Evaluating hydrology of the Soil and Water

doi:10.2489/jswc.67.6.513

Evaluating hydrology of the Soil and Water Assessment Tool (SWAT) with new tile drain equations D.N. Moriasi, C.G. Rossi, J.G. Arnold, and M.D. Tomer

Key words: Hooghoudt—Kirkham—simulation accuracy—Soil Water Assessment Tool— streamflow—tile drainage

Agricultural subsurface drainage, commonly known as tile drainage, is a common water management practice in agricultural regions with seasonal high water tables, such as the midwestern United States. Agricultural drainage systems are mainly installed to allow timely seedbed preparation, planting, harvesting, and other field operations and to protect field crops from extended periods of saturated soil conditions in the soil root zone. Tile drains intercept percolating waters and route them, along with chemical pollutants, directly to surface waters (Baker et al. 1975; Logan et al. 1994; Arnold and Allen 1996). In particular, subsurface drainage is known to expedite the

JOURNAL OF SOIL AND WATER CONSERVATION

transport of nitrate nitrogen (NO3-N) and pesticides, such as herbicides, insecticides, and fungicides, to surface waters (Thomas et al. 1992; Randall et al. 1997; Zucker and Brown 1998; Dinnes et al. 2002). Therefore, agricultural production policy and decision makers in the midwestern United States must balance the act of ensuring agricultural production while minimizing the impact of nutrient transport into water bodies. One of the watershed-scale models that contain a tile drainage component is the Soil and Water Assessment Tool (SWAT) model (Arnold et al. 1998; Arnold and Fohrer 2005). SWAT is a continuous and physically based watershed model developed to predict

Copyright © 2012 Soil and Water Conservation Society. All rights reserved. Journal of Soil and Water Conservation 67(6):513-524 www.swcs.org

Abstract: Although subsurface drainage is a water management system widely used to maximize crop production in regions with seasonal high water tables, such as the midwestern United States, it is also a major source of nutrients into water bodies. Recently, physically based Hooghoudt and Kirkham tile drain equations were incorporated into the Soil and Water Assessment Tool (SWAT) model (herein referred to as Modified SWAT) as alternative tile flow simulation methods and a tool to design cost-effective and environment-friendly tile drain water management systems. The goal of this study was to determine a range of values for the new tile drain parameters and to use measured streamflow data from the South Fork Watershed (SFW) in Iowa to evaluate the capability of the Modified SWAT to simulate water balance components for this tile-drained watershed. This was accomplished by reviewing literature of tile drainage studies and by comparing measured streamflow with that predicted by the Modified SWAT using the Nash-Sutcliffe efficiency (NSE) and percent bias (PBIAS [%]) statistical methods in addition to hydrographs. During the calibration period, the Modified SWAT simulated streamflow very well (monthly NSE = 0.85 and PBIAS = ±2.3%). During the validation period, the Modified SWAT model simulated streamflow well (monthly NSE = 0.70 and PBIAS = ±2.5%). Simulated water balance results indicated that the soil water with tile drainage (260 mm [10 in]) was significantly (p-value = 0.00) lower than soil water without tile drainage (355 mm [14 in]), while streamflow with (205 mm [8 in]) tile drainage was significantly (p-value = 0.03) greater than streamflow without (128 mm[ 5 in]) tile drainage. This shows that the Hooghoudt steady-state and Kirkham tile drain equations are potential alternative tile flow simulation methods and tile drainage design tools in SWAT.

the impact of land management practices on water, sediment, and agricultural chemical yields in watersheds with varying soils, land use, and management conditions across space and time. SWAT has become an effective means for evaluating nonpoint source water resource problems (flow, sediment, nutrients) for a large variety of water quality applications nationally and internationally (Gassman et al. 2007). SWAT is the primary model being used in the Conservation Effects Assessment Project (CEAP), which is a USDA program designed to quantify the environmental benefits of conservation practices used by private landowners participating in USDA conservation programs (Mausbach and Derick 2004; Duriancik et al, 2008; Richardson et al. 2008). CEAP is comprised of a national assessment (CEAP NA) and a series of watershed assessment studies (CEAP WAS). The CEAP NA is undertaken by the USDA Natural Resources Conservation Service. The CEAP WAS are conducted by the USDA Agricultural Research Service on 14 selected benchmark watersheds, including the South Fork Watershed (SFW), to help verify and improve models used in the CEAP NA. Arnold et al. (1999) enhanced SWAT2000 with a subsurface tile flow component whose equations assume that the tile systems have already been designed with regard to tile spacing and size. The enhanced SWAT2000 was tested at field scale and yielded satisfactory results (Arnold et al. 1999). But when applied on a watershed scale, SWAT2000 did not accurately simulate subsurface flow and stream discharge because the incorporated tile algorithms did not accurately represent the water table dynamics at the watershed scale. Also, the effects of large depressional storage areas (potholes), prevalent within the watershed of interest, were not included in the algorithms (Arnold et al. 1999). Later, SWAT2000 was modified to simulate water table dynamics and linked with the subsurface Daniel N. Moriasi is a hydrologist at the USDA Agricultural Research Service Grazinglands Research Laboratory. Colleen G. Rossi is a soil scientist, and Jeffrey G. Arnold is a supervisory agricultural engineer at the USDA Agricultural Research Service Grassland Soil and Water Research Laboratory. Mark D. Tomer is a soil scientist at the USDA Agricultural Research Service National Laboratory for Agriculture and the Environment.

NOV/DEC 2012—VOL. 67, NO. 6

513

514


et al. (2007a). Whereas Moriasi et al. (2007a) successfully incorporated the new tile drain equations into SWAT (Modified SWAT) and verified them, they did not include details on the range of values for the new tile drainage equations parameters described to guide model users. In addition, the Modified SWAT model was not evaluated to determine how well it simulated the water budget. The objective of this study was to determine a range of values for the new tile drain parameters and to use measured data from the SFW to evaluate the capability of the Modified SWAT to simulate water balance components for the tile-drained SFW. Materials and Methods Tile flow occurs when the water table height exceeds the height of the tile drains. The water table depth is computed using the approach by Du et al. (2005). In this approach, a restrictive layer, which simulates a confining layer and the maximum water table depth, is set at the bottom of the soil profile. Beginning with the bottom soil layer, the soil profile above the confining layer is allowed to fill with water to field capacity. When the bottom soil layer reaches field capacity, additional water is allowed to fill the profile from the bottom of the soil layer upward from which the height of the water table above the restrictive layer and hence the water table depth from the ground surface is computed. Du et al. (2005) provides more details and sample calculations. Hooghoudt’s and Kirkham Tile Drain Equations. In the Modified SWAT, tile flow is computed using the Hooghoudt’s (1940) steady-state and Kirkham (1957) tile drain equations that have been successfully used in the DRAINMOD model (Skaggs 1978). The Hooghoudt (1940) steady-state equation is used to compute both drainage and subirrigation flux. These equations use parallel tile drain systems and are sensitive to the depth and spacing of the drains. The rate of subsurface water movement into drain tubes or ditches depends on the lateral saturated hydraulic conductivity (K) of the soil; drain spacing, size, and depth; soil profile depth; and water table elevation (Skaggs 1980). Water moves toward drains in both the saturated and unsaturated zones. The drainage rates are computed by assuming that the lateral water movement occurs mainly in the saturated region. Effective horizontal saturated hydraulic conductivity (Ke ) is used by

