Selection

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

4th International SWAT Conference

Rapid Geomorphic Assessment of Watershed Sediment Budgets for Water


Supply Reservoirs Using SWAT and Sub-Bottom Acoustical
Profiling
P. M. Allen1/, B. Narasimhan2/, J. Dunbar1/, S. Capello1/ and R. Srinivasan2/
2 Spatial

1 Department of Geology, Baylor University, Waco, TX 76798


Sciences Laboratory, Texas A&M University, College Station, TX 77843.

Reservoirs in North Texas supply over 99 percent of the water to the Dallas Fort
Worth Metroplex or to over 5.8 million people. The rate of sedimentation to these reservoirs
has been estimated to be in the range of 2 cm a year. This is equivalent to an annual loss of
future water supply for up to 3000 people per reservoir (lost storage capacity). The costs of
new reservoirs as well as the cost of dredging makes options other than watershed
management (BMPs) unrealistic. In order to best manage sedimentation into the reservoir,
the source of sediment into the reservoir must be quantified. This study represents the
approach used by the Tarrant Regional Water District to study reservoir sedimentation. It
consists of three integrated steps: (1) rapid geomorphic assessment by sub-watershed and
physiographic province to ascertain the magnitude of active processes; sheet/rill; gully;
stream erosion supplemented with historic air photographic inventory (2) profiling the
reservoir with differential GPS and a 5 transducer sub-bottom acoustical system which gives
sediment thickness within the reservoir which is verified with vibracore and Cesium 137
analysis, and (3) modeling the watershed with the SWAT model calibrated to the total
volume of sediment as detailed in the survey, watershed erosion processes, any gaged data,
climate and land use. Failure to follow each step is shown to produce large errors in the
model and potential effectiveness of BMPs.
INTRODUCTION
A sediment budget is an accounting of the sources and disposition of sediment from
its point of origin to its eventual exit from a watershed. Although erosion from upland
sources is most often cited as a major source of sediment load; channel/streambank erosion
can also be a major source in some watersheds contributing from 17% to 93% of the total
sediment load (Trimble 1997 and Sekely et al. 2002). Watershed scale models such as
SWAT, HSPF and GWLF are widely used to assess sediment and nutrient budgets within the
watersheds. According to the EPAs recent National Water Quality Inventory report (USEPA
2002), 31% of the river and stream miles are impaired due to sediment; sediment/siltation is
the second leading cause of water quality impairment, next only to pathogens. In this
research, watershed scale model SWAT is used to quantify the sediment budget of the Cedar
Creek reservoir watershed, located in North-Central Texas.
More specifically, the objectives of this study are:
1)
2)

To test SWAT channel sediment routing routine against field geomorphic data, and
To illustrate the use of rapid geomorphic assessment of watersheds for use in SWAT
SWAT sediment and water routing routines

UNESCO-IHE
Delft, The Netherlands

63

4th International SWAT Conference

In SWAT a simplified stream power approach developed by Bagnold (1977) and


adopted by Williams (1980) is used for estimating channel degradation or deposition. The
maximum amount of sediment that can be transported in a reach segment is calculated as:
sp exp
concsed , mx = spcon vch
(1)
where concsed,mx is the maximum concentration of sediment that can be transported by
the water (ton/m3), spcon is a coefficient defined by the user, vch is the peak channel velocity
(m/s), and spexp is an exponent defined by the user.
If the maximum concentration of sediment calculated with Eq.1 is lesser than the
concentration of the sediment entering the reach then deposition is the dominant process in
the reach segment. Conversely, if the maximum concentration of the sediment calculated
with Eq.1 is higher than the concentration of the sediment entering the reach then channel
degradation is the dominant process in the reach segment. Sediment degradation is calculated
as:
sed deg = (conc sed ,mx conc sed ,ch )Vch K ch C ch
(2)

