Parameter Sensitivity in Hydrologic Modeling
Implementation by
David R. Maidment, Ph.D., and
Tanya Nicole Hoogerwerf
Technical Report
The University of Texas at Austin
May 2002
Abstract
Parameter Sensitivity in Hydrologic Modeling
David R. Maidment, Ph.D. and Tanya Nicole Hoogerwerf
The University of Texas at Austin, May 2002
The computation of discharge from a watershed depends on the lag time
between rainfall and runoff, while in turn the lag time depends on watershed
parameters, such as length of the longest flow path, watershed slope, and the SCS
curve number describing the effects of land use and soils. This research explores
the variation in lag time and discharge resulting from traditional and automated
methods of calculating hydrologic parameters. Four levels of extracting
hydrologic parameters are explored: (1) measurement from paper maps, (2) on-
screen extraction from raster maps, (3) using GIS and two different resolutions of
grid-based digital elevation models (DEMs), and (4) using a triangulated irregular
network (TIN). Results show that variations in watershed area and curve number
most directly impact the computed discharge, while variations in slope and flow
path length are relatively insignificant.
ii
Table of Contents
List of Tables..........................................................................................................vi
List of Figures ......................................................................................................viii
Chapter 1: Introduction ........................................................................................... 1
1.1 Background of GIS use in Water Resources Applications ................... 2
1.2 Project History....................................................................................... 2
1.3 Objectives.............................................................................................. 3
1.4 Organization .......................................................................................... 5
Chapter 2: Literature Review .................................................................................. 6
2.1 Introduction ........................................................................................... 6
2.2 Summary of Current Hydrologic Practices of TxDOT ......................... 6
2.3 Extracting Watershed Parameters from Digital Models........................ 8
2.3.1 Digital Elevation Models ............................................................. 9
2.3.2 Triangulated Irregular Networks................................................ 11
2.4 Lumped Versus Distributed Models ................................................... 14
2.5 Scale Dependency of Hydrologic Parameters..................................... 15
2.6 Modeling Urban Areas and Future Conditions ................................... 24
Chapter 3: Methodology........................................................................................ 26
3.1 Introduction ......................................................................................... 26
3.2 Case I: Traditional Methods............................................................... 27
3.2.1 The Planimeter ........................................................................... 28
3.2.2 The Map Wheel.......................................................................... 33
3.2.3 Parameter Extraction .................................................................. 34
3.3 Case II: On-Screen Digitizing of Raster Graphic Maps.................... 34
3.3.1 Digitizing in ArcView GIS 3.2 .................................................. 37
3.3.2 Parameter Extraction.................................................................. 39
iii
3.4 Case III: Automated Methods using a DEM...................................... 40
3.4.1 Overview of Automated Methods using a DEM........................ 41
3.4.2 ArcView GIS 3.2........................................................................ 45
3.4.2.1 CRWR-PrePro............................................................. 46
3.4.2.2 Parameter Extraction using ArcGIS and CRWR-
PrePro................................................................................ 49
3.4.3 Watershed Modeling System (WMS) ........................................ 51
3.4.3.1 TOPAZ ........................................................................ 54
3.4.3.2 Parameter Extraction................................................... 55
3.5 Case IV: Automated Methods using a TIN........................................ 57
3.5.1 TIN Development and Processing.............................................. 58
3.5.2 TIN Parameter Extraction .......................................................... 60
Chapter 4: Application .......................................................................................... 61
4.1 Site Selection....................................................................................... 61
4.2 Case I: Traditional Methods............................................................... 64
4.3 Case II: Automated Methods by Digitizing ....................................... 67
4.4 Case III: Automated Methods Using a DEM ..................................... 70
4.4.1 ArcView GIS 3.2 and CRWR-PrePro ........................................ 71
4.4.1.1 10 Meter DEM ............................................................ 75
4.4.1.2 30 Meter DEM ............................................................ 76
4.4.2 Automated Methods Using WMS .............................................. 78
4.4.2.1 10 Meter DEM ............................................................ 78
4.4.2.2 30 Meter DEM ............................................................ 81
4.5 Case IV: Automated Methods Using a TIN........................................ 84
Chapter 5: Results ................................................................................................. 89
5.1. Case Study Parameter Results Summary ............................................ 89
5.2 Elasticity Analysis............................................................................... 95
5.2.1 Step 1: Sensitivity of SCS Lag Time to Parameter Variations . 98
5.2.1.1 Longest Flow Path....................................................... 99
iv
5.2.1.2 Slope.......................................................................... 102
5.2.1.3 Curve Number........................................................... 103
5.2.1.4 Step 1 Results........................................................... 105
5.2.2 Step 2: Sensitivity of Flow to Variations in Lag Time ........... 106
5.2.2.1 HEC-1 Model ............................................................ 108
5.2.2.2 Lag Time Variation ................................................... 112
5.2.2.3 Drainage Area Variation ........................................... 115
5.2.2.4 Step 2 Results ............................................................ 118
5.2.3 Results from Elasticity Analysis .............................................. 119
Chapter 6: Conclusions and Recommendations.................................................. 124
6.1 Errors Sources Among Parameter Extraction Methods .................... 125
6.1.1 Parameter Extraction using Paper Maps .................................. 125
6.1.2 On-Screen Digitizing of Raster Graphic Maps ........................ 126
6.1.3 Parameter Extraction Using Automated Methods.................... 127
6.1.4 Parameter Extraction Using TINs ............................................ 128
6.2 Errors Among Extracted Hydrologic Parameters.............................. 129
6.2.1 Drainage Area .......................................................................... 129
6.2.2 Slope......................................................................................... 129
6.2.3 Longest Flow Path.................................................................... 130
6.2.4 Perimeter .................................................................................. 130
6.2.5 Curve Number.......................................................................... 130
6.3 Significance of Errors........................................................................ 131
6.4 Additional Sources of Error .............................................................. 134
6.5 Recommendations ............................................................................. 135
Appendix A: Online Internet Resources ............................................................ 137
Appendix B: Unmodified HEC-1 Model ........................................................... 138
Appendix C: Modified HEC-1 Model in HEC-HMS......................................... 141
Bibliography........................................................................................................ 150
v
List of Tables
Table 4.1: Hand Delineation Results.................................................................... 66
Table 4.2: On-Screen Digitizing Results.............................................................. 70
Table 4.3: 10 Meter DEM Watershed Data.......................................................... 75
Table 4.4: 10 Meter DEM in ArcView Results.................................................... 76
Table 4.5: 30 Meter DEM Watershed Data.......................................................... 77
Table 4.6: 30 Meter DEM in ArcView Results.................................................... 78
Table 4.7: 10 Meter DEM in WMS Results......................................................... 81
Table 4.8: 30Meter DEM in WMS Results.......................................................... 83
Table 4.9: TIN Results ......................................................................................... 88
Table 5.1: Area 1 Results ..................................................................................... 90
Table 5.2: Area 2 Results ..................................................................................... 91
Table 5.3: Area 3 Results ..................................................................................... 91
Table 5.4: Curve Number Values......................................................................... 95
Table 5.5: Parameter Base Values........................................................................ 99
Table 5.6: Longest Flow Path Calculations ....................................................... 100
Table 5.7: Slope Calculations............................................................................. 102
Table 5.8: Curve Number Measurements .......................................................... 104
Table 5.9: Curve Number Calculations.............................................................. 104
Table 5.10: Results from Analytical Calculations ............................................. 106
Table 5.11: HEC-1 Modifications...................................................................... 110
Table 5.12: TxDOT IDF (Intensity-Duration-Frequency) Curves..................... 112
Table 5.13: HEC-1 Lag Input.............................................................................. 113
vi
Table 5.14: Discharge Variations Due to Lag Variations .................................. 114
Table 5.15: Discharge Elasticity Due to Parameter Variation ........................... 115
Table 5.16: HEC-1 Area Input ........................................................................... 116
Table 5.17: Area Variation................................................................................. 117
Table 5.18: Discharge Elasticity due to Area Variations ................................... 118
Table 5.19: Parameter Coefficients of Variation (CV) ...................................... 122
Table 5.20: Discharge Variations for Area 3 ..................................................... 122
Table 6.1: Elasticity Analysis............................................................................. 132
vii
List of Figures
Figure 1.1: Lag Time Results (Anderson, 2000).................................................... 4
Figure 2.1: DEM Grid Cells................................................................................... 9
Figure 2.2: 30 Meter DEM................................................................................... 10
Figure 2.3: TIN Up-Close .................................................................................... 11
Figure 2.4: Channel Representation with a TIN .................................................. 12
Figure 2.5: Frequency Distribution of Subcatchment Length (Garbrecht et al.,
1999b)............................................................................................... 20
Figure 2.6: Frequency Distribution of Subcatchment Slope (Garbrecht et. al.,
1999b)............................................................................................... 23
Figure 3.1: Watershed Parameters ....................................................................... 27
Figure 3.2: Planimeter .......................................................................................... 29
Figure 3.3: Tracing Arm Movement (Kunkle, 2001)........................................... 30
Figure 3.4: Area Calculation (Kunkle, 2001)....................................................... 31
Figure 3.5: Final Area Calculation (Kunkle, 2001) ............................................. 33
Figure 3.6: Map Wheel......................................................................................... 34
Figure 3.7: USGS Digital Raster Graphic (DRG)................................................ 36
Figure 3.8: On-Screen Digitizing from a Raster Graphic Map ............................ 37
Figure 3.9: New Theme Creation in ArcView ..................................................... 38
Figure 3.10: Map-Based Model in ArcView........................................................ 40
Figure 3.11: 8-Direction Pour Point Model (Anderson, 2000) ............................ 42
Figure 3.12: The D-? Model (Tarboton, 1997) ................................................... 43
Figure 3.13: 30 Meter Delineated DEM............................................................... 45
viii
Figure 3.14: The CRWR-PrePro Menu Bar ......................................................... 47
Figure 3.15: WMS Drainage Toolbar .................................................................. 52
Figure 3.16: WMS Module .................................................................................. 53
Figure 3.17: TOPAZ Grids................................................................................... 55
Figure 3.18: Display Options in WMS................................................................. 56
Figure 3.19: WMS Calculators............................................................................. 57
Figure 3.20: WMS TIN Module........................................................................... 58
Figure 3.21: Vertex Options in WMS .................................................................. 59
Figure 4.1: Location of Travis County, Texas ..................................................... 62
Figure 4.2: Location of Buttermilk and Little Walnut in Travis County ............. 63
Figure 4.3: Area 1, Area 2, and Area 3 ................................................................ 64
Figure 4.4: Hand Delineation of Area 2............................................................... 65
Figure 4.5: Area 1 Digitized................................................................................. 68
Figure 4.6: Area 2 Digitized................................................................................. 69
Figure 4.7: Area 3 Digitized................................................................................. 69
Figure 4.8: Land Use/Land Cover (LULC) Data ................................................. 73
Figure 4.9: STATSGO Data................................................................................. 74
Figure 4.10: Delineated DEM Study Area ........................................................... 77
Figure 4.11: Area 1 Delineated in WMS.............................................................. 79
Figure 4.12: Area 2 Delineated with WMS........................................................... 80
Figure 4.13: Area 3 Delineated with WMS........................................................... 80
Figure 4.14: Area 1 Delineated with WMS.......................................................... 82
Figure 4.15: Area 2 Delineated with WMS.......................................................... 82
ix
Figure 4.16: Area 3 Delineated with WMS.......................................................... 83
Figure 4.17: TIN Development in WMS ............................................................. 85
Figure 4.18: Points, Breaklines, and Bounding Polygon in WMS....................... 86
Figure 4.19: Interpolated TIN for Area 2............................................................. 87
Figure 4.20: Area 2 TIN Delineated..................................................................... 88
Figure 5.1: Three Study Areas ............................................................................. 89
Figure 5.2: Coefficients of Variation ................................................................... 94
Figure 5.3: Concept of Elasticity.......................................................................... 96
Figure 5.4: Discharge Sensitivity Calculation...................................................... 97
Figure 5.5: Step 1 in Calculating the Discharge Sensitivity ................................ 98
Figure 5.6: Analytical Calculation of Gradients ................................................ 101
Figure 5.7: Slope Gradient Calculation.............................................................. 103
Figure 5.8: Curve Number Gradient Calculation............................................... 105
Figure 5.9: Step 2 in Calculating the Discharge Sensitivity .............................. 107
Figure 5.10: HEC-1 Study Area......................................................................... 108
Figure 5.11: Unmodified HEC-1 model (displayed in HEC-HMS)................... 109
Figure 5.12: Modified HEC-1 Model (displayed in HEC-HMS)....................... 111
Figure 5.13: Discharge Results for Lag Variations............................................ 114
Figure 5.14: Discharge Results for Drainage Area Variations........................... 117
Figure 5.15: Final Discharge Sensitivity Calculation ........................................ 119
Figure 5.16: Elasticity Diagram ......................................................................... 120
Figure 5.17: Elasticity Results ........................................................................... 121
x
Chapter 1: Introduction
Hydrology, as defined by the 2001 Texas Department of Transportation
(TxDOT) Hydraulic Design Manual, ?deals with estimating flood magnitudes as
the result of precipitation.? Estimating the magnitude of an extreme flood event is
essential in the design of highway drainage facilities such as culverts, bridges,
storm drain systems, and detention storage facilities. The time-dependent
determination of the quantity of water expected to be conveyed by each structure
is used as a guide when designing the structure so that peak flows associated with
an extreme flooding event do not cause flooding in areas adjacent to the structure
and the road. With proper design of these facilities, damages resulting from an
extreme flood event are minimized.
The principal factors affecting flood magnitudes in a watershed include
runoff (influenced by precipitation and abstractions), watershed area information
(slope, longest flow path, area), land use, and soil type. Detention storage
systems, flow diversions, channelization, and impervious cover from urban
development also influence the magnitude of an extreme flood event. Currently,
TxDOT relies heavily on manual techniques to locate drainage divides and to
estimate hydrologic parameters such as flow path length, watershed area, slope,
and abstrations. These parameters are necessary in determining the peak discharge
at an area outlet, although many runoff estimation techniques assume the size of
the contributing watershed and the watershed slope as the principal variables.
1
1.1 BACKGROUND OF GIS USE IN WATER RESOURCES APPLICATIONS
Recently, the use of automated methods in water resources engineering
applications has proven to be a viable alternative to more traditional hand
calculation methods for many engineering agencies. The Texas Department of
Transportation (TxDOT) is one of many agencies interested in using automated
methods to aid in the hydrologic parameter development required for highway
drainage facility design.
As a large portion of the cost associated with highway projects is
attributed to the design and construction of drainage facilities, research has been
dedicated to exploring a more economical and time efficient means of design.
Geographic Information Systems (GIS) are a means of simplifying this process in
that a GIS is capable of computing spatially derived hydrologic parameters such
as watershed area, SCS curve number (for runoff and lag time computation),
gridded precipitation, flow length, and slope for each watershed in a relatively
short time.
1.2 PROJECT HISTORY
This implementation project follows an investigation conducted by
Anderson (2000) for the Texas Department of Transportation (TxDOT). Anderson
implemented GIS-based tools developed by CRWR at two sites in Texas:
Castleman Creek (McClennan County, TX) and Pecan Bayou (Brown County,
TX). The methodology used in Anderson?s investigation utilized tools for
hydrologic analysis and parameter extraction (CRWR-PrePro), terrain data
2
development and floodplain delineation (CRWR-FloodMap and HEC-GeoRAS),
and lumped parameter hydrologic modeling and steady flow hydraulic analysis
(HEC-HMS and HEC-RAS). Anderson?s investigation was a first attempt for
TxDOT at automating this entire process and for representing the spatial
variability of the watershed characteristics, integrating hydrologic and hydraulic
modeling processes with GIS, and displaying an accurate floodplain map of the
project site.
1.3 OBJECTIVES
The initial scope of the author?s project was to implement a study in a
different area of Texas and apply the same methodology as Anderson (2000).
However, after Anderson?s study was completed TxDOT noted that the watershed
lag time values calculated using CRWR-PrePro for input into HEC-HMS for the
Castleman Creek (in McLennan County, Texas) watershed were over four times
greater than the values calculated previously for a TxDOT hydrologic model of
the area. At this point it became evident that some assumptions used in
Anderson?s automated hydrologic study need to be further investigated. Figure
1.1 gives an example of the large differences calculated by Anderson (2000) using
automated methods and TxDOT using traditional, paper map-based methods.
3
Figure 1.1: Lag Time Results (Anderson, 2000)
Discussions with the TxDOT project supervisor, David Stolpa, in March
2001 pertaining to the use of methods for hydrologic modeling led to other
uncertainties and doubts about the automated process. Questions arose regarding
scale effects of digital elevation models; as well as mechanical processes behind
calculating slope, area, and longest flow path for a watershed. This uncertainty in
the parameter values led TxDOT to question how variations in hydrologic
parameters would ultimately affect watershed lag time values and discharge.
Following these conversations, the scope of the research project was
redefined to explore the variation in lag time and discharge resulting from paper
map and automated methods of calculating hydrologic parameters. In order to
evaluate parameter uncertainty, four levels of extracting hydrologic parameters
are explored: (1) measurement from paper maps; (2) on-screen extraction from
raster maps; (3) using GIS and two different resolutions of grid-based digital
elevation models (DEMs); and (4) using a triangulated irregular network (TIN).
This report attempts to quantify the errors associated from parameter
variation in a two-step process. The first step is to quantify the error in lag time
4
(using the SCS lag equation) based on hydrologic parameter variation. Secondly,
a HEC-1 model supplied by TxDOT is modified to quantify the error of discharge
based on lag time and drainage area variations.
Although the drainage areas under investigation are urbanized, the scope
of this project does not fully take into account small-scale man-made structures
such as street gutters, inlets, and drainage ditches and culverts that control surface
drainage patterns.
1.4 ORGANIZATION
This implementation project is divided into five principal parts. Chapter 2
explores previous work that has been conducted in the issue of scale dependency
of some hydrologic parameters. Chapter 3 provides a general overview of the
steps taken in the four case studies, while Chapter 4 gives a step-by-step account
of the hydrologic parameter computation process for each of the four case studies.
Chapter 5 analyzes the results using the concept of elasticity to determine the
overall effect of parameter variation on the discharge. The final chapter, Chapter
6, discusses the findings from this implementation project.
5
Chapter 2: Literature Review
2.1 INTRODUCTION
From 1996 to 1999, the Center for Research in Water Resources (CRWR)
at the University of Texas at Austin developed hydrologic modeling tools for the
purpose of floodplain delineation at highway river crossings for TxDOT (Olivera
et al., 1999).
From 1999 to 2000, TxDOT funded CRWR to implement these tools and
to investigate the possibility of combining existing GIS tools, lumped parameter
hydrologic and one-dimensional hydraulic models, and the visual display
capabilities of GIS to overcome the historical limitations of floodplain mapping.
Anderson (2000) implemented these tools at two existing TxDOT drainage
structures. Apart from evaluating the feasibility of implementing the tools
developed at CRWR, Anderson (2000) set out to determine if existing digital data
are sufficient to produce an accurate representation of the floodplain.
Since the work of Anderson (2000), many issues have arisen regarding the
work of GIS in water resources. This chapter reviews some of these problems
and work that has been conducted in this area.
2.2 SUMMARY OF CURRENT HYDROLOGIC PRACTICES OF TXDOT
Currently TxDOT relies heavily on manual methods of watershed
parameter extraction. Measurements are taken by hand from paper maps and then
watershed parameters are entered into a hydrologic model.