these equations to evaluate flux in terms of the water table elevation midway between the drains and the water level or hydraulic head in the drains (figure 1). The tile drain equation used to compute tile flow, at any given time, is determined by the proximity of the water table to the ground surface and the amount of ponding in the maximum Sd (figure 2). Surface depressional storage, which is assumed constant in DRAINMOD (Skaggs 1978), varies as a function of dynamic soil random roughness (Onstad 1984) in the Modified SWAT model. The dynamic random roughness is a function of tillage type and intensity and amount of rainfall (Saleh and Fryrear 1999). Moriasi et al. (2007a) give detailed information on the equations used to compute the dynamic Sd. In brief, the drainage flux is calculated using a three-step approach as follows. First, for water tables below the surface and for ponded surface depressional depths less than S1, a depressional storage depth threshold when surface water cannot move freely toward the drains, the Hooghoudt (1940) steady-state equation is used to compute flux as represented by equation 1: q=

8 Ke d e m + 4 K e m 2

, (1) CL 2 where q is drainage flux (mm h–1), m is midpoint water table height above the drain (mm), Ke is effective lateral saturated hydraulic conductivity (mm h–1), L is distance between drains (mm), C is the ratio of the average flux between the drains to the flux midway between the drains (assumed to be one in DRAINMOD model and in this study), and de is substituted for d (height of the drain from the impervious layer) in order to correct for convergence near the drains (mm). The equivalent depth (de) is obtained using the equations developed from Hooghoudt’s solutions by Moody (1966) as a function of L, d, and tube radius (r). In reality, rather than completely open drain tubes, there is an additional loss of hydraulic head due to convergence as water approaches the finite number of openings in the tube and hence an effective drain tube radius, re (Skaggs 1978). Skaggs and Fernandez (1998) give a table of re for various tile tubing sizes, which can be used to determine de and the g-factor used in equation 2. In the modified SWAT, the value of S1 is estimated as 20% of Sd. For ponded depths greater than S1, when water table rises to completely fill the surface



tile flow equation in addition to inclusion of pothole algorithms (Du et al. 2005). These modifications resulted in improved predicted pattern and amount of monthly flow and subsurface drainage (Du et al. 2005). The modifications by Du et al. (2005) were later incorporated into the SWAT version 2005 (herein referred to as SWAT). In a CEAP WAS in the SFW in Iowa, Green et al. (2006) evaluated the SWAT model’s accuracy to simulate streamflow. Results revealed that SWAT is a promising tool to evaluate streamflow in tile-drained regions. However, the tile equation in SWAT assumes that the tile systems have already been designed with regard to tile spacing and size. Later, Moriasi et al. (2007a) incorporated the steady-state Hooghoudt’s (1940) and Kirkham (1957) tile drain equations into SWAT (herein referred to as Modified SWAT) alternative tile flow simulation methods, which take into account tile spacing and drain tube size in addition to depth to tile drain. These equations were intended to determine the impacts of tile spacing and size on water quantity and quality at the watershed scale and to design cost-effective and environment-friendly tile drain water management systems. These equations are specifically used on fields containing a network of parallel drainage ditches or subsurface drains. Despite this limitation, these equations can easily be adapted to simulate streamflow, subsurface drainage, and water table depth for study areas with different tile drain orientations by assuming an equivalent network of parallel drains (He et al. 2002). If known, different spacing, drain size, and depth to drain values can be assigned for different fields or hydrologic response units (HRUs). The Hooghoudt’s (1940) steady-state and Kirkham (1957) tile drain equations depend on the maximum surface depressional storage (Sd). The drainage flux or daily tile flow volume is calculated using a three-step approach: (1) Hooghoudt steady-state equation is used when the water table is below the Sd or the surface; (2) Kirkham equation is used to compute the ponded surface drainage flux when the water table rises to completely fill the surface; and (3) when the drainage flux predicted by the appropriate equation is greater than the design drainage capacity (also known as drain coefficient [DC]), then the flux is set equal to DC. Complete details of the new equations and the associated modifications are presented by Moriasi

Figure 1 Schematic of water table drawdown to and subirrigation from drain tubes. In this study, equivalent lateral hydraulic conductivity (Ke) is determined for soil profiles with any number of layers (modified from Skaggs [1980]). d1 is the distance from the bottom of semisaturated layer n to the water table elevation midway between the tile drains; yo is the height of the water table elevation at the minus the equivalent depth (ho – de).

R

D1 m

yo ho

D2 de

D3

m

K1

d1

K2

hm

K3

D4

K4 Impervious layer

Water table elevation during subsurface irrigation Water table elevation during drainage Layer depth bottom between surface and impervious layer

Figure 2 Schematic of drainage from a ponded surface. Water will move over the surface to the tile vicinity until the ponded depth becomes less than S1 (redrawn from Skaggs [1980]). Sd is the maximum depressional storage, h is the actual depth of the soil profile to impervious layer, d is the actual height of the tile drain from the impervious layer, b is the depth of the tile drain from the soil surface, r is the radius of the tile drain, and t is the average depressional storage depth.

