Technical Report
CRWR 266
WATER QUALITY AND QUANTITY IMPACTS OF HIGHWAY
CONSTRUCTION AND OPERATION:
SUMMARY AND CONCLUSIONS
by
MICHAEL E. BARRETT, M.S.
Project Manager
JOSEPH F. MALINA, JR., P.E.
Principal Investigator
and
RANDALL J. CHARBENEAU, P.E.
GEORGE H. WARD, Ph.D.
Co-Principal Investigators
November 1995
CENTER FOR RESEARCH IN WATER RESOURCES
Bureau of Engineering Research
The University of Texas at Austin
Austin, TX 78712
iii
ACKNOWLEDGEMENTS
This research was funded by the Texas Department of Transportation under grant
number 7-1943, ?Water Quantity and Quality Impacts Assessments of Highway
Construction in Austin, Texas.? Numerous graduate students in the Department of Civil
Engineering at the University of Texas contributed to the research summarized in this
volume.  They include Adam Bogusch, Trey Collins, David Cox, Tom Heathman, Lyn
Irish, John Kearney, Terry McCoy, Sean Tenney, and Rob Zuber.  Center for Research in
Water Resources staff who performed laboratory analyses included Hanne Nielsen, Ty
Lehmen, and Tess Kallick.  Members of the technical review committee who offered
suggestions and guidance included Tom Word, Sharon Barta, Wes Burford, Rob Stuard,
Bill Couch, Ron Fieseler, and Don Rauschuber.
v
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ...............................................................................................iii
TABLE OF CONTENTS .................................................................................................... v
LIST OF FIGURES...........................................................................................................vii
LIST OF TABLES ............................................................................................................. ix
1. INTRODUCTION........................................................................................................... 1
2. TEMPORARY EROSION CONTROLS: USE AND EFFECTIVENESS..................... 3
3. EFFECTS OF HIGHWAY CONSTRUCTION AND OPERATION............................. 9
3.1 Construction Effects.................................................................................................. 9
3.2 Highway Operation Effects ..................................................................................... 10
4. QUALITY OF HIGHWAY RUNOFF IN AUSTIN, TEXAS ...................................... 13
5. FACTORS AFFECTING THE QUALITY OF HIGHWAY RUNOFF ....................... 19
5.1 Model of Highway Runoff Quality ......................................................................... 20
5.2 Distribution of Highway Runoff EMCs .................................................................. 23
6. PERFORMANCE OF PERMANENT RUNOFF CONTROLS................................... 27
6.1 Pollutant Removal Effectiveness of a Grassy Swale............................................... 27
6.2 Field Performance of Vertical Sand Filter Systems ................................................ 29
6.3 Laboratory Filtration Experiments .......................................................................... 31
7. CONCLUSIONS........................................................................................................... 33
BIBLIOGRAPHY ............................................................................................................. 35
vii
LIST OF FIGURES
Figure 2.1 Testing Silt Fence Performance in a Flume....................................................... 5
Figure 2.2 TSS Removal Efficiency as a Function of Detention Time (Flume Studies) .... 6
Figure 2.3 Typical Relationship Between Head and Flow Rate.......................................... 6
Figure 5.1 Rainfall Simulator in Operation....................................................................... 19
Figure 5.2 Probability Plots of TSS Data.......................................................................... 25
Figure 6.1 Grassy Swale, MoPac at Walnut Creek ........................................................... 27
Figure 6.2 Typical Vertical Sand Filter System ................................................................ 29
Figure 6.3 Drainage of Six Runoff Controls after Storm on 5/16/94................................ 30
ix
LIST OF TABLES
Table 2.1 Properties of Tested Silt Fence Fabrics............................................................... 5
Table 3.1 EMC?s in Danz Creek During Highway Construction...................................... 10
Table 3.2 EMC?s in Danz Creek During Highway Operation .......................................... 12
Table 4.1 EMC?s of Constituents in Highway Runoff...................................................... 14
Table 4.2 Comparison of Median EMC?s from High Traffic Sites................................... 14
Table 4.3 Comparison Low Traffic Site to Residential Land Use (Mean EMC).............. 15
Table 4.4 Comparison of High Traffic Site to Commercial Land Use (Mean EMC) ....... 15
Table 4.5 Estimated Annual Pollutant Loadings (kg/ha) .................................................. 16
Table 5.1 Variables Affecting Pollutant Runoff Loads..................................................... 21
Table 6.1 Pollutant Removal Efficiency of a Grassy Swale.............................................. 28
Table 6.2 Pollutant Removal Efficiencies of Alternative Media ...................................... 32
1
 1.  INTRODUCTION
Environmental regulatory agencies recently have focused attention on nonpoint
sources of pollution such as urban and highway runoff.  Enforcement of the National
Pollutant Discharge Elimination System (NPDES) regulations regarding stormwater
runoff by the U.S.  Environmental Protection Agency (EPA) is evidence of this effort.  In
Texas, the Barton Springs/Edwards Aquifer Conservation District (District) and several
environmentally oriented organizations became concerned about the potential for
contamination of the Edwards aquifer as a result of proposed highway construction
activities over the recharge zone of the aquifer.  The proposed highway construction
corridor crosses and parallels three creeks and overlies a portion of the recharge zone of
the Edwards aquifer which feeds Barton Springs.  This concern resulted in litigation
involving the Texas Department of Transportation (TxDOT) and the Federal Highway
Administration (FHA), which temporarily halted construction activities on the project
site.
Prior to this halt in construction, the District and TxDOT negotiated a settlement,
the Consent Decree, which was approved by the U. S. District Court.  The District
removed itself from the litigation and TxDOT began implementing certain actions and
practices to answer the concerns of the District.  The cooperative efforts of the two
agencies have been effective in preventing and reducing pollution from both point and
nonpoint sources during roadway construction activities.  Many improvements and
innovations have been developed for structural and non-structural Best Management
Practices (BMPs) which have gained local, state, and national recognition as leaders in
the field of pollution prevention and mitigation for both agencies.
The Consent Decree also ordered a study of the water quality and quantity of
highway runoff and the effects of highway construction and operation on the quality of
receiving waters.  TxDOT and the District agreed to have the study conducted by the
Center for Research in Water Resources (CRWR) at The University of Texas at Austin.
CRWR, TxDOT and the District developed the objectives and scope of the study and
identified specific project tasks.  A technical review committee consisting of three
representatives of the District, two from TxDOT, and two from the CRWR met quarterly
to review activities and progress reports.  The committee provided input and guidance to
the CRWR project personnel dealing with the overall study, its procedures, equipment,
and future work efforts.
2
The construction of the new highway allowed the evaluation of the hydrologic
changes to creeks in the recharge zone.  Effectiveness of temporary runoff control
measures were evaluated in the field during the construction process.  In addition, field-
scale laboratory experiments were conducted to determine the hydraulic properties and
sediment removal effectiveness of silt fences under realistic operating conditions.
