 cells for June
and July, 1997; ?mod10? indicates that the data for
the July 10
th
 storm has been adjusted by a factor of
0.64 to reflect Micronet observations (see Section 5.3)
ascii 06to0797prec.csv StageIII hourly rainfall estimates in 55 cells for June
and July, 1997
ascii exnexrad.bat UNIX batch file used to process NEXRAD StageIII
data described in Appendix B.3.
ascii jul10storm.txt hourly StageIII data for July 10, 1997 storm
transposed so that rows are spatial units and columns
are hours; used to create Figure 5.1
shape lwxgeoccprj.shp,.dbf,.shx shapefile of NEXRAD cells before intersection with
the Little Washita watershed (shown in Figure 3.18)
ncdump executable program compiled on a Sun Workstation
used to convert netCDF files to ASCII files (see
Section B.1 for further description)
ascii rdump.c C source code for program used to automate the
conversion of hundreds of netCDF files to ASCII files
(see Table B1)
224
ascii readvals2.f FORTRAN source code for program used to extract
hourly StageIII estimates for only cells within the
Little Washita watershed from larger files (see Table
B1)
ascii replace2.c C source code for a program that reformats StageIII
ASCII files created using ncdump (see Table B1)
Dir prj contains projection files
ascii demtook.prj projection file from decimal seconds to the projection
defined in Table 3.2
ascii geook.prj projection file from decimal degrees to the projection
defined in Table 3.2
ascii utmtook.prj projection file from UTM to the projection defined in
Table 3.2
ascii utmtook7.prj projection file from UTM (NAD27) to the projection
defined in Table 3.2
ascii natok.prj projection file from national Albers Equal-Area
parameters to the projection defined in Table 3.2
Dir readdata contains source code for time series data
processing programs
readdata.frm, .vbp, .vbw Visual Basic time series data processing programs
described in Appendix B
Dir referenc data used for spatial reference
cover hucsalb USGS 8 digit HUCs for the Washita River
cover okalb state of Oklahoma boundary
cover rf1 EPA River Reach File 1 for the Washita River
shape arm7nd8.shp, .dbf, .shx point shapefile of ARM stations EF7 and EF8
shape kansok.shp, .dbf, .shx shapefile of the Kansas and Oklahoma state
boundaries
shape lawhyd3p.shp, .dbf, .shx 1:100,000 scale hydrography DLG for the Lawton-3
quadrangle
shape lawhyd4p.shp, .dbf, .shx 1:100,000 scale hydrography DLG for the Lawton-4
quadrangle
shape lawt3.shp, .dbf, .shx 1:100,000 scale transportation DLG for the Lawton-3
quadrangle
shape lawt4.shp, .dbf, .shx 1:100,000 scale transportation DLG for the Lawton-4
quadrangle
shape okalb.shp, .dbf, .shx shapefile of Oklahoma boundary
shape okcitalb.shp, .dbf, .shx shapefile of cities in Oklahoma
shape okcntalb.shp, .dbf, .shx shapefile of counties in Oklahoma
shape paulval1.shp, .dbf, .shx 1:100,000 scale transportation DLG for the Paul
Valley-1 quadrangle
shape pvhyd1p.shp, .dbf, .shx 1:100,000 scale hydrography DLG for the Paul
Valley-1 quadrangle
Dir sgp97_grav_soil_data contains observed gravimetric soil-water content
data for the Little Washita from SGP97
ascii bulkdens.txt bulk density data for SGP sampling sites
cellscompare.txt SGP sites and cells where a comparison between
modeled and observed soil-water content will be made
figure11.gif scanned map of Little Washita area
figure11.tif scanned map of Little Washita area
225
ascii lwsmweb.txt gravimetric soil moisture data downloaded from the
web (Little Washita data only)
(daac.gsfc.nasa.gov/campaign_docs/sgp97/gravi_mois
t.html)
shape smsites.shp, .dbf, .shx approximate location of soil moisture measurement
sites with points located visually
ascii smweb.txt gravimetric soil moisture data downloaded from the
web (all sites)
(daac.gsfc.nasa.gov/campaign_docs/sgp97/gravi_mois
t.html)
Dir single_layer contains program files and data files for
singl_layer simulation runs
ascii cellscompare.txt SGP sites and cells where a comparison between
modeled and observed soil-water content will be made
ascii senssminit.txt soil-water content initialization file for soil moisture
sensitivity runs
ascii sminit.csv soil-water content initialization file used with 1:250K
runs
ascii sminit1to24.csv soil-water content initialization file used with 1:24K
runs
swbms1lay3.frm, .log, .vbp,
.vbw
Visual Basic code for 1-layer model
ascii swbms1to24cc.csv cell parameter file used with 1:24K scale input data
ascii swbms1to250.txt cell parameter file used with 1:250K scale input data
ascii swbmssil.txt cell parameter file used when the watershed is
assumed to be entirely silt loam (watershed level
resampling with 1:24K soils data)
ascii swbmssll.csv cell parameter file used when the watershed is
assumed to be entirely sandy loam (watershed level
resampling with 1:250K soils data)
Dir cal1to24 VB-SWBMS calibration files using 1:24K input
data
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii flowm3.txt flowfile.txt converted to cms
excel flownoadj.xls spreadsheet used to plot flow results
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
excel soil_moist0.xls spreadsheet used to compare observed and modeled
soil moisture
ascii summary.txt VB-SWBMS output summary file
Dir cal1to24adj VB-SWBMS calibration files using 1:24K input
data and adjusted rainfall for July 10 storm
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii flowm3.txt flowfile.txt converted to cms
excel flowc.xls spreadsheet used to plot flow results
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
excel soil_moist.xls spreadsheet used to compare observed and modeled
226
soil moisture
ascii summary.txt VB-SWBMS output summary file
Dir cal1to250 VB-SWBMS calibration files using 1:250K input
data
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir cal1to250adj VB-SWBMS calibration files using 1:250K input
data and adjusted rainfall for July 10 storm
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir calsiladj VB-SWBMS calibration files using watershed level
resampling, 1:24K input data, and adjusted
rainfall for July 10 storm
ascii 0805ctl.txt control file
excel flow.xls spreadsheet used to plot streamflows
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii flowm3.txt flowfile.txt converted to cms
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir calsladj VB-SWBMS calibration files using watershed level
resampling, 1:250K input data, and adjusted
rainfall for July 10 storm
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir senshb VB-SWBMS files used for sensitivity analysis with
h
b
 described in Section 5.6.
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii p_hbhigh.csv parameter lookup table with high h
b
 values
ascii p_hblow.csv parameter lookupt table with low h
b
 values
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir senslam VB-SWBMS files used for sensitivity analysis with
? described in Section 5.6.
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii p_lamhigh.csv parameter lookup table with high ? values
ascii p_lamlow.csv parameter lookupt table with low ? values
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
227
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir senspor VB-SWBMS files used for sensitivity analysis with
porosity described in Section 5.6.
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii p_porhigh.csv parameter lookup table with high porosity values
ascii p_porlow.csv parameter lookupt table with low porosity values
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir sensrain VB-SWBMS files used for sensitivity analysis with
rain described in Section 5.6.