Water table Surface

Sd

t S1

b 2r h d

and the ponded water remains at the surface for relatively long periods of time, flux drainage is computed using the Kirkham (1957) equation: q=

4π Ke (t + b − r ) , (2) gL

where t, b, and r are as shown in figure 2 and g is a dimensionless factor which is determined using an equation developed by Kirkham (1957). Factor g is computed as a function of


L

d, L, actual depth of the profile (h [mm]), m, and the radius of the tile tube (r [mm]). Thirdly, when the flux predicted by the appropriate equation is greater than the drainage coefficient (DC [mm d–1]), then the flux is set equal to the DC as q = DC. (3) The DC is typically 10 to 20 mm d–1 (0.4 to 0.8 in day–1) depending on the geographic


Legend

location and crops grown (Skaggs 1980).The DC may also be estimated as a function of the surface inlet types, soil type, and crop type (Wrighr and Sands 2001). Although subsurface irrigation is incorporated in the Modified SWAT, it was not considered in this study. Moriasi et al. (2007a) give detailed information on the equations used to compute subsurface irrigation. The orientation and the exact location of subsurface tile drains in the agricultural fields


515

Figure 3

516


Popejoy 10

12

1 2

4

13

3

11

Iowa Falls

9

14

Williams

16 15

40 N

7 39

Jewell

8

19

29 28 30 31 43 41

Beaver Creek

17

5

6

27

25 18

22

20

23

32

Buckeye Tipton Creek South Fork Iowa River

24

21

44 26

38

37 33

1

Raingauge Subbasin Digitized stream

10

what poorly drained Aquic Hapludolls, and poorly drained Typic Haplaquolls (NCSS 1986; USDA NRCS 2003). Beginning about 100 years ago, artificial drainage was installed to enable agricultural production in this watershed. According to Green et al. (2006), approximately 80% of the agricultural watershed is tile drained. This estimate includes all the soils in fields that are dominated by poorly or somewhat poorly drained soils. An assessment of the water quality and agricultural land use in this watershed is given by Tomer et al. (2008a; 2008b). The SFW was selected because it is located in the midwestern United States, where tile drainage systems are installed due to seasonal high water tables in order to allow for agricultural production. The SFW was also selected due to availability of data to evaluate the Modified SWAT model. The Modified SWAT model can be used in any watershed with tile drainage system and with good datasets for model evaluation. Weather and Streamflow Data. SWAT requires daily minimum and maximum temperature, relative humidity, solar radiation, and wind speed as inputs, while measured streamflow data are used to calibrate and validate model streamflow simulation. Daily weather data were obtained from the National Oceanic and Atmospheric

0

Eldora 35

36

42

Hubbard

Legend

Steamboat Rock

34

USGS gauge 10 km

Administration National Climatic Data Center (NOAA 2012) from eight rain gauge stations within and adjacent to the SFW (figure 3). These stations are located in Buckeye, Hubbard, Jewell, Steamboat Rock, Williams, Eldora, Iowa Falls, and Popejoy. Daily maximum and minimum temperatures were also obtained from Eldora, Iowa Falls, and Popejoy from 1998 to 2004. Solar radiation, wind speed, and humidity values were simulated by the model. Potential evapotranspiration was simulated using the Penman–Monteith (Monteith 1965) method. Observed daily streamflow data were obtained from the USGS gauging station (site 05451210; figure 3) established in 1995 covering a watershed area of 58,050 ha (143,445 ac). The length of the streamflow data record at the SFW was 9 years, starting in October of 1995 through September of 2004. Input Data. In addition to the weather data inputs, the SWAT model requires three geographic information system data layers, namely digital elevation model (DEM), soils, and land use and agricultural management data. A 30 m (98 ft) DEM obtained from the US Geological Survey (2001) was used in this study. The DEM is used by the interface to calculate subbasin parameters, such as slope and slope length, and to define the stream network. The resulting stream network was



within the SFW are not known. Therefore, a system of equivalent network of parallel tile drains was assumed in order to use the Modified SWAT model to simulate tile flow. Tile drain spacing and size for each drained field then become parameters that can be adjusted systematically during the calibration process to simulate streamflow that is fitted to measured streamflow as closely as possible. Soil Water Assessment Tool Tile Drainage Variables. In addition to tile drainage prediction inputs for SWAT, which include depth to subsurface tile (DDRAIN), drain tile lag time (GDRAIN), and depth to impervious layer (DEP_IMP), the Modified SWAT model requires several new inputs. These include DC (mm d–1), tile drainage flag/code used to switch between the old and new incorporated tile drain algorithms, multiplication factor (LATKSATF) used to determine K using the SWAT saturated hydraulic conductivity (Ks) input value for each soil layer and each soil type, random roughness for a particular tillage operation (RANRNS [mm]), effective radius of drains (RE [mm]), and distance between the midpoints of two tile tubes/drains (SDRAIN [mm]). In the event subirrigation is being simulated, then pump capacity for subsurface irrigation (PC [mm h–1]) must be defined. These inputs are defined for each HRU or for the entire basin in projects where all fields are tile drained. The recommended range of values for the new tile drain parameters, determined based on a review of literature on past tile drainage simulation studies, are given in the Results and Discussion section of this manuscript. Watershed Description. Evaluation of the Modified SWAT model used the same data and supporting information that was used in an earlier SFW study to evaluate how the tile drainage component of SWAT could improve simulations of streamflow (Green et al. 2006). The SFW covers 775 km2 (299 mi2) and includes the tributaries of Tipton and Beaver Creeks (figure 3). It is a part of the Des Moines Lobe, which is a recently glaciated landform region that covers most of north-central Iowa. Natural stream incision and development of alluvial valleys have been limited because the terrain is young (about 10,000 years since glacial retreat). The soils are highly productive, with the Clarion-Nicollet-Webster soil association being dominant, forming a sequence of well-drained Typic Hapludolls, some-

Distribution of rain gauges, US Geological Survey gauge (site 05451210), and subbasins in the South Fork Watershed. Temperature gauges are located at Eldora, Iowa Falls, and Popejoy gauging sites only (Green et al. 2006).

Table 1 Land use classification for the South Fork Watershed (Green et al. 2006).


Land use

Percentage of watershed (%)

Soybean–corn, manure 23.6 Soybean–corn, no manure 18.1 Corn–soybean, manure 17.7 Corn–soybean, no manure 14.0 Continuous corn, manure 8.3 Continuous corn, no manure 4.1 Urban 7.8 Pasture 4.1 Forest 1.9 Wetland 0.2 Water 0.2

