Professional Documents
Culture Documents
Hydrologic Model For Design and Constructed Wetlands
Hydrologic Model For Design and Constructed Wetlands
Hydrologic Model For Design and Constructed Wetlands
167–178
䉷 2001, The Society of Wetland Scientists
2
Geology Department
Baylor University
Waco, Texas, USA 76798
Halff Associates
3
Abstract: The Trinity River Mitigation Bank was proposed to develop and use a mature, contiguous, diverse
riparian corridor along the West Fork of the Trinity River near Dallas, Texas, USA. In the proposed wetland
design, water would be diverted from Walker Creek as necessary to maintain wetland function. Therefore,
assessment of the magnitude and continuity of the flow from Walker Creek was paramount to successful
wetland operation. The Soil and Water Assessment (SWAT) model was used to assess whether the sustained
flow (storm flow and base flow) from the Walker Creek Basin could maintain the proposed bottomland
wetland ecosystem. For this study, SWAT was modified to allow ponded water within the prescribed wetland
to interact with the soil profile and the shallow aquifer. The water budget was prepared for the wetland
based on a three-step process. First, data required to run the model on Walker Creek, including soils,
topographic, land-use, and daily weather data were assembled. Next, data required to validate the model
were obtained. Since stream flow was not available at the proposed site, flow from a nearby watershed with
similar soils, land use and topography were used. In the final step, the model was run for 14 years and
compared to the measured water balance at the nearby watershed. The model results indicate that the wetland
should be at or above 85 percent capacity over 60 percent of the time. The wetland did not dry up during
the entire simulated time period (14 years) and reached 40 percent capacity less than one percent of the time
during the simulation period. The advantages of the continuous simulation approach used in this study include
(1) validation of wetland function (hydroperiod, soil water storage, plant water uptake) over a range of
climatic conditions and (2) the ability to assess the long-term impact of land-use and management changes.
Key Words: watershed model, base flow filter, water balance, hydroperiod
167
168 WETLANDS, Volume 21, No. 2, 2001
lands (Carter 1986, Owen 1995). Koob et al. (1999) compilation and availability of regional data bases
cites several problems with current methodologies, from which to run and calibrate the models. Water-
which include oversimplification and extrapolation of budget models often used in evaluating wetlands in-
limited amounts of data to predict design flows into clude DRAINMOD (Skaggs 1984), FLATWOODS
wetlands. Dexler et al. (1999), in a detailed study of (Sun et al. 1998), a Wetlands Dynamic Water Budget
a small peatland, concluded that there was a wide mar- Model (Walton et al. 1996), and a Soil Water Balance
gin of error in all components of the water budget, Model (Bidlake and Boetcher 1996). This paper de-
with the exception of precipitation. In the same study, scribes the SWAT (Soil and Water Assessment Tool)
error in other similar water budgets of lakes and wet- model (Arnold et al. 1998) for use in preliminary de-
lands were compared, and estimates ranged from 5 to sign of constructed wetlands. The structure and testing
10% for precipitation, 20 to 40% for evapotranspira- of the model is illustrated, and a case study of using
tion, from 5 to 60% for surface runoff, and from 25 the model for a proposed constructed wetland is pre-
to 100% for ground-water flow. In addition to these sented.
problems, very few of the water budgets have been
done for extended periods of time (through several
PROJECT: CASE STUDY
wet/dry cycles and perhaps through drought periods).
Skaggs et al. (1991) illustrates problems with moni- The Trinity River Mitigation Bank was proposed to
toring even for periods of up to 3–5 years. He states develop and use a mature, contiguous, diverse riparian
that in Wilmington, North Carolina, there are instances corridor along the West Fork of the Trinity River as a
in the 40 years of record that the water table may not Mitigation Bank for U. S. Army Corps of Engineers
meet the criteria for 3–4 years. If the monitoring was authorized wetland impacts (Federal Register 1995).