The quality of highway stormwater runoff was measured at three sites along an
existing segment of highway to determine runoff characteristics, the probable impact of
the new highway segments, and to identify treatment systems to mitigate adverse water
quality impacts.  A rainfall simulator was operated along a section of active highway to
determine the factors which affect the quality of highway runoff.  The effectiveness of
sand and other media for filtration of runoff was evaluated in laboratory experiments and
the performance of permanent runoff treatment systems was monitored after completion
of the highway.  A comprehensive review of literature pertaining to highway runoff was
also performed and reported in CRWR Technical Report # 239 (Barrett et al., 1995a) and
Center for Transportation Technical Report #1943-1.
3
 2.  TEMPORARY EROSION CONTROLS: USE AND EFFECTIVENESS
Most of the more important environmental impacts of highway construction
results from erosion of topsoil during storms.  Two strategies for minimizing these
impacts of stormwater runoff from construction sites are erosion control and sediment
control.  Erosion control is a source management method and is usually accomplished
with slope coverings.  Techniques for slope stabilization include temporary and
permanent vegetation, plastic sheeting, straw and wood fiber mulches, matting, netting,
chemical stabilizers, or some combination of the above.  Sediment control may be
considered as the second line of defense.  Sedimentation ponds, silt fences, and rock
berms commonly are used for sediment control once erosion has occurred.  These devices
are designed to diminish solids loading through short term retention and/or velocity
reduction and filtration.
An inventory of temporary runoff controls installed on TxDOT construction sites
indicated that rock berms and silt fences were the most commonly used erosion and
sediment controls on construction sites.  Rock berms were used to treat the drainage from
53% of the area of the six construction sites in the study area.  Silt fences and
sedimentation ponds were the next most common runoff controls, treating the runoff from
23% and 22% of the total area, respectively.  Sediment ponds were the most inexpensive
control on a cost per area basis and were used more frequently in the earlier stages of
construction.  Erosion control blankets were the most expensive controls and tended to be
used in the later phases of construction.
Field evaluation of the efficiency of silt fences in removing sediment carried in
runoff from highway construction sites, showed that the median removal resulting from
filtration was 0%.  Additional removal occurred as a result of particle settling, but was
not quantified in the field portion of the study.  The median concentration of solids
discharged from the silt fence controls was approximately 500 mg/L.  Geotextile silt
fences also proved to be ineffective in reducing turbidity.  The median turbidity
reductions for the sites monitored was about 2%.  The average pore size of geotextile
fabrics may exceed 600 ?m, which is considerably larger than the maximum size of silt
and clay particles (75 ?m) which cause turbidity.  Monitoring of a single rock berm also
showed negligible Total Suspended Solids (TSS) removal.
The poor filtration performance of the geotextile fabrics alone indicates the
disparity between test efficiency and actual field performance.  The bulk of the difference
could be attributed to an unrealistic particle size distribution in the slurry mixtures of
previous laboratory studies.  Silt and clay size particles were the primary constituents of
4
construction site generated sediment in this study.  The observed data indicated that silt
and clay size particles comprised 92% of the total suspended solids.
The field efficiency of silt fences appears to be dependent mainly on the detention
time of the runoff behind the control.  The detention time is controlled by the geometry of
the upstream pond, hydraulic properties of the fabric, and maintenance of the control.
Despite comments by project supervisors that little maintenance of controls was required,
numerous installation and maintenance deficiencies were noted during the study.  Holes
in the fabric and inadequate ?toe-ins? that result in under-runs, reduced the detention time
available for particle settling.  In addition, the openings released the discharge in a
concentrated flow, which promoted erosion below the structure and resulted in short
circuiting in the ponded area.
In contrast to the field monitoring, high removal efficiencies were achieved with
silt fences in the flume studies (Figure 2.1).  Four types of silt fences and a rock berm
were subjected to cycles of simulated runoff events.  The silt fences tested were
constructed of geotextile fabrics.  Properties of the fabrics, as reported by the
manufacturers, are summarized in Table 2.1.  The geometry of the flume resulted in the
creation a large ponded area behind the control section, resulting in long detention times
and settling of particles even with the fine-grained sediment used in the tests.  Mean
sediment removal efficiency in the flume ranged from 68 to 90% and was highly
correlated with the detention time of the runoff (Figure 2.2).  This observation indicates
that silt fences should be sited in the field, so as to maximize the ponded volume behind
the fence.
The flow rates of sediment-laden runoff through the control sections were two
orders of magnitude less than those typically specified by transportation agencies.  The
flow rate of a sediment slurry through geotextile fences is a function of apparent opening
size as well as permittivity (or other measures of clean water flow rates).  The fabric with
the longest detention times in this series of flume tests had the highest reported
permittivity, but it also had the smallest apparent opening size, suggesting that clogging
of the fabric with sediment affected the hydraulic performance.  A comparison of the
reported and measured permittivities is shown in Table 2.1.
In addition, tests of silt fence fabrics in a modified permeameter showed that flow
rates through the fabric were not linearly related to head as is assumed in the ASTM
definition (Figure 2.3).  The discrepancy was due to the fact that flow is turbulent at the
heads on the fabrics in silt fence applications resulting in much lower flow rates.
5
Figure 2.1 Testing Silt Fence Performance in a Flume
Table 2.1 Properties of Tested Silt Fence Fabrics
Type of Fabric AOS, # sieve
size(?m)
Permittivity, sec
-1
reported/observed
No. of Tests
Belton woven #30 (600) 0.4/0.015 9
Exxon woven #30 (600) 0.1/0.002 5
Mirafi non-woven #100 (150) 1.5/0.0004 5
Amoco woven #20 (850) 0.2/0.012 5
6
0
20
40
60
80
100
10 100 1000 10000 100000
Detention Times, min
TSS Removal Efficiency, %
Belton w
Exxon w
Mirafi A nw
Mirafi B nw
Amoco w
Rock Berm
Figure 2.2 TSS Removal Efficiency as a Function of Detention Time (Flume Studies)
y = 22.322x
0.457
R
2
 = 0.9981
0
10
20
30
40
0 0.1 0.2 0.3 0.4 0.5 0.6
Level, m
Flow per Area, cm
/s
Figure 2.3 Typical Relationship Between Head and Flow Rate
Flow rates through rock berms greatly exceeded the rates typically recommended
in guidelines promulgated by regulatory agencies.  The short detention times and large
pore size of the berms resulted in only a slight reduction in the suspended solids load in
the flume test.
Despite the high sediment removal efficiency of silt fences under optimum
conditions, water quality in lakes and streams below construction sites which rely on silt
fences for sediment control will continue to be adversely impacted under most conditions.
7
The suspended sediment concentration in untreated construction runoff may exceed 3000
mg/L.  A correctly installed and maintained system of silt fences might reasonably
achieve a sediment removal efficiency of 85%, which would result in runoff discharged
from the construction site with a sediment concentration of 450 mg/L.  This concentration
is higher than occurs naturally in most lakes and many small streams and would result in
water quality impacts which would be apparent to even the most casual observer.
Occasional large storm events which would overwhelm the storage capacity of the
temporary controls could be expected to produce even more severe water quality impacts.
Use of slope coverings to minimize erosion could improve the quality of
construction runoff; however their use would not be practical in an area of active
construction.  Complete elimination of water quality impacts would require the retention
of all runoff from the construction site.  In environmentally sensitive areas, large
sedimentation ponds could be constructed to retain all the runoff from most storm events.