where seddeg is the amount of sediment degraded/reentrained from the reach segment
(metric tons), concsed,ch is the initial sediment concentration in the reach (ton/m3), Vch is the
volume of water in the reach sediment (m3), Kch is the channel erodibility factor and Cch is the
channel cover factor.
As can be seen from the above equations, flow velocity (vch) in Eq.1 is an important
variable in determining the streambank erosion or deposition from a reach segment. In
SWAT water can be routed from upstream to downstream using either the variable storage or
Muskingum routing methods. In both these methods a bucket type approach is used for
calculating the amount of water entering the reach, i.e. the water is moved from one reach to
another by volume basis. In the bucket type approach the flow rate in each segment is also a
function of stream length (Narasimhan et al. 2007). Hence, the flow velocity predicted is
artificially higher at a smaller reach downstream of two big reaches (due to the nature of
watershed delineation). Hence, an iterative approach developed by (Narasimhan et al. 2007)
was used (Figure 1).
q ch =

Vch
(24 60 60)

Do While (qit < qch)


Dch = Dch + 0.01
Ach = (Wbtm + zch * Dch) * Dch
Pch = Wbtm + 2. * Dch * Sqrt(1. +zch2)
Rch = Ach / Pch
v ch =

1 2 / 3 1/ 2
Rch S ch
n

qit = vch * Ach


end do
Figure 1. Modification to water routing algorithm
where qch is the flow rate of water entering the channel (m3/s) and Vch is the volume
of water entering the channel during the day (m3/day).
The flow velocity calculated by this modified approach is no longer a function of
length and hence the flow velocity will vary gradually independent of length as the water
moves downstream. In this research, the modifications discussed above were made to the
variable storage routing method for hydrology simulation in the Cedar Creek watershed.

UNESCO-IHE
Delft, The Netherlands

64

4th International SWAT Conference

Study Area
Water quality is a growing concern in the Cedar Creek reservoir (Figure 2). In
addition to nutrient enrichment and reservoir siltation from upland and streambank erosion
affects the quality of drinking water and the life of this reservoir. On an average 1032 ac-ft
of reservoir volume (679,200 ac-ft) is lost every year due to siltation. Cedar Creek is one
among the five major reservoirs managed by Tarrant Regional Water District (TRWD) in the
Trinity River Basin. Currently TRWD supplies water to about 1.6 million people across 11
counties from its network of five reservoirs and are expected to serve 2.66 million by 2050.
Due to this projected population increase in North Central Texas over the next 50 years,
TRWD is collaborating with the Texas Water Resources Institute and Texas A&M University
to assess the watershed condition and develop watershed protection plans for these five major
reservoir catchment. Cedar creek reservoir has a contributing drainage area of about 2611 sq.
km. The average annual rainfall is about 991 mm. The western half of the watershed is
underlain by shale and residual clay soils; the eastern half of the watershed is underlain by
interbbedded sands and shales and mostly loam to sandy soils. The land cover is
predominantly pasture (62%), followed by forest (16%), Urban (7%), cropland (6%) and rest
occupied by water (6%) and wetlands (3%) types.

Figure 2. Land cover classes, sub-basin and stream network of Cedar Creek watershed.

UNESCO-IHE
Delft, The Netherlands

65

4th International SWAT Conference

WATERSHED AND LAKE SURVEY


In order to assess the accuracy of SWAT model output, various Rapid Geomorphic
Methods can be utilized. The methods found to be useful in the Cedar Creek Watershed are
shown in Table 1.
I

II

Rapid Method
Point Monitoring with
erosion pins/scour
gages; sediment
monitoring
Volumetric Monitoring
of Downstream
reservoirs

III Photo Interpretation

IV Literature Rates and


Equations

Visual site geomorphic


Assessment

Time Frame
Comments
Individual storm- Costly for large watersheds; gage sites
years
very costly to install and maintain;
cheaper duratrac (Figure 10.) can give
quantitative cumulative data
Repeat
Volumetric not real good for
bathymetric
sedimentation amounts; acoustic methods
surveys
are superior and sediment flux possible
(decades) or
with Cesium-137 (for reservoirs older
Sub-bottom
than 1964)
acoustical
decadal
Problems for small streams which make
up to 70% of watersheds owing to riparian
vegetative cover.
Not applicable.
Due to wide scatter of data, empirical
equations require a lot skill to interpret
results; land use impacts hard to
discriminate; typically short term data.
Individual storm Gives qualitative estimate of recent trends
or trend
and rates