(HRUs) was simulated with a SDRAIN of 24 m (79 ft) at a DDRAIN of 1.0 m (3.3 ft), operated with free drainage at the drain outlet (CES 1987; Schilling and Helmers 2008). In the Modified SWAT, the Sd can vary as a function of dynamic soil random roughness as explained earlier in this study. However, a constant Sd value of 12.5 mm (0.5 in) (Singh and Helmers 2008; Schilling and Helmers 2008) for the tile-drained HRUs was used in this study in order to determine the impact of the new tile drainage algorithms without the confounding effects of the new dynamic Sd routine. Follow-up studies will determine the impact of the dynamic Sd on tile flow and streamflow simulation by comparing model simulation accuracy when dynamic and constant Sd approaches are used. Model Calibration and Validation. The Modified SWAT model was manually calibrated (1995 to 1998) and validated (1999 to 2004) for streamflow discharged from the SFW. The calibration and validation periods, chosen because of data availability, were representative of the catchment with rainfall ranging from 653 mm (26 in) in 2003 to 955 mm (38 in) in 1999. These periods were selected such that each period contained dry, average, and wet years. Initial automatic sensitivity analysis conducted in this study indicated that ET, surface runoff, and subsurface flow were the most sensitive processes. Although the new tile drain parameters DC, LATKSATF, RE, and SDRAIN can be determined by calibration, realistic values based on several tile drain studies in central Iowa (Singh et al. 2006; Singh et al. 2007; Schilling and Helmers 2008; Singh and Helmers 2008) and the recommendations in the Iowa Drainage Guide (CES 1987) were used to evaluate the Modified SWAT. Therefore, only a few

parameters related to ET and surface runoff were adjusted during calibration in this study. The only calibration parameters varied in this study include the plant uptake compensation factor (EPCO) for ET, curve number coefficient (CNCOEF), surface runoff lag coefficient (SURLAG), and initial soil water storage expressed as a fraction of field capacity water content (FFCB) for surface runoff. The values of the other streamflow calibration parameters obtained by Green et al. (2006) in a previous SFW simulation study were used in this study. All other parameters (Neitsch et al. 2002) were kept at the SWAT default values. EPCO adjusts plant water uptake and varies between 0.01 and 1.00, inclusive. As EPCO approaches 0.0, the model limits uptake of water by the plant to the upper portions of the root zone. CNCOEF is the weighting coefficient used to calculate the retention coefficient for daily curve number calculations dependent on plant evapotranspiration. CNCOEF varies from 0 to 1, and as CNCOEF increases, surface runoff increases. SURLAG is the surface runoff lag coefficient and provides a storage factor in the model that allows runoff to reach a subbasin outlet when the time of concentration is greater than one day. As SURLAG decreases, the amount of water reaching the outlet decreases. FFCB determines how dry or wet the soil profile is at the start of the model simulation and varies between 0 and 1, inclusive. The initial soil water increases from completely dry (FFCB = 0) to completely saturated (FFCB = 1). Based on Green et al. (2006) and Tomer et al. (2008a) model outputs, between 1995 and 2004 the average annual rainfall was 768 mm (30 in), with average annual ET being 74% of the total rainfall. Average annual tile flow was



used by the AVSWAT-X interface to define a layout of 44 subbasins. The DEM is also used to obtain the stream network characteristics, such as channel slope, length, and width. The Soil Survey Geographic data sets obtained from the National Cooperative Soil Survey for the Hardin, Hamilton, Franklin, and Wright counties (NCSS 1985, 1986) were used to generate the required soil physical and hydraulic input parameters. Corn and soybean or continuous corn rotations occupy about 86% of the watershed, while urban, pasture, forest, wetland, and water occupy about 14% of the watershed (table 1).The combination of land use and soil type resulted in 727 HRUs. Although parameter input values can be determined by model calibration within the recommended range of values, known parameter values from previous studies can be used as well. In this study, a 2.5 m (8.2 ft) DEP_IMP and a DDRAIN of 1.0 m (3.3 ft) (Green et al. 2006) were used throughout the SFW cropped fields to account for tile flow. The GDRAIN was set to 96 hours (Green et al. 2006). The initial Soil Conservation Service runoff curve number to moisture condition II (CN2) values calibrated using the SWAT model ranged from 66 to 78 (Green et al. 2006) using the curve number (CN) method which bases CN on plant evapotranspiration (ET) (CN method flag [ICN] = 1). The total number of heat units to bring corn and soybeans to maturity (PHU_PLT) was calibrated using SWAT by Green et al. (2006) and was set at 1,800 in this study.The PC was set to 0.0 mm h–1 (0.0 in hr–1) because there was no subsurface irrigation in the SFW.The RANRNS values for each tillage operation are included in the till. dat input file. Based on several tile drain studies in central Iowa, DC (Singh et al. 2006; Singh et al 2007; Schilling and Helmers 2008), LATKSATF (Singh et al. 2006; Singh and Helmers 2008; Schilling and Helmers 2008) and RE (Singh and Helmers 2008; Schilling and Helmers 2008) were set at 35 mm d–1 (1.4 in day–1), 1.4, and 20 mm (0.8 in), respectively. The Iowa Drainage Guide (CES 1987) recommends a drain spacing of 18 to 24 m (59 to 79 ft) at a drain depth of 0.9 m (3.0 ft) or a drain spacing of 24 to 30 m (59 to 98 ft) at a drain depth of 1.2 m (3.9 ft). These systems are generally operated with an approach of free drainage at the drain outlet. In this study, the recommended subsurface drainage system for cropped fields

517

518


than zero. Simulating the SFW without tile drainage was accomplished by setting DDRAIN values to zero in all tile-drained HRUs. Student’s t-test for the significance of the difference (α = 0.05) between the annual means of simulated streamflow with and without tile drainage was carried out. Results and Discussion The range of parameter values and the calibrated and used values for this study are presented in table 2. The range of values for the Modified SWAT tile drain parameters in table 2 are based on previous studies (Skaggs 1980; Skaggs and Fernandez 1998; Neitsch et al. 2002; Haan and Skaggs 2003; Jin and Sands 2003; Burchell et al. 2005; Singh et al. 2006; Salazar et al. 2008; Sands et al. 2008). The range of values for the other parameters in table 2 is suggested by SWAT model developers. The calibrated EPCO, CNCOEF, SURLAG, and FFCB parameter values (table 2) were based on the water balance component proportions determined from results of previous SFW studies (Green et al. 2006; Tomer et al. 2008a) in addition to optimizing simulated streamflow NSE and PBIAS statistics. These calibration parameter values and the rest of the parameters used in this study resulted in an average simulated ET value of 564.8 mm (22.2 in), which was 74% of the average annual precipitation of 768.0 mm (30.2 in). These values also resulted in simulated surface runoff of 46.7 mm (1.8 in), baseflow of 13.4 mm (0.5 in), and tile flow of 145.7 mm (5.7 in). These simulated averages were 23% for surface runoff, 6% for baseflow, and 71% for tile flow. The Modified SWAT tile flow equations’ parameters were not calibrated in this study because realistic values, based on previous tile drainage studies in central Iowa and recommendations in the Iowa Drainage Guide (CES 1987), were used. However, in future studies where the values of these parameters are not known, best parameter values can be determined through model calibration and validation processes. Model Calibration and Validation Performance for Streamflow. The calibration and validation model performance results for the daily and monthly time steps are presented in table 3 while the time-series graphical plot of monthly streamflow is illustrated in figure 4. During the calibration period, the daily and monthly NSE values were 0.76 and 0.85, respectively, while the PBIAS value at both the daily and monthly