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir senssm VB-SWBMS files used for sensitivity analysis with
soil moisture described in Section 5.6.
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir sensthr VB-SWBMS files used for sensitivity analysis with
residual water content described in Section 5.6.
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii p_thrhigh.csv parameter lookup table with high porosity values
ascii p_thrlow.csv parameter lookupt table with low porosity values
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir val0507sl VB-SWBMS files used for second validation run
with watershed level resampling (assuming the
entire watershed is sandy loam)
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii flowm3sl.txt flowfile.txt converted to cms
excel flowv1.xls spreadsheet used to plot flows
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir val1to250 VB-SWBMS validation files using 1:250K input
data
ascii 0805ctl.txt control file
excel flowv1.xls spreadsheet used to plot flows
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
228
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir val1to250adj VB-SWBMS validation files using 1:250K input
data
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir val1to24 VB-SWBMS validation files using 1:24K input
data
ascii 0805ctl.txt control file
excel flow0.xls spreadsheet used to plot flows
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii flowm3.txt flowfile.txt converted to cms
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir val1to240507 VB-SWBMS files for the second validation period
using 1:24K input data
ascii 0805ctl.txt control file
excel flowv1.xls spreadsheet used to plot flows
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii flowm3.txt flowfile.txt converted to cms
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir val1to24adj VB-SWBMS validation files using 1:24K input
data
ascii 0805ctl.txt control file
excel flowfig.xls spreadsheet used to plot flows
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii flowm3.txt flowfile.txt converted to cms
excel flowv1.xls spreadsheet used to plot flows
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir valsiladj VB-SWBMS validation files using watershed level
resampling and 1:24K input data
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii flowm3.txt flowfile.txt converted to cms
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir valsladj VB-SWBMS calibration files using watershed level
resampling and 1:250K input data
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
229
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir single_layer_base VB-SWBMS files used for one-layer simulations
with baseflow
swbms1base.vbw, .vbp, .frm Visual Basic code for 1-layer model which uses
baseflow simulation
Dir cal1to24adj VB-SWBMS calibration files using 1:24K input data
and adjusted rainfall for July 10 storm
ascii 0805ctl.txt control file
excel flowbig_cms.xls spreadsheet plotting flows for example with a big
baseflow storage constant (see discussion in Section
5.4)
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii flowm3_small.txt VB-SWBMS output flow file converted to cms for the
case of a small baseflow storage constant
ascii flowm3big.txt VB-SWBMS model flow output converted to cms for
case with big baseflow storage constant
excel flowsmall_cms.xls spreadsheet plotting flows for example with a small
baseflow storage constant (see discussion in Section
5.4)
ascii smdailavg.txt VB-SWBMS daily average soil moisture output file
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir single_layer_pred VB-SWBMS files used for one-layer simulations
and the predominant soil type option (looping in
the code is slightly different for this case)
swbms1lay4, .frm, .frx,
.vbp, .vbw
Visual Basic code for 1 layer model and the
predominant soil type option
ascii swbms1to24pred.csv cell parameter file for VB-SWBMS for cases using
1:24K input data
ascii swbms1to250pred.txt cell parameter file for VB-SWBMS for cases using
1:250K input data
ascii wcinit1to24pred.csv soil-water content initialization file for the case of
1:24K input data and NEXRAD cell level resampling
ascii wcinit1to250pred.csv soil-water content initialization file for the case of
1:250K input data and NEXRAD cell level
resampling
Dir cal1to24adj VB-SWBMS calibration files using 1:24K input
data and adjusted rainfall for July 10 storm
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smf.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir cal1to250adj VB-SWBMS calibration files using 1:250K input
data and adjusted rainfall for July 10 storm
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smf.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir val1to24adj VB-SWBMS validation files using 1:24K input
230
data
ascii 0805ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smfinal.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir val1to250adj VB-SWBMS validation files using 1:250K input
data
ascii 0824ctl.txt control file
ascii flowfile.txt VB-SWBMS output flow file (cfs)
ascii smf.txt VB-SWBMS final soil moisture file
ascii summary.txt VB-SWBMS output summary file
Dir soils contains soil maps and associated data tables
30mcprop2.dbf summary of component properties for 30-m soils grid
created using AV-SWBMS
compjoin.dbf summary of component properties for Oklahoma
STATSGO data created using AV-
SWBMS
ascii lookup.txt lookup table to translate STATSGO texture names
into the 12 basic USDA texture classes, created at
Penn State and slightly modified here (see discussion
in Section 3.2.2.2.2)
lwcommod.dbf modified version of lwcomp.dbf that contains only
one entry for each ?Gisclass? (a.k.a. grid-code)
lwcomp.dbf component attribute table associated with 30-m soils
grid
lwlayer.dbf layer attribute table associated with 30-m soils grid
lws30utm.bil, lws30utm.hdr original spatial data for 30-m soils grid obtained from
the ARS Grazinglands Research Laboratory in El
Reno, OK (in the UTM coordinate system)
lwsoils.wk4 original tabular data for 30-m soils grid obtained from
the ARS Grazinglands Research Laboratory in El
Reno, OK (in Lotus 1-2-3 format)
ascii statusda.txt lookup table (in ArcView readable, comma delimited,
format) used to translate STATSGO texture names
into USDA texture class names
ascii usdapar.txt parameter value lookup table based on Carsel and
Parrish (1988) paper
ascii usdapar_r.csv parameter value lookup table based on Rawls,
Brakensiek, and Saxton (1982) paper
Dir info contains Arc/Info ?info? files (attribute tables)
grid lws30alb 30-m soils grid in Arc/Info format and Albers
projection (see Table 3.2)
Dir statsgo contains the STATSGO soil map and associated
attribute files for the state of Oklahoma (organized
into three directories that are created when the data is
downloaded from the NRCS web site)
Dir header contains text files describing the mapextent and
projection of the soils data
Dir metadata contains a text file with metadata (data about data)
Dir spatial contains 1:250K STATSGO soil maps and attribute
231
tables
cover Ok STATSGO map in original national Albers Equal-area
map projection
cover oknat STATSGO map in orignal national Albers Equal-Area
projection with arc topology only
cover okok STATSGO map projected into an Albers Equal-Area
projection with parameters used for making state-wide
maps of Oklahoma
comp.dbf STATSGO component attribute table
compyld.dbf STATSGO table that stores crop yield information
forest.dbf STATSGO table that stores information for plant
cover
interp.dbf STATSGO table
layer.dbf STATSGO layer table
mapunit.dbf STATSGO mapunit table
shape oknat.shp, .dbf, .shx shapefile with same data as the coverage ?Ok?