Table 1. Rapid Geomorphic Assessment Methods


The first problems to assess are the basic elements of the sediment budget within the
watershed. A more detailed list is given by Reid and Dunne (1996). For the Cedar Creek
watershed, the major contributors found in the field were sheet, rill and ephemeral gully
erosion, and concentrated flow erosion (gully erosion and channel erosion). In order to verify
the output of the SWAT model, various methods were utilized to assess the magnitude of
each contributing sources. It was found easiest to work backward from the most inclusive
source of data or the reservoir sediment survey. The sediment survey was done to estimate
the density and volume of the post-impoundment reservoir sediment. A Vibracoring
technique was used for obtaining undisturbed cores of unconsolidated sediment at saturated
or nearly saturated conditions. Five cores were collected within the reservoir. The core
locations were determined with the aid of acoustic sub-bottom profiling, Figure 3.

UNESCO-IHE
Delft, The Netherlands

66

4th International SWAT Conference

Figure 3. Core Locations Cedar Creek Reservoir


The cores were used to determine the thickness, density and dry weight of postimpoundment sediment in the reservoir, Figure 4. Acoustic sub-bottom profiling of the lake
showed that the sediment is uniformly distributed along the length of the lake with an average
thickness of 37.4 cm. The average dry density of the sediment was estimated to be 344.5
kg/m3. A reservoir volumetric survey conducted by Texas Water Development Board
(TWDB) in 2005 showed that the lake has lost a total of 41,276 acre-ft during the 40 year
period since its construction in 1965. (Cesium 137 could not be used in this reservoir). The
volume loss is attributed to the deposition of sediment from the supplying watershed.
Multiplying this sediment volume and the average dry weight density determined in the
current study and by assuming a reservoir sediment trap efficiency of 98%, indicates that the
total dry weight of the post-impoundment sediment in the lake is 17,862,337 Metric Tons
over the 40 year period or an average sedimentation rate of 446,558 Metric Tons/year. In
order to quantify the individual contribution of sheet and rill versus concentrated flow
erosion, supplemental methods were employed.

Figure 4. Cores for assessing sediment thickness and density


For sheet and rill erosion, regional estimates based on regression equations for the
same physiographic provinces from sediment gage data, past USLE and sediment delivery
calculations and visual assessment of soil loss by NRCS personnel were utilized. For channel

UNESCO-IHE
Delft, The Netherlands

67

4th International SWAT Conference

and gully erosion, a comprehensive field survey was done at 56 sites across the watershed to
identify eroding stream segments in the watershed. At each site a survey form was used to
categorize the level of channel incision as well as degree of erosion in addition to several
other parameters, Figure 5.
Streambank erosion from the watershed is estimated to be 152,572 Metric Tons per
year. This is based on five different methods of channel erosion assessment: (1) erosion
assessment made for the basin based on NRCS field evidence (Griener, 1982), (2) field
assessment of channel erosion and SWAT generated channel lengths and dimensions using
theRapid Assessment Point Method (RAP-M) used by NRCS (Windhorn, 2001), (3) using
power functions utilized in SEDNET (Wilkinson, et. al. 2004), (4) comparison of erosion
rates to gage data by Ecoregion after Simon, et. al. (2004) and (5) literature review of channel
erosion rates. Based on these methods, channel erosion could be assigned an average annual
loss rate. Volumetric loss was calculated by the product of channel length times the erosion
rate and the soil density and eroded channel height. In addition to the field survey, historical
aerial photographs were also analyzed to identify historical trends in erosion across the
watershed. Based on land use, stream condition at surveyed locations, and using historical air
photographs, the streambank erosion categories were extended between field locations to the
entire watershed area (Figure 6). The streambank erosion estimate along with the streambank
erosion category mapped from site visits and field photographs were used for both assessing
original SWAT model output, and then in calibrating the SWAT Streambank erosion
parameters erodibility factor (Kch) and channel cover factor (Cch).
Based on this reservoir sedimentation rate (446,558 Metric Tons/year) and the
Streambank erosion rate (152,572 Metric Tons/year) from the field survey, the overland
erosion rate was inferred as 293,986 Metric Tons/year. These erosion rates were used as
target numbers for calibrating the SWAT model parameters.
Results
Flow calibration and validation
SWAT was calibrated for flow by adjusting appropriate inputs that affect surface
runoff and base flow. Adjustments were made to runoff curve number, soil evaporation
compensation factor, shallow aquifer storage, shallow aquifer re-evaporation, and channel
transmission loss until the simulated total flow and fraction of base flow were approximately
equal to the measured total flow and base flow, respectively. Flow calibration was performed
from 1963 through 1987. For this period predicted flow matched measured flow very well at
the two USGS stream gages in the watershed. With the same calibration inputs, flow was
validated from 1980 through 2002 using the measured mass balance of Cedar Creek
Reservoir for comparison to predicted inflow values (Figure 7). The predicted inflow match
measured inflow very well (r2 = 0.76 and Nash-Sutcliffe COE of 0.80).