In most areas, the impacts to receiving waters are usually transitory and extreme measures
(i.e., 100% runoff detention) are generally not necessary.
Development of a new test or series of tests to characterize the expected
performance of geotextile fabrics, when used as silt fences, is urgently needed.  Field
performance can not be determined from current parameters used to characterize the
hydraulic properties of these fabrics.  The use of these parameters results in an over-
estimate of the area that can be treated without over-topping.  Knowledge of the
performance under field conditions would allow the development of rational guidelines
governing the installation of these controls.  Testing of fabrics using sediment with
particle size distributions more characteristic of construction site sediment should serve
as the foundation for future studies.  A detailed presentation of the results of the
temporary control evaluation is contained in CRWR Technical Report #261 (Barrett et
al., 1995b).
9
 3.  EFFECTS OF HIGHWAY CONSTRUCTION AND OPERATION
 3.1 Construction Effects
Grab samples from 10 storms were collected at monitoring sites on Danz Creek in
southwestern Travis County, Texas.  Nine of the storms were sampled at both locations
and one storm at each site was not sampled at the other location.  These samples were
collected during the period from June 11, 1992 through October 10, 1993, coincident with
the construction of the new highway.  The data indicate that suspended solids are the
most important constituent of storm water runoff from the construction corridor.  The
concentration of suspended solids in Danz Creek increased at least 5-fold during and
immediately after storm events despite the presence of an extensive system of temporary
controls (primarily silt fences).  Other solid related parameters such as turbidity and iron
also increased.  A summary of concentrations above and below the highway crossing is
shown in Table 3.1.  The concentrations shown are the means and medians of the event
mean concentrations (EMC?s).  The EMC is the average concentration of a constituent for
a single storm event.
The increase in suspended solids concentration is consistent with the results of the
temporary control monitoring performed as part of this overall study and described by
Barrett et al. (1995b).  Dilution of the construction runoff with water from upstream,
easily could account for the median observed concentration downstream of the highway
right-of-way of approximately 180 mg/L.  This site was subject to weekly inspections,
which insured adequate maintenance and installation of the temporary controls.  The
contractor also was required to keep all heavy equipment out of the creek crossings.
Trees in the stream channels were removed by hand, where necessary, to minimize
disruption to the creeks.  Consequently, the reduction in sediment load to the creek was
the best obtainable with the selected controls.  The amount of sediment discharged to the
creek would probably have been three to four times as large without the use of temporary
controls and restrictions on the use of heavy equipment at the creek crossings.
Despite the high concentrations of suspended solids, no permanent change in the
channel resulting from runoff during construction was obvious.  The effects of
construction on Danz Creek were temporary, and similar to those reported in the literature
for other rivers and streams.  Of particular concern in this area are the effects of
construction on the water quality in the Edwards Aquifer.  Approximately 85% of the
recharge flow into the Barton Springs portion of the Edwards Aquifer occurs in the beds
of the creeks that cross the recharge zone.  The portion of Danz Creek affected by this
10
construction project lies on the recharge zone; therefore, higher concentrations of
suspended solids could be expected to enter the aquifer during the period when runoff
from the construction site occurred.  Estimating the effect of higher sediment loads on
water quality and storage in the aquifer was beyond the scope of this study project.
Table 3.1 EMC?s in Danz Creek During Highway Construction
Parameter
Upstream Conc.
(mg/L)
Downstream Conc.
(mg/L)
Increase
Mean Median Mean Median (%)
Total Coliform (CFU)
a
2853 1875 2385 2050 9
Fecal Coliform (CFU) 2533 1825 10225 1750 -4
Fecal Strep (CFU) 6078 3900 4234 4200 8
TSS 34.8 13.9 179 79 470
VSS 4.8 2.1 20 4 88
Turbidity (NTU)
b
28 6 112 72 1100
BOD
5
2.8 2.7 3.2 2.2 -19
COD 16.3 11.2 14.8 13.3 18
Tot. Organic Carbon 22.9 19.1 22.1 18.4 -4
NO
3
-N 0.78 0.48 0.67 0.65 35
Oil and Grease ND ND ND ND NA
Zinc <0.029 0.023 <0.048 0.043 85
a) Colony forming units
b) Nephelometric turbidity units
 3.2 Highway Operation Effects
The volume of water which runs off a given highway crossing will be relatively
small and changes in water quality will probably not be measurable for bodies of water
with large catchments.  Monitoring of a small ephemeral stream near Austin, Texas,
allowed documentation of changes in water quality and quantity resulting from highway
runoff.  Water quality samples were collected during 34 runoff events after the opening of
a new highway.  Fourteen of these events were sampled both above and below the
Cadmium <0.009 0.004 <0.007 0.004 0
Chromium <0.012 0.007 <0.019 0.007 0
Copper <0.043 0.041 <0.048 0.046 11
Iron 0.699 0.358 2.697 2.489 595
Lead <0.154 0.042 <0.126 0.042 0
Nickel ND ND ND ND NA
11
highway right-of-way.  In this case, storm water runoff from the highway caused
significant changes in both the quantity and quality of water in Danz Creek.
The paved surfaces and storm sewer system combined to increase the total volume
and the maximum flow rate of the creek.  Small storm events were sufficient to generate
runoff below the highway right-of-way.
Storm water runoff from the highway caused increases in suspended solids, oil
and grease, and zinc in Danz Creek (Table 3.2).  These constituents commonly are found
in highway runoff.  The increases were substantial; however, the resulting water quality
was well within levels appropriate for aquatic life or at concentrations commonly
reported for streams during the elevated flows following storm events in undeveloped
watersheds.  Ambient concentrations of many constituents in the creek were higher than
those in the runoff from the new highway, because of the nature of the surrounding land
use.  High concentrations of fecal bacteria in the creek probably were derived from
livestock and wildlife in the watershed upstream of the highway right-of-way.  Nutrients
(nitrate and phosphorus) in the creek may have been the result of fertilization of the golf
course, through which the highway passes.  Dilution of the creek with highway runoff
reduced the concentrations of these constituents.  A detailed discussion of the surface
water monitoring at Danz Creek is contained in CRWR Technical Report #262 (Barrett et
al., 1995c).