plantcom.dbf STATSGO table that stores plant symbols and percent
of plant composition
plantnm STATSGO table with the common and scientific
names for plants
rsprod.dbf STATSGO table that stores range site productivity
information for soils
taxclass.dbf STATSGO table that stores taxonomic classification
for soil components
windbrk.dbf STATSGO table
wlhabit.dbf STATSGO table
woodland.dbf STATSGO table
woodmgt.dbf STATSGO table
yldunits.dbf STATSGO table
Dir streamflow_data contains hourly streamflow data for the gaging
station (Ninnekah) shown in Figure 3.2
Dir info contains Arc/Info attribute tables
cover usgssel point coverage of stations 07327447 and 07327550;
attributes from the file usgssel.txt can be joined to this
coverage
ascii 0507to052497sf.txt observed stramflow and separated baseflow for
5/7/97-5/24/97; units are cfs, baseflow separation
parameter = 60 hours
ascii 0530to0731b.txt observed stramflow and separated baseflow for
5/30/97-7/31/97; units are cfs, baseflow separation
parameter = 60 hours
ascii 0601to0701.txt observed stramflow and separated baseflow for
6/01/97-7/01/97; units are cfs, baseflow separation
parameter = 60 hours
ascii 0601to0801.txt observed stramflow and separated baseflow for
6/01/97-8/01/97; units are cfs, baseflow separation
parameter = 60 hours
ascii 601to18base.txt observed stramflow and separated baseflow for
6/01/97-6/18/97; units are cfs, baseflow separation
parameter = 60 hours
232
ascii 709to718b15.txt observed stramflow and separated baseflow for
7/09/97-7/18/97; units are cfs, baseflow separation
parameter = 15 hours
ascii 709to718b30.txt observed stramflow and separated baseflow for
7/09/97-7/18/97; units are cfs, baseflow separation
parameter = 30 hours
ascii 709to718b60.txt observed stramflow and separated baseflow for
7/09/97-7/18/97; units are cfs, baseflow separation
parameter = 60 hours
ascii bfj18tojl20.txt observed stramflow and separated baseflow for
6/18/97-7/30/97; units are cfs, baseflow separation
parameter = 60 hours
shape chick_sf.shp, .dbf, .shx locations of ARS streamflow stations in lat/lon
ascii ninnekah 15 minute flows for the USGS gaging station East of
Ninnekah (07327550) ; 10/1/95 ? 10/29/97
excel ninnflow2_s.xls hourly flows for the USGS gaging station East of
Ninnekah (07327550) ; 10/1/95 ? 10/29/97 (cfs);
computed from ?ninnekah?; also includes separated
baseflow with n = 60 hours
ascii ninnflow2_s.txt hourly flows for the USGS gaging station East of
Ninnekah (07327550) ; 10/1/95 ? 10/29/97 (cfs);
computed from ?ninnekah?
excel ninnflowjj97.xls hourly flows for the USGS gaging station East of
Ninnekah (07327550) ; June and July 1997 (cfs)
shape on1sec.shp, .dbf, .shx USGS gaging station locations shifted onto streams
defined by the 30 m DEM
shape statalb.shp, .dbf, .shx locations of ARS streamflow stations projected into
an Albers Equal-area projection with parameters
defined in Table 3.2
ascii usgs.txt attribute information for USGS gaging stations
ascii usgssel.txt attribute information that can be joined to the stations
in the coverage usgssel
ascii whatidid.txt text file containing information about how the point
coverage usgssel was created
233
Appendix D:  Table used to reclassify 719 STATSGO texture names
into 12 basic USDA classes and the "other" category.
STATSGO_name,USDA_t
exture_class
BY-C,C
BY-CL,CL
BY-COS,S
BY-COSL,SL
BY-FS,S
BY-FSL,SL
BY-L,L
BY-LCOS,LS
BY-LS,LS
BY-LVFS,LS
BY-S,S
BY-SC,SC
BY-SCL,SCL
BY-SI,SI
BY-SIC,SIC
BY-SICL,SICL
BY-SIL,SIL
BY-SL,SL
BY-VFS,S
BY-VFSL,SL
BYV-C,C
BYV-CL,CL
BYV-COS,S
BYV-COSL,SL
BYV-FS,S
BYV-FSL,SL
BYV-L,L
BYV-LCOS,LS
BYV-LFS,LS
BYV-LS,LS
BYV-LVFS,LS
BYV-S,S
BYV-SC,SC
BYV-SCL,SCL
BYV-SI,SI
BYV-SIC,SIC
BYV-SICL,SICL