initiated during such a period, wetland function would Wetland banking allows developers to purchase credits
not have been properly assessed. at large regional wetland ‘‘banks’’ rather than creating
Due to the problems associated with monitoring and smaller less efficient ‘‘pocket’’ wetlands. This pro-
the continued need for such budgets in predicting the posed project encompasses four tracts totaling approx-
viability of a constructed wetland, other methods have imately 560 ha in Fort Worth, Tarrant County, Texas,
been proposed to predict components of the budget for USA (Figure 1). The site consists of a unique array of
constructed wetland design. There seem to be three riparian habitats, sloughs, deep-water ponds, bottom-
general approaches used (Koob et al. 1999): (1) single land forests, and upland forests and grasslands. The
event models, (2) stochastic models involving one pa- site is located on the edge of the Cross Timbers/Black-
rameter, and (3) complete water-budget models. Single land Prairies vegetation zones of Texas. Precipitation
event models use rainfall of a given duration, frequen- averages 865 mm per year. The proposed 560 ha bank
cy, and amount, coupled with site soils and land-use is composed of approximately 30 ha of jurisdictional
conditions to predict runoff. Such estimates of runoff Wetlands/Waters of the United States and 102 ha of
are then used with general design criteria to construct active gravel mining pits and depressions.
the wetlands. For example, Schueler (1992) advocate A water-budget model was prepared for the pro-
that wetland volume should be based on hydraulic re- posed wetland site. In the proposed wetland design,
tention time and the runoff volume from the 90% water would be diverted from Walker Creek as nec-
storm. Pierce (1992) detailed use of standard meth- essary to maintain wetland function. Therefore, as-
odologies for assessing such event-based predictions sessment of the magnitude and continuity of the flow
as well as averages of different components of the wa- from Walker Creek was paramount to successful wet-
ter budget. land operation. The SWAT model was used to assess
The second approach is forwarded by Koob et al. whether the sustained flow (storm flow and base flow)
(1999) in which they propose a stochastic framework from the Walker Creek Basin could maintain the pro-
to examine the probability of extended dry periods and posed bottomland wetland ecosystem. The proposed
daily seepage and evapotranspiration losses in order to wetlands were to have a combined surface area of
give some level of assurance to maintaining a specified 15.4 ha. The average depth of the wetlands was 0.46
wetland water volume. This method strives to provide m, as calculated from topographic maps. The site lies
the planner with some level of acceptable risk to as- within the computed 100-year floodplain of the Trin-
sociate with design. ity River.
The final method involves use of models that at- Walker Creek, which flows along the east side of
tempt to simulate all or the majority of the components the proposed wetlands, drains 51.8 km2 of urbanized
of the water budget. These models have been growing uplands. The basin is not monitored by a U.S. Geo-
rapidly based on the decreasing costs of computer logical Survey gaging station. The basin lies almost
memory, increased computational speed, as well as the totally within the outcropping area of Cretaceous
Arnold et al., HYDROLOGIC MODEL FOR CONSTRUCTED WETLANDS 169
Figure 1. Location of Walker Creek watershed and the proposed wetlands. Dark shaded areas are deep wetlands (⬎0.5m)
and light shaded areas are shallow wetlands (⬍0.5m).
Woodbine Sand, a minor aquifer consisting of inter- north of Walker Creek basin, was chosen as a com-
bedded sands and shale. The main channel length of parison basin. Big Bear Creek (U.S. Geological Survey
the basin is 16.1 km, and the channel slope is about Stream Gage Station Number 08049550) has been
0.0039 m/m. The soils consist of 14% B Hydrologic monitored from 1966 to 1979. This basin lies within
Group Bastil Series and about 86% D Hydrologic the same geological formation and contains the same
Group Navo, Crosstel, and Radar Series soils (USDA soils, land use, topography, and climate as Walker
1977). The dominant land use in the basin is residen- Creek basin. Big Bear Creek basin has a slightly larger
tial. Since measured flow is not available on Walker drainage area of 76.7 km2, a similar slope of 0.0032
Creek, Big Bear Creek basin, directly adjacent to the m/m, and a main channel length of 23.8 km.
170 WETLANDS, Volume 21, No. 2, 2001
Figure 2. SWAT Hydrology Model Flowchart for the Walker Creek and the proposed wetlands.