12
Table 3.2 EMC?s in Danz Creek During Highway Operation
Zinc <0.009 0.006 <0.024 0.020 235
Upstream
Concentration
(mg/L)
Downstream
Concentration
(mg/L)
Increase
Parameter Mean Median Mean Median %
Total Coliform (CFU) 25727 27805 16343 12000 -57
Fecal Coliform  (CFU) 12380 12380 8650 2000 -84
Fecal Streptococcus (CFU) 35192 36000 23779 20500 -43
TSS 64 36 109 70 93
VSS 12 5 14 8 47
Turbidity (NTU) 25 20 34 29 43
BOD
5
5.6 6.0 4.3 4.0 -33
COD 45 41 30 26 -37
Total Carbon 45 46 34 26 -44
Dissolved Total Carbon 43 43 27 21 -52
NO
3
-N 0.30 0.28 0.17 0.11 -60
Total Phosphorus 0.36 0.26 0.12 0.13 -50
Oil and Grease 0.8 0.8 1.2 1.3 62
Cadmium <0.001 0.001 <0.002 0.001 0
Chromium <0.005 0.003 <0.009 0.002 -13
Copper <0.005 0.002 <0.003 0.002 -14
Iron 1.756 1.184 2.634 1.365 15
Lead <0.024 0.014 <0.049 0.014 0
Nickel <0.007 0.005 <0.006 0.005 0
13
 4.  QUALITY OF HIGHWAY RUNOFF IN AUSTIN, TEXAS
Water quality of highway runoff in the Austin, Texas, area was determined by
monitoring runoff at three locations on Loop 1 (MoPac), which represented different
daily traffic volumes, surrounding land uses, and types of highway drainage systems.
MoPac at West 35
th
 Street is a high traffic site (60,000 vehicles per day) located in central
Austin.  The land use of the area is mixed residential and commercial.  MoPac at Convict
Hill is a low traffic site (8700 vehicles per day) located on the southwestern edge of
Austin.  The land use around Convict Hill is mostly residential and rural undeveloped.
The area draining to the catchment basin at these two sites were 100 % impervious.  The
Walnut Creek site is located in north Austin and consists of a combination of paved
highway and grassy shoulder and median.  The land use classification of the area is
mostly commercial and high density residential, and approximately 47,000 vehicles per
day pass this location.  At Walnut Creek, the highway runoff crosses a large grassy
median before entering the storm sewer system where the samples were collected.
During rainfall events the runoff flow rates were measured, and samples were
collected automatically.  Water quality parameters analyzed in the laboratory for all
runoff samples included: turbidity, total and volatile suspended solids (TSS and VSS), 5-
Day Biochemical Oxygen Demand (BOD
5
), Chemical Oxygen Demand (COD), Total
Organic Carbon (TOC), oil and grease (O&G), nutrients (nitrate and total phosphorus),
heavy metals (iron, lead, cadmium, nickel, lead, zinc, and copper), and bacteria (total
coliform, fecal coliform, and fecal streptococcus).
Table 4.1 contains the means and medians of the EMC?s from individual storms at
the three sites.  The highest concentrations of all constituents were measured at the high
traffic site at 35
th
 Street.  The lowest concentrations were found at the Walnut Creek
monitoring site.  The concentrations at all sites were similar, with the exception of metals,
to median values compiled in a nationwide study of highway runoff quality (Driscoll et
al., 1990).  Concentrations of metals measured during this study are lower than those
which were measured several years ago.  The elimination of leaded gasoline has resulted
in an especially large reduction in the concentration of lead in highway runoff.  A
comparison of selected median EMC?s for 35
th
 Street and nationwide is shown in Table
4.2.
The concentrations of pollutants measured in highway runoff are also similar to
those commonly reported in urban runoff.  The City of Austin (COA) has conducted an
extensive monitoring program to characterize the water quality of urban runoff from
14
specific land uses (City of Austin, 1995).  A comparison between the storm water quality
from an area of single family homes and the low traffic site on MoPac is shown in Table
4.3.  A comparison between water quality from areas with commercial/industrial land use
and the high traffic site on MoPac is shown in Table 4.4.
Table 4.1 EMC?s of Constituents in Highway Runoff
Parameter 35
th
 Street Convict Hill Walnut Creek Rainfall
Median Mean Median Mean Median Mean Median
Total Coliform (CFU/100ml) 13000 48000 4200 7900 189000 145000 0
Fecal Coliform (CFU/100ml) 5800 13000 1000 22000 102000 116000 0
Fecal Streptococcus (CFU/100ml) 12000 16000 3800 17000 78000 89000 0
pH 7.15 6.94 5.61 6.14 6.51 7.16
TSS (mg/L) 131 202 118 142 19 27 0
VSS (mg/L) 36 41 20 22 7 7 0
BOD
5
 (mg/L) 12.2 16.5 5.0 6.3 3.5 4.1 ND
a
COD (mg/L) 126 149 40 48 35 33 6
Total Carbon (mg/L) 47 58 21 24 16 18 ND
Dissolved Tot. Carbon (mg/L) 25 31 11 14 13 15 ND
NO
3
-N (mg/L) 1.03 1.25 0.73 0.96 0.28 0.36 0.52
Total Phosphorus (mg/L) 0.33 0.42 0.11 0.13 0.10 0.10 0.05
Oil & Grease(mg/L) 4.1 6.5 1.7 2.2 0.5 0.5 ND
Cu (mg/L) 0.034 0.038 0.007 0.010 0.008 0.007 0.003
Fe (mg/L) 2.606 3.537 1.401 2.437 0.361 0.442 0.079
Pb (mg/L) 0.050 0.099 0.016 0.041 0.007 0.009 ND
Zn (mg/L) 0.208 0.237 0.050 0.077 0.022 0.019 0.019
a) ND = not detected
Table 4.2 Comparison of Median EMC?s from High Traffic Sites
Parameter MoPac at 35
th
 Street
(mg/L)
Driscoll et al. (1990)
(mg/L)
TSS 131 142
VSS 36 39
COD 126 114
NO
2
+NO
3
1.03
a
0.76
Copper 0.034 0.054
Lead 0.050 0.400
Zinc 0.208 0.329
a) NO
3
 only
15
Table 4.3 Comparison Low Traffic Site to Residential Land Use (Mean EMC)
Parameter Single Family Residential
(COA, 1995)
Convict Hill
Fecal Coliform (CFU/100ml) 34970 22000
Fecal Streptococcus (CFU/100ml) 44000 17000
TSS (mg/L) 171 142
BOD
5
 (mg/L) 9 6.3
COD (mg/L) 46 48
NO
3
-N (mg/L) 1.14 0.96
Total Phosphorus (mg/L) .29 0.13
Cu (mg/L) .010 0.010
Pb (mg/L) .016 0.041
Zn (mg/L) .046 0.077
Table 4.4 Comparison of High Traffic Site to Commercial Land Use (Mean EMC)
Parameter Industrial/Commercial
(COA, 1995)
West 35th
Fecal Coliform (CFU/100ml) 79850 13000
Fecal Streptococcus (CFU/100ml) 79130 16000
TSS (mg/L) 221 202
BOD
5
 (mg/L) 12 16.5
COD (mg/L) 93 149
NO
3
-N (mg/L) 1.82 1.25
Total Phosphorus (mg/L) 0.45 0.42
Cu (mg/L) 0.022 0.038
Pb (mg/L) 0.034 0.099
Zn (mg/L) 0.149 0.237
These comparisons show the general similarity between storm water runoff from
highway surfaces and from urban areas.  Much of the runoff generated in urban areas is
from streets, parking lots and other paved surfaces and may therefore, contain the same
pollutants as runoff from highways.  Because of the similarity in water quality, the storm
water control systems appropriate for improving the water quality of urban runoff would
be equally effective for treating highway runoff.  The water quality impacts of highway
storm water runoff would be similar to those produced by runoff from an urban area of
comparable size.