BYV-SIL,SIL
BYV-SL,SL
BYV-VFS,S
BYV-VFSL,SL
BYV-SL,SL
BYX-C,C
BYX-CL,CL
BYX-COS,S
BYX-COSL,SL
BYX-FS,S
BYX-FSL,SL
BYX-L,L
BYX-LCOS,LS
BYX-LFS,LS
BYX-LS,LS
BYX-LVFS,LS
BYX-S,S
BYX-SC,SC
BYX-SCL,SCL
BYX-SI,SI
BYX-SIC,SIC
BYX-SICL,SICL
BYX-SIL,SIL
BYX-SL,SL
BYX-VFS,S
BYX-VFSL,SL
C,C
CB-C,C
CB-CL,CL
CB-COS,S
CB-COSL,SL
CB-FS,S
CB-FSL,SL
CB-L,L
CB-LCOS,LS
CB-LFS,LS
CB-LS,LS
CB-LVFS,LS
CB-S,S
CB-SC,SC
CB-SCL,SCL
CB-SI,SI
CB-SIC,SIC
CB-SICL,SICL
CB-SIL,SIL
CB-SL,SL
CB-VFS,S
CB-VFSL,SL
CBA-C,C
CBA-CL,CL
CBA-COS,S
CBA-COSL,SL
CBA-FS,S
CBA-FSL,SL
CBA-L,L
CBA-LCOS,LS
CBA-LS,LS
CBA-LVFS,LS
CBA-S,S
CBA-SC,SC
CBA-SCL,SCL
CBA-SI,SI
CBA-SIC,SIC
CBA-SICL,SICL
CBA-SIL,SIL
CBA-SL,SL
CBA-VFS,S
CBA-VFSL,SL
CBV-C,C
CBV-CIND,O
CBV-CL,CL
CBV-COS,S
CBV-COSL,SL
CBV-FS,S
CBV-FSL,SL
CBV-L,L
CBV-LCOS,S
CBV-LFS,LS
CBV-LS,LS
CBV-MUCK,OM
CBV-S,S
CBV-SC,SC
CBV-SCL,SCL
CBV-SIC,SIC
CBV-SICL,SICL
CBV-SIL,SIL
CBV-SL,SL
CBV-VFSL,SL
CBX-C,C
CBX-CL,CL
CBX-COS,S
CBX-COSL,SL
CBX-FS,S
CBX-FSL,SL
CBX-L,L
CBX-LCOS,LS
CBX-LFS,LS
CBX-LS,LS
234
CBX-LVFS,LS
CBX-S,S
CBX-SC,SC
CBX-SCL,SCL
CBX-SI,SI
CBX-SIC,SIC
CBX-SICL,SICL
CBX-SIL,SIL
CBX-SL,SL
CBX-VFS,S
CBX-VFSL,SL
CE,O
CEM,O
CIND,O
CL,CL
CN-C,C
CN-CL,CL
CN-COS,S
CN-COSL,SL
CN-FS,S
CN-FSL,SL
CN-L,L
CN-LCOS,LS
CN-LFS,LS
CN-LS,LS
CN-LVFS,LS
CN-S,S
CN-SC,SC
CN-SCL,SCL
CN-SI,SI
CN-SIC,SIC
CN-SICL,SICL
CN-SIL,SIL
CN-SL,SL
CN-VFS,S
CN-VFSL,SL
CNV-C,C
CNV-CL,CL
CNV-COS,S
CNV-COSL,SL
CNV-FS,S
CNV-FSL,SL
CNV-L,L
CNV-LCOS,LS
CNV-LFS,LS
CNV-LS,LS
CNV-LVFS,LS
CNV-S,S
CNV-SC,SC
CNV-SCL,SCL
CNV-SI,SI
CNV-SIC,SIC
CNV-SICL,SICL
CNV-SIL,SIL
CNV-SL,SL
CNV-VFS,S
CNV-VFSL,SL
CNX-C,C
CNX-CL,CL
CNX-COS,S
CNX-COSL,SL
CNX-FS,S
CNX-FSL,SL
CNX-L,L
CNX-LCOS,LS
CNX-LS,LS
CNX-LVFS,LS
CNX-S,S
CNX-SC,SC
CNX-SCL,SCL
CNX-SI,SI
CNX-SIC,SIC
CNX-SICL,SICL
CNX-SIL,SIL
CNX-SL,SL
CNX-VFS,S
CNX-VFSL,SL
COS,S
COSL,SL
CR-C,C
CR-CL,CL
CR-COS,S
CR-COSL,SL
CR-FS,S
CR-FSL,SL
CR-L,L
CR-LCOS,LS
CR-LS,LS
CR-LVFS,LS
CR-S,S
CR-SC,SC
CR-SCL,SCL
CR-SI,SI
CR-SIC,SIC
CR-SICL,SICL
CR-SIL,SIL
CR-SL,SL
CR-VFS,S
CR-VFSL,SL
CRC-C,C
CRC-CL,CL
CRC-COS,S
CRC-COSL,SL
CRC-FS,S
CRC-FSL,SL
CRC-L,L
CRC-LCOS,LS
CRC-LS,LS
CRC-LVFS,LS
CRC-S,S
CRC-SC,SC
CRC-SCL,SCL
CRC-SI,SI
CRC-SIC,SIC
CRC-SICL,SICL
CRC-SIL,SIL
CRC-SL,SL
CRC-VFS,S
CRC-VFSL,SL
CRV-C,C
CRV-CL,CL
CRV-COS,S
CRV-COSL,SL
CRV-FS,S
CRV-FSL,SL
CRV-L,L
CRV-LCOS,LS
CRV-LS,LS
CRV-LVFS,LS
CRV-S,S
CRV-SC,SC
CRV-SCL,SCL
CRV-SI,SI
CRV-SIC,SIC
CRV-SICL,SICL
CRV-SIL,SIL
CRV-SL,SL
CRV-VFS,S
CRV-VFSL,SL
CRX-C,C
CRX-CL,CL
CRX-COS,S
CRX-COSL,SL
CRX-FS,S
CRX-FSL,SL
CRX-L,L
CRX-LCOS,LS
CRX-LS,LS
CRX-LVFS,LS
CRX-S,S
CRX-SC,SC
CRX-SCL,SCL
CRX-SI,SI
235
CRX-SIC,SIC
CRX-SICL,SICL
CRX-SIL,SIL
CRX-SL,SL
CRX-VFS,S
CRX-VFSL,SL
DE,O
FB,OM
FL-C,C
FL-CL,CL
FL-COS,S
FL-COSL,SL
FL-FS,S
FL-FSL,SL
FL-L,L
FL-LCOS,LS
FL-LS,LS
FL-LVFS,LS
FL-S,S
FL-SC,SC
FL-SCL,SCL
FL-SI,SI
FL-SIC,SIC
FL-SICL,SICL
FL-SIL,SIL
FL-SL,SL
FL-VFS,S
FL-VFSL,SL
FLV-C,C
FLV-CL,CL
FLV-COS,S
FLV-COSL,SL
FLV-FS,S
FLV-FSL,SL
FLV-L,L
FLV-LCOS,LS
FLV-LS,LS
FLV-LVFS,LS
FLV-S,S
FLV-SC,SC
FLV-SCL,SCL
FLV-SI,SI
FLV-SIC,SIC
FLV-SICL,SICL
FLV-SIL,SIL
FLV-SL,SL
FLV-VFS,S
FLV-VFSL,SL
FLX-C,C
FLX-CL,CL
FLX-COS,S
FLX-COSL,SL
FLX-FS,S
FLX-FSL,SL
FLX-L,L
FLX-LCOS,LS
FLX-LS,LS
FLX-LVFS,LS
FLX-S,S
FLX-SC,SC
FLX-SCL,SCL
FLX-SI,SI
FLX-SIC,SIC
FLX-SICL,SICL
FLX-SIL,SIL
FLX-SL,SL
FLX-VFS,S
FLX-VFSL,SL
FRAG,O
FS,S
FSL,SL
G,O
GR,O
GR-C,C
GR-CL,CL
GR-COS,S
GR-COSL,SL
GR-FS,S
GR-FSL,SL
GR-L,L
GR-LCOS,LS
GR-LFS,LS
GR-LS,LS
GR-LVFS,LS
GR-MARL,O
GR-MUCK,OM
GR-S,S
GR-SC,SC
GR-SCL,SCL
GR-SI,SI
GR-SIC,SIC
GR-SICL,SICL
GR-SIL,SIL
GR-SL,SL
GR-VAR,O
GR-VFS,S
GR-VFSL,SL
GRC-C,C
GRC-CL,CL
GRC-COS,S
GRC-COSL,SL
GRC-FS,S
GRC-L,L
GRC-LCOS,LS
GRC-LS,LS
GRC-LVFS,LS
GRC-S,S
GRC-SC,SC
GRC-SCL,SCL
GRC-SI,SI
GRC-SIC,SIC
GRC-SICL,SICL