6
Hydrologic modeling practices used by TxDOT are outlined in the 2001
TxDOT Hydraulic Design Manual. Methods used by TxDOT primarily include
unit hydrographs, the Rational Method, statistical analysis of stream flow data,
and regional regression equations.
The NRCS (National Resource Conservation Service) dimensionless unit
hydrograph along with the NRCS Runoff Curve Number Method (also known as
the SCS curve number method) is the primary unit hydrograph technique used by
TxDOT. The unit hydrograph, a method for estimating storm runoff, was first
proposed by L. K. Sherman in 1932. The unit hydrograph is defined as the
watershed response to a unit depth of excess rainfall uniformly distributed over
the entire watershed and applied at a constant rate for a given period of time
(Chow et al., 1988).
Unit hydrograph techniques consider the time distribution of rainfall, the
initial rainfall losses, and an infiltration rate that decreases during the course of
the storm. Variables include the drainage area, time of concentration, curve
number (if the SCS curve number method is used), rainfall distribution, and total
design rainfall. Equations to calculate the time of concentration can also consider
watershed parameters such as watershed length and slope. Popular unit
hydrograph application programs used by TxDOT have included the TR20 (and
its TR55 variant) and HEC-1. TxDOT is now moving towards the use of HEC-
HMS. (Stolpa, 2002)
The Rational Method is very simple, and is best suited for small urban and
rural watersheds. The statistical analysis of stream gauge data method is applied
7
when long records (greater than 25 years) are available. Statistical analysis
provides peak discharge estimates using annual peak stream flow data. Regional
regression equations based on the hydrologic parameters area and slope, are
commonly used for calculating flows at ungauged sites. These equations were
developed by the USGS in 1993 to estimate the magnitude and frequency of
floods at ungauged sites in six separate regions in Texas (Jennings et al., 1993).
2.3 EXTRACTING WATERSHED PARAMETERS FROM DIGITAL MODELS
Over the last twenty years, digital representations of topographic
information have become increasingly available in the form of digital terrain
models (DTMs). Using computers and extracting watershed data from digital
terrain models is faster, and provides more reproducible measurements than
traditional manual techniques using topographic maps (Garbrecht et al., 1999a).
GIS uses a DTM to describe the spatially distributed attributes of the
terrain that are classified as topologic and topographic data. The topography of an
area describes its elevation and land surface shape, while also important are the
spatial distribution of terrain attributes other than elevation such as the land cover,
soil type, and connectivity of the features. A DTM may be used to represent both
topographic and/or physiographic features in the format of raster or grid data,
triangulated irregular network (TIN) data or vector (point, line, and polygon) data
(Olivera et al., 2000).
Using automated methods a surface may be analyzed so that drainage
basin boundaries are defined, stream networks created, and drainage basin data
8
computed in a relatively quick time. Once the drainage basin data have been
computed, geometric modeling parameters can be extracted automatically and
entered into a hydrologic model (Nelson et al., 1997).
2.3.1 Digital Elevation Models
Digital terrain models are commonly found as grids, referred to in this
document as digital elevation models or DEMs (in other documents DEM data
includes TIN data). DEMs are composed of identical square cells arranged in
rows and columns, each with a unique value to represent the terrain elevation at
that point, as shown in Figure 2.1.
Figure 2.1: DEM Grid Cells
9
DEMs are provided at 1 km grid cell size for the entire world and 30 meter
cell sizes for the entire United States. Figure 2.2 shows an example of a 30 meter
DEM terrain representation in Austin, Texas.
Figure 2.2: 30 Meter DEM
DEMs are used in water resources engineering to identify drainage related
features such as ridges, valley bottoms, channel networks, and surface drainage
patterns. DEMs are also useful in quantifying subcatchment and channel size,
length, and slope. The accuracy of this topographic information is a function of
the quality and resolution of the DEM, in addition to the DEM processing
algorithms used to extract this information (Garbrecht et al., 1999).
One solution to reduce the errors associated with DEMs, as described by
Garbrecht et al. (1999a), is to use a high resolution DEM produced by more
10
advanced methods or to customize a DEM. Another factor affecting accuracy is
whether the data are integer or floating points. When data are integers, less
memory is required than when data are floating points; however floating point
data are more accurate (Olivera et al., 1999). For this reason quality and
resolution must be considered when selecting a DEM for hydrologic modeling so
that both are consistent with the scale of the model and the objectives of the study
being made (Garbrecht et al., 1999a).
2.3.2 Triangulated Irregular Networks
Another form of digital data is the triangulated irregular network or TIN.
TINs, however, are currently not as widely used as grid DEMs, but may be used
to serve the same purpose. TINs consist of a set of representative irregularly
distributed points connected by straight lines to produce triangles, as shown in
Figure 2.3.
Figure 2.3: TIN Up-Close
11
Figure 2.4 shows the ability of a TIN to conform to complex terrain and
identify channel features.
Figure 2.4: Channel Representation with a TIN
Delauney triangulation, based on the concept of maximizing the minimum
angle of all triangles produced by connector lines to their nearest neighbor points,
is most often used to generate TINs. Breaklines are used to control the
smoothness and continuity of the surface such as streamlines or roads by forcing
triangles to conform to these lines. For this reason, TINs are generally used for
surface representation of stream channels in hydraulic modeling since complex
land surface details may accurately be represented (Lee et al., 1980).
12
TINs have the advantage compared to grid-based elevation models in that
they require less memory than grids. In addition, linear features are more
accurately represented with TINs than with DEMs. When using grids to model
channels and other linear features, edges must be always oriented along the
horizontal, vertical, or diagonal directions. TINs eliminate this data redundancy
and are thus better suited for modeling streams and other linear features. TINs
can be constructed so that triangle edges conform to features and are not restricted
to lie in the horizontal, vertical, and diagonal directions. The TIN data structure is
also often more efficient because the terrain model can be adapted readily to the
surface being modeled. In areas where the terrain is flat, only few a points need
to be utilized (Nelson et al., 1999).
Grid-based watershed modeling is advantageous over TIN-based
watershed delineation in that grids have a simpler data structure than TINS, grid-
based data is very abundant, and grid-based models are reproducible. Other
disadvantages of TINs result when inserting breaklines. Inserting breaklines may
result in small or long thin triangles which, in turn, will cause difficulties in
numerical round off or tolerance problems. TINs have the major disadvantage in
that large TINs are difficult to work with, and editing pits and flat triangles can be
a very time consuming process, especially when areas are large. Lastly,
determining an appropriate resolution for a TIN can be a difficult task (Nelson et
al., 1999a).
TIN-based watershed delineation is based on the process of tracing a flow
across triangle surfaces. Because each triangle has a flat surface, the mathematics
13
behind determining the path of maximum downward gradient is straightforward
as described by Jones et al. (1990). Watershed boundaries are delineated with
TINS by identifying outlet locations. Once outlets have been selected, flow paths
are traced along the path of steepest descent and by combining together triangles
whose flow paths pass through a common outlet point.
2.4 LUMPED VERSUS DISTRIBUTED MODELS
Creating an accurate hydrologic model of an area is a difficult task. As
most hydrologic systems are spatially variable, distributed models may be
required to fully describe the system. Distributed models require that calculations
be made on a point-to-point basis within the model, and that flow be calculated as
a function of time and space throughout the system. Lumped models on the other
hand provide a unique representative value for the entire subcatchment. In a
lumped model, flow is calculated as a function of time alone (Nelson et al.,
1997).
Olivera et al. (1999) discuss how there have been attempts to account for
spatially distributed terrain attributes based on lumped models, as the boundary
between lumped models and distributed models is not clearly defined. For
example, models such as HEC-1, developed by the US Army Corps of Engineers,
are neither purely lumped nor purely distributed. HEC-1 may be used to partition
the hydrologic system into subsystems and to apply lumped models to each of the
subsystems. HEC-1 then routes the responses from each subbasin to the
watershed outlet.
14
Nelson et al. (1997) suggest that although distributed models are the focus
of current research, lumped models are still more common and preferred because
regulatory agencies have not accepted distributed models due to the effort
involved in calibrating and verifying them. Models known as data reduction
(DR) models are one way of converting distributed properties of an area, such as
slope and subcatchment length, into lumped parameters by reducing distributed
properties into a representative value for each subcatchment (Garbrecht et al.,
1999a). GIS is a tool that allows the user to jump from strictly lumped models to
more spatially distributed models, in that a GIS may be used to generate input
files for lumped models based on a distributed interpretation of the terrain
(Nelson et al., 1997).
If the rainfall-runoff response of a watershed is linear, a unit hydrograph
can be used to relate rainfall to runoff. Most lumped models are based on either
synthetic or derived unit hydrographs. Once a unit hydrograph is determined for a
watershed, then one can determine the flood hydrograph resulting from any
measured or design rainfall. For both traditional and automated processes, the
unit hydrograph method is commonly used to model rainfall-runoff processes.
Since the systems are linear, the overall response time can be calculated as the
sum of the sub-area responses (Nelson et al., 1997).
2.5 SCALE DEPENDENCY OF HYDROLOGIC PARAMETERS
Although using DEMs provides for quick analysis, there are several
disadvantages to using DEMs, which include the effect of grid size on some
15
certain computed topographic parameters (such as longest flow path), and the
inability to adjust the grid size to the dimensions of topographic land surface
features (Garbrecht et al., 1999).
Miller et al. (1999) explored the effects of spatial resolution and accuracy
of DEMs on hydrologic characterization using GIS. The study analyzed the area,
slope, drainage density, and surface variation for watersheds ranging in size from
0.0016 km
2
to 146 km
2
, using 2.5, 10. 30, and 40 meter DEMs. Miller et al.
(1999) noted an overall reduction in slope with increasing cell size. In Miller?s
study, both the mean and standard deviation of watershed slopes are highest for
IFSAR DEMs (highest resolution DEMs). A reduction in slope standard
deviation implies that much of the natural surface has been simplified to a more
continuous smooth surface (Miller et al., 1999).
Another observation drawn by Miller et al. (1999) is that the high
resolution DEMs create more tortuous flow paths, more complex routing, and
longer drainage networks. Total drainage lengths were found to be considerably
different among four DEMs of different cell sizes on smaller watersheds as
described by the drainage density (length/area). Mean drainage density is higher
for watersheds and channels created with high resolution DEMs than for other
surfaces (0.0104 m for the 2.5 m IFSAR DEM as compared to 0.0085 m for the
40 m DEM).
Miller et al. (1999) found that the high resolution IFSAR DEM provided
significantly different results at small scales when compared to other surfaces,
while the differences among DEMs at larger scales were reduced. The final
16
conclusion from this study was that the suitability of various digital elevation data
is primarily a function of the research objectives and scale of application (Miller
et al., 1999).
In a study conducted by Garbrecht et al. (1994), the accuracy of drainage
features extracted from DEMs as a function of DEM resolution is evaluated. The
horizontal resolution of a DEM with an original grid spacing of 30 meters is
decreased by cell aggregation. Selected drainage features for several hypothetical
channel network configurations were extracted for a range of DEM resolutions
using TOPAZ software.
The study by Garbrecht et al. (1994) concluded that the dependency of
physical characteristics on grid resolution ?was introduced by the inability of a
DEM to accurately reproduce drainage features that are at the same scale as the
spatial resolution of the DEM.? In other words, the number of channels, the size
of the drainage area, and the channel network pattern from a low resolution DEM
may depart considerably from the one obtained by a high resolution DEM. For
sinuous channels, the use of a low resolution DEM results in shorter channel
lengths. For networks with a high drainage density, the use of a low resolution
DEM leads to channel and drainage area capturing, This being the point at which
the DEM resolution can no longer resolve the separation between channels or
drainage boundaries. If small drainage features are important, then resolution
must be selected relative to the size of these features (Garbrecht et al., 1994).
Garbrecht et al. (1999b) discuss the extraction of drainage properties from
DEMs. This study compares methods of extracting length and slope values using
17
both automated and traditional methods from 177 subcatchments located in the
USDA-ARS Walnut Gulch Experimental Watershed in Tombstone, Arizona.
For the manual method in Garbrecht?s study, length is measured
subjectively as the distance between the upslope subcatchment drainage boundary
and downslope channel. Slope is calculated as a lumped parameter by converting
variable slope into a straight-line profile (Gray?s method) while maintaining the
horizontal distance and area under the profile. For the automated methods, DEMs
were processed using the TOPAZ software, producing 183 subcatchments. Length
and slope values were extracted using data reduction (DR) models. (Garbrecht et
al., 1999b)
Subcatchment length is important in hydrologic modeling applications
because it is used to estimate runoff travel distance or flow routing distance.
Garbrecht et al. (1999b) describe two methods for calculating this length using
data reduction (DR) models. One method is the average travel distance, and the
other is the average flow path length, or the distance of overland flow within a
subcatchment.
The average travel distance traveled by surface runoff is calculated as the
average distance from every point in the subcatchment to the first downstream
channel that the flow reaches at this point, or the arithmetic mean of all travel
distances within a subcatchment. For subcatchments that are rectangular in shape
the average travel distance is about half the subcatchment length, and twice the
average travel distance corresponds to length from the drainage divide at the
upstream boundary to the downstream channel, as can be seen in Equation 2.1.
18
?
?
=
=
=
s
s
n
i
i
n
i
ii
t
k
kD
L
1
1
Equation 2.1
L
t
is the average travel distance, n
s
is the total number of cells in the
subcatchment, D
i
is the travel distance of cell i to the adjacent channel, and k
i
is a
weighting factor with a value of 1 for travel distances originating at subcatchment
cells, and ? for travel distances originating at channel cells. The weighting factor
accounts for the fact that channel cells contain a channel and the cell area is
evenly split between the right and left subcatchments adjacent to the channel. No
adjustments are needed for subcatchment cells, so their weighting factor is 1
(Garbrecht et al., 1999b).
The second method, the average flow path length, shown in Equation 2.2,
is different in that not all points in the subcatchment are considered in the length
calculations. The flow path in this method is considered as the distance from a
divide to the first adjacent downstream channel. Only the cells in the drainage
divide are considered in this calculation. Drainage divides are not only located at
the upstream boundary of the subcatchment, but also within the subcatchment as
defined by local ridges in the topography. For this reason the flow path length is
generally shorter than the average travel distance method to the drainage divide.
(Garbrecht et al., 1999b)
?
=
=
i
n
i
f
i
i
f
l
n
L
1
1
Equation 2.2
19
Under this method L
f
is the average flow path length, n
i
is the number of
flow paths in the subcatchment, i is the flow path counter, and l
f
the length of
individual flow paths. Figure 2.5 shows the results by Garbrecht et al. (1999b) for
the 177 subcatchments, using manual and automated methods of length
extraction.
Figure 2.5: Frequency Distribution of Subcatchment Length (Garbrecht et al.,
1999b)
Figure 2.5 is an example of parameter value variations, in this case
subcatchment length, that occur using different methods of parameter extraction
(automated and manual) and different models for data reduction. The distance to
divide length is twice the average travel distance for rectangular-shaped
subcatchments and 1.33 times the average travel distance for triangular-shaped
subcatchments. The distance to divide and the manual method closely resemble
each other since they represent the same length from the watershed boundary to
20
the downstream channel. The travel distance consistently gives the smallest
length value since the this method accounts for all the cells in the watershed
(Garbrecht et al., 1999b).
Moglen et al. (2001) show that DEMs at a 30 meter resolution are not
sufficiently dense for analyzing flat areas; thus a higher resolution grid must be
used regardless of the quality of the 30 meter grid. Garbrecht et al. (1999a) note
the reason that the lower resolution grid will not work is due to the fact that as
some DEMs (with the exception of NED DEMs, as NED is in floating point
meters) are reported in meters or feet, the computed slope can only take on a
limited number of values. For example, a 30 meter DEM in meters could have a
slope value of zero, or a multiple of 0.033 (for a 1 meter change in elevation).
These increments may be suitable to model terrain in mountainous terrain with
large slopes, but insufficient to provide accurate values in flat areas (Garbrecht et
al; 1999a).
Subcatchment slope, similar to subcatchment length, is an important
variable for runoff calculations. Garbrecht et al. (1999b) present four DR
methods of slope calculation. These are the average terrain slope, the average
travel distance slope, the average flow path slope, and the global slope.
Equation 2.3 shows the calculation for average terrain slope is the average
of the local slope value at every point in the subcatchment.
?
=
=
s
n
i
t
i
s
t
s
n
S
1
1
Equation 2.3
21
For the average terrain slope, S
t
is the average terrain slope, n
s
is the
number of subcatchment cells, and s
t
i
is the terrain slope at cell i (Garbrecht et al.,
1999b).
The average travel distance slope, Equation 2.4, is the average of the slope
from each point in the subcatchment to the next adjacent downstream channel.
?
=
=
s
n
i
c
i
s
c
s
n
S
1
1
Equation 2.4
S
c
is the average travel distance slope, n
s
is the number of cells in the
subcatchment, and s
c
i
is the slope of the travel distance that starts at cell i. The
travel distance slope of cell i is the mean of all the slopes along the travel distance
between subcatchment i and the adjacent channel (Garbrecht et al., 1999b).
The average flow path slope, Equation 2.5, is the average slope of all the
flow paths in the subcatchment, as defined as the route followed by the runoff
starting at the divide and ending at the first adjacent downstream channel
(Garbrecht et al., 1999b).
?
=
=
f
n
i
f
i
f
f
s
n
S
1
1
Equation 2.5
S
f
is the average flow path slope, n
f
is the number of flow paths in the
subcatchment, and s
f
i
is the flow path slope of the flow path starting at cell i. The
22
flow path slope, s
f
i
, is the mean of all slopes along the flow path between divide
cell i and the adjacent channel (Garbrecht et al., 1999b).
The global slope, Equation 2.6, is calculated as the average elevation of
the subcatchment minus the average elevation of the receiving channel divided by
the average travel distance (Garbrecht et al., 1999b).
t
cs
g
L
EE
S
?
= Equation 2.6
S
g
is the global slope for the subcatchment, E
s
is the mean elevation of the
subcatchment, and E
c
is the mean elevation of the adjacent channel and L
t
is the
average travel distance of the subcatchment (Garbrecht et al., 1999b). Figure 2.6
shows the results from Garbrecht?s study in terms of slope.
Figure 2.6: Frequency Distribution of Subcatchment Slope (Garbrecht et. al.,
1999b)
23
Figure 2.6 further demonstrates parameter variations among methods of
parameter extraction. The average travel distance-based slope method produces
the smallest slope because it accounts for the flatter slopes in the lower part of the
subcatchment area, thus emphasizing areas that are more subject to higher
discharges. The terrain slope method results in the steepest slope values as this
method equally emphasizes each maximum local slope value at each cell. The
average flow path slope method is steeper than the average travel distance-based
slope method because there are fewer divides in the lower part of the catchment.
The global slope and manual methods resemble each other in that the models used
to calculate these slope values are similar (Garbrecht et al., 1999b).
Garbrecht et al. (1999b) conclude that each method is equally valid, and
the user should select a method that is most appropriate for the user?s application.
For instance, if the user is most interested in calculating runoff, the average travel
distance based slope method and the average flow path slope method are better
suited for this calculation than the terrain slope method (Garbrecht et al., 1999b).
2.6 MODELING URBAN AREAS AND FUTURE CONDITIONS
Although digital data in the form of DEMs is readily available and easy to
work with, it does not accurately describe terrain in urban areas. Barrett (2000)
suggests in larger areas, where it is not feasible to digitize these drainage systems,
to delineate the watershed under undeveloped conditions. The errors associated
from water entering and leaving the watershed should cancel out, at least with
larger areas.