 The concentrations of pollutants measured in highway runoff in the Austin area
are below levels which would be toxic to aquatic life.  Numerous bioassays and toxicity
tests have been performed with highway runoff in other studies and acute toxicity
16
generally has not been demonstrated (Barrett et al., 1995a).  The impacts of highway
runoff alone, like many other nonpoint sources of pollution generally are not significant
when considered singly, but may result in degradation of water quality when combined
with other sources such as urban runoff.  The potential chronic effects of highway runoff
on aquatic species has not been adequately documented.  Constituents such as polycyclic
aromatic hydrocarbons (PAH) which were not monitored during this study, may also pose
a threat.
The total load of pollutant discharged into many receiving waters is more
important for estimating water quality impacts than is concentration.  Pollutant load is the
product of concentration and volume of runoff.  Table 4.5 shows the estimated annual
loads based on the mean EMC at each site.  Normalized for surface area, the greatest
loads were generated at 35th Street, while the lowest amounts were found at the Walnut
Creek monitoring site.  The monitored watershed at Walnut Creek had a runoff
coefficient of only about 10% while the other two sites had runoff coefficients of
approximately 90%.  The lower concentrations at Walnut Creek, combined with the much
lower flows at this site, were responsible for the low loads at this site.  Annual constituent
loads in untreated runoff from the new highways constructed in the recharge zone can be
estimated by multiplying the area of the new highways by the storm water loads at the
Convict Hill monitoring site.  As development continues and traffic density increases the
loads in untreated runoff will approach those measured at 35
th
 Street.
Table 4.5 Estimated Annual Pollutant Loadings (kg/ha)
Pollutant 35
th
 Street Convict Hill Walnut Creek
TSS 229 145 3.3
VSS 46 23 0.8
BOD
5
18.7 6.5 0.5
COD 169 49 4.0
Total Carbon 66 25 2.2
Dissolved Tot. Carbon 35 14 1.8
NO
3
-N 1.42 0.98 0.04
Total Phosphorus 0.48 0.13 0.01
Oil & Grease 7.36 2.25 0.06
Cu 0.043 0.010 0.001
Fe 4.008 2.497 0.053
Pb 0.112 0.042 0.001
Zn 0.269 0.079 0.002
17
A first flush effect (i.e., higher pollutant concentrations at the beginning of an
event) was very evident during selected events, but was generally limited to a small
volume.  The overall effect of first flush was small or negligible when all monitored
events were considered.  The concentrations appeared to be affected by changes in traffic
volume, rainfall intensity, and other factors.  In addition, vehicles provided a continuous
input of pollutants to the road surface and runoff for the duration of runoff events.  In
considering the potential effectiveness of storm water treatment systems, constant
concentrations for individual storm events should be assumed.  A detailed discussion of
this portion of the research project is contained in CRWR Technical Report 263 (Barrett
et al., 1995d)
19
 5.  FACTORS AFFECTING THE QUALITY OF HIGHWAY RUNOFF
One goal of this research project was identification of the variables which affect
the build-up and wash-off of constituents from highways in the Austin, TX area and
develop a water quality model which incorporates these variables.  The isolation of the
variables which influence highway runoff quality is facilitated during ?steady-state? storm
conditions (e.g., a constant rate of constituent input from rainfall and traffic).  A unique
rainfall simulator was used to produce steady-state storm events during this research.  The
rainfall simulator provided a uniform rainfall over a 230 meter length of 3-lane highway
during periods of active traffic (Figure 5.1).  The entirety of the runoff drained to a single
curb inlet where water quality samples were collected throughout the simulation.  The
length of highway exposed to the artificial rainfall allowed for the collection of water
which had washed from the bottoms of the moving vehicles.  This project marked the first
scientific use of a rainfall simulator in conjunction with active traffic.
Figure 5.1 Rainfall Simulator in Operation
Thirty-five rainfall simulations were conducted between July 6, 1993 and July 14,
1994.  Additionally, twenty-three natural storm events were sampled at the same location
during the same period of time.  Statistical analysis showed no significant difference
between the runoff generated by the rainfall simulator and the natural runoff.  The
samples collected during simulated and natural storm events combined to provide 423
storm water runoff observations.  Furthermore, twenty-one variables were identified for
20
each storm event and multiple regression analysis was used to determine the relationship
of each variable to the quality of the highway runoff.  The variables found to be
statistically significant were retained for use in a constituent-specific regression model.  A
comprehensive discussion of the methodology and results of this portion of the research is
contained in CRWR Technical Report 264 (Irish et al., 1995).
 5.1 Model of Highway Runoff Quality
The identification of the causal variables that significantly influence constituent
loading is among the more important findings of this study.  There are two major
applications of this knowledge.  First, recognition of the specific variables that influence a
given constituent load may suggest constituent-specific mitigation procedures, and
second, the applicability of the model is directly reflected in the causal variables.
The majority of variations observed in highway storm water loading in the Austin
area may be explained by causal variables measured during the rain storm event, the
antecedent dry period, and the previous rain storm event.  Significant causal variables
during the rainfall event include the duration of the event (min), the volume of runoff per
area of watershed (L/m
2
), the intensity of the runoff per area of watershed (L/m
2
-min),
and the average volume of traffic per lane.  The significant causal variables from the
antecedent dry period include the duration of the dry period (hr) and the average volume
of traffic per lane during the dry period.  The significant causal variables from the
preceding storm event include the duration of the event (min), the volume of runoff per
area of watershed (L/m
2
) and the intensity of the runoff per area of watershed (L/m
2
-min).
The dependent variable in the regression analysis is expressed as load (g/m
2
);
therefore, the total volume of flow during the storm event will appear in every constituent
model.  Similarly, the intensity of the runoff and the duration of the runoff also will
frequently appear in the models.  The variables flow, intensity and storm duration,
therefore, offer little diagnostic information in the interpretation of the model
specification.  However, the appearance of the other variables in the model, such as the
number of vehicles during the storm, the duration of the antecedent dry period, and the
volume of runoff during the previous storm event, are variables which ?control? the
constituent loading.  The examination of the controlling variables in each model adds
insight into the applicability of the model and the mitigation of constituent loading.
Variables related to season, such as day of the year, and air and water temperature, were
also tested for significance, but do not appear to affect pollutant loads.  A summary of
21
selected water quality constituents and their relevant causal variables is presented in
Table 5.1.