time steps was ±2.3% (table 3). According to Moriasi et al. (2007b), a model is considered calibrated for streamflow if monthly NSE ≥ 0.65 and PBIAS ≤ ±10%. Therefore, the Modified SWAT model was well calibrated as shown by the statistics in table 3 and supported by the monthly hydrograph (figure 4). The statistical results during the validation period were not as good as during the calibration period (table 3 and figure 4). With exception of the monthly streamflow peak in 2004, most of the peaks were overpredicted.This could be due to the typical (CES 1987; Singh and Helmers 2008; Schilling and Helmers 2008) maximum Sd and tile drain parameters (DDRAIN, RE, and SDRAIN) used. This could also be due to adherence to the water budget component proportion constraints determined from the simulated hydrologic budget outputs and the measured precipitation and streamflow in previous SFW studies (Green et al. 2006; Tomer et al. 2008a). Adjustments of the tile drain parameters for the drained HRUs and water budget component proportions within reasonable range (say ±5% of the used value but within the recommended range) could minimize or eliminate the overprediction of these peaks as shown in the sensitivity analysis of these tile drain parameters. The deviation between measured and simulated streamflow varied between 0% in 1998 and ±126% in 2000 (table 4). SWAT tends to overpredict low-flows (baseflow) especially during the summer months (Bouraoui et al. 2002). One possible cause for this occurrence may be due to the routing delay (Watson et al. 2003). SWAT routes only channel flow. For the overland flow, a simple surface runoff storage feature is used to lag a portion of the surface runoff released to the main channel (Neitsch et al. 2002). According to Watson et al. (2003), this simple approach may be affecting recession rates during baseflow.This may explain the unusually high overprediction in the year 2000 with mostly low-flows caused by more frequent but low intensity rainfall events (especially during the summer months) with an annual amount of 723.2 mm (28.5 in) compared, for example, to 1997 with an annual rainfall of 681.7 mm (26.8 in) (figure 5). Green et al. (2006) had similar result for the year 2000 in their SFW study that sought to evaluate SWAT’s accuracy in simulating streamflow. To address this issue, Arnold et al. (2010) have developed routines to route flow across the landscape



estimated to be about 71%, baseflow (combined groundwater and lateral flow) was 6%, and surface runoff was 23% of the total streamflow measured at the SFW outlet.This information was also used to constrain the calibration parameter values to ensure that approximately the same proportions of the water budget were obtained during calibration. For the remainder of this paper, the sum of groundwater and lateral flow will simply be referred to as baseflow. Model Performance Evaluation Methods. In addition to graphical methods such as the hydrographs, the Nash-Sutcliffe efficiency (NSE; Nash and Sutcliffe 1970), and percent bias (PBIAS [%]; Gupta et al. 1999) statistical methods were used to evaluate the performance of the Modified SWAT model to predict streamflow. Percent bias measures the average tendency of the simulated data to be larger or smaller than their observed counterparts. The optimal value of PBIAS is 0.0 with low magnitude values indicating accurate model simulation. Positive values indicate model underestimation bias, and negative values indicate model overestimation bias. In this study, monthly hydrographs were used to show model bias and differences in the timing and magnitude of peak flows. Sensitivity Analysis of Tile Drain Parameters. After the calibration process, sensitivity of the three physical tile drain parameters, namely the DDRAIN, RE, and spacing (SDRAIN), were analyzed in order to determine their potential effect on streamflow simulation at the watershed outlet. In this study, sensitivity analysis was accomplished by varying DDRAIN from 0.50 to 1.50 m (1.64 to 4.92 ft) at 0.10 m (0.33 ft) increments, RE from 3.0 to 40.0 mm (0.12 to 1.57 in) at 5 mm (0.20 in) increments, and SDRAIN from 10 to 50 m (33 to 164 ft) at 5 m (16 ft) increments (Singh et al. 2007). The rest of the parameters were kept constant while varying one of these tile drain parameters. The impact of each of these tile parameters on daily and monthly streamflow simulation accuracy was investigated using NSE and PBIAS. Impacts of Tile Drainage on the Water Balance. Finally, the impact of tile drainage on water balance (ET, surface runoff, baseflow, streamflow, and soil water) was determined by comparing simulated water balance with and without tile drainage. In the Modified SWAT, tile drainage is simulated by setting DDRAIN to values greater

Table 2 Calibrated values of adjusted parameters for streamflow calibration of the Modified Soil and Water Assessment Tool model for the South Fork Watershed (A dash signifies an unknown parameter range). Parameter Description Range

Calibrated/ used value

ESCO

Soil evaporation compensation factor

0.01 to 1.00

1.00

EPCO

Plant uptake compensation factor

0.01 to 1.00

0.62

FFCB

Initial soil water storage expressed as a fraction of field capacity water content

0.00 to 1.00

0.95

SURLAG

Surface runoff lag coefficient (d)

0.00 to 4.00

0.20

ICN

Curve number (CN) method flag: 0 use traditional SWAT method which bases CN on 0 or 1 soil moisture, 1 use alternative method which bases CN on plant evapotranspiration

CNCOEF

Curve number coefficient

0.00 to 2.00

0.20

DDRAIN

Depth to subsurface tile (mm)

500 to 1,500

1,000

DEP_IMP

Depth to impervious layer (mm)

—

2,500

1

Additional tile drain parameters for the Modified SWAT Drainage coefficient (mm d–1)

10 to 51

35

ITDRN

Tile drainage routines flag/code: 1 = DRAINMOD tile equations (Subroutine DRAINS); 0 = Original SWAT tile equations (Subroutine ORIGTILE)

0 or 1

1

LATKSATF

Multiplication factor to determine lateral hydraulic conductivity (ksat) from 0.01 to 4.00 SWAT ksat input value for HRU

1.40

PC

Pump capacity (default value = 1.042 mm h–1 or 25 mm d–1) (mm h–1)

—

0.00

RANRNS

Random roughness for a given tillage operation (mm) — SWAT till.dat input file

—

till.dat

RE

Effective radius of drains (mm)

3.0 to 40.0

20.0

SDRAIN

Distance between two drain or tile tubes (mm)

7,600 to 30,000

24,000

Table 3 Streamflow simulation performance: daily and monthly streamflow calibration and validation statistics of the measured and simulated data for the South Fork Watershed at the US Geological Survey gauge (site 05451210). NSE is Nash-Sutcliffe efficiency, and PBIAS is percent bias.