24
In some cases the digital data in the form of DEMs must be edited to
reduce the error that occurs from building and road features that are captured in
the DEM. Barrett (2000) used a 30 foot grid created from a TIN to help resolve
the difference in elevation of roads and bridges from the surrounding terrain, as
these features act as dams when performing flow accumulation. The initial grid
was also edited manually to improve the stream representation by creating
openings in the dams, roads and bridges.
25
Chapter 3: Methodology
3.1 INTRODUCTION
Case studies were conducted as a means to compare traditional methods of
parameter calculation to automated methods of parameter calculation. Case
studies were conducted on each of the four levels of current model development.
The first case study uses purely traditional methods on paper maps. The second
level involves using a computer and ArcView GIS 3.2 to digitize the watershed
boundaries and channels from scanned USGS quadrangle maps. The third level
involves using ArcView GIS 3.2 (with CRWR-PrePro) and WMS (Watershed
Modeling System) to compute hydrologic parameters from two different
resolution DEMs, a 10 meter DEM and a 30 meter DEM. The purpose of using
two different resolution DEMs is to quantify cell scale effects on channel length,
watershed slope, and watershed area. The final method uses a triangulated
irregular network (TIN) and the TIN processing capabilities of WMS.
Figure 3.1 illustrates the watershed parameters that are the focus of this
investigation. The longest flow path (LFP) is the longest length a drop of water
will travel in the watershed. The area of the watershed encompasses all the water
that will flow to the watershed outlet, and the slope of the watershed is the
difference in a representative watershed elevation divided by a representative
watershed length. Chapter 2 outlines several methods used to calculate slope,
depending on the application. The soil type and land use are used to derive a
26
curve number, which along with longest flow path and slope, is used to calculate
the lag time of the watershed.
Figure 3.1: Watershed Parameters
3.2 CASE I: TRADITIONAL METHODS
Traditional hydrologic modeling involves calculating watershed
parameters based on a paper map. Two instruments, a planimeter and a map
wheel, aid in this process.
Traditional methods of computing terrain-based hydrologic data involve
delineating the watershed by hand using map contours as guidelines. This is a
very time consuming process, as drainage divides may be hard to locate. Pencil
27
lines are drawn perpendicular to contour lines to indicate drainage divides. Once
the perimeter of the watershed has been established, a planimeter is used to
measure its area. Perimeter and length of the longest flow path are measured
using a map wheel. Slope is calculated by taking the difference in elevation
between map contours or from field survey data.
3.2.1 The Planimeter
A planimeter is an instrument used to trace around the perimeter of an
object to determine its area. Planimeters are useful tools in determining surface
areas from maps and aerial photos. A planimeter mechanically integrates an area,
and records this area as a tracing point moves along the boundary of the figure to
be measured. This number can be converted to an area by multiplying the
planimeter reading by a constant called the planimeter constant. This constant
varies from planimeter to planimeter.
A planimeter is composed of a graduated drum and disk, vernier, tracing
arm and point, and anchor arm and point (anchored to table). An elbow connects
the tracing arm and anchor arm, and bends and slides freely. Parallel to the
elbow, which slides and bends freely, is a wheel with a scale (consisting of a disk,
drum and vernier) that records how far the wheel has turned. Figure 3.2 is a
picture of the planimeter used in this study (Kunkle, 2001).
28
Figure 3.2: Planimeter
Calculating the planimeter constant involves plotting out a square and a
circle of known areas, placing the anchor base outside of the area to be measured,
and inserting the anchor arm into the drum assembly. Once the planimeter
reading is set to zero (or the initial reading is recorded), the perimeter is traced in
a clockwise motion and the final readings from the disk, drum, and vernier
recorded. The disk reads whole numbers, the drum tenths and hundredths of a
unit, and the vernier thousandths of a unit. Three readings of each of the two
areas should be taken. Finally, the values derived from areas of the square and
circle are averaged.
The planimeter reading from the average of circle and square value
calculations is used to solve for the constant C= N/A. The value N is the
planimeter reading, A is the known area of the object and C is the planimeter
constant. Once the planimeter constant is determined, new areas may be
29
measured using the formula A= C*N. For large areas, individual polygons should
be solved separately (Knill, 2000).
Kunkle (2001) describes the basic fundamentals behind this mathematical
operation. In Figure 3.3, the blue arm (AB) is the anchor arm. It is anchored to the
table at Point A, located outside the object to be measured. Point B is the location
of the measuring wheel (drum, disk and vernier). The green arm (BC) is the
tracing arm, and C is the tracing point. The movement of the anchor arm is
restricted to a circle. As the tracing arm moves in a clockwise direction from C to
C1, the area dw is measured.
Figure 3.3: Tracing Arm Movement (Kunkle, 2001)
In Figure 3.4, point E divides the area dw into a trapezoid and a triangle to
separate the components of the sliding and turning motion of the measurement
wheel. Area BCE represents the area covered while the tracing arm is pivoting.
During this pivoting motion d? is the rotation of the arm. Area EC?B?B is the area
covered as the tracing arm slides, and dm is the change in the scale reading while
30
the arm is sliding. Both d? and dm are arbitrarily small, and reflect the rotation of
the scale in a plane perpendicular to the tracing arm.
Figure 3.4: Area Calculation (Kunkle, 2001)
The area of the pentagon BCEC?B? can be expressed as the sum of area
BCE and area EC?B?B as shown in Equation 3.1. The area of BCE is that of a
triangle, while area EC?B?B is that of a rectangle. As mentioned earlier, k is the
length of the tracing arm.
()
kdmdkdw
BBECareaBCEareadw
BBCECareadw
+=
+=
=
?
2
2
1
)''()(
''
Equation 3.1
The region traced by the tracing arm is determined by integrating Equation
3.1. The formula for W, as shown in Equation 3.2, describes the area of the entire
region EC?B?B.
31
??
?
+=
=
dmkdkW
dwW
?
2
2
1
Equation 3.2
Since the planimeter begins and ends in the same position, the net change
in the angle d? is zero. The net change in the scale reading is due only to the
change in the scale reading. Integration of dm over the entire circuit is M, the
final reading in the scale as also shown in Equation 3.3.
kMW
dmkW
d
=
=?
=
?
?
0?
Equation 3.3
In Figure 3.5, Kunkle (2001) shows how the final area is determined.
When the arm is sliding backwards, the area covered is subtracted from the total
area. The result is only the area crossed by the tracing arm, k.
32
Figure 3.5: Final Area Calculation (Kunkle, 2001)
3.2.2 The Map Wheel
The map wheel, shown in Figure 3.6, is a simple device that measures
length. A map wheel is comprised of a wheel connected to a scale that measures
the distance the wheel has traced. This measure is then multiplied by a factor to
correct for the map scale.
33
Figure 3.6: Map Wheel
3.2.3 Parameter Extraction
Using traditional methods, hydrologic parameters are extracted from paper
maps. Slope is manually calculated based on the difference in contour line
elevations from the upper point of the longest flow path and from the outlet. The
difference in elevation is divided by the length of the longest flow path. Perimeter
and longest flow path length are determined by using the map wheel, and area is
calculated using the planimeter.
3.3 CASE II: ON-SCREEN DIGITIZING OF RASTER GRAPHIC MAPS
The process of on-screen digitizing of raster graphic maps is the next
closest method to the traditional, paper map based methods of using a map wheel
and a planimeter. The methodology is analogous; with the exception that the
34
process of digitization involves using a computer-aided mouse to draw drainage
divides on a scanned map and to trace along lines to determine their length.
Geospatial input to watershed models can be described with vector data.
Points or nodes can be used to represent outlet points, arcs (polylines) can be used
to represent streams, and enclosed polygons used to represent the watershed. This
is usually done by digitizing the streams and an approximate boundary for a
watershed with an image of the site in the background. In this study a digital
raster graphic, or DRG, is used as the basis for digitizing as shown in Figure 3.7.
This is a scanned image of a USGS topographic map produced at a 1:24,000
scale, obtained by the USGS (Appendix A). DRGs are frequently used to edit and
revise other digital data. Once the USGS map is scanned, the digital image is
georeferenced to the true ground coordinates and projected into the Universal
Transverse Mercator (UTM) for projection consistency (USGS, 1999).
35
Figure 3.7: USGS Digital Raster Graphic (DRG)
Figure 3.8 shows watershed delineation using raster maps. The arrows
point in the direction of steepest slope, while the lines divide the drainage
boundaries. Arcs may also be used to define canals, railroads, streets, or other
features that tend to act like streams during a rainfall/runoff event (Nelson et al.,
1997).
36
Figure 3.8: On-Screen Digitizing from a Raster Graphic Map
3.3.1 Digitizing in ArcView GIS 3.2
The first step in digitizing is to create a new theme as selected from the
ArcView menu, as shown in Figure 3.9.
37
Figure 3.9: New Theme Creation in ArcView
A polyline is used to represent the longest flow path. Once the polyline is
drawn (or polygon for area), ArcWorkstation is used to convert the new themes to
coverages. After the polylines and polygons are converted to coverages, they are
processed using the ArcWorkstation commands build and clean. Following this
process the coverages are opened once again in ArcView GIS 3.2 and converted
back to shapefiles.
38
3.3.2 Parameter Extraction
Figure 3.10 shows an example of a map-based model along with its length,
area, and perimeter attributes. Information provided in the attribute tables for
each shapefile give the length, area, and perimeter for each shape. Slope,
determined using the same method as with paper maps, is calculated by
subtracting the contour value at the outlet from the contour value at the start of the
longest flow path. The difference in elevation is then divided by the length of the
longest flow path.
39
Figure 3.10: Map-Based Model in ArcView
3.4 CASE III: AUTOMATED METHODS USING A DEM
In Case III, hydrologic parameters are extracted from a grid-based digital
elevation model (DEM) using both 10 meter and 30 meter DEM data. Two
different processes are used to extract hydrologic parameters from elevation data.
The first method involves using ArcView GIS 3.2 and CRWR-PrePro, while the
second method extracts hydrologic parameters using the Watershed Modeling
System (WMS). Both methods involve the same fundamental steps. These
include 1) a raster-based terrain analysis, 2) raster-based sub-basin and stream
network delineation, and 3) vectorization of the sub-basins and stream segments.
40
3.4.1 Overview of Automated Methods using a DEM
Topographic, topologic, and hydrologic information from digital spatial
data can be extracted using a DEM and automated methods. This section
describes the fundamental steps that must be taken to process a DEM and
accurately extract information.
Errors in DEMs are usually classified as sinks (pits) or peaks. Removing
the pits is a standard operation when working with grid-based DEMs. A peak is
less detrimental to the calculation of flow direction and is usually ignored. A
DEM free of sinks is termed a depressionless DEM, and is the ideal input to
calculate flow direction for watershed delineation.
The 8-Direction Pour Point Model as shown below in Figure 3.11
represents the direction water flows based on the neighboring cell with the lowest
elevation. This method is the most widely used raster DEM processing method.
41
Figure 3.11: 8-Direction Pour Point Model (Anderson, 2000)
Most basin delineation techniques for grids are based around the 8-
Direction Pour Point Model concept. However, the 8-Direction Pour Point Model
has been criticized because it permits only one flow direction leaving a cell.
Garbrecht et al. (1999a) note that this is satisfactory if being used for large
drainage areas with well-defined channels, but the model may be less appropriate
for overland flow analysis on hill slopes. A second method for determining flow
direction is the D-? Model developed by Tarboton (1997). Figure 3.12 below
depicts this model.
42
Figure 3.12: The D-? Model (Tarboton, 1997)
The D-? Model assigns a flow direction based on steepest slope on a
triangular facet (Tarboton, 1997). The automated methods covered in this report
do not use the D-? approach, and so this method of watershed delineation is not
further investigated.
Before computing filling pits and computing flow direction, the user may
want to ?burn-in? streams. This process raises the land surface cells off the
stream cells by an arbitrary elevation so that streams delineated from the DEM
match those produced in the National Hydrography Dataset (NHD) produced by
the USGS. This step is not necessary for high resolution DEMs (Olivera et al.,
1999).
Once flow direction has been calculated, a flow accumulation grid is
computed. This is done by counting the number of contributing cells to each cell
43
in the grid. Based on the flow accumulation grid, a raster-based stream network
can be developed based on a user-defined cell threshold (Olivera et al., 2000).
Threshold, a term used to describe the number of cells that drain to a
point, also defines the number of cells that constitute the beginning of a stream.
The selected threshold creates a drainage (stream) network based on all the cells
with a catchment area greater than the threshold. The choice of threshold is
complicated. Two methods for determining a threshold are the constant area
threshold method and the slope-dependent critical support area method.
The constant threshold area method represents the change in sediment
transport from sheet flow to concentrated flow, rather than a spatial transition in
longitudinal slope profiles. Using the slope-dependent critical support area
method the drainage density is greater in steeper portions of the catchment, as
found in natural landscapes (Tarboton et al., 1991). The constant threshold area
method is considered more practical in application than the slope-dependent
critical support area method, and is the only method considered in this report. In
this study, both CRWR-PrePro and WMS use the constant threshold area method.
(Garbrecht et al., 1999)
Once the DEM has been processed by filling pits, flow direction and flow
accumulation grids have been computed and a threshold chosen, the user may add
streams and watershed outlets to the model. The final step for DEM processing is
to vectorize the streams and watersheds. While a grid is convenient for the
development of hydrologic data, it is inefficient for data storage. This is because
with a grid there is a one-to-many relationship for most basin parameters. A more
44
convenient and efficient way to store both stream and basin data is through
vectors and polygons (Nelson et al., 1997).
Figure 3.13 below shows an example of a watershed delineated from a
grid-based digital elevation model (DEM).
Figure 3.13: 30 Meter Delineated DEM
The following paragraphs describe this process using both ArcView GIS
3.2 (with CRWR-PrePro) and WMS.
3.4.2 ArcView GIS 3.2
The Environmental Systems Research Institute, Inc. (ESRI) developed
ArcView GIS 3.2. This has now been replaced by ESRI?s ArcGIS software
package. ArcView GIS 3.2 is still very widely used in the professional world, and
is examined in this study for the reason that the Texas Department of
45
Transportation (TxDOT) has this software package, and also that the preprocessor
under investigation, CRWR-PrePro, operates only with ArcView GIS 3.2.
3.4.2.1 CRWR-PrePro
CRWR-PrePro was developed by several investigators at CRWR. It is the
combination of several pieces of work developed by different researchers over a
period of time. The first person to work on CRWR-PrePro was Ferdi Hellwager
(1997). That same year, the Watershed Delineator ArcView extension was
developed by Dean Djokic and Zichuan Ye of ESRI. This work was followed by
the Flood Flow Calculator ArcView GIS 3.2 extension in 1997 by CRWR to
estimate flood peak flows according to regional regression equations using the
spatial data extraction capabilities of GIS. Combining these three ArcView
extensions led to development of a hydrologic modeling tool that prepares an
input file for the HEC-HMS basin component (Olivera et al., 1999). CRWR-
PrePro, shown in Figure 3.14, is the predecessor to the more commonly used
HEC-GeoHMS, as described in the forward of the HEC-HMS User?s Manual
USACOE-HEC, (2000).
46
Figure 3.14: The CRWR-PrePro Menu Bar
CRWR-PrePro conducts a raster-based terrain analysis that includes
burning-in streams, filling sinks, and calculating the flow direction (using the 8-
Direction Point Pour Method as described earlier) and flow accumulation grids.
Following a raster-based terrain analysis (Fill Sinks, Flow Direction, and
Flow Accumulation commands), a raster-based sub-basin and reach network
analysis creates stream definition grids based on threshold (Stream Definition),
stream segmentation grids (Stream Segmentation), an outlet grid (Outlets from
Links), and delineated watershed grid (Sub-Watershed Delineation). Finally sub-
basins and reach segments are vectorized (Vectorize Streams and Watersheds).
47
CRWR-PrePro calculates losses using the SCS curve number method or
the initial plus constant loss method. To calculate curve number, soils data must
be obtained. This may be done easily free of charge by downloading the data
from the USDA-NRCS website (Appendix A). Attribute tables that relate to the
soils data include the mapunit.dbf and comp.dbf tables, which should be added to
the working directory in ArcView. The tables are modified and the CRWR-
PrePro function Soil Group Percentages may be executed. This creates a table
called muidjoin.dbf. Next Land Use/Land Cover (LULC) data must be obtained.
Similar to the soils data, these data may be downloaded free of charge from the
USGS website (Appendix A). By combining the LULC data and the soils data, a
curve number grid (CN) may be generated using the Curve Number Grid
command in CRWR-PrePro. An additional table that may be obtained off the
CRWR-PrePro webpage, rcn.txt, is used as a look-up table to link curve number
values to land use and soils data. Table wshpar.txt is necessary when the initial
plus constant loss method is used (fields in table include Initial_Loss,
Const_LossRate, and Wsh_Velocity). The final CN for each sub-basin is
calculated as the average of the curve number values within the sub-basin polygon
(Olivera et al., 1999).
Once the data are in vector format, hydrologic parameters of sub-basins
and reaches such as reach length, reach routing method (Muskingum or lag), and
either number of sub-reaches into which the reach is subdivided (for Muskingum
routing), or the flow time (for pure lag) are computed. The other reach
parameters such as flow velocity and Muskingum X cannot be computed from
48
spatial data and must be supplied by the user. The attributes are generated using
the CRWR-PrePro command Calculate Attributes (Olivera et al., 1999).
3.4.2.2 Parameter Extraction using ArcGIS and CRWR-PrePro
CRWR-PrePro uses the SCS unit hydrograph method for sub-basin
routing, and lag time may be calculated either with the SCS lagtime formula or
length over velocity. The SCS equation is most frequently used owing to the fact
that all the parameters for this equation can easily be extracted using GIS. The
SCS Lag Time Equation (1972) is shown below in Equation 3.4. In Equation 3.4
TLAG is the lag time in hours, L is the length of the longest flow path in feet, S is
the slope of the longest flow path, and CN is the average curve number in the sub-
basin.
()[]
5.0
7.08.0
*67.31
9/1000
S
CNL
TLAG
?
= Equation 3.4
In the case of the length over velocity equation, flow velocities must be
known for each cell in the subwatershed, and generally this is not readily known.
As shown in Equation 3.5, TLAG is the lag time in hours, L the length of the
longest flow path in feet, and v (m/s) is the representative velocity in the longest
flow path. As CRWR-PrePro only operates in metric units, all English units are
converted into metric units.
v
L
TLAG
*60
*3048.0
*6.0= Equation 3.5
49
Once the watershed of interest has been clipped, the user is prompted to
select a time step. The time step is the interval that determines the resolution of
model
rate method, and lag time with only
the SC
high slope
of the
results during a run in HEC-HMS. Gauge data is linearly interpolated to
the time interval, which may be chosen to range from 1 minute to 24 hours
(USACOE-HEC, 2000). The Muskingum and lag methods are used for flow
routing in the reaches, depending on the travel time. If the longest flow path
(LFP) divided by the reach velocity is less than the time step the pure lag equation
is used; otherwise Muskingum routing is used. It is apparent that time step
selection and velocity have a very large effect on the outcome and should be
modified accordingly (Olivera et al., 1999).