Table 5.1 Variables Affecting Pollutant Runoff Loads
Storm
Duration
Storm
Volume
Storm
Intensity
Vehicles
During
Storm
Length of
Antecedent
 Dry Period
Antecedent
Traffic
Count
Previous
Storm
Duration
Previous
Storm
Volume
Previous
Storm
Intensity
Iron * * *
TSS * * * *
Zinc * * * * * *
COD * * * * *
Phosphorus * * * *
Nitrate * * *
BOD
5
** * *
Lead * * * *
Copper * * *
Oil & Grease * *
As an example, 93% of the variation observed in the storm water loadings of total
suspended solids (TSS) are explained by the total volume of storm water runoff (L/m
2
),
intensity of the runoff (L/m
2
-min), total duration of the antecedent dry period (hr), and the
intensity of the runoff during the previous storm event (L/m
2
-min).  This model
formulation suggests that the conditions during the antecedent dry period (e.g., dust fall,
pavement/right-of-way maintenance activities, etc.) and the intensity of the preceding
storm event (e.g., the thoroughness of the previous wash-off) have a greater influence on
TSS storm water loadings than any of the other variables examined, including the traffic
volume during the storm event.  Therefore, efforts to mitigate the storm water loading of
TSS should be directed at activities during the antecedent dry period which deposit dirt
and debris on the highway surface.  Consequently, street sweeping was found to be
effective at reducing TSS loads.  Street sweeping on a once every two-week schedule, as
compared to no street sweeping, significantly reduced the average loads of TSS observed
in the highway storm water runoff.  However, no other constituent showed a significant
change in loading during the street sweeping period.
Highway runoff constituents, in general, fall into one of three categories: (1) those
constituents, such as TSS, that are influenced by conditions during the dry period and
may be mitigated by dry period activities such as street sweeping and others; (2) those
22
constituents that are most influenced by conditions during the rainfall event and may only
be mitigated through the use of runoff controls; and (3) those constituents that are
influenced equally by both periods.  The constituents that are significantly affected by
conditions during the preceding storm event, generally are those constituents that are
controlled by the dry period variables.
The variables which significantly affect the other highway runoff constituents are
detailed below:
? Nutrients:  The total duration of the storm event (min), total volume of storm water
runoff (L/m
2
), intensity of the runoff (L/m
2
-min), and the total volume of traffic
during the antecedent dry period (a measure of the length of the dry period) combine
to explain 95% of the variation in nitrate load, and 90% of the variation in total
phosphorus load, observed in the highway runoff.  This regression formulation is
strongly influenced by the quantity of these nutrients contained in the rainfall.  The
concentrations of nutrients observed in rainfall accounted for 50% to 100% of the
nitrate load, and up to 22% of the total phosphorus load, observed in the highway
runoff.  The mitigation of nutrients in highway runoff requires the use of runoff
controls.
? Organics:  The total duration of the storm event (min), total volume of storm water
runoff (L/m
2
), runoff intensity (L/m
2
-min), total volume of traffic during the storm,
and the total volume of traffic during the antecedent dry period combine to explain
86% of the BOD
5
 load, 95% of the COD load, 94% of the total carbon load, and 91%
of the dissolved total carbon load observed in the highway runoff.  The mitigation of
organics must be accomplished with runoff controls.
? Oil and Grease:  The total volume of storm water runoff (L/m
2
) and the total volume
of traffic during the storm combine to explain 94% of the variation in the oil and
grease loads observed in the highway runoff.  The mitigation of oil and grease must
be accomplished with runoff controls.
? Copper:  The total duration of the storm event (min), total volume of storm water
runoff (L/m
2
), and total volume of vehicles during the storm combine to explain 90%
of the variation in the copper load observed in the highway runoff.  The mitigation of
copper must be accomplished with runoff controls.
 
23
? Lead:  The total volume of storm water runoff (L/m
2
), runoff intensity (L/m
2
-min),
total volume of vehicles during the storm, and the intensity of the previous storm
runoff (L/m
2
-min) combine to explain 68% of the variation in the lead load observed
in the highway runoff.  The mitigation of lead must be accomplished with runoff
controls.
 
? Iron:  The total volume of storm water runoff (L/m
2
), runoff intensity (L/m
2
-min) and
the total duration of the antecedent dry period (hr) combine to explain 92% of the
variation in the iron load observed in the highway runoff.  The mitigation of iron must
be accomplished with dry period practices.
 
? Zinc:  The total duration of the storm event (min), total volume of storm water runoff
(L/m
2
), volume of vehicles during the antecedent dry period, total duration of the
previous storm (min), and the total volume of storm water runoff in the previous
storm (L/m
2
) combine to explain 92% of the variation in the zinc load observed in the
highway runoff.  The mitigation of zinc must be accomplished with both runoff
controls and dry period practices.
Traffic volume during the storm does not appear as a statistically significant
variable in every model formulation; however, traffic nevertheless, is an influential factor
in all constituent loading.  The storm water constituent wash-off patterns for high speed
highway pavements were found to be different during periods when traffic is using the
highway than during periods when there is no traffic.  The runoff from pavements with
high speed traffic does not exhibit as pronounced a first flush of constituent mass as the
runoff of pavements without traffic.  The continuous input of material from traffic insures
a continual increase in the cumulative constituent load throughout the duration of the
storm event.  As a result, highway watersheds which contain large shoulder areas or other
non-traffic bearing surfaces (e.g., > 35% of the total watershed) can be expected to
produce less constituent loading per unit of surface area than other highway pavements.
 5.2 Distribution of Highway Runoff EMCs
The quality of highway runoff varies from storm to storm and is highly variable.
The impact of highway runoff on receiving waters can be estimated using probability
distributions of runoff quality, which are normally developed from limited measurements
so that the effects of extreme events can be predicted.  The log-normal distribution is the
24
most commonly used probability density model for environmental contaminant data.  The
event mean concentrations (EMCs) of constituents in urban runoff, and highway runoff in
particular, have been described by the log-normal distribution.  The shape of the
underlying distribution must be known in order to select the statistics which will best
estimate the parameters of the population.  Methods that are used to evaluate
distributional shape include (1) probability plotting, (2) examination of the coefficient of
variation, (3) skewness, (4) kurtosis, and (5) normality testing with the Jarque-Bera
statistic.
Probability plotting commonly is used to determine the shape of an underlying
distribution.  Probability plotting can provide a quick determination of whether the data
are likely to have come from a specific type of distribution, however, the principal
application of the method is the determination of the mean and variance of the
distribution once the shape is known.
Normal and log-normal probability plots were constructed for all highway runoff
constituents in this study.  The results indicate that each constituent is best represented by
a skewed distribution.  However, reliance only on the ?straightness? of the plotted points
to determine the normality or non-normality of the distribution is risky at best.  A
probability plot can detect a skewed distribution; however, the plot cannot evaluate the
amount of skewness, which is imperative in the selection of descriptive statistics.
Therefore, probability plots have limited value for the determination of distributional
shape.  The normal and lognormal probability plots for the TSS data collected at the West
35
th
 Street site are presented in Figure 5.2.  
The Jarque-Bera statistic tests whether a series is normally distributed.  The
arithmetic mean, median, and variance of a sample is statistically an unbiased estimator
of the true population parameters regardless of the shape of the underlying distribution;
however, it is the minimum variance unbiased (MVU) estimator only if the underlying
distribution is normal.  The results of the Jarque-Bera test indicate that the skew in the
underlying distributions of the highway constituent EMCs is not statistically
distinguishable from a normal distribution.