GRC-SIL,SIL
GRC-SL,SL
GRC-VFS,S
GRC-VFSL,SL
GRF-C,C
GRF-CL,CL
GRF-COS,S
GRF-COSL,SL
GRF-FS,S
GRF-FSL,SL
GRF-L,L
GRF-LCOS,LS
GRF-LS,LS
GRF-LVFS,LS
GRF-S,S
GRF-SC,SC
GRF-SCL,SCL
GRF-SI,SI
GRF-SIC,SIC
GRF-SICL,SICL
GRF-SIL,SIL
GRF-SL,SL
GRF-VFS,S
GRF-VFSL,SL
GRV-C,C
GRV-CL,CL
GRV-COS,S
GRV-COSL,SL
GRV-FS,S
GRV-FSL,SL
GRV-L,L
GRV-LCOS,LS
GRV-LFS,LS
GRV-LS,LS
GRV-LVFS,LS
GRV-S,S
GRV-SC,SC
GRV-SCL,SCL
GRV-SI,SI
GRV-SIC,SIC
GRV-SICL,SICL
236
GRV-SIL,SIL
GRV-SL,SL
GRV-VFS,S
GRV-VFSL,SL
GRX-C,C
GRX-CL,CL
GRX-COS,S
GRX-COSL,SL
GRX-FRAG,O
GRX-FS,S
GRX-FSL,SL
GRX-L,L
GRX-LCOS,LS
GRX-LFS,LS
GRX-LS,LS
GRX-LVFS,LS
GRX-S,S
GRX-SC,SC
GRX-SCL,SCL
GRX-SI,SI
GRX-SIC,SIC
GRX-SICL,SICL
GRX-SIL,SIL
GRX-SL,SL
GRX-VFS,S
GRX-VFSL,SL
GYP,O
HM,OM
ICE,O
IND,O
L,L
LCOS,LS
LCOS,LS
LFS,LS
LS,LS
LVFS,LS
MARL,O
MI-SIL,SIL
MK-C,C
MK-CL,CL
MK-COS,S
MK-COSL,SL
MK-FS,S
MK-FSL,SL
MK-L,L
MK-LCOS,LS
MK-LFS,LS
MK-LS,LS
MK-LVFS,LS
MK-MARL,O
MK-PEAT,OM
MK-S,S
MK-SC,SC
MK-SCL,SCL
MK-SI,SI
MK-SIC,SIC
MK-SICL,SICL
MK-SIL,SIL
MK-SL,SL
MK-VFS,S
MK-VFSL,SL
MPT,OM
MUCK,OM
NONE,O
PEAT,OM
PT-SIC,SIC
PT-SIL,SIL
RB-C,C
RB-CL,CL
RB-COS,S
RB-COSL,SL
RB-FS,S
RB-FSL,SL
RB-L,L
RB-LCOS,LS
RB-LS,LS
RB-LVFS,LS
RB-S,S
RB-SC,SC
RB-SCL,SCL
RB-SI,SI
RB-SIC,SIC
RB-SICL,SICL
RB-SIL,SIL
RB-SL,SL
RB-VFS,S
RB-VFSL,SL
S,S
SC,SC
SCL,SCL
SG,O
SH-C,C
SH-CL,CL
SH-COS,S
SH-COSL,SL
SH-FS,S
SH-FSL,SL
SH-L,L
SH-LCOS,LS
SH-LS,LS
SH-LVFS,LS
SH-S,S
SH-SC,SC
SH-SCL,SCL
SH-SI,SI
SH-SIC,SIC
SH-SICL,SICL
SH-SIL,SIL
SH-SL,SL
SH-VFS,S
SH-VFSL,SL
SHV-C,C
SHV-CL,CL
SHV-COS,S
SHV-COSL,SL
SHV-FS,S
SHV-FSL,SL
SHV-L,L
SHV-LCOS,LS
SHV-LS,LS
SHV-LVFS,LS
SHV-S,S
SHV-SC,SC
SHV-SCL,SCL
SHV-SI,SI
SHV-SIC,SIC
SHV-SICL,SICL
SHV-SIL,SIL
SHV-SL,SL
SHV-VFS,S
SHV-VFSL,SL
SHX-CL,CL
SHX-L,L
SI,SI
SIC,SIC
SICL,SICL
SIL,SIL
SL,SL
SP,OM
SR,O
SR-,O
SR-SIL,SIL
ST-SIL,SIL
ST-C,C
ST-CL,CL
ST-COS,S
ST-COSL,SL
ST-FS,S
ST-FSL,SL
ST-L,L
ST-LCOS,LS
ST-LFS,LS
ST-LS,LS
237
ST-LVFS,LS
ST-MUCK,OM
ST-S,S
ST-SC,SC
ST-SCL,SCL
ST-SI,SI
ST-SIC,SIC
ST-SICL,SICL
ST-SIL,SIL
ST-SIL-,SIL
ST-SL,SL
ST-VFS,S
ST-VFSL,SL
STV-C,C
STV-CL,CL
STV-COS,S
STV-COSL,SL
STV-FS,S
STV-FSL,SL
STV-L,L
STV-LCOS,LS
STV-LFS,LS
STV-LS,LS
STV-LVFS,LS
STV-MPT,OM
STV-MUCK,OM
STV-S,S
STV-SC,SC
STV-SCL,SCL
STV-SI,SI
STV-SIC,SIC
STV-SICL,SICL
STV-SIL,SIL
STV-SL,SL
STV-VFS,S
STV-VFSL,SL
STX-C,C
STX-CL,CL
STX-COS,S
STX-COSL,SL
STX-FS,S
STX-FSL,SL
STX-L,L
STX-LCOS,LS
STX-LFS,LS
STX-LS,LS
STX-LVFS,LS
STX-MUCK,OM
STX-PEAT,OM
STX-S,S
STX-SC,SC
STX-SCL,SCL
STX-SI,SI
STX-SIC,SIC
STX-SICL,SICL
STX-SIL,SIL
STX-SL,SL
STX-VFS,S
STX-VFSL,SL
SY-C,C
SY-CL,CL
SY-COS,S
SY-COSL,SL
SY-FS,S
SY-FSL,SL
SY-L,L
SY-LCOS,LS
SY-LS,LS
SY-LVFS,LS
SY-S,S
SY-SC,SC
SY-SCL,SCL
SY-SI,SI
SY-SIC,SIC
SY-SICL,SICL
SY-SIL,SIL
SY-SL,SL
SY-VFS,S
SY-VFSL,SL
SYV-C,C
SYV-CL,CL
SYV-COS,S
SYV-COSL,SL
SYV-FS,S
SYV-FSL,SL
SYV-L,L
SYV-LCOS,LS
SYV-LS,LS
SYV-LVFS,LS
SYV-S,S
SYV-SC,SC
SYV-SCL,SCL
SYV-SI,SI
SYV-SIC,SIC
SYV-SICL,SICL
SYV-SIL,SIL
SYV-SL,SL
SYV-VFS,S
SYV-VFSL,SL
SYX-C,C
SYX-CL,CL
SYX-COS,S
SYX-COSL,SL
SYX-FS,S
SYX-FSL,SL
SYX-L,L
SYX-LCOS,LS
SYX-LS,LS
SYX-LVFS,LS
SYX-S,S
SYX-SC,SC
SYX-SCL,SCL
SYX-SI,SI
SYX-SIC,SIC
SYX-SICL,SICL
SYX-SIL,SIL
SYX-SL,SL
SYX-VFS,S
SYX-VFSL,SL
UWB,O
VAR,O
VFS,S
VFSL,SL
WB,O
_
238
REFERENCES
Allen, P.B., and J.W. Naney, ?Hydrology of the Little Washita River Watershed,
Oklahoma: Data and Analyses,? U.S. Department of Agriculture, Agricultural
Research Service, ARS-90, 74 pp., 1991.
ASCE, Evapotranspiration and Irrigation Water Requirements, ASCE Manuals and
Reports on Engineering Practice No. 70, Jensen, M..E., R.D. Burman, and R.G.
Allen, eds., ASCE, 1990.
Anderson, M.G., and T.P. Burt, ?Automatic Monitoring of Soil Moisture Conditions
in a Hillslope Spur and Hollow,? Journal of Hydrology, 33, 27-36, 1977.