In addition to the fact that losses may only be calculated using the SCS
curve number or the initial plus constant loss
S lagtime formula or length over velocity, another constraint of CRWR-
PrePro is the slope calculation. CRWR-PrePro calculates slope based on the
DEM cell elevations at 1% and 99% of the longest flow path divided by their
distance along the longest flow path, and allows the user no flexibility in
calculating watershed slope unless the user changes the Avenue code.
The user may want to calculate the slope at 10% and 85% of the longest
flow path, not 1% and 99% percent of this flow path. In this way the
shallow concentrated flow is not included in the calculations for lag time.
CRWR-PrePro calculates the longest flow path of a sub-basin in the set of cells
50
for which the sum of the downstream flow length to the outlet plus the upstream
flow length to the drainage divide is maximum (Olivera et al., 1999).
Preparation of the HMS basin file (an ascii file readable by HEC-HMS) is
the final step of CRWR-PrePro. In order to do thi, transfer and parameter tables
(wshpar.txt and streamp.txt as mentioned above) are needed and are posted on the
CRWR-PrePro webpage (Appendix A). All units must be in SI (Syst?me
International d'unit?s) (Olivera et al., 1999). In this study only the parameters
curve number, slope, and longest flow path were extracted from the delineated
watershed, and thus the HEC-HMS ascii file was not created.
3.4.3 Watershed Modeling System (WMS)
WMS was developed by the Environmental Modeling Research
Laboratory (EMRL) at Brigham Young University for use in hydrologic
computation and modeling. WMS is a set of modeling codes along with grid and
mesh generation utilities, and post-processing and visualization tools in both two
and three dimensions (Nelson et al., 1997). WMS is different from using
ArcView GIS 3.2 in conjunction with CRWR-PrePro in that it may be used to run
different hydrologic modeling programs in addition to computing hydrologic
parameters. Similar to a GIS, WMS automates watershed delineation and
parameter calculation from digital elevation terrain data, importing GIS data, and
extracting watershed information from the GIS database.
As shown in Figure 3.15, the WMS drainage menu is similar to the
CRWR-PrePro menu. Like CRWR-PrePro, the WMS drainage menu bar contains
51
basic DEM processing algorithms to compute flow direction and flow
accumulation, although in WMS this is done with a program called TOPAZ. The
drainage menu in WMS also gives the user the ability to vectorize data
(Basins->Polygons), and compute the basin data (Compute Basin Data).
Figure 3.15: WMS Drainage Toolbar
Figure 3.16 shows the editing tools that are included in the DEM module.
Among these tools are a draw flow path tool, in which a flow path will be traced
across the DEM from point to point according to the flow direction grid.
Additional tools include point, vertex, arc and polygon selection and creation
tools, a measuring tool, and an upstream arc selection tool.
52
Figure 3.16: WMS Module
Additional modules available in WMS include a map module (which
allows the import of GIS vector-based data to be used in WMS for watershed
definition and hydrologic attribute mapping), hydrologic modeling with HEC-1,
NFF (National Flood Frequency), TR-20, TR-55, Rational Method, and HSPF
(Hydrologic Simulation Program Fortran), and hydrologic/hydraulic calculators.
Unlike a GIS system, WMS can run hydrologic modeling programs without
having to export or import data.
53
3.4.3.1 TOPAZ
The Watershed Modeling System (WMS) can be used to automatically
delineate watersheds and sub-basin boundaries from digital elevation models
(DEMs). WMS uses TOPAZ to delineate grid elevation data. However, flow
directions and accumulations may also be computed by ArcInfo, ArcView, or
GRASS and imported into WMS in grid ascii format. In this study, TOPAZ is
run to determine the flow direction and flow accumulation grids for WMS.
TOPAZ (Topographic Parameterization) is an automated digital landscape
analysis tool for topographic evaluation, drainage identification, watershed
segmentation, and sub-catchment parameterization, and is similar to CRWR-
PrePro, which also uses the 8-Direction Point Pour Model.
TOPAZ, shown in Figure 3.17, creates three grids: a new elevation grid
(relief.dat), a flow direction grid (flovec.dat) and a flow accumulation grid
(uparea.dat).
54
Figure 3.17: TOPAZ Grids
3.4.3.2 Parameter Extraction
Once a watershed has been delineated in WMS and all its hydrologic data
has been defined using the Compute Basin Data command, a hydrologic model
may be run. Hydrologic parameters were extracted using the Display Options
tool, shown in Figure 3.18, so that the parameters of interest as calculated by
WMS are displayed on the map.
55
Figure 3.18: Display Options in WMS
Because WMS has tools to automatically calculate parameters such as the
Green-Ampt parameters, time of concentration and other modeling parameters in
a single application, it can be used to model watersheds more efficiently than
traditional methods (Nelson et al., 1999a). Figure 3.19 shows a very handy tool in
56
WMS to compute time of travel, using user-defined equations or a variety of
predefined equations which include the Tulsa District Lag Time Equation, the
Denver Lag Time Equation, and SCS Lag Time Equation.
Figure 3.19: WMS Calculators
3.5 CASE IV: AUTOMATED METHODS USING A TIN
In addition to working with DEM terrain data, WMS also has the
capability to work with TIN (triangulated irregular network) data. Similar to grid-
based DEMs, TINs have also been used to characterize watersheds. TINs may be
created in many ways, such as from gridded data, raw survey data, and digitizing
contour data. Working with TINs is useful in that TINs provide a more precise
description of the landscape than grids; however TINs are not as widely available
and used as grids. TINs consist of a set of vertex points, connected by triangles,
that represent scattered X, Y, and Z locations. (Nelson et al., 1999b)
For TINs to be used efficiently for basin delineation and hydrologic
modeling, they need to be constructed from readily available data such as USGS
57
DEMs (Nelson et al., 1999b). According to Nelson et al. (1999b), TINs should be
constructed around streams, canals, streets, and other linear features so that the
TIN conforms to these features.
3.5.1 TIN Development and Processing
Figure 3.20allows the user to create and edit TINs.
Figure 3.20: WMS TIN Module
58
The TIN is first created using Vertex Options under the TINs menu bar.
By selecting an elevation, and a location on the background map (such as a digital
raster graphic, or DRG), a point is added to the map at the specified elevation as
shown in Figure 3.21. Alternatively elevation data can be imported from a
LIDAR or aerial land survey.
Figure 3.21: Vertex Options in WMS
Once the user is satisfied with the amount of points, TIN breaklines may
be added to the model to reflect streams and other terrain features. When
breaklines are added, the triangles created conform to these lines. So that the TIN
conforms to the terrain, breaklines must be created where the water drains. This
is done by selecting the conceptual (map) model icon and turning off the TIN.
59
Using the rough boundary and stream network defined in the conceptual
model, WMS can create a triangulated irregular network (TIN) using the adaptive
tessellation algorithm. The adaptive tessellation algorithm generates a mesh of
triangular elements. The mesh lies within the model domain, and honors the
conceptual model set up of point elevations and channel and watershed
boundaries. The boundary defines the TIN extents, and the stream and ridge arcs
are forced into the TIN as breaklines to ensure that the triangle edges will be
enforced along all streams and ridges. A stream is created from the triangle edges.
Following the development of the TIN based on background contour information,
a DEM may be used as a background elevation source (Nelson et al., 1999b).
After the TIN has been created, the final step before the TIN can be
delineated is to condition the TIN. Conditioning is necessary because there are
usually flat triangles and pits in the TIN. WMS comes with tools to condition the
TIN. These tools include elevation smoothing, edge swapping, adjusting
elevations, and adding or deleting vertices (Nelson et al., 1999b).
3.5.2 TIN Parameter Extraction
Once the basins have been defined, geometric attributes such as stream
lengths, stream slopes, basin areas, basin slopes, and maximum drainage distance
within a basin, are automatically computed from the TIN model. These attributes
may be combined with runoff curve numbers to generate a runoff model.
60
Chapter 4: Application
This implementation study applies traditional methods in hydrologic
modeling and compares those traditional processes to those using GIS for three
differently sized watershed areas. Four different case studies were performed:
(1) measurement from paper maps; (2) on-screen extraction from raster maps; (3)
using GIS and two different resolutions of grid-based digital elevation models
(DEMs); and (4) using a triangulated irregular network (TIN).
4.1 SITE SELECTION
The study areas were determined in a meeting Friday, February 9
th
, 2001
with David Stolpa of TxDOT. During this meeting Mr. Stolpa mentioned that he
was interested in comparing analyses for certain area sizes, and these watersheds
each met this area size criteria. Buttermilk Creek watershed was chosen initially
as this is an area in which TxDOT has conducted an analysis for a highway
project.
All three study areas are located in Travis County, Texas, as shown in
Figure 4.1.
61
Figure 4.1: Location of Travis County, Texas
Within Travis County are two watersheds, Buttermilk and Little Walnut,
located in northern Austin. A shapefile of these two watersheds, provided by the
City of Austin, is shown in Figure 4.2. Little Walnut is the larger watershed, and
Buttermilk is the smaller watershed.
62
Figure 4.2: Location of Buttermilk and Little Walnut in Travis County
Area 1, the small area, is a subwatershed of Buttermilk Creek. This area is
located at the upper end of Buttermilk. Area 2 is Buttermilk watershed, and Area
3 is Little Walnut creek above its confluence with Buttermilk Creek. The stream
lines shown are from the National Hydrography Dataset (NHD). Figure 4.3 gives
a general idea of the shape and location of the three study areas.
63
Figure 4.3: Area 1, Area 2,and Area 3
4.2 CASE I: TRADITIONAL METHODS
Two USGS quad maps, Austin East (3097-242) and Pflugerville West
(3097-243), were required to cover the study area of Buttermilk and Little Walnut
watersheds. Measurements were all done by hand for the three differently sized
areas, Area 1, Area 2, and Area 3.
The first watershed to be delineated was Area 2 (Buttermilk). The first
step to determine Area 2?s boundary was to locate the outlet point from the USGS
map. This point marks the confluence of Buttermilk with Little Walnut Creek
watershed. Using a pencil, the watershed of Buttermilk was traced on the USGS
maps. Secondly, the portion of Area 3 (Little Walnut) above its confluence with
Area 2 was delineated. The subwatershed of Buttermilk was delineated last (Area
1). For each area the longest flow path was determined and traced with the pencil.
Figure 4.4 shows the outlet point for Area 2 and Area 3.
64
Figure 4.4: Hand Delineation of Area 2
65
The areas were measured by first determining the planimeter constant.
The constant was determined by measuring a square area of 4 in
2
. Three readings
were taken that varied by only 0.002 units. These three measurements were
averaged, and a constant of 13.730 calculated. Following this procedure, a circle
of area 4 in
2
was measured three times and the average of these readings
calculated. For the circle, an average constant of 13.029 was calculated. The final
step in determining the overall planimeter constant was by averaging these two
constants. In this way, both straight lines and curved lines are taken into account
in the calculation of area. The final averaged planimeter constant was 13.380.
The perimeter and longest flow path length were determined using a map
wheel. Similar to measurements taken with the planimeter, each area was
measured three times, and the average measurement taken. Lastly, slope was
calculated by subtracting the elevation of the longest flow path at the outlet of the
subbasin from the elevation at the start of the longest flow path, divided by the
length of the longest flow path. Table 4.1 outlines the results of each calculation
for area, perimeter, longest flow path, and slope as derived by traditional methods.
Table 4.1: Hand Delineation Results
Watershed
Area
(mi
2
)
Perimeter
(mi)
Longest Flow Path
(mi) Slope (%)
Area 1 0.55 3.31 1.69 1.27
Area 2 1.65 6.17 3.14 1.47
Area 3 8.49 13.83 6.28 0.77
66
4.3 CASE II: AUTOMATED METHODS BY DIGITIZING
In Case II watershed boundaries are digitized using ArcView GIS 3.2 from
digital raster maps (DRGs) assuming undeveloped conditions. With the DRG
placed on the computer screen, new shapefiles are created using the digitizing
capabilities of ArcView GIS 3.2 and ArcWorkstation.
A USGS digital raster graphic (DRG) was used as a background in
ArcView GIS 3.2. The DRGs used were Austin West (o30097c6.tif and
o30097d6.tif), produced by the USGS in 1988.
The first step in digitizing is to create a new theme as selected from the
ArcView menu. As discussed in Chapter 3, watersheds are delineated in the same
fashion as with the paper maps, except with a computer-aided mouse. For each
area a polygon was used to represent area, and a polyline to represent the longest
flow path. Once the polyline was drawn (or polygon for area), ArcWorkstation
was used to convert the new themes to coverages using the shapearc command.
After the polylines and polygons were converted to coverages, they were
processed using the ArcWorkstation commands build and clean. Once they were
built and cleaned, they were opened once again in ArcView and converted back to
shapefiles.
During the process of digitization, it became apparent that some areas
were difficult to digitize because the contour lines could not easily be seen in
areas with highway overpasses. In an effort to incorporate parameter variations
that might be encountered due to highway construction, two delineations were
67
made for each study area--one incorporating infrastructure and one disregarding
the infrastructure.
Figure 4.5 below shows the digitization results for Area 1, both for
highways excluded and highways included. The largest difference occurs at the
upper end of Area 1, where it is very difficult to distinguish the contour lines.
Figure 4.5: Area 1 Digitized
The delineation of Area 2 yielded the following, as shown in Figure 4.6.
As Area 2 lies within Area 1, the digitization of the upper end of Area 2 replicates
the digitization process for Area 1.
68
Figure 4.6: Area 2 Digitized
The delineation of Area 3 is shown in Figure 4.7. The differences
between With Highways as opposed to Without Highways are hard to note in this
figure due to the large area that Area 3 encompasses.
Figure 4.7: Area 3 Digitized
Watershed parameters are calculated using the GIS (for area, perimeter
and longest flow path). From the new shapefile the parameters area and perimeter
69
can be calculated automatically. A longest flow path is also digitized from the
DRG and its length calculated by the GIS.
Table 4.2 outlines the results of Area 1, Area 2, and Area 3 from
digitization.
Table 4.2: On-Screen Digitizing Results
Watershed
No Hwy/
Hwy
Area
(mi
2
)
Perimeter
(mi)
Longest
Flow Path
(mi) Slope (%)
No Hwy 0.55 3.42 1.79 1.27 Area 1
Hwy 0.52 3.24 1.57 1.32
No Hwy 1.68 6.57 3.28 1.47 Area 2
Hwy 1.65 6.50 3.09 1.44
No Hwy 8.85 14.75 6.63 0.77 Area 3
Hwy 8.87 15.53 6.58 0.78
4.4 CASE III: AUTOMATED METHODS USING A DEM
Case III is divided into two parts. The first part describes using ArcView
GIS 3.2 and CRWR-PrePro to delineate the watersheds and extract watershed
parameters using two different spatial resolution DEMs. The second part explores
using the Watershed Modeling System (WMS) to also delineate the watersheds
and extract watershed parameters at the same two spatial resolutions as with
ArcView GIS 3.2.
The most widely available grid DEMs in the US are distributed by the
USGS and are of 7.5 minute resolution (same coverage as a standard USGS 7.5
minute map series quadrangle), with a grid spacing of 30 x 30 meters. These
70
USGS digital elevation models are produced from models based on aerial
photographs and satellite remote sensing images. The USGS 7.5 minute DEM
data are georeferenced using the Universal Transverse Mercator (UTM)
coordinate system. For this study, the DEMs used were developed by the USGS
and distributed by the Texas Natural Resources Information System (TNRIS).
The USGS is also producing a 10 meter DEM; however the regional availability
of these elevation models is less than that of the 30 meter DEMs.
4.4.1 ArcView GIS 3.2 and CRWR-PrePro
Case III explores using ArcView GIS 3.2 and CRWR-PrePro to delineate
the watersheds using two different spatial resolutions. As mentioned previously,
this part of the study entails using two different resolution grid DEMs: a 10 meter
DEM and a 30 meter DEM. All data are projected into the TCMS Albers Equal
Area georeferencing system. The Texas Geographic Information Council (TGIC)
defines a statewide mapping system for use in projected geospatial datasets that
cover Texas in order to facilitate overlay and integration of datasets. The Texas
Centric Mapping System/ Albers Equal Area or TCMS/AEA is defined below:
Mapping System Name: Texas Centric Mapping System/Albers Equal Area
Mapping System Abbreviation: TCMS/AEA
Projection: Albers Equal Area Conic
Longitude of Origin: 100 degrees West (-100)
Latitude of Origin: 18 degrees North (18)
71
Lower Standard Parallel: 27 degrees, 30 minutes (27.5)
Upper Standard Parallel: 35 degrees (35.0)
False Easting: 1,500,000 meters
False Northing: 6,000,000 meters
Datum: North American Datum of 1983 (NAD83)
The DEMs were clipped by first converting a box theme (as digitized in
ArcView around the study area) to a coverage using the ArcWorkstation
command shapearc box boxcov. Then the boxcov coverage was processed using
the ArcWorkstation build boxcov command. The cell size was set to that of the
DEM (setcell grid), and the window to that of the box coverage (setwindow
boxcov grid). The coverage was turned into a grid (temp = polygrid [boxcov]),
set to one (temp2 = temp / temp) and then clipped (clpgrd = temp2 * dem30m).
Once the DEM was clipped, it was projected into the TCMS Albers
projection. This was done in ArcWorkstation. Each DEM was processed as
outlined under CRWR-PrePro in Chapter 3.
The development of a curve number (CN) grid was necessary to estimate
the spatial variability of runoff from a rainfall event. After the DEMs were
clipped, the same was done with the USGS Land Use/Land Cover (LULC)
coverage and the USDA/NRCS STATSGO soils coverage for the same area.
Figure 4.8 shows the Land Use/Land Cover data clipped to the area of interest.
72
Figure 4.8: Land Use/Land Cover (LULC) Data
Figure 4.9 below shows the USDA/NRCS STATSGO soils data clipped to
the study area. Similar to the USGS LULC coverage, the STATSGO coverage is
a 1:250,000-scale map product. Both coverages are coarse compared to the
resolutions of the 10 and 30 meter DEMs.
73
Figure 4.9: STATSGO Data
The DEMs were clipped to a small area, and delineated separately for the
three areas. For each area, a threshold was chosen so that the areas of interest
would be delineated as separate units.
74
4.4.1.1 10 Meter DEM
As each 10 meter DEM covers an area of about 66 mi
2
, the DEMs were
first merged and clipped so that the study areas were covered completely by the
DEM.
For the large watershed an outlet was placed at the confluence with
Buttermilk. The steps taken were those described earlier for watershed
delineation using CRWR-PrePro. Table 4.3 below outlines these data, and the
thresholds chosen for each area.
Table 4.3: 10 Meter DEM Watershed Data
Watershed Threshold Cells in Watershed Area/Cell
Area 1 4500 13895 100
Area 2 22500 43504 100
Area 3 90000 228578 100
In a study conducted by Garbrecht et al. (1999b) on an agricultural
watershed, digital elevation models were applied to the study areas with cells
ranging from 5 to 500 meters. In this study the ratio of average subcatchment
area to grid cell area was used as an indicator of spatial resolution. An overall
catchment-to-grid ratio of 100 was found to be an acceptable threshold of spatial
resolution for reasonable model result. For this reason this parameter has been
incorporated into the analysis. Table 4.4 below outlines the physical parameters
determined from this study using a 10 meter DEM.
75
Table 4.4: 10 Meter DEM in ArcView Results
Watershed
Area
(mi
2
)
Perimeter
(mi)
Longest Flow
Path (mi) Slope (%)
Area 1 0.54 4.08 1.56 1.36
Area 2 1.68 8.38 3.21 1.43
Area 3 8.83 20.04 6.96 0.74
4.4.1.2 30 Meter DEM
The 30 meter DEMs were delineated in a similar fashion to the 10 meter
DEMs. The DEMs used for this study are Pflueastm (Pflugerville East),
Pflugwestm (Pflugerville West), Manorm (Manor), Jollyvillm (Jollyville),
Austinwestm (Austin West), and Austinestm (Austin East).