There are two possible cases that can occur when the surface of the wetland, a daily water balance is
simulating a wetland: ponded water and no ponded maintained with the equation:
water. If the wetland is dry (no ponded water), soil
䉭Vw ⫽ I⫺et⫺sp⫺rf (1)
water processes and plant growth are simulated as de-
scribed in Arnold et al. (1998). Water movement for where 䉭Vw is the change in wetland water volume, I
the ‘‘no ponded water’’ case can be tracked in Figure is inflow, et is evapotranspiration, sp is seepage into
1 starting with precipitation on the watershed. In this the soil, and rf is rainfall on the wetland, with all units
case, surface runoff is zero. When water is ponded on in cubic meters. Inflow is diverted from Walker Creek
172 WETLANDS, Volume 21, No. 2, 2001
Water
Drainage Yield/ Sur- GW Plant
Area Stream- Soil face Base Soil GW Re- Bio-
Location Reference (km2) flow Water Runoff Flow ET ET charge mass
1. Middle Bosque River,
Texas Arnold et al. (1993) 471 X X X X
2. Coshocton, OH Arnold and Williams (1985) lysimeter X
3. Bushland, TX Arnold and Williams (1985) field plot X X
4. Riesel, TX Savabi et al. (1989) 1.3 X X X
Sonora, TX Savabi et al. (1989) 4.1 X X
5. Seco Creek, TX Srinivasan and Arnold (1994) 114 X
6. Neches River Basin, TX King et al. (1999) 25,032 X
7. Colorado River Basin, TX King et al. (1999) 40,407 X
8. Lower Colorado, TX Rosenthal et al. (1995) 8,927 X
9. White Rock Lake, TX Arnold and Williams (1987) 257 X
10. North Carolina Jacobsen et al. (1995) 4.6 X X
11. Goose Creek, IL Arnold and Allen (1996) 246 X X X X X X X
12. Hadley Creek, IL Arnold and Allen (1996) 122 X X X X X X X
13. Panther Creek, IL Arnold and Allen (1996) 188 X X X X X X X
14. Goodwin Creek
Watershed, MS Bingner et al. (1996) 21.3 X
15. Watersheds in:
OK, OH, GA, ID, MS, VT,
AZ Arnold and Williams (1987) 9.0–538 X
16. Bushland, TX; Logan, UT;
Temple, TX Arnold and Stockle (1990) field plot X
on days when Vw drops below the maximum value. capabilities, SWAT has been linked to MODFLOW
Evapotranspiration is calculated with the equation: (Perkins and Sophocleus 1999) and, with additional
inputs, is capable of simulating wetland and ground-
et ⫽ 10. * etco * sa * eo (2)
water interactions. The proposed wetland bank lies
where etco is an evapotranspiration coefficient (0.6), within the 100-year flood plain of the West Fork of
sa is surface area in ha, and eo is potential evapotrans- the Trinity River and, while not simulated in this
piration in mm. The 10 multiplier in Equation 2 con- study, water derived from periodic flooding of the
verts units from ha-mm to m3. Potential evapotrans- Trinity River can be input to the system.
piration is computed using the Penman-Monteith equa-
tion (Monteith 1965), and the value of etco (0.6) was
METHODOLOGY
taken from the Handbook of Hydrology (Shuttleworth
1992). When the water table rises into the soil profile The water budget was prepared for the wetland us-
(saturated conditions), seepage from the wetland into ing the three main steps described below. A detailed
the soil is set to zero. When the water table falls below chart of all processes used in preparing the water bud-
the soil profile, unsaturated flow through the soil layers get is shown in Figure 3.
occurs according the soil water routing method de-
scribed in Arnold et al. (1998). Surface area of the
Assemble Data Required to Run Model
wetland below capacity is estimated as a non-linear
function of storage. Drainage area, channel slopes, and channel lengths
The version of SWAT used in this study does not were derived from the USGS 7.5 Minute Hurst Quad-
simulate interactions between rivers and regional water rangle. Soils and land use were delineated for the basin
tables beneath the proposed wetland. In this case using the Tarrant County Soil Survey and air photo-
study, the assumption of no ground-water inflow to the graphs of the watershed. The watershed was subdivid-
wetland was justified due to the extremely low per- ed to account for differences in soils and land use.