25
TSS Normal Probability Plot
-200
0
200
400
600
800
1000
-3 -2 -1 0 1 2 3
Z Score
EMC (mg/l)
(a)
TSS Log Probability Plot
0
1
2
3
4
5
6
7
8
-3 -2 -1 0 1 2 3
Z Score
Ln(EMC)
(b)
Figure 5.2 Probability Plots of TSS Data
27
 6.  PERFORMANCE OF PERMANENT RUNOFF CONTROLS
The effectiveness of storm water controls for improving the quality of runoff from
operating highways was investigated in the field and laboratory.  A detailed discussion of
the methodology and results of the field and laboratory filtration studies is contained in
CRWR Technical Report 265 (Tenney et al., 1995), and the monitoring of the grassy
swale is described in Technical Report 263 (Barrett et al., 1995d).
 6.1 Pollutant Removal Effectiveness of a Grassy Swale
The effectiveness of grassy swales for treating highway runoff was evaluated by
comparing the runoff at Walnut Creek, before and after passing across a swale.  The
drainage system for this portion of MoPac consists of a grassy median which is underlain
by a storm sewer system.  Water enters the storm sewer through drop inlets after traveling
along the grassy swale for as much as 100 meters.  Runoff was collected from the storm
sewer system at the discharge to Walnut Creek.  The grassy swale is shown in  Figure 6.1.
Figure 6.1 Grassy Swale, MoPac at Walnut Creek
A second sampler at this site collected runoff directly from the paved surface of
the highway.  The combined system of a grassy swale and storm sewer provided
28
unexpected benefits even though the system was not designed specifically for water
quality improvement.  The removal of water via the drop inlets helped maintain a shallow
water depth in the swale.  The piping system also acted as an underdrain for the swale;
infiltration into the pipe supplied base flow to Walnut Creek event during the longest dry
periods.
The grassy swale proved effective for reducing the concentrations of most
constituents in runoff (Table 6.1).  The low runoff coefficient due to infiltration of runoff
into the swale produced an even larger reduction (90%) in pollutant load discharged
(Table 4.5).  This reduction of runoff volume effectively reduces the impact of
constituents whose concentrations are not reduced by the swale.  Large increases in
bacteria counts occurred in either the swale or the storm sewer system; however, the
bacteria probably do not indicate the presence of a significant human health threat.  A
constant seep of water is carried by the drain pipe.  This condition is conducive for the
growth of bacteria in the pipe.  This phenomenon may be associated with in increase in
the number of indicator organisms.
Table 6.1 Pollutant Removal Efficiency of a Grassy Swale
Parameter Roadway Grassy Swale Removal
(%)
Total Coliform (CFU/100mL) 3678 188197 -
Fecal Coliform (CFU/100mL) 1934 101545 -
Fecal Streptococcus (CFU/100mL) 6909 89482 -
TSS (mg/L) 104 27 74
VSS  (mg/L) 23 7 72
BOD
5 
 (mg/L) 7.5 4.1 46
COD  (mg/L) 51 33 35
Total Carbon  (mg/L) 34 18 48
Dissolved Tot. Carbon  (mg/L) 17 15 9
NO
3
-N  (mg/L) 0.88 0.36 59
Total Phosphorus  (mg/L) 0.15 0.10 31
Oil & Grease  (mg/L) 3.9 0.5 88
Cu  (mg/L) 0.014 0.007 49
Fe  (mg/L) 2.066 0.442 79
Pb  (mg/L) 0.014 0.009 35
Zn  (mg/L) 0.074 0.019 74
29
 6.2 Field Performance of Vertical Sand Filter Systems
A number of permanent runoff controls were constructed along the new highways
in the Edwards aquifer recharge zone and their performance has been monitored since the
highways opened.  The control systems consist of a hazardous material trap, a
sedimentation basin, and a vertical sand filter.  The filter is constructed as part of the wall
of the basin and held in place with filter fabric and rock gabions.  A typical system is
shown in Figure 6.2.
Figure 6.2 Typical Vertical Sand Filter System
Numerous problems were documented with these systems, mostly in conjunction
with the performance of the vertical sand filter.  Drainage rates observed for the control
systems varied from 30 to 50 hours for the faster draining systems, to several days for the
systems that drained slowly (Figure 6.3).  Channeling of the runoff through the filter
section may wash out the sand, resulting in inadequate detention times and no filtration.
In other systems, the filters clogged almost immediately creating permanent storage in the
sedimentation basin so that all subsequent runoff bypasses the control.  Accurate
determination of the pollutant removal effectiveness of these systems was not possible
because of these hydraulic problems.
The use of sand and geotextile fabrics in the vertical sand filters makes it difficult
to predict the drainage rate of these runoff control systems.  Drainage of the runoff from
the control system through the vertical filters is not controlled solely by the sand but also
is affected by the geotextile fabric that is used to support the sand between the rock
Hazardous
Material Trap
Vertical Sand
Filter
Detention Pond
Drain to
Creek
30
gabions.  Therefore, control systems designed based only on the hydraulic behavior of the
sand may not drain in 24 hours as called for in the design.
0
10
20
30
40
50
60
70
80
90
100
0 50 100 150 200 250
Hours After Runoff Event
Percent Remaining
N
M
L
K
A
B
Figure 6.3 Drainage of Six Runoff Controls after Storm on 5/16/94
The hazardous material trap (HMT) retains the first flush of runoff during a
rainfall event.  Therefore, the HMT cannot function as a hazardous materials collection
basin during runoff events when the control system is full of water, the roads are wet and
the chance for an accident is higher.
Sedimentation is the most important mechanism for pollutant removal in the
runoff control systems.  Removal of solids as a result of sedimentation was high in
control ?N? which provided minimal detention time.  Modifications of runoff control
systems which focus on extending the detention time of the basins may be more effective
in controlling suspended solids in runoff than enhancing the filter performance.  Scour
and resuspension of sediments was observed in the detention basins.  This phenomenon
causes increased suspended solids loadings on the filters resulting in discharge of higher
concentrations of suspended solids in the filter effluent and in clogging of the sand filter.
Sediment and suspended solids removal efficiencies can be increased and maintenance
requirements reduced by the installation of rock gabions, baffles or another device which
reduces resuspension of solids.
The estimate of the pollutant load contributed by the new highways ?pre? and
?post? control was an original goal of the study.  Because of the variability in the
31
performance and the numerous modifications of the controls installed on the new
highways no estimate of overall water quality improvement was possible.
 6.3 Laboratory Filtration Experiments
Laboratory, bench-scale filtration columns using various media were investigated
at the Center for Research in Water Resources.  The performance of filtration media and
adsorptive media was evaluated.  The bench-scale horizontally-bedded vertical-flow
filtration systems were dosed with stormwater runoff collected from an area highway.
Media selected for these experiments include a well-sorted medium grain size
sand, a fine aggregate, grade 5 gravel, compost, and zeolites.  The well sorted sand is
typical of that used in sand filtration systems in the Austin area.  The compost was
obtained from a company in Oregon which has used it successfully in runoff controls.
The zeolites were obtained locally and were tested because of their adsorption capability.
The zeolites were tested in combination with the fine sand.  In the latter case the column
was constructed with four inches of sand on top of four inches of zeolites.
The results of laboratory studies indicate that high removal efficiencies for
constituents in highway runoff can be achieved in horizontal (vertical flow) sand filter
columns (Table 6.2).  The data indicate that the compost is a very effective medium.  It
out performed the other media for the removal of TSS, oil and grease, and metals.