In the same process as done for the 10 meter DEM, LULC and STATSGO
data were prepared for the 30 meter DEM. The results for the three areas are
shown below. In both cases (10 meter and 30 meter DEM delineations) an outlet
was added to delineate the large watershed at the confluence with Buttermilk.
Figure 4.10 shows the three areas delineated separately, and the longest flow path
for each.
76
Figure 4.10: Delineated DEM Study Area
For the large watershed an outlet was placed at the confluence with
Buttermilk, as with the 10 meter DEM data. The steps taken were those described
earlier for watershed delineation using CRWR-PrePro. Table 4.5 outlines these
data, and the thresholds chosen for each area.
Table 4.5: 30 Meter DEM Watershed Data
Watershed Threshold Cells in Watershed Area/Cell
Area 1 500 1568 899
Area 2 2500 4835 900
Area 3 10000 25427 900
77
The following table, Table 4.6, outlines the DEM derived parameters from
this study using a 30 meter DEM.
Table 4.6: 30 Meter DEM in ArcView Results
Watershed
Area
(mi
2
)
Perimeter
(mi)
Longest Flow
Path (mi) Slope (%)
Area 1 0.54 4.06 1.50 1.26
Area 2 1.68 8.16 3.06 1.39
Area 3 8.84 19.08 6.78 0.76
4.4.2 Automated Methods Using WMS
The second part of the DEM-based analysis examines using WMS to
delineate both the 10 meter and 30 meter DEMs. Results from the WMS analysis
are expected to give slightly different results than with ArcView GIS 3.2 and
CRWR-PrePro, because different algorithms are used to extract the hydrologic
parameters.
4.4.2.1 10 Meter DEM
The 10 meter digital elevation data used for the WMS study are the same
as those used for the CRWR-PrePro study. The same grid that was clipped for the
CRWR-PrePro study was imported into WMS as a grid ascii file using the import
data command. Once this was done the DRG for the area was placed in the
background to locate the study area. Flow direction and flow accumulation grids
78
were computed using TOPAZ. Figure 4.11 shows the WMS delineating of Area
1, and its area value as 0.53 mi
2
.
Figure 4.11: Area 1 Delineated in WMS
Figure 4.12 shows the area of Area 2 to be 1.66 mi
2
. The white line is the
longest flow path.
79
Figure 4.12: Area 2 Delineated with WMS
Figure 4.13 shows the area of Area 3 to be 8.76 mi
2
.
Figure 4.13: Area 3 Delineated with WMS
80
Table 4.7 shows the results from the study using a 10 meter DEM to
extract watershed parameters in WMS.
Table 4.7: 10 Meter DEM in WMS Results
Watershed
Area
(mi
2
)
Perimeter
(mi)
Longest Flow
Path (mi)
Slope
(%)
Area 1 0.53 4.09 1.54 1.25
Area 2 1.66 8.53 3.15 1.41
Area 3 8.76 19.85 6.81 0.77
4.4.2.2 30 Meter DEM
The 30 meter digital elevation data used for the WMS study are also the
same as those used for the CRWR-PrePro study. The same grid that was clipped
nto WMS as a grid ascii file, using the
import
TOPAZ computed the flow direction and
flow accumulation.
Figure 4.14 shows the area of Area 1 to be 0.53 mi
2
.
for the CRWR-PrePro study was imported i
data command. Once this was done the DRG for the area was placed in
the background to locate the study area.
81
Figure 4.14: Area 1 Delineated with WMS
Figure 4.15 shows the area of Area 2 to be 1.67 mi
2
.
Figure 4.15: Area 2 Delineated with WMS
Figure 4.16 shows the area of Area 3 to be 8.84 mi
2
.
82
Figure 4.16: Area 3 Delineated with WMS
Table 4.8 shows the results from the study using a 30-meter DEM to
extract watershed parameters in WMS.
Table 4.8: 30 Meter DEM in WMS Results
Watershed
Area
(mi
2
) Perimeter (mi)
Longest Flow Path
(mi) Slope (%)
Area 1 0.53 3.76 1.38 1.36
Area 2 1.67 7.89 2.87 1.46
Area 3 8.84 18.72 6.66 0.78
83
4.5
MS for Area 2, and the watersheds are
delineated in WMS. The TIN is made using the conceptual model approach and
WMS has the ability to create and
delineate TINs. Once the TIN has been delin attributes may
.
ptual model appr involve
digitizing points from a DRG (the same DRG as used in Case II). This was done
y opening the TIN module in WMS, and selecting edit vertex points from Vertex
enu. Once the Vertex Options box is open, point elevations
are entered based on contour lines. Figure 4.17 shows the point elevations
entered using DRG contour lines as a reference of elevation.
CASE IV: AUTOMATED METHODS USING A TIN
In Case IV a TIN is created in W
interpolating points from a 30 meter DEM.
eated, geometric basin
be computed
The steps to create a TIN using the conce oach
b
Options in the TIN m
84
Figure 4.17: TIN Development in WMS
Points were added until the study area was densely populated with point
elevations. Breaklines were added to the model to reflect streams and other
terrain features as shown on the DRG. In this way, when the triangles are created
they are forced to conform to these lines. To add breaklines, the conceptual
model icon was selected and the TIN was turned off. A new coverage of type
drainage was created with attributes typed as generic to form the bounding
polygon around Area 2. Generic arcs were used to cut around the TIN. Under
Feature Objects, arcs were converted to polygons.
To create the streams, stream arcs (feature type is stream) were generated
to make the triangle edges conform to the streamlines as shown in Figure 4.18.
The feature objects were then edited to redistribute spacing of points every 75
meters (this number was chosen fairly arbitrarily).
85
Figure 4.18: Points, Breaklines, and Bounding Polygon in WMS
Once the triangles were turned on again, the polygon marking the
boundary around Area 2 was selected, and under feature objects the create TIN
command was selected. The TIN options selected for this study are linear
interpolation from DEM with a size bias of 0.5.
Once the TIN was conditioned and free of pits and flat triangles, the TIN
was used to delineate the drainage network. Figure 4.19 shows the TIN for Area
2 after it has been interpolated off the 30 meter DEM surface. Once the TIN is
ready to be delineated, the user must add an outlet so that the basins may be
defined.
86
Figure 4.19: Interpolated TIN for Area 2
Once the TIN is delineated, basin data may also be calculated by WMS for
input into a hydrologic model as shown in Figure 4.20.
87
Figure 4.20: Area 2 TIN Delineated
TINs may be modified very easily, and for this reason are useful data
sources for urban areas. This study was not replicated for Area 3, because after
the time it took to work with Area 2 it was deemed impractical to proceed to the
larger area. Table 4.9 shows the results for Area 2.
Table 4.9: TIN Results
Watershed
Area
(mi
2
)
Perimeter
(mi)
Longest Flow
Path (mi) Slope (%)
Area 2 1.68 7.4 3.14 1.53
88
Chapter 5: Results
5.1. CASE STUDY PARAMETER RESULTS SUMMARY
The results from the case study are evaluated for the following parameters:
slope, longest flow path, area, perimeter, and curve number. Figure 5.1 shows
the three drainage areas used in this study.
Figure 5.1: Three Study Areas
89
Methods used to extract the watershed parameters (as discussed in detail
in Chapter 4) include (1) measurement from paper maps; (2) on-screen extraction
from raster maps; (3) using GIS and two different resolutions of grid-based digital
elevation models (DEMs); and (4) using a triangulated irregular network (TIN).
Table 5.1, Table 5.2, and Table 5.3 summarize the results found for Area 1, Area
2, and Area 3. It should be noted that perimeter is not a hydrologic parameter. It
is evaluated in this study to examine trends in line computation among methods.
Table 5.1: Area 1 Results
Area 1
Area 1
Case Number
Area
Miles
2
Perimeter
Miles
LFP
Miles
Slope
%
Hand Case I 0.55 3.31 1.69 1.27
Digitized-no hwy Case II 0.55 3.42 1.79 1.27
Digitized-hwy Case II 0.52 3.24 1.57 1.32
PrePro-30 Case III 0.54 4.06 1.50 1.26
PrePro-10 Case III 0.54 4.08 1.56 1.36
WMS-30 Case III 0.53 3.76 1.38 1.36
WMS-10 Case III 0.53 4.09 1.54 1.25
Mean 0.54 3.71 1.58 1.30
Std.Dev. 0.01 0.38 0.13 0.05
CV(%) 2.07 10.28 8.42 3.65
90
Table 5.2: Area 2 Results
Area 2
Area 2
Case Number
Area
Miles
2
Perimeter
Miles
LFP
Miles
Slope
%
Hand Case I 1.65 6.17 3.14 1.47
Digitized-no hwy Case II 1.68 6.57 3.28 1.47
Digitized-hwy Case II 1.65 6.50 3.09 1.44
PrePro-30 Case III 1.68 8.16 3.06 1.39
PrePro-10 Case III 1.68 8.38 3.21 1.43
WMS-30 Case III 1.67 7.89 2.87 1.46
WMS-10 Case III 1.66 8.53 3.15 1.41
TIN Case IV 1.68 7.40 3.14 1.53
Mean 1.67 7.45 3.12 1.45
StdDev 0.01 0.93 0.12 0.04
CV(%) 0.81 12.48 3.90 2.97
Table 5.3: Area 3 Results
Area 3
Area 3
Case Number
Area
Miles
2
Perimeter
Miles
LFP
Miles
Slope
%
Hand Case I 8.49 13.83 6.28 0.77
Digitized-no hwy Case II 8.85 14.75 6.63 0.77
Digitized-hwy Case II 8.87 15.53 6.58 0.78
PrePro-30 Case III 8.84 19.08 6.78 0.76
PrePro-10 Case III 8.83 20.04 6.96 0.74
WMS-30 Case III 8.84 18.72 6.66 0.78
WMS-10 Case III 8.76 19.85 6.81 0.77
Mean 8.78 17.40 6.67 0.77
Std.Dev. 0.13 2.61 0.22 0.01
CV(%) 1.52 14.99 3.22 1.80
Table 5.1, Table 5.2, and Table 5.3 without further analysis provide
important information regarding variations of hydrologic parameters as a function
of area size and method of parameter extraction. Coefficient of variation (CV),
91
the standard deviation of each parameter divided by its mean, is used in addition
to the standard deviation to evaluate each of the watershed parameters. Assuming
that errors are normally distributed, in 68% of the cases the observed variation
will fall within one standard deviation above or below the mean, or ? CV
expressed as a percent of the mean.
The standard deviation for area and slope is small for all three areas. For
both Area 1 and Area 2, the drainage area can be determined within
approximately ? 0.01 square miles (2.07% for Area 1 and 0.81% for Area 2), and
? 0.13 square miles (1.52%) for Area 3. For Area 3, using traditional delineation
techniques on paper maps produces an area smaller than any of the other
techniques (0.29 square miles below the mean). This is most likely attributed to
area measurement using the planimeter.
Similar to area, the standard deviation (as well as the CV) for slope is
small for all three areas. The standard deviation for slope is highest for Area 1, at
a value of 0.05%. The highest coefficient of variation for slope, as calculated for
Area 1, is 3.65%.
Longest flow path and perimeter show different results from area and
slope, in that there is much more parameter variation among extraction methods.
Longest flow path was determined within ? 0.13 miles for Area 1, ? 0.12 miles
for Area 2, and ? 0.22 miles for Area 3. Precision in determining the longest flow
path increases as the area becomes larger, with a CV for Area 1 of 8.42% and a
CV for Area 3 of 3.22%. This is most likely because the flow path for Area 1 is
small, and hence there is more error in this measurement. Secondly, for Area 3 a
92
larger portion of the flow path was already marked with a blue line on paper maps
and on the DRG, whereas for Area 1 the blue line was not drawn on the map. It
should also be noted that longest flow path using traditional paper map-based
methods for Area 3 (6.28 miles) is 0.39 miles below the mean of 6.67 miles, well
below the measurements using other methods of longest flow path extraction.
Perimeter measurements become less precise as the drainage area
increases (the CV for Area 1 is 10.28%, while the CV for Area 3 is 14.99%). In
addition, the perimeter values found using traditional methods and on-screen
digitizing of raster graphic maps are smaller than those found using DEMs. Part
of the reason for this large variation in perimeter could be due to the subjective
nature of traditional paper map-based methods and on-screen digitizing of raster
graphic maps, in addition to the more tortuous path created by digital data.
Table 5.3 shows that automated methods using CRWR-PrePro, WMS, and
on-screen digitizing produce more consistent results than using manual hand
delineation techniques on paper maps to determine area, longest flow path and
perimeter for large areas. The values for area, longest flow path, and perimeter
determined using traditional paper map-based methods for Area 3 are much
smaller than those determined using more automated methods. Figure 5.2
outlines the CV results from Table 5.1, Table 5.2, and Table 5.3.
93
Figure 5.2: Coefficients of Variation
In Figure 5.2, SCS curve number is shown to give the reader an idea of the
curve number variation that will be applied to the remainder of the study. For
Area 2, a range of curve numbers was evaluated using a curve number selected by
TxDOT as a low value (85.8), and a curve number derived by STATSGO soils
data along with Land Use/Land Cover data as a high value (91.7). TxDOT did
not provide curve number values for Area 1 and Area 3. Instead, a range the same
as used in Area 2 was applied, with the curve number derived by STATSGO soils
data along with Land Use/Land Cover data as a high value. These values, shown
in Table 5.4, were used to reflect reasonable variations of curve number that one
might encounter in hydrologic modeling.
94
Table 5.4: Curve Number Values
Area 1 Area 2 Area 3
CN- 10m 94.0 91.7 90.7
CN- 30m 93.9 91.7 90.7
CN-HEC-1 - 85.82 -
Range 87-95 85-93 84-92
Mean 91 89 88
StdDev 2.74 2.74 2.74
CV (%) 3.01 3.08 3.11
Figure 5.2 shows that for the areas analyzed, there is a general trend for
both longest flow path (LFP) and slope. The trend in variations in slope reflects
the trend in variations in longest flow path, as it is calculated from the longest
flow path measurement. Both show higher variations with Area 1 and lower
variations for Area 3. Another key point shown by Figure 5.2 is that for the small
area (Area 1), there is more variation among parameters than for the larger area,
Area 3, except in the case of perimeter and curve number.
5.2 ELASTICITY ANALYSIS
In this study ?elasticity? is used as a measure of the influence of one
variable on another. While gradient describes the relationship between two
variables, it is a unit-dependent quantity. Elasticity, on the other hand, is a
dimensionless quantity. In simple terms, elasticity is the percent change in output
per percent change in input. If the elasticity is greater than one, the parameter is
?elastic? in that the dependent variable is more sensitive to the independent
95
variable. If the elasticity is less than one, the parameter is ?inelastic? in that the
dependent variable is less sensitive to the independent variable. Figure 5.3 below
shows the concept of elasticity graphically:
dx
dy
gradient =
Y*
dy
input
output
change
change
Figure 5.3: Concept of Elasticity
The goal of this analysis is to determine the effect of parameter variations
on the lag time, and subsequently the effect of lag time variations due to
parameter variations on discharge. This analysis is conducted only for Area 2.
The reason for this is that Area 2 is the only study area for which TxDOT has
created a hydrologic model. After an analytical calculation of gradients to
determine lag time elasticity for each parameter, the hydrologic model created by
TxDOT is used to numerically calculate the elasticity of discharge with lag time
as shown in Figure 5.4.
X*
X
dx
Y
dy
elasticity
%
%
*
*
==
dx
96
Figure 5.4: Discharge Sensitivity Calculation
Equation 5.1, Step 1 in calculating the discharge sensitivity, describes the
calculation of lag time elasticity. The first term in this equation is the gradient.
The gradient is multiplied by the average parameter value for Area 2, and divided
by the lag time calculated at the average parameter.
)(
*
)(
)(
avg
avg
parameterlag
parameterlag
parameter
parameter
lag
elasticity
?
?
=
?
Equation 5.1
Equation 5.2, Step 2 in calculating the discharge sensitivity, describes the
calculation of discharge elasticity based on changes in lag values. Similar to
Equation 5.1 the gradient is multiplied by the average lag time value and divided
by the discharge calculated at the average lag time.
)(
*
)(
)(
avg
avg
lagQ
lagQ
lag
lag
Q
elasticity
?
?
=
?
Equation 5.2
97
The final step approximates the overall discharge sensitivity as a result of
variation in parameter values. Equation 5.3 outlines this calculation. The results
from Equation 5.1 describing the lag time elasticity with respect to each
parameter are multiplied by the discharge elasticity with respect to lag time as
calculated by Equation 5.2.
?
?
?
?
?
?
?
?
=
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=
?
)(
*
)(
)(
)(
*
)(
)(
*
)(
*
)(
)(
parameterQ
parameter
parameter
Q
parameterlag
parameter
parameter
lag
lagQ
lag
lag
Q
elasticity
parameterQ
Equation 5.3
5.2.1 Step 1: Sensitivity of SCS Lag Time to Parameter Variations
Figure 5.5 describes the process analyzed in this section. Variations in
parameter values are used to calculate resulting variations in SCS lag times.
Figure 5.5: Step 1 in Calculating the Discharge Sensitivity
98
Analytical calculation of the gradient involves plugging in each of the
parameters derived from each level of parameter extraction into the SCS lag
equation, shown in Equation 5.4, while holding the other parameters constant.
s
CN
L
t
w
lag
67.31
9
1000
(min)
7.0
8.0
?
?
?
?
?
?
?
= Equation 5.4
In Equation 5.4, L
w
is longest flow path in units of feet, and slope is
expressed as a percentage. Curve number is symbolized by CN, and t
lag
is the lag
time in minutes. Table 5.5 outlines the base values used for these calculations.
These numbers are the mean values calculated for Area 2, as shown previously in
Table 5.2 and Table 5.4.
Table 5.5: Parameter Base Values
Parameter Base
Value
LFP 3.12 Miles
Slope 1.45 %
CN 89
5.2.1.1 Longest Flow Path
To evaluate changes in SCS lag time that result from changes in the
longest flow path, the curve number and average slope were held constant while
99
longest flow path was allowed to vary. Table 5.6 gives the lag time values
calculated for each longest flow path measurement.
Table 5.6: Longest Flow Path Calculations
LFP
Miles
% Slope
(mean)
CN
(mean)
SCS
Lag, Min
1.45 89.00
Hand 3.14 109.8
Digitized-no hwy 3.28 113.7
Digitized-hwy 3.09 108.4
PrePro-30 3.06 107.5
PrePro-10 3.21 111.7
WMS-30 2.87 102.2
WMS-10 3.15 110.1
TIN 3.14 109.8
MEAN 3.12 MEAN 109.0
?X
max-min
/X
mean
13.2% ?Y
max-min
/Y
mean
10.6%
Subtracting the highest longest flow path measurement from the lowest
longest flow path measurement, and then dividing by the mean flow path length
determines the longest flow path variation. Figure 5.6 below demonstrates this
process for the calculation of longest flow path.