meability of the clayey flood-plain materials (USDA Daily rainfall from the Arlington rain gauge (National
1977). Although the standard version of SWAT (Ar- Weather Service—10 km southeast of the proposed
nold et al. 1998) has limited ground-water-interaction wetland) for the period of 1966 to 1979 was used to
Arnold et al., HYDROLOGIC MODEL FOR CONSTRUCTED WETLANDS 173
run the model. While the model is capable of gener- balance of the watershed (Table 2). Surface runoff was
ating local rainfall, the model runs used actual rainfall separated from base flow using the previously de-
to more accurately verify the results of the model out- scribed digital filter technique on daily stream flow.
put against the water budget derived from the gauged Realistic simulation of base flow is critical for hydro-
data at Big Bear Creek. Daily temperature, radiation, logic design of wetlands during droughts. Total evapo-
wind speed, and humidity were generated from month- transpiration (ET) was determined as the difference be-
ly statistics that are readily available. tween precipitation and stream flow (neglecting deep
aquifer recharge).
Assemble Data Required to Validate Model
As is common in practice, measured stream-flow Run and Validate Model
data were not available at the proposed wetland site.
After the soil, land-use, and watershed characteris-
Big Bear Creek, as noted, has been monitored from
tics of Walker Creek were put into the model, it was
1966 to 1979. This site is comparable in drainage area,
run for the period of record (1966–1979) to assess the
soils, and geology to the Walker Creek site. Daily
impact of climatic variability on wetland hydrology at
stream flow records from USGS gauge 08049550 on
the project site. A decade or more is usually advocated
Big Bear Creek were obtained to estimate the water
in such model runs so that both the wet and dry pe-
riods can be adequately assessed. Without measured
Table 2. Water budget for Big Bear Creek from U.S.G.S. Gage data on Walker Creek, we assumed that the water bal-
Records. Water balance—average annual (1966–1979). ance and stream flow were similar to that derived from
the Big Bear Creek gauge analysis (Table 2). While
Component Amount Source all input parameters are physically based and the mod-
Precipitation 869 mm National Weather Service (Daily) el does not require calibration, there is always some
Stream Flow 78 mm USGS Gage 08049550 (Daily) uncertainty in model inputs, and measurement errors
Surface Runoff 69 mm Base Flow Filter (88% surface can occur. Therefore, after initial model runs, minor
runoff) changes were made to the available soil water capacity
Base Flow 9 mm Base Flow Filter (12% base to calibrate the water balance of the Walker Creek sim-
flow) ulation to that of Big Bear Creek gauge analysis. The
Total ET 791 mm (Precip ⫺ Flow)
changes made to the available soil water capacity were
174 WETLANDS, Volume 21, No. 2, 2001
Figure 4. Average Annual Water Balance of Walker Creek Watershed and the proposed wetland simulated by the SWAT
model.
within general uncertainty ranges shown in the Tarrant ground-water flow about 12 percent of total flow. Pre-
County Soil Survey (USDA 1977). cipitation minus stream flow is often used to calculate
loss due to evaporation and transpiration or ET. In Big
Bear Creek, ET averages 91 percent of rainfall.
RESULTS AND DISCUSSION
The SWAT simulated water budget for the Walker
The water budget prepared for Big Bear Creek for Branch site is shown in Figure 4. SWAT allows as-
the period of 1966–1979 is shown in Table 2. Record- sessment of each component of the hydrologic budget.
ed stream flow averages about 8.9 percent of precipi- For the same time period and rainfall, Walker Creek
tation. The regional average is 15 percent of precipi- shows results similar to the gauged basin (Big Bear
tation (Gebert et al. 1987). In this basin, storm (sur- Creek). ET is 89 percent of the rainfall. Total stream
face) flow contributes 88 percent of total flow and flow is about 11 percent of rainfall. Of this total, sur-
Arnold et al., HYDROLOGIC MODEL FOR CONSTRUCTED WETLANDS 175
Figure 6. Wetland hydroperiod—daily wetland volume simulated by the SWAT model from 1966 to 1977.