However, the compost decomposes and subsequent breakthrough occurs.  The medium-
sized sand performed well for the removal of TSS and most of the metals.  Clogging of
the 20-cm column of sand occurred prior to breakthrough; therefore, clogging is expected
to precede breakthrough in the field, where the filters are 90 cm across.  The column with
the medium-sized sand outperformed the column with the fine sand plus zeolites,
showing that the zeolites are not a promising medium for enhancing removal via
adsorption.  Negative removals were obtained for nitrate in all of the columns, and
indicate possible nitrification of ammonia which may occur in the columns.
Similar removal efficiencies were measured using concrete aggregate sand and the
Brady sand.  Pea gravel and grade 5 gravel are not effective filtration media.  The gravel
medium contained a significant fine portion which continued to wash out of the column
for the duration of the experiment, resulting in negative removal for TSS and associated
metals.  Grade 5 gravel installed in runoff controls serves only as a hydraulic control
device and not as a filtration media.
32
Table 6.2 Pollutant Removal Efficiencies of Alternative Media
Constituent Medium
Sand
Fine Aggregate Grade 5
Gravel
Compost Sand plus
Zeolites
TSS 75 55 -91 97 46
COD 38 23 7 36 35
Nitrate -66 -1 -17 -314 -269
Total Carbon 30 -4 0 -27 27
Oil & Grease 26 44 NA 59 20
Zinc 59 37 -6 86 60
Iron 48 53 -27 78 34
Copper 39 59 -17 61 13
The most effective runoff control system monitored during this study was a grassy
swale.  The swale reduced the concentrations of all pollutants monitored.  In addition, the
loss of runoff through infiltration to the subsurface resulted in an even greater reduction
of pollutant loads.  The vertical sand filters monitored on the new highway were not
effective because of problems associated with the sand filter.  Horizontal sand filters
which have been used for many years in this area have been shown to be relatively
effective for improving the quality of storm water runoff.  Sand filter systems may be the
best choice in urban areas where there is insufficient right-of-way for grassy swales.  The
use of alternative media in the filter systems appears to be promising.  In laboratory tests,
compost demonstrated greater removal of pollutants, except for nutrients, than all other
media tested.
33
 7.  CONCLUSIONS
The negative impact of highway construction on the quality of surface waters is
often significant despite improvements in the technologies used for erosion and sediment
control.  Monitoring of water quality in one of the creeks below a highway construction
site indicates that even an extensive system of temporary controls is not adequate to
prevent large amounts of suspended sediment from entering receiving waters.
Deterioration in water quality has certainly been less than would have occurred in the
past; however, the performance of even the best controls is often compromised by
inadequate installation and maintenance.
The development of adequate guidelines for the placement of temporary sediment
controls has been hampered by the lack of knowledge about the performance of silt fences
and other controls in a field setting.  In addition, parameters commonly used to
characterize geotextile fabrics were found to have little relevance for estimating their
sediment removal abilities or hydraulic characteristics under field conditions.  Accurate
prediction of the hydraulic properties would allow the estimation of appropriate drainage
areas to minimize over-topping of these temporary controls.
The quality of highway runoff was determined by monitoring the quality of storm
water runoff from three sites along the MoPac Expressway.  The quality of the runoff was
similar to that reported in other highway studies across the United States.  Little adverse
impact would be expected for all but the most sensitive receiving waters based on the
quantity and quality of highway of runoff generated during storms.  Concentrations of
many of the constituents in runoff were higher at the beginning of storm events (first
flush effect); however, the continued input of pollutants from vehicles and other sources
minimized the overall effect.  The concentrations of pollutants appeared to be affected by
changes in traffic volume, rainfall intensity, and other factors.
The use of the rainfall simulator allowed collection of sufficient data to formulate
a computer model that will predict the quality of runoff from operating highways in the
Austin, Texas area.  The majority of the variation observed in highway stormwater
loading could be explained by causal variables measured during the storm event, the
antecedent dry period, and the previous storm event.  This model allows prediction of
potential impacts of new highways as well as provides a screening tool to identify
existing highway segments that may threaten nearby receiving waters.
Permanent water quality controls may be required if stormwater runoff from a
highway is identified as a threat to environmental quality.  Unfortunately, structural
controls built on the new highway segments to protect the Edwards Aquifer have not
34
performed effectively.  The hydraulic performance of the vertical sand filters has been
uneven, resulting in little apparent improvement in runoff quality.  Testing of new filter
configurations is continuing with the goal of improving the performance of these systems.
 A grassy swale was found to be effective for reducing runoff volumes and
pollutant concentrations.  These grassy areas provide a low maintenance alternative to
structural controls where sufficient land is available and the topography is appropriate.
35
BIBLIOGRAPHY
Barrett, Michael E.,. Zuber, Robert D, Collins, III, E. R., Malina, Jr., Joseph F.,
Charbeneau, Randall J., and Ward, George H., 1995a, A Review and Evaluation of
Literature Pertaining to the Quantity and Control of Pollution from Highway Runoff and
Construction, 2nd Edition, Center for Research in Water Resources Technical Report
239, The University of Texas at Austin.
Barrett, Michael E., Kearney, John, E., McCoy, Terry G., Malina, Jr., Joseph F.,
Charbeneau, Randall J., and Ward, George H., 1995b, An Evaluation of the Performance
of Temporary Sediment Controls, Center for Research in Water Resources Technical
Report 261, The University of Texas at Austin.
Barrett, Michael E., Malina, Jr., Joseph F., Charbeneau, Randall J., and Ward, George H.,
1995c, Effects of Highway Construction and Operation on Water Quality and Quantity in
an Ephemeral Stream in the Austin, Texas Area, Center for Research in Water Resources
Technical Report 262, The University of Texas at Austin.
Barrett, Michael E., Malina, Jr., Joseph F., Charbeneau, Randall J., and Ward, George H.,
1995d, Characterization of Highway Runoff in the Austin, Texas Area, Center for
Research in Water Resources Technical Report 263, The University of Texas at Austin.
City of Austin (COA), 1995, Characterization of Stormwater Pollution for the Austin,
Texas Area, Environmental Resource Management Division, Environmental and
Conservation Services Department.
Driscoll, Eugene D., Shelley, Philip E., and Strecker, Eric W., 1990, Pollutant Loadings
and Impacts from Highway Stormwater Runoff, Vol. I: Design Procedure, Federal
Highway Administration, Office of Research and Development Report No. FHWA-RD-
88-006.
Irish, Lynton B., Jr., Lesso, William G., Barrett, Michael E., Malina, Jr., Joseph F.,
Charbeneau, Randall J., and Ward, George H., 1995, An Evaluation of the Factors
Affecting the Quality of Highway Runoff in the Austin, Texas Area, Center for Research
in Water Resources Technical Report 264, The University of Texas at Austin.
Tenney, Sean, Barrett, Michael E., Malina, Jr., Joseph F., Charbeneau, Randall J., and
Ward, George H., 1995, An Evaluation of Highway Runoff Filtration Systems, Center for
Research in Water Resources Technical Report 265, The University of Texas at Austin.