Anderson, J.R., E.E. Hardy, J.T. Roach, and R.E. Witmer, 1976, ?A Land Use and
Land Cover Classification System for Use with Remote Sensor Data,? U.S.
Geological Survey, Professional Paper 964, Reston, VA.
Atmospheric Radiation Measurement Program (ARM), www.arm.gov, 1998.
Bennett, T., ?Development of the Continuous Soil Moisture Accounting Model for
HEC-HMS ? DRAFT,? unpublished material obtained through personal
communication, January, 1998.
Betson, R.P., ?What is watershed runoff?,?  Journal of Geophysical Research, 69, 8,
1541-1552, 1964.
Beven, K.J, and Kirkby, M.J., "A Physically-based Variable Contributing Area
Model of Basin Hydrology," Hydrol. Sci. Bull., 24, 1, pp. 43-69, 1979.
Beven, K.J., "Runoff Production and Flood Frequency in Catchments of Order n:
An Alternative Approach," Scale Problems in Hydrology, Gupta, V.K., I.
Rodriquez-Iturbe, and E.F. Wood, eds., D. Reidel Publishing, Dordrecht, Holland,
1986.
Bouwer, H., "Rapid Field Measurement of Air Entry Value and Hydraulic
Conductivity of Soil as Significant Parameters in Flow System Analysis,"  Water
Resources Research, 2, 4, pp. 729-738, 1966.
Bouwer, H., ?Infiltration of Water into a Non-uniform Soil,? Journal of the
Irrigation and Drainage Division, ASCE, 451-462, December, 1969.
239
Brutsaert, W., Evaporation into the Atmosphere: Theory, History, and Applications,
D. Reidel Publishing Company, Dordrecht, Holland, 1982.
Burnash, R.J., R.L. Ferral, and R.A., McGuire, ?A Generalized Streamflow
Simulation System Conceptual Modeling for Digital Computers,? U.S. National
Weather Service and State of California Department of Water Resources, March,
1973.
Burt, T.P., and D.P. Butcher, "Topographic Controls of Soil Moisture
Distributions," Journal of Soil Science, 36, pp. 469-486, 1985.
Canadell, J., R.B. Jackson, J. Ehleringer, H.A. Mooney, O.E. Sala, and E.D.
Schulze, "Maximum rooting depth of vegetation types at the global scale,"
Oecologia, 108, 583-595, 1996.
Carsel, R.F., and R.S. Parrish, ?Developing Joint Probability Distributions of Soil
Water Retention Characteristics,? Water Resources Research, 24, 5, 755-769, 1988.
CRWR-ESRI, GISHydro98 CD-ROM,
www.crwr.utexas.edu/gis/gishyd98/GisHyd98.htm, 1998.
Charbeneau, R.J., "Liquid Moisture Redistribution:  Hydrologic Simulation and
Spatial Variability," Unsaturated Flow in Hydrologic Modeling Theory and
Practice, H.J. Morel-Seytoux (ed.), Kluwer Academic Publishers, 127-160, 1989.
Chow, V.T., D.R. Maidment, and L.W. Mays, Applied Hydrology, McGraw-Hill,
Inc., New York, 1988.
Clapp, R.B., and G.M. Hornberger, ?Empirical Equations for Some Soil Hydraulic
Properties,? Water Resources Research, v. 14, no. 4, August, 1978.
Coles, N.A., M. Sivapalan, J.E. Larsen, P.E. Linnet, and C.K Fahrner, ?Modeling
Runoff Generation on Small Agricultural Catchments:  Can Real World Runoff
Responses be Captured?? Hydrologic Processes, 11, 111-136, 1997.
Dawdy, D.R., R.W. Lichty, and J.M. Bergmann, "A Rainfall-Runoff Simulation
Model for Estimation of Flood Peaks for Small Drainage Basins," Professional
Paper 506B, U.S. Geological Survey, pp. B1-B28, 1972.
DeVries, J.J., and T.V. Hromadka, ?Computer Models for Surface Water,?
Handbook  of Hydrology, Ch. 21, ed. D.R. Maidment, McGraw-Hill, Inc., 1993.
240
Dickinson, R.E., ? Modeling Evapotranspiration for Three-dimensional Global
Climate Models,? Climate Processes and Climate Sensitivity, Geophys. Monogr.
Ser., J.E. Hansen and T. Takahashi, eds., 29, 58-72, AGU, Washington, D.C., 1984.
Dingman, S.L., Physical Hydrology, Prentice Hall, Englewood Cliffs, NJ, 1994.
Doorenbos and Pruitt, ?Guidelines for Predicting Crop Water Requirements,? Food
and Agriculture Organization of the United Nations Irrigation and Drainage Paper
24, Rome, 1977.
Dunne, T., "Field studies of hillslope flow processes," Hillslope Hydrology, Chapter
7, M.J. Kirkby (ed.), 1978.
Dunne, T., T.R. Moore, and C.H. Taylor, "Recognition and Prediction of Runoff-
Producing Zones in Humid Regions," Hydrological Sciences Bulletin, 20, 305-327,
1975.
Dyck, S., ?Overview on the Present Status of the Concepts of Water Balance
Models,? IAHS Publ. 148, Wallingford, pp. 3-19, 1983.
Entekhabi, D., and P.S. Eagleson, "Land Surface Hydrology Parameterization for
Atmospheric General Circulation Models Including Subgrid Scale Spatial
Variability," Journal of Climate, 2, 816-831, 1989.
Famiglietti, J.S., and E.F. Wood, "Multiscale Modeling of Spatially Variable Water
and Energy Balance Processes," Water Resources Research, 30, 11, 3061-3078,
November 1994.
Finnerty, B., and D. Johnson, "Comparison of National Weather Service
Operational Mean Areal Precipitation Estimates Derived from NEXRAD Radar vs.
Rain Gage Networks," presented at International Association for Hydraulic
Research (IAHR), XXVII Congress, San Francisco, CA, August 10-15, 1997.
Franchini, M., and M. Pacciani, ?Comparative Analysis of Several Conceptual
Rainfall-Runoff Models,? Journal of Hydrology, v. 122, 161-219, 1991.
Fukami, K., E.T. Engman, P. Kumar, A. Terakawa, ?A Study on Basin-wide Soil-
layer Modeling in the Little Washita Basin in Oklahoma, USA,?  Proceedings of the
Consultative Meeting on Soil Moisture and Ocean Salinity Measurement
Requirements and Radiometer Techniques, ESA WPP-87, ESTEC, Noordwijk, The
Netherlands, 35-40, April 20-21, 1995.
241
Goldman, D.M., ?Loss Rate Representation in the HEC-1 Watershed Model,?
Unsaturated Flow in Hydrologic Modeling Theory and Practice, H.J. Morel-
Seytoux (ed.), Kluwer Academic Publishers, 1989.
Govindaraju, R.S., D. Or, M.L. Kavvas, D.E. Rolston, and J. Biggar, ?Error
Analyses of Simplified Unsaturated Flow Models Under Large Uncertainty in
Hydraulic Properties,? Water Resources Research, v. 28, no.11, 2913-2924,
November 1992.