176 WETLANDS, Volume 21, No. 2, 2001
Figure 7. Frequency Diagram of Daily SWAT simulated Wetland Volume from 1966 to 1977.
low ground-water system. This water eventually bration period. Seepage maximum predominates dur-
moves toward the Trinity River and emerges as base ing periods of high water, which correspond to periods
flow. of high runoff. Evaporation is highest during the sum-
Simulated flow for the twelve-year time period was mer months.
routed to the wetland site and allowed to fill the wet-
land volume (Figure 4). The maximum wetland vol-
SUMMARY AND CONCLUSIONS
ume used in this calculation was 70,000 cubic meters.
Water in the wetland was allowed to seep into the soil The water budget of the proposed wetland was de-
and evaporate using equation 2. Monthly simulated rived and calibrated. The results indicate that the wet-
stream flow from Walker Creek is shown in Figure 5. land should be at or above 85 percent capacity over
Monthly flow is indicated in mm over the entire wa- 60 percent of the time. The wetland did not dry up
tershed, and approximately 1.3 mm are required to fill during the entire modeled time period (14 years) and
the proposed wetland site (from zero storage). reached 40 percent capacity less than one percent of
The hydroperiod of the wetland is illustrated in Fig- the modeled time period. The majority of water to the
ure 6a and Figure 6b for the twelve-year period. The wetland will be derived from storm flow from Walker
graphs depict a winter maximum water level with a Creek, principally from residential areas. Analysis re-
late summer minimum. This trend reflects the predom- vealed that simulated wetland stream flows can be cal-
inant rainfall pattern in the area. The decline during ibrated to local gauged watershed data for more real-
the summer is more gradual and due to both a lack of istic estimates of the local hydrologic budget. Advan-
rainfall during the summer, increased consumptive use tages of the modeled wetland response using the cal-
by plants, and increased surface evaporation off the ibrated SWAT model includes validation of wetland
ponded wetland surface. The rapid rise in the late fall function over long-term weather conditions.
is consistent with lower ET and increased frequency A model that accurately simulates the water budget
of frontal storms. for over a decade increases confidence that the sys-
Figure 7 indicates that the wetland is about 93 per- tem’s hydrology is understood (Skaggs et al. 1991).
cent full for 35 percent of the simulated 12-year time Modeling is also a useful approach in studying inter-
period and 85 percent full about 57 percent of the time. action of one or more hydrologic processes often nec-
The lowest the wetland reaches is about 14 percent essary in choosing between alternative management
full during 1975 (Figure 6b). This level occurs less scenarios. In addition, such simulated events can give
than 5 percent of the time. The mean storage volume a more realistic description of influent to a storm-wa-
is approximately 50–55,000 cubic meters. ter-treatment wetland than do time-averaged flow
Figure 8 show seasonal trends in major components models, which can predict only average flow rate and
of the water budget for the wetland during the cali- average concentrations (Werner and Kadlec 2000).
Arnold et al., HYDROLOGIC MODEL FOR CONSTRUCTED WETLANDS 177
Figure 8a. Daily seepage simulated by the SWAT model into the soil under the wetland in cubic meters from 1966 to 1978.
Figure 8b. Daily evaporation simulated by the SWAT model from the wetland in cubic meters from 1966 to 1978.
Bidlake, W. R. and P. F. Boetcher. 1996. Simulation of the soil techniques for baseflow and recession analysis. Water Resources
water balance of an undeveloped prairie in West Central Florida. Research 26:1465–1473.
U.S. Geological Survey Water Supply Paper 2472. Owen, C. 1995. Water budget and flow patterns in an urban wetland.
Carter, V. 1986. An overview of hydrologic concerns related to wet- Journal of Hydrology 169:171–187.
lands in the United States. Canadian Journal of Botany 64:364– Perkins, S. P. and M. Sophocleus. 1999. Development of a compre-
374. hensive watershed model applied to study stream yield under
Dexler, J. Z., B. L. Bedford, A. T. DeGactano, and D. I. Siegel. drought conditions. Groundwater 37:418–426.