100
Figure 5.6: Analytical Calculation of Gradients
As the longest flow path varies by 13.2%, the lag varies by 10.6% when
evaluated at the base value for longest flow path (3.12 miles). The elasticity is
then 10.6 ? 13.2 or 0.80. As this number is less than one, the relationship
between longest flow path and lag time is considered inelastic, signifying that lag
time is not sensitive to changes in longest flow path length. These results show
that a 1% increase in longest flow path will cause a 0.8% increase in SCS lag
time.
101
5.2.1.2 Slope
The lag time elasticity with respect to slope was calculated by holding the
longest flow path and curve number constant and varying the slope. The results
from this calculation are shown in Table 5.7.
Table 5.7: Slope Calculations
%
Slope
LFP
(mean),Miles
CN
(mean)
SCS
Lag, Min.
3.12 89.00
Hand 1.47 107.8
Digitized-no hwy 1.47 107.8
Digitized-hwy 1.44 108.9
PrePro-30 1.39 110.9
PrePro-10 1.43 109.3
WMS-30 1.46 108.2
WMS-10 1.41 110.1
TIN 1.53 105.7
MEAN 1.45 MEAN 108.6
?X
max-min
/X
mean
9.7% ?Y
max-min
/Y
mean
-4.8%
As can be seen from Figure 5.7, there is an inverse relationship between
lag time and slope.
102
Figure 5.7: Slope Gradient Calculation
The elasticity of lag time with respect to slope is calculated as ?4.8/9.7, or
?0.50. In other words, a 1 % increase in slope will cause a 0.5% decrease in SCS
lag time.
5.2.1.3 Curve Number
The lag time elasticity with respect to curve number was determined for
Area 2 by holding slope and longest flow path constant, and varying the curve
number. Curve numbers for Area 2 are shown in Table 5.8. These are the same
values shown earlier in Table 5.4.
103
Table 5.8: Curve Number Measurements
10m 30m HEC-1
CN range 91.7 91.7 85.8
Table 5.9 outlines the curve number values. As mentioned earlier, this
range is used to reflect a range of curve numbers that could likely result using
traditional methods and derived mathematically using digital Land Use/ Land
Cover and STATSGO soils data.
Table 5.9: Curve Number Calculations
CN
%Slope
(mean)
LFP
(mean), Miles
SCS
Lag, Min.
1.45 3.12
85 126.4
86 122.0
87 117.6
88 113.3
89 108.9
90 104.6
91 100.4
92 96.1
93 91.9
MEAN 89 MEAN 109.0
?X
max-min
/X
mean
9.0% ?Y
max-min
/Y
mean
-31.7%
This curve is plotted in Figure 5.8. The graph shows the inverse
relationship between curve number and lag time.
104
Figure 5.8: Curve Number Gradient Calculation
The elasticity of the lag time to the curve number, -31.7/9.0, is ?3.52.
These results show that a 1% increase in CN will result in a 3.52% decrease in lag
time.
5.2.1.4 Step 1 Results
Table 5.10 summarizes the results from the lag time elasticity study.
Results show that the only elastic parameter is curve number. Both the longest
flow path and slope are inelastic parameters.
105
Table 5.10: Results from Analytical Calculations
Variation-
LFP
Variation-
Lag
Elasticity
Lag-LFP
13.2% 10.6% 10.6/13.2=0.80
Variation-
Slope
Variation-
Lag
Elasticity
Lag-Slope
9.7% -4.8% -2.8/5.6=-0.50
Variation-
CN
Variation-
Lag
Elasticity
Lag-CN
9.0% -31.7% -31.7/9.0=-3.52
5.2.2 Step 2: Sensitivity of Flow to Variations in Lag Time
Now that the relationships between physical parameters and SCS lag time
have been quantified, the next step is to quantify the relationship between SCS lag
time and discharge. Figure 5.9 below shows the process analyzed in this section
in which variations in SCS lag time values are used to calculate resulting
variations in discharge.
106
Figure 5.9: Step 2 in Calculating the Discharge Sensitivity
In 1992, a contractor for TxDOT developed a HEC-1 model of Buttermilk
watershed (Area 2) as an undeveloped area. This pre-developed HEC-1 model
included areas outside of Buttermilk watershed. These areas were edited, and the
new, slightly modified HEC-1 model was imported into HEC-HMS.
The U.S. Corps of Engineers? Hydrologic Modeling System (HMS) is a
?next generation? software for the precipitation-runoff simulation that will
supersede HEC-1. HEC-HMS contains most of the watershed runoff and routing
capabilities of HEC-1, in addition to continuous hydrograph simulation over long
periods of time, and distributed runoff computation using a grid cell depiction of
the watershed (HEC-HMS User?s Manual, 2000).
There are two principal parameters that are evaluated in this study with
respect to discharge: lag time and drainage area. After modification of the HEC-
1 model, lag times are varied in each sub-basin. For each lag time variation,
107
discharge is calculated. Next, the drainage area of each sub-basin is varied, and
discharge calculated once again for each area variation.
5.2.2.1 HEC-1 Model
Figure 5.10 shows the area delineated by TxDOT for the HEC-1 model.
The area in blue is included in the HEC-1 model although it does not pertain to
Buttermilk (Area 2). For this analysis, the blue area is deleted from the HEC-1
model.
Figure 5.10: HEC-1 Study Area
The corresponding unmodified HEC-HMS diagram (the HEC-1 model
was imported into HEC-HMS) is shown below in Figure 5.11.
108
Figure 5.11: Unmodified HEC-1 model (displayed in HEC-HMS)
As mentioned previously, the HEC-1 model had to be modified for this
analysis. Table 5.11 below gives the name of each sub-basin along with its
drainage area, SCS lag time, and SCS curve number. In the modification process,
109
the area downstream of COMB14 was deleted. This area includes 183105, LW14
and 183106 as shown in red italicized print.
Table 5.11: HEC-1 Modifications
Sub-basin Area mi
2
SCS Lag (min) SCS Curve
Number
Tb711 0.444 15.30 89
Tb710 0.285 17.88 88
Tb709 0.052 7.74 80
Tb708 0.241 15.72 86
Tb707 0.042 9.36 87
Tb706 0.121 16.02 84
Tb1003 0.150 6.66 77
183101 0.011 - 85
Tb1001 0.025 11.1 85
183103 0.004 - 83
Tb705 0.102 12.3 83
Tb704 0.046 7.74 87
Tb702 0.097 9.72 86
183104 0.004 - 85
Tb703 0.052 8.22 85
Tb701 0.088 8.40 86
183105 0.001 - 86
Lw14 0.127 7.5 80
183106 0.006 - 80
Total Area 1.898
The new area, after modification, is 1.764 mi
2
. The modified model is
shown in Figure 5.12. The weighted curve number after modification is 85.82.
110
Figure 5.12: Modified HEC-1 Model (displayed in HEC-HMS)
For rainfall-runoff modeling HEC-HMS requires a basin component, a
precipitation component, and a control component. The basin and precipitation
components are heavily dependent on spatial factors.
The HEC-1 model is imported as the basin component. For the
precipitation component, six different return periods are analyzed using the
111
TxDOT IDF (Intensity-Duration-Frequency) curves and an alternating block
cumulative precipitation hyetograph for Travis county and a three-hour design
storm. Equation 5.4 and Table 5.12 are used to calculate the cumulative rainfall
for a three-hour storm in six-minute increments for each sub-basin. A six-minute
increment is chosen because this is the value TxDOT used.
()
c
d
bT
a
I
+
= Equation 5.4
Table 5.12: TxDOT IDF (Intensity-Duration-Frequency) Curves
Return
Period
2 Years 5 Years 10 Years 25 Years 50 Years 100 Years
a 56 69 77 87 91 103
b 8.1 8.6 8.6 8.6 8.6 8.1
c 0.796 0.780 0.775 0.766 0.751 0.752
5.2.2.2 Lag Time Variation
The next step was to determine the discharge elasticity from changes in
lag time. This was done by proportionally changing the lag time values for each
of the sub-basins in the HEC-1 model. Table 5.13 gives the original SCS lag
value for each sub-basin, followed by 5%, 10%, and 20% increases in lag time,
then 5%, 10%, and 20% decreases in lag time.
112
Table 5.13: HEC-1 Lag Input
Subbasin Lag, min Lag +5%
Lag +10% Lag +20% Lag -5% Lag -10% Lag -20%
TB711 15.30 16.07 16.83 18.36 14.54 13.77 12.24
TB710 17.88 18.77 19.67 21.46 16.99 16.09 14.30
TB709 7.74 8.13 8.51 9.29 7.35 6.97 6.19
TB708 15.72 16.51 17.29 18.86 14.93 14.15 12.58
TB707 9.36 9.83 10.30 11.23 8.89 8.42 7.49
TB706 16.02 16.82 17.62 19.22 15.22 14.42 12.82
TB1003 6.66 6.99 7.33 7.99 6.33 5.99 5.33
TB1001 11.10 11.66 12.21 13.32 10.55 9.99 8.88
TB705 12.30 12.92 13.53 14.76 11.69 11.07 9.84
TB704 7.74 8.13 8.51 9.29 7.35 6.97 6.19
TB702 9.72 10.21 10.69 11.66 9.23 8.75 7.78
TB703 8.22 8.63 9.04 9.86 7.81 7.40 6.58
TB701 8.40 8.82 9.24 10.08 7.98 7.56 6.72
Figure 5.13 shows the discharge values in cubic feet per second (cfs) for
each return period. The relationship between discharge (Y-axis) and lag time (X-
axis) is given by the linear equations shown in Figure 5.12 for 2, 5, 10, 25, 50,
and 100-year events.
113
Figure 5.13: Discharge Results for Lag Variations
To determine the percent change in discharge resulting from a percent
change in lag time (due to parameter variations), the discharge (Q) for each
corresponding lag variation is calculated using the linear equations relating lag
time to discharge. Table 5.14 shows the percent changes in discharge for each
return period corresponding to each variation in SCS lag time.
Table 5.14: Discharge Variations Due to Lag Variations
Lag (From LFP Variations) Q
2
Q
5
Q
10
Q
25
Q
50
Q
100
10.6% -3.04% -2.87% -2.27% -3.14% -3.18% -3.17%
Lag (From Slope Variations) Q
2
Q
5
Q
10
Q
25
Q
50
Q
100
-4.8% 1.38% 1.30% 1.03% 1.42% 1.44% 1.44%
Lag (From CN Variations) Q
2
Q
5
Q
10
Q
25
Q
50
Q
100
-31.7% 9.09% 8.58% 6.79% 9.38% 9.50% 9.48%
114
For each return period, the elasticity of discharge with respect to SCS lag
time is calculated by dividing the percent change in discharge by the percent
change in SCS lag time. As shown in Table 5.15, there are slight variations in
discharge elasticity for each return period. These variations were averaged to
establish a relationship between SCS lag time and discharge. The averaged value
is ?0.28.
Table 5.15: Discharge Elasticity Due to Parameter Variation
Elasticity
Q
2
-Lag
Elasticity
Q
5
-Lag
Elasticity
Q
10
-Lag
Elasticity
Q
25
-Lag
Elasticity
Q
50
-Lag
Elasticity
Q
100
-Lag
Elasticity
Q
avg
-Lag
-0.29 -0.27 -0.21 -0.30 -0.30 -0.30 -0.28
Results from this analysis show that a 1% increase in SCS lag time yields
a -0.28% decrease in discharge.
5.2.2.3 Drainage Area Variation
The second component involving analysis of the HEC-1 model is to
evaluate changes in area, and how these changes affect discharge. Similar to the
lag time study, the areas of each sub-basin were increased by 5%, 10%, and 20%
and also decreased by 5%, 10%, and 20%. Table 5.16 shows these values.
115
Table 5.16: HEC-1 Area Input
Sub-basin
Area
Miles
2
Area +5% Area +10% Area +20% Area -5% Area -10% Area -20%
TB711 0.444 0.466 0.488 0.533 0.422 0.400 0.355
TB710 0.285 0.299 0.314 0.342 0.271 0.257 0.228
TB709 0.052 0.055 0.057 0.062 0.049 0.047 0.042
TB708 0.241 0.253 0.265 0.289 0.229 0.217 0.193
TB707 0.042 0.044 0.046 0.050 0.040 0.038 0.034
TB706 0.121 0.127 0.133 0.145 0.115 0.109 0.097
TB1003 0.150 0.158 0.165 0.180 0.143 0.135 0.120
TB1001 0.025 0.026 0.028 0.030 0.024 0.023 0.020
183101 0.011 0.012 0.012 0.013 0.010 0.010 0.009
TB705 0.102 0.107 0.112 0.122 0.097 0.092 0.082
183103 0.004 0.004 0.004 0.005 0.004 0.004 0.003
TB704 0.046 0.048 0.051 0.055 0.044 0.041 0.037
TB702 0.097 0.102 0.107 0.116 0.092 0.087 0.078
183104 0.004 0.004 0.004 0.005 0.004 0.004 0.003
TB703 0.052 0.055 0.057 0.062 0.049 0.047 0.042
TB701 0.088 0.092 0.097 0.106 0.084 0.079 0.070
Figure 5.14 shows the discharge values in cubic feet per second (cfs) for
each return period. The relationship between discharge (Y-axis) and area (X-axis)
is given by the linear equations, also shown in Figure 5.14 below, for 2, 5, 10, 25,
50, and 100-year events.
116
Figure 5.14: Discharge Results for Drainage Area Variations
Table 5.17 gives the variation in area for Area 2 (Buttermilk). The
variation in drainage area among methods is 1.80%.
Table 5.17: Area Variation
Area, Miles
2
Hand 1.65
Digitized-no hwy 1.68
Digitized-hwy 1.65
PrePro-30 1.68
PrePro-10 1.68
WMS-30 1.67
WMS-10 1.66
TIN 1.68
MEAN 1.67
?X
max-min
/X
mean
1.8%
117
The same steps were taken to determine the effect of drainage area on
discharge as were taken to determine the effects of lag time on discharge. For
each return period, the percent variation in discharge was calculated based on the
percent variation in drainage area (1.80%). There were slight variations in
percent discharge for each return period. Discharge elasticity for each return
period was calculated by dividing each discharge variation by 1.80% (the
variation in drainage area). There were slight differences in elasticity for each
return period. These numbers were averaged to provide an approximate discharge
elasticity, as shown in Table 5.18.
Table 5.18: Discharge Elasticity due to Area Variations
Q
2
Q
5
Q
10
Q
25
Q
50
Q
100
2.05% 1.74% 1.60% 1.96% 2.09% 1.99%
Elasticity
Q
avg
-Area
Elasticity
Q
2
-Area
Elasticity
Q
5
-Area
Elasticity
Q
10
-Area
Elasticity
Q
25
- Area
Elasticity
Q
50
- Area
Elasticity
Q
100
- Area
1.14 0.97 0.89 1.09 1.16 1.11 1.91/1.80=1.06
The resulting elasticity of discharge with respect to area is 1.06. Hence, a
1% increase in area yields approximately a 1% increase in discharge, as one
would expect.
5.2.2.4 Step 2 Results
Results from the numerical calculations of gradient show that there is a
constant, inversely proportional, relationship between lag time and discharge.
118
This elasticity value, -0.28, shows the relatively inelastic effect of lag time
changes on discharge, e.g., a 1% increase in lag time results in a 0.28% decrease
in discharge. Area, as one would expect, is purely elastic and directly
proportional to discharge. In this case a 1% increase in area would result in a 1%
increase in discharge.
5.2.3 Results from Elasticity Analysis
The effect of parameter variation on discharge is determined by
multiplying elasticity results from Step 1 (sensitivity of SCS lag time to parameter
variations) by Step 2 (sensitivity of flow to variations in discharge). The overall
picture is shown once again in Figure 5.15.
Figure 5.15: Final Discharge Sensitivity Calculation
The effect of each parameter (longest flow path, slope, and curve number)
on discharge was evaluated using the elasticity for both discharge (due to
119
variations in lag) and lag (due to variations in each parameter). These two
elasticity values are multiplied together. Figure 5.16 outlines the results from the
elasticity study.
Figure 5.16: Elasticity Diagram
The following relationships can be drawn from Figure 5.16:
? A 1% increase in longest flow path length decreases peak
discharge by 0.22%;
? A 1% increase in slope increases peak discharge by 0.14%;
? A 1% increase in curve number increases flow by 1%;
? A 1% increase in drainage area increases flow by 1%.
LFP
Slope
CN
Area
Discharge
E=0.80
E=-0.50
E=-3.52
E=-0.28
E=1.06
SCS Lag Time
E=-3.52*-0.28=0.99
E=-0.50*-0.28=0.14
E=0.80*-0.28=-0.22
Parameter
Lag
Change
Change
%
%
Parameter
Q
Change
Change
%
%
Lag
Q
Change
Change
%
%
120
Both curve number and drainage area have a direct effect on discharge,
while longest flow path length and slope have minimal influence on discharge.
Figure 5.17 shows the relative discharge elasticities of longest flow path, slope,
curve number, and drainage area.
Figure 5.17: Elasticity Results
An analysis of coefficient of variation (CV) values in Table 5.19, as
derived for each parameter in Table 5.1, Table 5.2, and Table 5.3, shows that the
error associated with determining drainage area is small compared to the error
associated with determining longest flow path, slope, and curve number.
121
Table 5.19: Parameter Coefficients of Variation (CV)
Study Area CV (%) Area CV (%) LFP CV (%) Slope CV (%) CN
Area 1 2.07 8.42 3.65 3.01
Area 2 0.81 3.90 2.97 3.08
Area 3 1.52 3.22 1.80 3.11
Drainage area CV values are the smallest of the parameter CV values for
Area 1, Area 2, and Area 3. Although the relationship between area and
discharge is elastic, the error associated with determining drainage area is small.
On the other hand, there is a larger error associated with determining longest flow
path and slope. However, since longest flow path and slope are inelastic
parameters, large errors in determining these values will cause minimal percent
changes in discharge.
Table 5.20 shows the percent discharge variations resulting from
parameter variations for Area 1, Area 2, and Area 3.
Table 5.20: Discharge Variations for Area 3
Parameter Discharge CV%
(Area 1)
Discharge CV%
(Area 2)
Discharge CV%
(Area 3)
Area 2.07*1.06=2.19% 0.81*1.06=0.86% 1.52*1.06=1.61%
LFP 8.42*-0.22=-1.85% 3.90*-0.22=-0.86% 3.22*-0.22=-0.71%
Slope 3.65*0.14=0.51% 2.97*0.14=0.42% 1.80*0.14=0.25%
CN 3.01*0.99=3.00% 3.08*0.99=3.05% 3.11*0.99=3.08%
122
The CV for discharge is calculated by multiplying the CV for each
parameter by the discharge elasticity with respect to that parameter. Table 5.20
shows that curve number will most influence discharge after taking into
consideration the error that occurs as a result of variations in extracting
hydrologic parameters.
123
Chapter 6: Conclusions and Recommendations
This implementation project was conducted to provide guidance to
TxDOT regarding parameter sensitivity in hydrologic modeling. The principal
factors affecting flood magnitudes in a watershed include runoff, and watershed
area information such as slope, flow path length, area, land use, and soil type.
These parameters are all important in determining the peak discharge at a
watershed outlet such as a culvert or bridge crossing. Currently TxDOT
predominantly uses traditional methods of hand delineation to extract hydrologic
parameters, although TxDOT is moving towards using more automated methods
for this process.