Greene, D.R., and M.D. Hudlow, "Hydrometeorologic Grid Mapping Procedures,"
AWRA International Symposium on Hydrometeorology, Denver, CO, June 13-17,
1982 (DRAFT).
HEC, ?GridParm: Procedures for Deriving Grid Cell Parameters for the ModClark
Rainfall-Runoff Model, User?s Manual,? U.S. Army Corps of Engineers Hydrologic
Engineering Center, Davis, CA, CPR-71, November 1996.
HEC, "HEC-HMS Hydrologic Modeling System User's Manual," U.S. Army Corps
of Engineers Hydrologic Engineering Center, Davis, CA, 1997.
HEC, "modClark User?s Manual," U.S. Army Corps of Engineers Hydrologic
Engineering Center, Davis, CA, 1995.
Hellweger, Ferdinand, HEC-Prepro: A GIS Pre-processor for HMS, Master?s
Thesis, University of Texas at Austin, May, 1997.
Hewlett, J.D., and A.R. Hibbert, "Factors Affecting the Response of Small
Watersheds to Precipitation in Humid Areas," International Symposium on Forest
Hydrology, W.E. Supper and H.W. Lull, eds., Pergamon Press, Oxford, 275-290,
1967.
Hydrocomp Simulation Programming Operations Manual, Hydrocomp
International, Inc., Palo Alto, CA, 1970.
Jackson, T.J., J. Schmugge, and E.T. Engman, "Remote Sensing Applications to
Hydrology: soil moisture," Hydrological Sciences, 41, 4, 1996.
Jenson, S.K., Domingue, J.O. : Extracting topographic structure from digital
elevation data for geographic information system analysis.  Photogrammetric
Engineering and Remote Sensing 54, 1593-1600, 1988.
242
Jury, W.A., Gardner, W.R., and Gardner, W.H., ?Soil Physics : Fifth Edition,? John
Wiley and Sons, Inc., New York, 1991.
Kanemasu, E.T., S.B. Verma, E.A. Smith, L.J. Fritschen, M. Wesely, R.T. Field,
W.P. Kustas, H. Weaver, J.B. Stewart, R. Gurney, G. Panin, and J.B. Moncrieff,
?Surface Flux Measurements in FIFE: An Overview,? Journal of Geophysical
Research, v. 97, No. D17, 18547-18555, November 30, 1992.
Leavesley, G.H., R.W. Lichty, B.M. Troutman, and L.G. Saindon, "Precipitation-
runoff Modeling System:  User's Manual," Water-Resources Investigations Report
83-4238, USGS, 1983.
Maidment, D.R., and Mizgalewicz, P. : Digital delineation of watersheds and stream
networks in the Allegheny basin.  Report prepared for the Hydrologic Engineering
Center, U.S. Army Corps of Engineers, Davis, CA, 1993.
McFadden, D.E., "Soil Moisture Accounting in Continuous Simulation Watershed
Models,"  Master's Project Report, University of California at Davis, 1994.
McCuen, R.H., W.J. Rawls, and D.L. Brakensiek, "Statistical Analysis of the
Brooks-Corey and the Green-Ampt Parameters Across Soil Textures," Water
Resources Research, 17, 4, 1005-1013, 1981.
Mein, R.G., and C.L. Larson, "Modeling Infiltration During a Steady Rain," Water
Resources Research, 9, 2, 384-394, 1973.
Miller, D.A. and R.A. White, ?A Conterminous United States Multi-Layer Soil
Characteristics Data Set for Regional Climate and Hydrology Modeling,?  Earth
Interactions, 2, 1998. (More information at:
www.essc.psu.edu/soil_info/index.cgi?soil_data&index.html).
Monteith, J.L., "The Development and Extension of Penman's Evaporation
Formula," Applications of Soil Physics, Chapter 10, D. Hillel (ed.), Academic Press,
247-253, 1980.
Monteith, J.L., Szeicz, G., and Waggoner, P.E., "The Measurement and Control of
Stomatal Resistance in the Field," J. Appl. Ecol., 2, 345-355, 1965.
Moore, I.D., Grayson, R.B., Ladson, A.R.:  Digital terrain modelling : a review of
hydrological, geomorphological, and biological applications.  Hydrological
Processes 5, 3-30, 1991.
243
Moore, I.D., E.M. O'Loughlin, and G.J. Burch, "A Contour-based Topographic
Model for Hydrologic and Ecological Applications," Earth Surface Process
Landforms, 13, pp. 305-320, 1988.
Moore, I.D., S.M. Mackay, P.J. Wallbrink, G.J. Burch, and E.M. O'Loughlin,
"Hydrologic Characteristics and Modelling of a Small Forested Catchment in
Southeastern New South Wales.  Pre-logging Condition,"  Journal of Hydrology,
83, 307-335, 1986.
Morgan, M.G., and M. Henrion, Uncertainty, A Guide to Dealing with Uncertainty
in Quantitative Risk and Policy Analysis, Cambridge University Press, 1990, p. 178.
Neuman, S.P., "Wetting Front Pressure Head in the Infiltration Model of Green and
Ampt," Water Resources Research, 12, 3, 564-566, 1976.
Noilhan, J., and S. Planton, ?A Simple Parameterization of Land Surface Processes
for Meteorological Models,? Monthly Weather Review, 117, 536-549, 1989.
O?Callaghan, J.F., Mark, D.M. : The extraction of drainage networks from digital
elevation data. Computer Vision, Graphics and Image Processing 28, 323-344
(1984).
Olivera, F., D.R. Maidment, and R.J. Charbeneau, Spatially Distributed Modeling of
Storm Runoff and Non-Point Source Pollution using Geographic Information
Systems, CRWR Online Report 96-4, December 1996.
Olivera, F., S. Reed, and D. Maidment, ?HEC-PrePro v. 2.0: An ArcView Pre-
Processor for HEC?s Hydrologic Modeling System,?  paper presentation at the 1998
ESRI User?s Conference, San Diego, CA, July 25-31, 1998
(www.ce.utexas.edu/prof/olivera/esri98/p400.htm).
O?Loughlin, E.M., ?Saturation Regions in Catchments and their Relations to Soil
and Topographic Properties,? Journal of Hydrology, 53, 229?246, 1981.
O'Loughlin, E.M., ?Prediction of Surface Saturation Zones in Natural Catchments
by Topographic Analysis,? Water Resources Research, 22, 5, pp. 794-804, May
1986.
Peters, J., personal communication, 1998.
Peters, J., and D. Easton, ?Runoff Simulation Using Radar Rainfall Data,? Water
Resources Bulletin, 32, 4, 753-760, 1996.
244
Peppler, R.A., Sisterson, D.L., and Lamb, P.J., "Site Scientific Mission Plan for the
Southern Great Plains CART Site: July-December 1997," ARM-97-002, prepared
for the U.S. Department of Energy under Contract W-31-109-Eng-38.
Peppler, R.A., Sisterson, D.L., and Lamb, P.J., "Site Scientific Mission Plan for the
Southern Great Plains CART Site: January-June 1998," ARM-98-002, prepared for
the U.S. Department of Energy under Contract W-31-109-Eng-38.