Daily

Monthly

Period

NSE

PBIAS (%)

NSE

PBIAS (%)

Calibration (1995 to 1998) Validation (1999 to 2004)

0.76 0.51

–2.3 2.5

0.85 0.70

–2.3 2.5

between HRU’s. Documentation and interfaces are being developed to guide users in parameterizing management scenarios. On average, simulated streamflow (205.4 mm [8.08 in]) was 99.5% of the measured average value (206.4 mm [8.13 in]) (table 4). In this study, model performance was determined based on the monthly timestep statistics during the validation period. According to Moriasi et al. (2007b), the streamflow trends (depicted by NSE) simulation performance rating was good (NSE = 0.70) while the streamflow magnitude (depicted by PBIAS) simulation performance was very good (PBIAS = ±2.5%). These results compare relatively well with a study by Green et al. (2006) in the SWF, where they obtained monthly NSE values


of 0.90 and 0.50 during the calibration and validation period, respectively. The simulated PBIAS values of between ±2.3% and ±2.5% were within the same order of magnitude as that found by Arnold and Allen (1996) and Green et al. (2006), who reported PBIAS values of ±5% and ±6%, respectively. Sensitivity Analysis of Tile Drain Parameters. Figures 6 to 8 illustrate the sensitivity of daily and monthly streamflow simulation accuracy to DDRAIN, RE, and SDRAIN. Streamflow simulation accuracy was least sensitive to RE both at the daily and monthly time steps (figure 7), which implies that if the model user does not know RE values, they can use any value between 3 mm (0.12 in) and 40 mm (1.57 in) without impacting streamflow simulation

results. However, both the daily and monthly NSE and PBIAS values for DDRAIN and SDRAIN varied moderately (figures 6 and 8), which indicates that streamflow simulation was somewhat sensitive to both of these parameters. The NSE values increased as the DDRAIN increased from 50 to 80 cm (20 to 31 in), then leveled off as DDRAIN increased (figure 6). However, as DDRAIN increased, PBIAS decreased. Therefore, the optimum DDRAIN to maximize streamflow NSE and minimize PBIAS was in the range of 80 to 150 cm (31 to 59 in). Also the NSE values increased as the SDRAIN increased from 10 to 20 m (33 to 66 ft), leveled off as SDRAIN increased to 30 m (98 ft), and then decreased as SDRAIN increased to 50 m (164 ft) (figure 8). As SDRAIN increased from 10 to 35 m (33 to 115 ft), PBIAS remained relatively stable, and then PBIAS increased with an increase in SDRAIN.Therefore, the optimum SDRAIN to maximize streamflow NSE was in the range of 20 to 30 m (66 to 98 ft). For each of the analyzed tile drain parameters, the daily and monthly NSE values of streamflow simulations followed similar patterns as the parameters were varied over their respective prestated ranges (table 2). In general, all the PBIAS values (figures 6 to 8)



DC

519

Figure 4 Measured and simulated monthly streamflow for the South Fork Watershed (October 1995 to September 2004).

180

Monthly streamflow (mm)

160 140

Calibration

Validation

120 100 80 60 40

Oc De t. Fe c. b Ap . Ju r. n Au e g Oc . De t. Fe c. b Ap . Ju r. n Au e g Oc . De t. Fe c. b Ap . Ju r. n Au e g Oc . De t. Fe c. b Ap . Ju r. n Au e g Oc . De t. Fe c. b Ap . Ju r. n Au e g Oc . De t. Fe c. b Ap . Ju r. n Au e g Oc . De t. Fe c. b Ap . Ju r. n Au e g Oc . De t. Fe c. b Ap . Ju r. n Au e g O . Dect. Fe c. b Ap . Ju r. n Au e g.

0

1996 1997 1998 1999 2000 2001 2002 2003 2004

Month and Year Legend Measured

Modified SWAT

Figure 5

10

80

5

100

0

120

Legend Streamflow 1997 Streamflow 2000

520


Precipitation 1997 Precipitation 2000

t. 1 Oc

pt.

Se

Au g

Jul

Jun

r. 1 Ap

Fe b

Jan

Date

No v. 1 De c. 1

60

1

15

.1

40

y1

20

e1

20

Ma y1

25

.1 Ma r. 1

0

.1

30

Precipitation (mm)

Measured streamflow (mm)

Rainfall distribution and measured streamflow in the South Fork Watershed for the year 1997 and 2000.



20

were within the satisfactory streamflow calibration criteria (PBIAS ≤ ±10%; Moriasi et al. 2007b). Also, virtually all NSE values were within the satisfactory streamflow calibration criterion (monthly NSE ≥ 0.65; Moriasi et al. 2007b). This result is in agreement with the recommended ranges obtained based on previous literature on tile drain modeling (Skaggs 1980; Skaggs and Fernandez 1998; Jin and Sands 2003; Burchell et al. 2005; Singh et al. 2006; Salazar et al. 2008; Sands et al. 2008). However, a validation of the sensitivity of DDRAIN, RE, and SDRAIN on streamflow simulations is needed on fields or watersheds where the impact of these parameters on streamflow has been measured individually and interactively. Impact of Tile Drainage on Water Balance. Based on the simulated average annual water balance (1995 to 2004) results (table 5), ET with tile drainage (565 mm [22 in]) was not significantly (p-value = 0.42) lower than the ET without tile drainage (582 mm [23 in]). The same average annual water balance results indicated that soil water (SW) with tile drainage (260 mm [10 in]) was significantly (p-value = 0.00) lower than soil water without tile drainage (355 mm [14 in]). With the exception of the surface runoff (p-value = 0.98), the other streamflow components, namely baseflow (p-value = 0.00) and tile flow (p-value = 0.00), were significantly different based on the simulated average annual water balance with and without tile drainage (table 5).The average annual surface runoff with (46.9 mm [1.8 in]) and without (47.1 mm [1.9 in]) tile drainage hardly changed because the CN method, which bases daily CN computation on plant ET (ICN = 1) (Neitsch et al. 2002), was used in this study. According to Neitsch et al. (2002), computation of the daily CN value as a function of plant evapotranspiration was added because the soil moisture method was predicting too much runoff in shallow soils. When ICN = 1, the amount of surface runoff is set using the CNCOEF. Depending on the CNCOEF determined during the calibration to meet the predetermined surface runoff proportion, simulated surface runoff remains same whether the watershed is tile drained (with tile drainage scenario) or not (without tile drainage scenario). In this study, the CNCOEF calibration value was 0.2 and yielded an annual average surface runoff of 43.6 mm (1.7 in), which is 23% of the mea-

Table 4 Comparison of measured and simulated annual streamflow for the South Fork Watershed (1995 to 2004).