1999. Quantification of the water budget and nutrient loading in Pierce, G. I. 1992. Planning hydrology for constructed wetlands.
a small peatland. Journal of American Water Resources Associ- Wetland Training Institute, West Clarksville, NY, USA.
ation 35:753–769. Restrepo, J. I., A. M. Montoya, and J. Obeysekera. 1998. A wetland
Federal Register. 1995. Federal guidance for the establishment, use, simulation module for the MODFLOW ground water model.
and operation of mitigation banks. November 28, 1995, 60(228): Groundwater 36:764–770.
58605–58614. Rutledge, A. and C. C. Daniel. 1994. Testing an automated method
Gebert, W. A., D. J. Graczyk, and W. R. Krug. 1987. Average an- to estimate groundwater recharge from streamflow records.
Groundwater 32:180–189.
nual runoff in the United States 1951–1980. U. S. Geological Sur-
Schueler, T. 1992. Design of stormwater wetland systems: guide-
vey Hydrologic Investigations. Atlas HA⫺ 710.
lines for creating diverse and effective stormwater wetland sys-
Hammer, D. A. 1989. Constructed Wetlands for Wastewater Treat-
tems in the mid Atlantic region. Metropolitan Washington Council
ment. Lewis Publishers, Chelsea, MI, USA. of Governments, Washington, DC, USA.
Kadlec, R. H. and R. L. Knight. 1996. Treatment Wetlands. CRC. Shuttleworth, W. J. 1992. Chapter 4. Evaporation. p. 4.1–4.53. In
Lewis Publishers, Boca Raton, FL, USA. Handbook of Hydrology. Maidment, D. A. (ed.) McGraw-Hill,
Koob, T., M. E. Barbe, and W. E. Hathorn. 1999. Hydrologic design Inc., New York, NY, USA.
consideration of constructed wetlands for urban stormwater run- Skaggs, R. W. 1984. DRAINMOD—A water management model
off. Journal of the American Water Resources Association 35: for shallow water table soils-user’s guide. North Carolina State
323–331. University, Raliegh, NC, USA.
Kusler, J. A. and H. E. Kentula (eds.). 1996. Wetland Restoration Skaggs, R. W., J. W. Gilliam, and R. O. Evans. 1991. A computer
and Creation- the Status of the Science. Island Press, Washington, simulation study of Pocosin hydrology. Wetlands 11:399–416.
DC, USA. Sun, G., H. Riekerk, and N. B. Comerford. 1998. Modeling the
Mau, D. P. and T. C. Winter. 1997. Estimating groundwater recharge forest hydrology of wetland-upland ecosystems in Florida. Journal
from stream flow hydrographs for a small mountain watershed in of the American Water Resources Association 34:827–841.
a temperate humid climate, New Hampshire, USA. Groundwater USDA—Soil Conservation Service. 1977. Soil survey of Tarrant
35:291–304. County, Texas. USDA-SCS in cooperation with Texas Agricul-
McCuen, R. H. 1989. Hydrologic Analysis and Design. Prentice tural Experiment Station. Temple, TX, USA.
Hall, Englewood Cliffs, NJ, USA. Walton, R., R. S. Chapman, and J. E. Lewis. 1996. Development
McMahon, M. A. and R. G. Mein. 1986. River and reservoir yield. and application of the wetlands dynamic water budget model.
Water Resources Publication, Littleton, CO, USA. Wetlands 16:347–357.
Mitsch, W. T. and J. G. Gosselink. 1993. Wetlands. Van Nostrand Werner, T. M. and R. H. Kadlec. 2000. Stochastic simulation of
Reinhold Co., New York, NY, USA. partially-mixed, event driven treatment wetlands. Ecological En-
gineering 14:223–267.
Monteith, J. L. 1965. Evaporation and environment. Symposia of
the Society for Experimental Biology 19:205–234. Manuscript received 13 June 2000; revisions received 5 October
Nathan, R. J. and T. A. McMahon. 1990. Evaluation of automated 2000 and 18 December 2000; accepted 5 January 2001.