The use of GIS in water resources engineering has proved to be aquick
and relatively simple means of computing hydrologic data. In 2000, however,
Anderson (2000) conducted a digital floodplain analysis for TxDOT which led to
doubts about the automated process. Anderson?s watershed lag time values were
significantly greater than values found in an earlier analysis conducted by
TxDOT. In an effort to determine possible causes of this error, four case studies
were developed to analyze three areas of different size: (1) measurement from
paper maps; (2) on-screen extraction from raster maps; (3) using GIS and two
different resolutions of grid-based digital elevation models (DEMs); and (4) using
a triangulated irregular network (TIN). This section of the report discusses
conclusions and recommendations as drawn from the implementation study.
124
6.1 ERROR SOURCES AMONG PARAMETER EXTRACTION METHODS
This study examines parameter extraction methods in three drainage areas
of different size in Austin, Texas. Area 1, the smallest area, is approximately 0.5
mi
2
. Area 2, the medium-sized area, is approximately 1.7 m
2
; and Area 3, the
largest area, is approximately 8.8 mi
2
. Errors associated with each of the four
levels of parameter extraction, in addition to the influence of drainage area size on
parameter extraction method, are evaluated.
6.1.1 Parameter Extraction using Paper Maps
Traditional hydrologic modeling involves calculating watershed
parameters from paper maps. Calculation of terrain-based hydrologic data
involves delineating the watershed by hand using map contours as guidelines.
Once the perimeter of the watershed has been established, a planimeter is used to
measure its area. Perimeter and length of the longest flow path are measured
using a map wheel. Slope is calculated by taking the difference in elevation
between map contours.
Variations in traditional, paper map-based methods were most apparent in
Area 3. This drainage area, the largest area under investigation, yielded hand
measurements for area and longest flow path, 0.29 mi
2
and 0.39 miles below the
mean, respectively. These area and longest flow path values were smaller than
any of the other area or longest flow path measurements for Area 3. The length of
perimeter, although not normally used as a hydrologic parameter, was 3.57 miles
125
below the mean using traditional methods of parameter extraction. For Area 1
and Area 2 the differences between paper map methods and digital automated
methods were not as apparent.
The reason for the large deviation from the mean using traditional
methods in area, longest flow path, and perimeter measurements is most likely
due to the error associated with 1) taping two topographic maps together to make
a map large enough to cover the whole drainage area, 2) difficulty in determining
flow path and drainage divides from the map contours and 3) the accuracy of
manual measurement techniques using a map wheel and a planimeter.
6.1.2 On-Screen Digitizing of Raster Graphic Maps
The process of on-screen digitizing of raster graphic maps closely
resembles paper map methods. The methodology is analogous, except that the
process of on-screen digitizing of raster graphic maps involves using a computer-
aided mouse and a scanned topographic map to draw in watershed boundaries and
flow paths. Once the watershed boundaries and flow paths have been determined,
a GIS (or similar system) computes drainage area, longest flow path length, and
perimeter. Slope is calculated by taking the difference in elevation between map
contours, as done with paper maps.
On-screen digitizing eliminates the error of taping maps together, and
interpreting lengths and areas from hand-held instruments. As with paper maps,
determining flow paths and drainage divides is a very time-consuming and
subjective process. On-screen digitizing also has the advantage over paper
126
map-based methods in that map features can be ?zoomed-in? or magnified to
better understand flow paths. In this study each area was digitized two ways: by
attempting to incorporate the effect of urban infrastructure such as major
highways, and by disregarding this infrastructure. The on-screen digitizing results
show that area and longest flow path measurements for Area 3 closely resemble
results derived from automated processes using DEMs, more so than results
derived using paper map-based methods. This result indicates that errors
associated with physically measuring parameters with a map wheel and
planimeter largely contribute to the variation in area and longest flow path values
for Area 3. For on-screen digitizing of larger areas, the measurement error due to
the subjective nature of the delineation process is small when compared to the
error due to hand-held measurement techniques applied to paper maps.
6.1.3 Parameter Extraction Using Automated Methods
DEM analysis in WMS (Watershed Modeling System) and CRWR-PrePro
presents errors associated with data resolution and accuracy in describing the
physical characteristics of a surface. DEM resolution affects the channel length,
area, slope, and perimeter measurements. If the DEM resolution is too low, small
changes in terrain are not observed and small areas may not accurately be
described. If the resolution of the DEM is very high, channel lengths become
larger due to more tortuous flow paths. This study showed that flow paths for the
10 meter DEM were longer than flow paths for the 30 meter DEM for all three
areas as determined in both WMS and CRWR-PrePro. For Area 1 there was a 4%
127
increase in longest flow path using CRWR-PrePro, and a 12% increase using
WMS. For the large area, Area 3, there was a 3% increase in longest flow path
using CRWR-PrePro and a 2% increase using WMS. Regardless, the longest
flow path standard deviation for Area 1 is 0.13 miles (with a mean of 1.58 miles),
and the longest flow path standard deviation for Area 3 is 0.22 miles (with a mean
of 6.67 miles). Variations are small, and results show that these variations do not
significantly influence lag time and discharge calculations. Distributed properties
such as slope and longest flow path require a method or model to reduce the
distributed information into a representative value for the entire subcatchment.
Variations in subcatchment slope and longest flow path values between WMS and
CRWR-PrePro may be attributed to differences in underlying models used for
extracting data.
6.1.4 Parameter Extraction Using TINs
TINs, triangulated irregular networks, consist of a set of vertex points
connected by triangles, that represent scattered X, Y, and Z locations. TINs have
many advantages over DEMs, in that TINs can describe a surface precisely and
are adaptive to different types of terrain. The major disadvantage found working
with TINs in this project is the large amount of time that is required to create, edit,
and condition the TIN. For this reason, TIN development was not conducted for
Area 3. Parameter values for area, perimeter, and longest flow path did not vary
significantly for the TIN when compared to the DEM methods and on-screen
digitizing. Slope measurement using a TIN, however, produced a greater value
128
than any of the other methods implemented. Probably this is the most accurate
slope value since the DEM methods tend to smooth out slopes.
6.2 ERRORS AMONG EXTRACTED HYDROLOGIC PARAMETERS
The parameter variation associated with each of the four levels of
parameter extraction is evaluated based on a simple statistical analysis using
standard deviation and coefficient of variation (CV). The coefficient of variation
(CV) is defined as the standard deviation divided by the mean for a group of
measurements.
6.2.1 Drainage Area
The results for drainage area showed little variation among the different
methods of parameter extraction. Results show that drainage area can be
approximated within ? 0.01 mi
2
for areas less than approximately 1.6 mi
2
, and
within ? 0.13 mi
2
for areas approximately 8.8 mi
2
. Area measurement, whether
obtained by using automated or traditional methods, will most likely not result in
a large error. This study shows that the maximum coefficient of variation (CV)
for area, determined by Area 1, is 2.07%.
6.2.2 Slope
Slope, similar to area, produced fairly consistent measurements for all
three study areas. For the largest area, Area 3, slope could be approximated
within ? 0.01 as percent slope (a CV of 1.80%). For the small area, Area 1, slope
129
could be approximated within ? 0.05 as percent slope (a CV of 3.65%). The
mean slope for Area 3 was 0.77%, while the mean slope for Area 1 was 1.30%.
6.2.3 Longest Flow Path
Longest flow path results produced errors greater than slope or area errors
for all three study areas. However, this error decreased from 8.42% to 3.22% as
the study area size increased from Area 1 to Area 3. For the largest area, longest
flow path was calculated within ? 0.22 miles, with a mean value of 6.67 miles.
For the smallest area, longest flow path was calculated within ? 0.13 miles, with a
mean value of 1.58 miles.
6.2.4 Perimeter
Watershed perimeter length, although not a hydrologic parameter, was
examined to evaluate trends in line variation among methods. Automated
methods using DEMs produced greater values for perimeter than traditional
methods for all three study areas. This resulted in consistently high CV values
for perimeter. For Area 1, the CV value was 10.28%, and for Area 3, the CV
value was 14.99%.
6.2.5 Curve Number
The SCS runoff curve number obtained by combining land use and soil
types in GIS for Area 2 was 91.7, while the curve number determined by an
independent TxDOT study was 85.8. A range of curve number values, using the
130
curve number developed by automated methods as an approximate upper limit,
and the curve number developed by TxDOT as an approximate lower limit, gave a
coefficient of variation of 3.08%, and a mean value of 89. This number is used to
reflect a reasonable error one might encounter in determining SCS curve number.
6.3 SIGNIFICANCE OF ERRORS
Following hydrologic parameter extraction, ranging from traditional
methods using paper maps to advanced methods using TINs, a sensitivity analysis
based on the concept of elasticity was conducted for Area 2.
Quantifying discharge elasticity with respect to parameter variation is a
two-step process. The first step involves determining the variation in SCS lag
time with slope, longest flow path, and curve number. Variations in parameter
values using the four levels of parameter extraction were plotted against resulting
SCS lag time values. The range in parameter values, divided by the mean
parameter value, was used to describe the parameter variation. The range of
resulting SCS lag times, divided by the mean SCS lag time, was used to quantify
the lag time variation. The lag time variation, divided by the parameter variation,
describes the lag time elasticity with respect to the parameter. In other words, this
describes the percent increase that will occur in lag time by a 1% increase in the
parameter. If this number is less than 1, then lag time is ?inelastic? with respect
to the parameter. If this number is greater than 1, then lag time is ?elastic? with
respect to the parameter. An elastic parameter will cause larger variations in lag
time.
131
The second step in quantifying the discharge elasticity with respect to
parameter variation involves determining the variation in discharge with variation
in SCS lag time. This step requires running a HEC-1 model developed by
TxDOT. For each return period, the HEC-1 model was run seven times. The lag
times were changed in each sub-basin by an equal percentage (5%, 10%, 20%,
-5%, -10%, and -20%) for each run. The resulting linear equations were used to
determine discharge elasticity, or the percent variation in discharge caused by
fluctuations in SCS lag time. Dividing the percent variation in discharge by the
percent variation in SCS lag time gives discharge elasticity with respect to SCS
lag time for each return period. This describes the percent change in discharge
resulting from a 1% change in SCS lag time.
Table 6.1 shows the overall discharge elasticity with respect to each
parameter. Results from the first step (sensitivity of lag time to parameter
variations) and the second step (sensitivity of discharge with respect to lag time
variations) are multiplied to produce discharge elasticity with respect to each
parameter.
Table 6.1: Elasticity Analysis
Parameter
Elasticity
Lag-Parameter
Elasticity
Q
avg
-Lag
Elasticity
Q
avg
-Parameter
LFP 0.80 -0.28 -0.22
Slope -0.50 -0.28 0.14
CN -3.52 -0.28 0.99
Area --- --- 1.06
132
Results show that although errors associated with parameter extraction
methods may cause significant changes in lag time values, the errors become less
significant once entered into a model to calculate discharge. Table 6.1 shows that
lag time is inelastic with respect to longest flow path (LFP), as the elasticity of lag
time with respect to this parameter is 0.8. Discharge is also inelastic with respect
to longest flow path, with an elasticity of ?0.22. This last number, discharge
elasticity with respect to longest flow path, was determined by multiplying the lag
time elasticity with respect to longest flow path by the discharge elasticity with
respect to lag time (-0.28). The discharge elasticity with respect to slope,
calculated using the same method as with longest flow path, is also inelastic at a
value of 0.14.
Flow path and slope values were originally thought to be large
contributors to lag time variations. This analysis shows that this is not necessarily
the case. Table 6.1 shows that SCS lag time acts inelastically with respect to
both slope and longest flow path (-0.50% and 0.80% respectively). Once entered
into a hydrologic model, variations in these parameters have minimal effect on
discharge, with discharge elasticities of 0.14% and ?0.22% respectively. Thus, a
1% variation in slope will cause a minimal 0.14% increase in discharge, and a 1%
variation in longest flow path will cause a 0.22% decrease in discharge.
Table 6.1 shows that curve number, unlike slope and longest flow path,
has an elastic effect on lag time. The lag time elasticity with respect to curve
number is -3.52. Discharge is also elastic with respect to curve number (at an
133
elasticity of 0.99) when lag times, based on curve number variations, are entered
into a hydrologic model.
The discharge elasticity with respect to drainage area, calculated to be
1.06, is also elastic. Although discharge is more sensitive to drainage area than to
curve number, errors associated with determining area are small. For instance,
Area 3 has a CV for area of 1.52%. A 1% increase in area will result in a 1.61%
increase in discharge. If curve number has a CV of 3%, the resulting discharge
for a 1% increase in curve number will be 3%.
6.4 ADDITIONAL SOURCES OF ERROR
This analysis of parameter sensitivity in hydrologic modeling is general
and meant to serve as a guide for TxDOT engineers. Several important elements
were not considered in this report when calculating discharge.
First of all, the elasticity analysis does not account for curve number
effects on excess rainfall. For each change in lag time within the HEC-1 model,
curve numbers used to calculate excess rainfall were held constant.
In highly developed areas, parameter extraction methods used in this
report may not apply. Without a thorough knowledge of the storm sewers and
other urban structures it is difficult to judge flow paths. Including buildings and
structures into the model does not account for the runoff flowing below the
ground surface or under a bridge. Storm sewers may enter and leave watersheds,
and water may flow along roads and enter into one watershed from another area.
134
The effects of using traditional methods as opposed to automated methods
were not fully evaluated for flat areas. In order to evaluate if the use of automated
methods still proves viable in areas of low slope, a more thorough investigation
using several study areas of the same shape and area, but different slopes, would
have to be implemented.
6.5 RECOMMENDATIONS
Automated methods, requiring the use of a computer to extract hydrologic
parameters, produce very consistent results for hydrologic parameters. Paper
map-based methods of parameter extraction tend to vary more than do automated
methods, especially for large areas.
As with traditional paper map-based methods, on-screen digitization from
raster graphic maps is a highly subjective process. However, measurements for
the large study area closely resemble results derived from automated processes
using DEMs, more so than results derived using paper maps. This result implies
that differences are most likely attributed to errors associated with physically
measuring parameters with a map wheel and a planimeter, rather than the
subjective nature of the application.
As paper map-based methods are more tedious than automated methods
(with the exception of the TIN), and more time consuming, moving to automated
methods would accelerate the design process. Parameters extracted using WMS,
CRWR-PrePro, different resolution DEMs, and on-screen digitization from raster
graphic maps do not vary significantly from one another, and any error associated
135
with parameter extraction using automated methods will not significantly
influence discharge calculations. The recommendation, therefore, is to move
towards any of the automated methods of hydrologic parameter extraction. On-
screen digitization of raster graphic maps for small and medium areas eliminates
the need to work with DEMs. For large areas, however, DEMs are recommended
for efficiency and precision.
The second observation is that discharge is most sensitive to curve
number, and less sensitive to slope or longest flow path. Discharge is also
sensitive to drainage area. However, this study shows that the error associated in
calculating drainage area is small compared to the error associated with
calculating curve number. Determining an accurate curve number will reduce lag
time and discharge calculation errors. The recommendation that results from this
study is to carefully evaluate soil type, vegetative cover, and land use in a given
area to obtain as accurate a curve number as possible.
136
Appendix A: Online Internet Resources
Data Type Source
USGS DRG http://mcmcweb.er.usgs.gov/drg/
30 Meter DEM http://edcnts12.cr.usgs.gov/ned/
http://www.tnris.org/DigitalData/DEMs/dem-a.htm.
STATSGO (USDA-
NRCS)
http://www.ftw.nrcs.usda.gov/stat_data.html
Land Use/ Land Cover http://edc.usgs.gov/doc/edchome/ndcdb/ndcdb.html
CRWR-PrePro
webpage
http://www.ce.utexas.edu/prof/olivera/prepro/prepro.htm
WMS Webpage http://www.ems-i.com/netpagz/wms.htm
TOPAZ Webpage http://duke.usask.ca/~martzl/topaz/
137
Appendix B: Unmodified HEC-1 Model
ID LITTLE WALNUT CREEK HEC-1 MODEL; CONVERSION FROM TR-20
ID ORIG. DEV'D MURPHEE ENG. 1992; L&M CONVERSION 5/97; FILE: LW100EX.H1
ID THIS FILE IS BASED ON ORIGINAL MTEST4.N INPUT.