UNESCO-IHE
Delft, The Netherlands

68

4th International SWAT Conference

Figure 5. Field Survey Form

Figure 6. Erosion Ratings and Sites

Average Monthly Inflow to Cedar Creek Reservoir


140

100

Measured Mean = 18.37 cms


Predicted = 19.18 cms
RMS Error = 11.70
R2 = 0.76; E: 0.80

Discharge (m /s)

120

80
60
40
20
0
1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

Year
Measured

Predicted

Figure 7. SWAT monthly streamflow validation with measured reservoir inflows.

UNESCO-IHE
Delft, The Netherlands

69

4th International SWAT Conference

Sediment Calibration
Annual overland and Streambank erosion rates determined from watershed survey and
lake survey were used as target rates for adjusting model parameters. Streambank erosion,
power function parameters, spcon (0.01) and spexp (1.5) in Eq.1 were adjusted based on
limited storm flow total suspended solids (TSS) data available at various stream segments.
The coefficients were chosen in such a way that the average simulated suspended sediment
concentration is two to three times higher than the measured TSS. This was done so because
SWAT does not simulate bed load transport and all the sediments are assumed to be in
suspension. If these (spcon and spexp) coefficients were tightly calibrated with measured
TSS without accounting for bed load transport, the model will considerably underestimate the
sediment transport power of the water.
Channel physical properties such as channel vegetation cover factor (Cch) (0.1 to 1.0)
and channel erodibility factor (Kch) (0.3 to 0.8) were adjusted for individual stream segments
based on field assessment and geological data. The higher these values are, greater is the
vulnerability of the channel for streambank erosion and vice versa (Eq.2).
Annual sediment contribution simulated from various sources is shown in Figure 8. Channel
degradation is the second major contributor of sediment (33%) next to cropland (44%). It is
important to note that SWAT predicted overland erosion was very close to the field
estimation; hence no adjustments were made to the SWAT overland erosion parameters. The
spatial distribution of Streambank erosion rates is shown in Figure 9. Overall there is a good
agreement between the field assessment (Figure 6) and the model estimates (Figure 9).
However, when comparing these maps it should be kept in mind that the field assessment
technique is only a qualitative measure of erosion based on visual analysis.
Cedar Creek (Sediment)

About 96.6% of the total sediment generated


in the watershed reaches the reservoir

SWAT Predictions:
Total Sediment load : 467,730 Metric Tons/yr
Channel Erosion
: 162,528 Metric Tons/yr
Overland Erosion
: 305,046 Metric Tons/yr
Total sediment reaching the reservoir : 451,600 Metric Tons/yr
(Trapping by upstream dams/PL655 structures: 16,130 Metric

WWTP
0.03%

Urban
7.40%
Forest
0.75%

Channel
34.75%

Wetland
0.10%

Cropland
40.78%

Range
0.39%

Baylor Field Estimation:


Pasture
15.78%

Total Sediment Load: 446,558 Metric Tons/yr


Channel Erosion
: 152,572 Metric Tons/yr
Overland Erosion : 293,986 Metric Tons/yr

Figure 8. Sediment contribution from channel and overland erosion.