Streamflow

Precipitation Year (mm)

Measured (mm)

1995* 51.2 4.3 1996 818.0 152.3 1997 681.7 251.3 1998 954.5 360.8 1999 766.7 281.8 2000 723.2 57.6 2001 811.0 235.7 2002 738.6 190.2 2003 653.0 146.4 2004* 664.5 163.9 Total 6,862.3 1,844.1 206.4

sured average annual streamflow of 206 mm (8.1 in). For areas where conditions allow use of the traditional method (ICN = 0), daily CN is computed based on the soil profile water content (Neitsch et al. 2002). In such cases, there could be significant changes in surface runoff with and without tile drainage because, as noted, tile drainage significantly

Deviation (%)

3.1 182.8 240.7 359.7 272.8 130.2 232.6 150.9 152.5 110.0 1,835.2

–27 20 –4 0 –3 126 –1 –21 4 –33 0

205.4

0

impacts soil water. The resultant average annual simulated streamflow with tile drainage (205 mm [8.1 in]) was significantly (p-value = 0.03) greater than streamflow without tile drainage (128 mm [5.0 in]). While this result is in agreement with the benefit of tile drainage increasing agricultural production due to drier soil profiles, this ben-

Figure 6 Effects (Nash-Sutcliffe efficiency [NSE] and percent bias [PBIAS] values) of the tile drain depth (DDRAIN) on daily and monthly streamflow simulation at the South Fork Watershed outlet.

1.00

1.5

0.90 0.80 0.70

NSE

0.69

0.84 0.75

0.73

0.85

0.84 0.76

0.84

0.76

0.84

0.76

0.76

0.83 0.75

0.82 0.83 0.75

0.75

0.5 0.0

0.65

–0.5

0.50 0.40

1.0

–1.0

–1.2

0.30

–1.7

–1.5

–1.5 –1.7

0.20

–2.2

0.10

–2.0 –2.3

–2.4

–2.5

–2.5

–2.6

0.00 50 60 70 80 90 100 110 120 130 140 150

DDRAIN (cm) Legend Daily NSE

Monthly NSE


PBIAS

–2.5 –3.0

PBIAS (%)

0.60

0.79 0.74

0.83

Summary and Conclusions In this study, the Modified SWAT model, which uses the Hooghoudt (1940) steadystate and Kirkham (1957) tile drain equations, was evaluated using measured streamflow from the SFW in central Iowa.The Modified SWAT model simulates daily tile flow as a function of lateral saturated hydraulic conductivity of the soil; drain spacing, size, and depth; profile depth; and water table elevation. Model calibration and validation were carried out based on measured daily streamflow for the period from October 1995 through September 2004. Calibration was also constrained such that the simulated average annual ET, tile flow, baseflow, and surface runoff proportions were equal or close to those obtained based on previous study simulation results in the SFW. During the calibration period, the daily and monthly NSE values were 0.76 and 0.85, respectively, while the PBIAS value at both the daily and monthly time steps was ±2.3%. During the validation period, the daily and monthly NSE values were 0.51 and 0.70, respectively, while the PBIAS value at both the daily and monthly time steps was ±2.5%. Analysis of the sensitivity of DDRAIN, RE, and SDRAIN on streamflow simulation revealed that RE had the least sensitivity to streamflow simulation both at the daily and monthly time steps. Streamflow was moderately sensitive to DDRAIN and SDRAIN. However, virtually all NSE and PBIAS values were within the predetermined satisfactory streamflow calibration criteria of monthly NSE ≥ 0.65 and PBIAS ≤ ±10%. Although this result is in agreement with the recommended ranges (table 2), a validation of the sensitivity of DDRAIN, RE, and SDRAIN on streamflow simulations is needed on fields or watersheds where the impact of these



Average 768.0 * Signifies incomplete years.

Simulated (mm)

efit is counteracted by increased streamflow. Increased streamflow, due to tile drainage, expedites nutrient transport into water bodies. Therefore, agricultural production policy and decision makers in tile-drained regions must balance the act of ensuring agricultural production while minimizing the impact of nutrient transport into water bodies by designing cost-effective water management systems. Finally, additional studies are needed to test tile flow and tile flow nutrient simulation accuracy using measured tile flow and tile flow water quality data.

521

Figure 7 Effects (Nash-Sutcliffe efficiency [NSE] and percent bias [PBIAS] values) of the tile drain size (effective radius [RE]) on daily and monthly streamflow simulation at the South Fork Watershed outlet.

1.00 0.90 0.80 0.70

0.83

0.84

0.84

0.84

0.85

0.85

0.85

0.85

0.84

0.75

0.75

0.76

0.76

0.76

0.76

0.76

0.76

0.76

Acknowledgements

1.00 0.50 0.00 -0.50

0.50

-1.00

0.40

PBIAS (%)

0.60

NSE

1.0

1.50

-1.50

0.30

-2.00

0.20

–2.2

0.10 0.00

–2.2

–2.2

–2.3

–2.3

–2.3

–2.3

–2.3

–2.3

3 5 10 15 20 25 30 35 40

-2.50 -3.00

RE (mm) Legend Daily NSE

Monthly NSE

PBIAS

Figure 8 Effects (Nash-Sutcliffe efficiency [NSE] and percent bias [PBIAS] values) of the tile drain spacing (SDRAIN) on daily and monthly streamflow simulation at the South Fork Watershed outlet.

1.00 0.90 0.80

0.80

0.70

NSE

0.71

0.75

0.84 0.76

0.84 0.78 0.75

0.71

0.65

0.74 0.66

0.6 0.69 0.63 –0.6

0.50

0.64 0.61

–1.6

0.30 0.20

–1.9

–1.9

–1.9 –2.3

0.10

–2.2

0.00 10 15 20 25 30 35 40 45 50

SDRAIN (m) Daily NSE


Agricultural Research Service Grassland Soil and Water

-1.5

Research Laboratory, Temple, Texas, for their invaluable assistance with incorporation of the Hooghoudt steady-state and Kirkham tile drain equations into SWAT and the model evaluation. Funding for this project was provided by the USDA Agricultural Research Service.

References Arnold, J.G., and P.M. Allen. 1996. Estimating hydrologic

-2.5

budgets for three Illinois watersheds. Journal of

-3.0

Arnold, J.G., P.M. Allen, M. Volk, J.R. Williams, and D.D.

Hydrology 176:57-77. Bosch. 2010. Assessment of different representations of spatial variability on SWAT model performance. Transactions of the American Society of Agricultural and Biological Engineers 53(5):1433-1443.