ID THIS FILE DIFFERS FROM EXIST.IN IN
ID
ID *** EXIST2.IN *** TXDOT MICRO MODEL (3 HR DIMENSIONLESS)
*DIAGRAM
* INITIALIZATION
IT 2 300
IO 5
* 2 YR 5 YR 10YR 25YR 50YR 100YR 3 HR DEPTHS
JR PREC 2.61 3.48 3.98 4.72 5.34 6.03
* BEGIN WITH LITTLE WALNUT (LW14)
KK TB711 * UPPER END OF BASIN
BA 0.444
PB 1.00
IN 6 0 0
PC0.0000 0.0077 0.0153 0.0230 0.0345 0.0460 0.0613 0.0766 0.0920 0.1111
PC0.1341 0.1571 0.1916 0.2375 0.3218 0.5824 0.7165 0.7816 0.8238 0.8544
PC0.8774 0.8966 0.9119 0.9272 0.9425 0.9540 0.9655 0.9770 0.9847 0.9923
PC1.0000
UD 0.255
LS 89
KKROUT43 * ROUTE TB711
RS 1 STOR 0
SV 0.00 0.59 1.19 1.93 2.54 3.31 4.07 4.57 5.59
SQ 0 180 450 900 1350 2000 2700 3200 4200
SE 643.0 645.2 646.2 647.2 647.9 648.8 649.6 650.1 650.8
KK TB710 * ORIGINAL MAP LABEL
BA 0.285
UD 0.298
LS 88
KKCOMB01 * COMBINE ROUTE43 + TB710
HC 2
KKROUT44 * ROUTE COMB01
RS 1 STOR 0
SV 0.00 2.93 5.97 9.63 12.70 16.56 20.36 22.87 27.96
SQ 0 180 450 900 1350 2000 2700 3200 4200
SE 643.0 645.2 646.2 647.2 647.9 648.8 649.6 650.1 650.8
KK TB709 * ORIGINAL MAP LABEL
BA 0.052
UD 0.129
LS 80
KKCOMB02 * COMBINE ROUTE44 + TB709
HC 2
KK TB708 * ORIGINAL MAP LABEL
BA 0.241
UD 0.262
LS 86
KKCOMB03 * COMBINE COMB02 + TB708
HC 2
KKROUT45 * ROUTE COMB03
RS 1 STOR 0
SV 0.00 2.90 6.51 11.81 16.32 23.25 30.62 35.87 45.79
SQ 0 180 450 900 1350 2000 2700 3200 4200
138
SE 603.0 605.5 606.8 608.1 608.9 609.9 610.8 611.3 612.1
KK TB707 * ORIGINAL MAP LABEL
BA 0.042
UD 0.156
LS 87
KKCOMB04 * COMBINE ROUTE45 + TB707
HC 2
KK TB706 * ORIGINAL MAP LABEL
BA 0.121
UD 0.267
LS 84
KKCOMB05 * COMBINE COMB04 + TB706
HC 2
KKTB1003 * NORWOOD
BA 0.150
UD 0.111
LS 77
KK STR80 * FLOW OF TB1003 THROUGH STR80
KM CHANNEL STORAGE ROUTING FOR RESERVOIR 80
RS 1 ELEV 628.5
SV 0.00 0.50 0.78 2.34 4.81 8.05 9.99 11.92 12.89 14.12
SV 16.31 21.09 25.86
SQ 0 0 5 30 110 250 289 325 340 470
SQ 900 1575 2400
SE 628.5 629.5 630.0 632.0 634.0 636.0 637.0 638.0 638.4 639.0
SE 640.0 641.0 642.0
KKROUT46 * ROUTE STR80
RS 1 STOR 0
SV 0.00 0.21 0.43 0.74 0.99 1.57 2.91 3.83 5.50
SQ 0 35 85 170 240 375 500 600 800
SE 606.0 606.9 607.3 607.7 607.9 608.4 609.4 610.0 610.9
KKTB1001 * FIRST MODIFIED AREA - IMPERVIOUS FROM 183
BA 0.025
UD 0.185
LS 85
KK183I01 * TB1001 AREA CONVERTED TO IMPERVIOUS
BA 0.011
UI 213.0 0
KKCOMB06 * COMBINE ROUTE46 + TB1001 + 183I01
HC 3
KKCOMB07 * COMBINE COMB06 + C0MB05
HC 2
KKROUT47 * ROUTE COMB07
RS 1 STOR 0
SV 0.00 1.16 2.39 4.15 5.66 8.74 17.13 20.26 28.33
SQ 0 180 450 900 1350 2000 2700 3200 5100
SE 587.8 590.8 592.1 593.4 594.4 595.8 598.5 599.3 601.0
KM
KK TB705 * ORIGINAL MAP LABEL
BA 0.102
UD 0.205
LS 83
KM
KK183I03 * TB705 AREA CONVERTED TO IMPERVIOUS
BA 0.004
UI 77.4 0
KKCOMB10 * COMBINE tb705 + 183i03
HC 2
KKCOMB08 * COMBINE ROUT47+ comb10
HC 2
KK TB704 * ORIGINAL MAP LABEL
BA 0.046
139
UD 0.129
LS 87
KKCOMB11 * COMBINE TB704 + COMB08
HC 2
KKROUT48 * ROUT COMB11
RS 1 STOR 0
SV 0.00 1.13 2.37 4.06 5.62 7.50 9.34 10.60 13.01 25.00
SQ 0 180 450 900 1350 2000 2700 3200 4200 8200
SE 570.0 572.8 574.1 575.4 576.3 577.4 578.3 578.9 580.0 585.0
KK TB702 * ORIGINAL MAP LABEL
BA 0.097
UD 0.162
LS 86
KKCOMB12 * COMBINE ROUT48 + tb702
HC 2
KK TB703 * ORIGINAL MAP LABEL
BA 0.052
UD 0.137
LS 85
KK183I04 * TB703 AREA CONVERTED TO IMPERVIOUS COVER
BA 0.004
UI 77.4 0
KKCOMB13 * COMBINE COMB12 + tb703 + 183I04
HC 3
KKROUT49 * ROUTE COMB71
RS 1 STOR 0
SV 0.00 3.01 5.67 8.99 11.77 15.22 18.50 20.73 25.72 35.00
SQ 0 180 450 900 1350 2000 2700 3200 4200 7000
SE 548.0 551.4 552.8 554.2 555.3 556.5 557.6 558.3 559.6 562.6
KK TB701 * ORIGINAL MAP LABEL
BA 0.088
UD 0.140
LS 86
KKCOMB14 * COMBINE ROUT49 + TB701
HC 2
KK183I05 * TB701 AREA CONVERTED TO IMPERVIOUS COVER
BA 0.001
UI 19.4 0
KKCOMB15 * COMBINE COM14 + 183I05
HC 2
KKLW14 * ORIGINAL MAP LABEL
BA 0.127
UD 0.125
LS 80
KKCOMB16 * COMBINE COMB15 + LW14
HC 2
KK183I06 * LW14 AREA CONVERTED TO IMPERVIOUS
BA 0.006
UI 116.2 0
KKCOMB17 * COMBINE 183I06 + COMB16
HC 2
ZZ
140
Appendix C: Modified HEC-1 Model in HEC-HMS
Basin: Exist2mod.in
Description: LITTLE WALNUT CREEK HEC-1 MODEL; CONVERSION FROM TR-20 ORIG.
DEV'D MURPHEE ENG. 1992; L&M CONVERSION 5/97; FILE: LW100EX.H1 THIS FILE IS BASED
ON ORIGINAL MTEST4.N INPUT. THIS FILE DIFFERS FROM EXIST.IN IN *** EXIST2.IN ***
TXDOT MICRO MODEL (3 HR DIMENSIONLESS)
Last Modified Date: 18 February 2002
Last Modified Time: 10:29
Version: 2.1.2
Default DSS File Name: D:\hmsproj\Exist.in_feb17\Exist.in_feb17.dss
Unit System: English
End:
Subbasin: TB711
Canvas X: -386.946
Canvas Y: 1940.607
Label X: 16
Label Y: 0
Area: 0.444
Downstream: ROUT43
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 89
Transform: SCS
Lag: 15.300
Show lag in hours: Yes
Baseflow: None
End:
Reservoir: ROUT43
Canvas X: -116.602
Canvas Y: 1794.804
Label X: 16
Label Y: 0
Downstream: COMB01
Route: Modified Puls
Routing Curve: Storage-Elevation-Outflow
Initial Storage: 0
Routing Table in DSS: Yes
End:
Subbasin: TB710
Canvas X: 86.915
Canvas Y: 1974.021
Label X: 16
Label Y: 0
Area: 0.285
Downstream: COMB01
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 88
Transform: SCS
141
Lag: 17.880
Show lag in hours: Yes
Baseflow: None
End:
Junction: COMB01
Canvas X: 32.239
Canvas Y: 1752.278
Label X: 8
Label Y: 5
Downstream: ROUT44
End:
Reservoir: ROUT44
Canvas X: 44.389
Canvas Y: 1636.850
Label X: 16
Label Y: 0
Downstream: COMB02
Route: Modified Puls
Routing Curve: Storage-Elevation-Outflow
Initial Storage: 0
Routing Table in DSS: Yes
End:
Subbasin: TB709
Canvas X: 354.221
Canvas Y: 1652.038
Label X: 16
Label Y: 0
Area: 0.052
Downstream: COMB02
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 80
Transform: SCS
Lag: 7.740
Show lag in hours: Yes
Baseflow: None
End:
Junction: COMB02
Canvas X: 44.389
Canvas Y: 1536.611
Label X: 16
Label Y: 0
Downstream: COMB03
End:
Subbasin: TB708
Canvas X: -356.570
Canvas Y: 1527.498
Label X: 16
Label Y: 0
Area: 0.241
Downstream: COMB03
142
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 86
Transform: SCS
Lag: 15.720
Show lag in hours: Yes
Baseflow: None
End:
Junction: COMB03
Canvas X: 59.577
Canvas Y: 1451.559
Label X: 16
Label Y: 0
Downstream: ROUT45
End:
Reservoir: ROUT45
Canvas X: 68.690
Canvas Y: 1384.732
Label X: 16
Label Y: 0
Downstream: COMB04
Route: Modified Puls
Routing Curve: Storage-Elevation-Outflow
Initial Storage: 0
Routing Table in DSS: Yes
End:
Subbasin: TB707
Canvas X: 354.221
Canvas Y: 1424.220
Label X: 16
Label Y: 0
Area: 0.042
Downstream: COMB04
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 87
Transform: SCS
Lag: 9.360
Show lag in hours: Yes
Baseflow: None
End:
Junction: COMB04
Canvas X: 71.727
Canvas Y: 1302.718
Label X: 16
Label Y: 0
Downstream: COMB05
End:
Subbasin: TB706
Canvas X: -222.917
Canvas Y: 1223.741
143
Label X: 16
Label Y: 0
Area: 0.121
Downstream: COMB05
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 84
Transform: SCS
Lag: 16.020
Show lag in hours: Yes
Baseflow: None
End:
Junction: COMB05
Canvas X: 32.239
Canvas Y: 1211.591
Label X: 16
Label Y: 0
Downstream: COMB07
End:
Subbasin: TB1003
Canvas X: 335.996
Canvas Y: 1278.417
Label X: 16
Label Y: 0
Area: 0.150
Downstream: STR80
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 77
Transform: SCS
Lag: 6.660
Show lag in hours: Yes
Baseflow: None
End:
Reservoir: STR80
Description: CHANNEL STORAGE ROUTING FOR RESERVOIR 80
Canvas X: 348.146
Canvas Y: 1090.088
Label X: 16
Label Y: 0
Downstream: ROUT46
Route: Modified Puls
Routing Curve: Storage-Elevation-Outflow
Initial Elevation: 628.5
Routing Table in DSS: Yes
End:
Reservoir: ROUT46
Canvas X: 384.597
Canvas Y: 1014.149
Label X: -23
Label Y: -21
144
Downstream: COMB06
Route: Modified Puls
Routing Curve: Storage-Elevation-Outflow
Initial Storage: 0
Routing Table in DSS: Yes
End:
Subbasin: TB1001
Canvas X: 162.854
Canvas Y: 935.172
Label X: 16
Label Y: 0
Area: 0.025
Downstream: COMB06
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 85
Transform: SCS
Lag: 11.100
Show lag in hours: Yes
Baseflow: None
End:
Subbasin: 183I01
Canvas X: 238.794
Canvas Y: 1187.290
Label X: 16
Label Y: 0
Area: 0.011
Downstream: COMB06
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 85
Transform: User-Specified UH
Unit Hydrograph Name: 183I01
Baseflow: None
End:
Junction: COMB06
Canvas X: 135.516
Canvas Y: 1068.825
Label X: 16
Label Y: 0
Downstream: COMB07
End:
Junction: COMB07
Canvas X: 32.239
Canvas Y: 1084.013
Label X: -66
Label Y: -8
Downstream: ROUT47
End:
Reservoir: ROUT47
145
Canvas X: 83.877
Canvas Y: 810.631
Label X: 16
Label Y: 0
Downstream: COMB08
Route: Modified Puls
Routing Curve: Storage-Elevation-Outflow
Initial Storage: 0
Routing Table in DSS: Yes
End:
Subbasin: TB705
Canvas X: -125.715
Canvas Y: 841.007
Label X: 16
Label Y: 0
Area: 0.102
Downstream: COMB10
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 83
Transform: SCS
Lag: 12.300
Show lag in hours: Yes
Baseflow: None
End:
Subbasin: 183I03
Canvas X: -92.302
Canvas Y: 698.241
Label X: 16
Label Y: 0
Area: 0.004
Downstream: COMB10
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 83
Transform: User-Specified UH
Unit Hydrograph Name: 183I03
Baseflow: None
End:
Junction: COMB10
Canvas X: 190.192
Canvas Y: 755.955
Label X: 16
Label Y: 0
Downstream: COMB08
End:
Junction: COMB08
Canvas X: 193.230
Canvas Y: 591.926
Label X: 21
Label Y: -1
146
Downstream: COMB11
End:
Subbasin: TB704
Canvas X: 354.221
Canvas Y: 546.363
Label X: 16
Label Y: 0
Area: 0.046
Downstream: COMB11
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 87
Transform: SCS
Lag: 7.740
Show lag in hours: Yes
Baseflow: None
End:
Junction: COMB11
Canvas X: 211.455
Canvas Y: 519.025
Label X: -10
Label Y: 37
Downstream: ROUT48
End:
Reservoir: ROUT48
Canvas X: 214.493
Canvas Y: 440.048
Label X: 15
Label Y: 2
Downstream: COMB12
Route: Modified Puls
Routing Curve: Storage-Elevation-Outflow
Initial Storage: 0
Routing Table in DSS: Yes
End:
Subbasin: TB702
Canvas X: -235.067
Canvas Y: 637.490
Label X: 16
Label Y: 0
Area: 0.097
Downstream: COMB12
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 86
Transform: SCS
Lag: 9.720
Show lag in hours: Yes
Baseflow: None
End:
147
Junction: COMB12
Canvas X: 223.606
Canvas Y: 364.109
Label X: 23
Label Y: -49
Downstream: COMB13
End:
Subbasin: TB703
Canvas X: 10.699
Canvas Y: 356.652
Label X: 16
Label Y: 0
Area: 0.052
Downstream: COMB13
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 85
Transform: SCS
Lag: 8.220
Show lag in hours: Yes
Baseflow: None
End:
Subbasin: 183I04
Canvas X: 351.184
Canvas Y: 382.334
Label X: 16
Label Y: 0
Area: 0.004
Downstream: COMB13
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 85
Transform: User-Specified UH
Unit Hydrograph Name: 183I04
Baseflow: None
End:
Junction: COMB13
Canvas X: 257.019
Canvas Y: 248.681
Label X: -10
Label Y: 29
Downstream: ROUT49
End:
Reservoir: ROUT49
Canvas X: 260.056
Canvas Y: 166.667
Label X: 16
Label Y: 0
Downstream: COMB14
Route: Modified Puls
Routing Curve: Storage-Elevation-Outflow
148
Initial Storage: 0
Routing Table in DSS: Yes
End:
Subbasin: TB701
Canvas X: 123.366
Canvas Y: 187.930
Label X: 16
Label Y: 0
Area: 0.088
Downstream: COMB14
LossRate: SCS
Percent Impervious Area: 0.0
Curve Number: 86
Transform: SCS
Lag: 8.400
Show lag in hours: Yes
Baseflow: None
End:
Junction: COMB14
Canvas X: 275.244
Canvas Y: 90.728
Label X: 16
Label Y: 0
End:
Default Attributes:
Default Basin Unit System: English
Default Meteorology Unit System: SI
Default Loss Rate: Initial+Constant
Default Transform: Modified Clark
Default Baseflow: Recession
Default Route: Muskingum
Enable Flow Ratio: No
Enable Evapotranspiration: No
Compute Local Flow At Junctions: No
Warning On Delete Component: Yes
Warning On Change Method: Yes
End:
149
Bibliography
Anderson, David James. 2000. CRWR Online Report 00-11: GIS-Based
Hydrologic and Hydraulic Modeling for Floodplain Delineation at
Highway River Crossings. Center for Research in Water Resources, The
University of Texas at Austin, Austin, Texas.
Barrett, Michael. 2000. LCRA/Marble Falls Regional Drainage Feasibility Study.
Center for Research in Water Resources, Bureau of Engineering Research,
Austin, Texas.
Chow, V. T., D. R. Maidment, and L. W. Mays. 1988. Applied Hydrology.
McGraw-Hill, New York, New York.
Djokic, D., Z. Ye, and A. Miller. 1997. Efficient Watershed Delineation Using
ArcView and Spatial Analyst. Proceedings, 17
th
Annual ESRI User?s
Conference, San Diego, California.
Garbrecht, J. and L. W. Martz. 1994. Grid Size Dependency of Parameters
Extracted from Digital Elevation Models. Computers and Geosciences,
20(1): 85-87.
Garbrecht, J. and L. W. Martz., 1999a. Digital Elevation Model Issues in Water
Resources Modeling. In Hydrologic and Hydraulic Modeling Support, ed.
by D. R. Maidment and D. Djokic, Environmental Systems Research
Institute, Redlands, California. 1-27. Internet website:
http://grl/ars/usda.gov/topaz/esri/paperH.htm.
Garbrecht, J., L. W. Martz, K. H. Syed, and D. Goodrich. 1999b. Determination
of Representative Catchment Properties from Digital Elevation Models.
1999 International Water Resources Engineering Conference, American
Society of Civil Engineers, Session HY-3, Geographic Information
Systems (GIS) Applications in Hydrologic Engineering, Seattle,
Washington. Published on CD-ROM. Internet Website:
http://grl.ars.usda.gov/topaz/ASCE99/ASCE99.HTM
Hellweger, F. and D. R. Maidment. 1997. HEC-PREPRO: A GIS Preprocessor
for Lumped Parameter Hydrologic Modeling Programs. CRWR Online
Report 97-8. Center for Research in Water Resources, The University of
Texas at Austin, Austin, Texas.
150
Jennings et. al., 1993. National Summary of U.S. Geological Survey Regional
Regression Equations for Estimating Magnitude and Frequency of Floods
for Unguaged Sites. Water Resources Investigations Report 94-4002.
Jones, N. L., S.G., Write, and D.R, Maidment. 1990. Watershed Delineation with
Triangle-Based Terrain Models. Journal of Hydrologic Engineering,
ASCE, 116(10), 1232-1251.
Knill, Oliver. 2000. The Planimeter and the Theorem of Green. Internet
Website: http://www.math.harvard.edu/~knill/math21a2000/planimeter/
Kunkle, Paul. 2002. Kunkel?s Mathematics Lessons: The Planimeter. Internet
Website: http://www.nas.com/~kunkel/planimeter/planimeter.htm
Lee, D. T., and B. J. Schacter. 1980. Two Algorithms of Constructing a Delauney
Triangulation. International Journal of Computer and Information
Sciences, ASCE, 9(3): 219-242.
Miller, Scott N., D. Phillip Guertin, Kamram H. Syed, and David C. Goodrich.,
1999. Using High Resolution Synthetic Aperture Radar for Terrain
Mapping: Influences on Hydrologic and Geomorphic Investigation. 1999
Annual Summer Specialty Conference Proceedings, Science Into Policy:
Water in the Public Realm/ Wildland Hydrology. Internet Website:
http://www.awra.org/proceedings/Montana99/Miller1/
Moglen, Glen. and Ginger L. Hartman. 2001. Resolution Effects on Hydrologic
Modeling Parameters and Peak Discharge. Journal of Hydrologic
Engineering. 6 (6): .
Nelson, James E.and Norman L. Jones. 1996. Utilizing the ArcInfo Data Model to
Build Conceptual Models for Environmental/Hydraulic/Hydrologic
Simulations. ESRI User?s Conference.
Nelson, James E., Norman L. Jones and Christopher M. Smemoe. 1997. From a
Grid or Coverage to a Hydrograph: Unlocking Your GIS Data for
Hydrologic Applications. ESRI User?s Conference, July 1997.
Published on CD ROM. Internet Website:
http://www.esri.com/library/userconf/proc97/proc97/to200/pap198/p198.h
tm.
Nelson, James E., Norman L. Jones, and Russell J. Barrett. 1999a. Adaptive
Tessellation Method for Creating TINs from GIS Data. ASCE Journal of
Hydrologic Engineering, 4(1)
151
Nelson, James E., Christopher M. Smemoe, and Bing Zhao. 1999b. A GIS
Approach to Watershed Modeling in Maricopa County, Arizona,
American Society of Civil Engineers, Water Resources Planning and
Management Conference, June 6-10, 1999, Tempe, AZ. Published on CD
ROM. http://emrl.byu/chris/documents/asceconf.pdf
Olivera, F. and D. R. Maidment. 1999. GIS-Based System of Hydrologic and
Hydraulic Applications for Highway Engineering. Research Report 1738-
6. Center for Transportation Research,The University of Texas at Austin,
Austin, Texas.
Stolpa, David. 2001-2002. Personal Interviews.
Tarboton, D. G. 1997. A New Method for the Determination of Flow Directions
and Upslope Areas in Grid Digital Elevation Models. Water Resources
Research, 33(2): 309-319.
Tarboton, D. G., R. L. Bras, and I. Rodriguez-Itube. 1991. On the Extraction of
Channel Networks from Digital Elevation Data. Water Resources
Research, 5(1): 81-100.
Texas Department of Transportation (TxDOT). 2001 Hydraulic Design Manual.
USACOE ? HEC. 2000. HEC-HMS Hydrologic Modeling System: User?s
Manual. Version 2.0. Hydrologic Engineering Center, Davis, California.
United States Geological Survey (USGS). Digital Raster Graphics (DRG). 1999.
Internet Webpage: http://mcmcweb.er.usgs.gov/drg/
Watershed Modeling System (WMS) Reference Manual, Version 6.0. 1999.
Environmental Modeling Research Lab., Brigham Young University.
152
153