UNESCO-IHE
Delft, The Netherlands

70

4th International SWAT Conference

Figure 9. Streambank erosion rate predicted by SWAT in terms of Metric Tons per unit crosssection area and per unit length of the channel.
In terms of streambank and gully erosion, a new method was created for longer term
assessment of erosion rates. In conjunction with erosion pins, a devise called the duratrac was
formulated (Figure 10). Basically, the device is a tachometer which records with a quartz clock
the water level at or above a selected height. In this way, the cumulative tractive force of a set of
erosion pins can be known and the SWAT model erosion coefficients adjusted accordingly. The
device costs about 35-50 US dollars and will run about six months or more under normal
conditions. The device could also be used to assess the length of time the water was over bankfull
stage. By placing these simple devices over the watershed, cheap accurate data can be collected
for any chosen time interval; the only cost after installation is associated with a trip to the field to
measure the erosion pins and visually download the cumulative time.

Figure 10. SWAT Duratrac recorder.


SUMMARY
A rapid geomorphic assessment in the field is useful to assess the SWAT model outputs
given the paucity of real gage data. Methods employed can be accomplished at about 250 sq.
km.a day. This would include field assessments of streams at road crossings and visual inspection

UNESCO-IHE
Delft, The Netherlands

71

4th International SWAT Conference

of surrounding terrain for evidence of landuse and erosion processes. A modification to the
velocity subroutine is proposed. Finally a simple device to be employed to assess cumulative
tractive force for stream or gully erosion assessment is introduced.
REFERENCES
Arnold, J. G., R. Srinivasan, R. S. Muttiah, and J. R. Williams. 1998. Large-area hydrologic
modeling and assessment: Part 1. Model development. J. American Water Res. Assoc.
34(1): 73-89.
Bagnold, R.A. 1977. Bedload transport in natural rivers. Water Resour. Res. 3:303-312.
Narasimhan, B., Allen, P.M., Srinivasan, R., Bednarz, S., Arnold, J. and Dunbar, J.A., 2007.
Streambank erosion and best management practice simulation using SWAT. ASABE
Fouth Conference on Watershed Management to Meet Water Quality Standards and
emerging TMDL, March 11-13, San Antonio, Texas.
Reid, L.M. and Dunne, T., 1996. Rapid evaluation of sediment budgets.
Reiskirchen:Germany, Catena Verlag (GeoEcology Press) 164p.
Di Luzio, M., R. Srinivasan, and J. G. Arnold. 2002. Integration of watershed tools and SWAT
model into basins. J. American Water Res. Assoc. 38(4): 1127-1141.
Neitsch, S. L., J. G. Arnold, J. R. Kiniry, J. R. Williams, and K. W. King. 2002. Soil and water
assessment tool. Theoretical documentation: Version 2000. TWRI TR-191. College
Station,Texas: Texas Water Resources Institute.
Sekely, A. C., Mulla, D. J., & Bauer, D. W. (2002). Streambank slumping and its contribution
to the phosphorus and suspended sediment load of the Blue Earth River Minnesota. J.
Soil and Water Cons. 57(5), 243250.
Trimble, S.W. 1997. Contribution of stream Streambank erosion to sediment yield from an
urbanizing watershed. Science. 278: 1442-1444.
USEPA. 2002. National Water Quality Inventory: 2000 Report. EPA 841-R-02-001. USEPA:
Washington, DC.
Williams, J.R. 1980. SPNM, a model for predicting sediment, phosphorus, and nitrogen
yields from agricultural basins. Water Resour. Bull. 16:843-848.
Windhorn, R.H., 2001. RAP-M: Rapid Assessment Point Method. NRCS, Champaign,Ill.
Misc. pages.

UNESCO-IHE
Delft, The Netherlands

72

You might also like