Pilgrim, D.H., and I. Cordery, "Flood Runoff," Handbook of Hydrology, Chapter 9,
D.R. Maidment (ed.), McGraw-Hill, Inc., 1993.
Rawls, W.J., Ahuja, L.R., Brakensiek, D.L., and A. Shirmohammadi, ?Infiltration
and Soil Water Movement,? Handbook of Hydrology, ed. D. Maidment, Chapter 5,
McGraw-Hill, Inc., 1993.
Rawls, W.J., D.L. Brakensiek, and N. Miller, "Green-Ampt Infiltration Parameters
from Soils Data," J. Hydraul. Div. Am. Soc. Civ. Eng., 109, No. 1, pp. 62 - 70, 1983.
Rawls, W.J., D.L. Brakensiek, K.E. Saxton, "Estimation of Soil Water Properties,"
Transactions of the ASAE, ASAE Paper No. 81-2510, pp. 1316-1320, 1982.
Rawls, W.J., and D.L. Brakensiek, "Prediction of Soil Water Properties for
Hydrologic Modeling,"  Watershed Management in the Eighties, ASCE, pp. 293-
299, 1985.
Reece, C.F., "Evaluation of a Line Heat Dissipation Sensor for Measuring Soil
Matric Potential," Soil Sci. Soc. Am. J., 60, 1022-1028, 1996.
Reed, S.M., and D.R. Maidment, A GIS Procedure for Merging NEXRAD
Precipitation Data and Digital Elevation Models to Determine Rainfall-Runoff
Modeling Parameters, CRWR Online Report 95-3, September 1995.
Reed, S.M., and D.R. Maidment, ?Coordinate Transformations for Using Nexrad
Data in GIS-based Hydrologic Modeling,? accepted for publication in the Journal of
Hydrologic Engineering, April 21, 1998.
Sellers, P.J., "Biophysical Models of Land Surface Processes," Climate System
Modeling, Chapter 14, K.E. Trenberth (ed.), Cambridge University Press, 1992.
Sellers, P.J., Y. Mintz, Y.C. Sud, and A. Dalcher, "A Simple Biosphere Model (SiB)
for Use with General Circulation Models," Journal of the Atmospheric Sciences, 43,
6, 1986.
245
Schneider, J.M., and D.K. Fisher, "Meeting GEWEX/GCIP Measurement Needs by
Adding Automated Measurement of Soil Water and Temperature Profiles to the
DOE ARM/CART Southern Great Plains Site," AMS 13th Conference on
Hydrology, 265-268, 1997.
Shedd, R.C., and Fulton, R.A., "WSR-88D Precipitation Processing and its Use in
National Weather Service Hydrologic Forecasting," Engineering Hydrology:
Proceedings of the Symposium Sponsored by the Hydraulics Division of the
American Society of Civil Engineers, San Francisco, CA, July 25-30, 1993.
Shuttleworth, J.W., "Evaporation," Handbook of Hydrology, D.R. Maidment Editor,
McGraw-Hill, Inc., 1993.
Sivapalan, M., K. Beven, and E.F. Wood, "On Hydrologic Similarity. 2.  A Scaled
Model of Storm Runoff Production, Water Resources Research, 23, 12, pp. 2266-
2278, 1987.
Sivapalan, M., and E.F. Wood, ?Spatial Heterogeneity and Scale in the Infiltration
Response of Catchments,? Scale Problems in Hydrology, V.K. Gupta et al. (eds.),
D. Reidel Publishing Company, 81-106, 1986.
Smith, J.A., D.J. Seo, M.L. Baeck, and M.D. Hudlow, ?An Intercomparison Study
of NEXRAD Precipitation Estimates,? Water Resources Research, v. 32, no. 7,
2035-2045, July, 1996a.
Smith, M.B., D.J. Seo, B.D. Finnerty, V. Koren, ?Distributed Parameter Hydrologic
Modeling and NEXRAD for River Forecasting:  Scale Issues Facing the National
Weather Service,? presented at the North American Water and Environment
Congress ?96, ASCE, Anaheim, California, June 24-28, 1996b.
Soil Conservation Service, National Engineering Handbook, Section 4, Hydrology,
U.S. Department of Agriculture, U.S. Government Printing Office, Washington,
D.C., 1972.
Southern Great Plains Hydrology Experiment (SGP 97),
daac.gsfc.nasa.gov/CAMPAIGN_DOCS/SGP97/sgp97.html, 1997.
United States Army Corps of Engineers (USACE),  Flood-runoff Analysis -
Engineer Manual, Manual No. 1110-2-1417, Washington, D.C., August 31, 1994.
246
United States Department of Agriculture, Soil Survey Manual, Handbook No. 18,
U.S. Government Printing Office, August, 1951.
United States Department of Agriculture, Soil Survey Manual, Handbook No. 18,
U.S. Government Printing Office, October, 1993.
United States Department of Agriculture Natural Resources Conservation Service,
"State Soil Geographic (STATSGO) Data Base Data Use and Information," Misc.
Publication No. 1492, December, 1994.
United States Department of Agriculture Natural Resources Conservation Service,
?Soil Survey Geographic Data Base Data Use and Information,? Misc. Publication
No. 1527, January, 1995.
U.S. Geological Survey Open-File Report 95-727 (CD-ROM), ?Watershed
Boundaries and Digital Elevation Model of Oklahoma Derived from 1:100,000-
scale Digital Topographic Maps,? 1995.
Watershed Modeling System, www.ecgl.byu.edu/software/wms/, 1998.
Wood, E.F., M. Sivapalan, K. Beven, and L. Band, ?Effects of Spatial Variability
and Scale with Implications to Hydrologic Modeling,? Journal of Hydrology, v.
102, 29-47, 1988.
Zahner, R., "Refinements in Empirical Functions for Realistic Soil-moisture
Regimes Under Forest Cover," International Symposium of Forest Hydrology, W.E.
Sopper and H.W. Lull, eds., Pergamon Press, New York, 261-274, 1967.
Zaslavsky, D., Sinai, G., ?Surface hydrology : 1 ? explanation of phenomena,?
Journal of the Hydraulics Division. Proceedings of the American Society of Civil
Engineers, 107, 1-16, 1981.
247
VITA
Seann Mischa Reed was born in Ithaca, New York on August 17, 1971, the son
of John and Kathleen Reed.  Seann graduated from Oak Ridge High School, Oak
Ridge, Tennessee, in 1989.  In the fall of 1989 he enrolled at Cornell University in
Ithaca, New York where he received a Bachelor of Science Degree in Civil
Engineering in May 1993.  From the fall of 1993 to the present, Seann has been a
graduate student at The University of Texas at Austin.  For his five-plus years at
The University of Texas at Austin, Seann has been a research assistant at the Center
for Research in Water Resources and has been a teacher and teaching assistant for
several short courses dealing with GIS, hydrology, and hydraulics.  Seann received
a Master of Science degree in Civil Engineering from The University of Texas at
Austin in May 1995.
Permanent Address:  2952 Cyrandall Valley Rd., Oakton, VA 22124
This dissertation was typed by the author.