Legend

522

-0.5

-2.0

–2.1

Sammons, information technology specialists at the USDA

0.0

-1.0

0.40

The authors are grateful to Georgie Mitchell and Nancy

0.5

PBIAS (%)

0.60

0.82

0.83

Monthly NSE

PBIAS

Arnold, J.G., and N. Fohrer. 2005. SWAT2000: Current capabilities and research opportunities in applied



1.5

parameters on streamflow has been measured individually and interactively. Results of the impact of tile drainage on water balance showed that the soil water with tile drainage was significantly (p-value = 0.00) lower than the soil water without tile drainage while streamflow with tile drainage was significantly (p-value = 0.03) greater than streamflow without tile drainage. Increased streamflow due to tile drainage expedites the nutrient transport into water bodies. Therefore, agricultural production policy and decision makers in tile-drained areas must balance the act of ensuring agricultural production while minimizing the impact of nutrient transport into water bodies by designing cost-effective water management systems. The SFW is located in the agricultural midwestern United States region where tile drainage is used as a common water management practice due to seasonal high water tables. The SFW is also representative of any region in the world where tile drainage is used as a water management system. These results showed that the Hooghoudt’s (1940) steady-state and Kirkham (1957) tile drain equations are potential alternative tile flow simulation methods for areas with or without tile drainage system design information and potential tile drainage design tools in SWAT. However, additional studies are needed to test the Modified SWAT on how well it simulates water quality components.

Table 5 Simulated impact of tile drainage on South Fork Watershed water balance (mm) for the period 1995 to 2004.

With tile

Without tile

Precip Year (mm)

Sflow

Sflow

SRO

Bflow

Tflow

ET

SW

SRO

Bflow

Tflow

ET

SW

1995* 51 3 2 1 0 36 278 3 2 1 0 36 278 1996 818 183 49 14 120 608 275 99 55 44 0 579 370 1997 682 241 59 17 166 517 224 153 58 96 0 564 320 1998 954 360 72 14 276 579 237 171 70 103 0 613 332 1999 767 273 49 15 210 563 173 153 49 104 0 577 278 2000 723 130 35 11 85 513 217 104 35 70 0 522 314 2001 811 233 50 14 169 587 240 140 49 91 0 629 323 2002 739 151 38 14 99 596 235 124 39 86 0 629 302 2003 653 152 32 15 106 533 204 107 32 75 0 534 306 2004* 665 110 32 7 71 516 241 89 33 57 0 520 350 Total 6,862 1,835 419 120 1,302 5,047 2,325 1,143 421 727 0 5,202 3,174 Average 768 205 47 13 146 565 260 128 47 81 0 582 355

watershed

modeling.

Hydrological

Processes

19(3):563-572.

Duriancik, L.F., D.A. Bucks, J.P. Dobrowolski, T. Drewes, S.D. Eckles, L. Jolley, R.L. Kellogg, D. Lund, J.R.

Climate. Transactions of the American Society of Agricultural Engineers 46(4):1011-1021.

Arnold, J.G., P.W. Gassman, K.W. King,A. Saleh, and U. Sunday.

Makuch, M.P. O’Neill, C.A. Rewa, M.R. Walbridge,

Kirkham, D. 1957. Theory of land drainage. In “Drainage

1999.Validation of the Subsurface Tile Flow Component

R. Parry, and M.A. Weltz. 2008. The first five years of

of Agricultural Lands Agronomy Monograph No. 7.

in the SWAT Model. Presented at the ASAE/CSAE-

the Conservation Effects Assessment Project. Journal

SCGR Annual International Meeting. American Society

of Soil and Water Conservation 63(6):185A-197A,

of Agricultural Engineers Paper No. 992138. St. Joseph, MI: American Society of Agricultural Engineers. Arnold, J.G., R. Srinivasan, R.S. Muttiah, and J.R. Williams.

doi:10.2489/jswc.63.6.185A.

Madison, WI: American Society of Agronomy. Logan, T.J., D.J. Eckert, and D.G. Beak. 1994. Tillage, crop, and climatic effects on runoff and tile drainage losses

Gassman, P.W., M.R. Reyes, C.H. Green, and J.G. Arnold. 2007. The soil and water assessment tool: Historical

of nitrate and four herbicides. Soil Tillage Research 30(1):75-103.

1998. Large-area hydrologic modeling and assessment:

development, applications, and

directions.

Mausbach, M.J., and A.R. Derick. 2004. The length we

Part I. Model development. Journal of American Water

Transactions of the American Society of Agricultural

go measuring environmental benefits of conservation

Resources Association 34(1):73-89.

Engineers 50(4):1211-1250.

practices. Journal of Soil and Water Conservation

future

Baker, J.L., K.L. Campbell, H.P. Johnson, and J.J. Hanway.

Green, C.H., M.D. Tomer, M. Di Luzio, J.G. Arnold. 2006.

1975. Nitrate, phosphorus, and sulfate in subsurface

Hydrologic evaluation of the Soil and Water Assessment

Monteith, J.L. 1965. Evaporation and the environment. In The

drainage water. Journal of Environmental Quality

Tool for a large tile-drained watershed in Iowa.

state and movement of water in living organisms, 205-

4(4):406-412.


234. XIXth Symposium of Society for Experimental

Bouraoui, F., L. Galbiati, and G. Bidoglio. 2002. Climate

Engineers 49(2):413-422.

59(5):96-103.

Biology, Swansea: Cambridge University Press.

change impacts on nutrient loads in the Yorkshire Ouse

Gupta, H.V., S. Sorooshian, and P.O. Yapo. 1999. Status

Moody, W.T. 1966. Nonlinear differential equation of drain

catchment (UK). Hydrology and Earth System Sciences

of automatic calibration for hydrologic models:

spacing. Journal of the Irrigation and Drainage Division,

6(2):197-209.

Comparison with multilevel expert calibration. Journal

Burchell II, M.R., R.W. Skaggs, G.M. Chescheir, J.W. Gilliam,

of Hydrologic Engineering 4(2):135-143.

American Society of Civil Engineers 92(IR2):1-9. Moriasi, D.N., J.G. Arnold, and C.H. Green. 2007a.

and L.A. Amold. 2005. Shallow subsurface drains to

Haan, P.K., and R.W. Skaggs. 2003. Effect of parameter

Incorporation of Hooghoudt’s steady state and

reduce nitrate losses from drained agricultural lands.

uncertainty on DRAINMOD predictions: I. Hydrology

Kirkham tile drain equations into SWAT. In The 4th


and yield. Transactions of the American Society of

International SWAT Conference Proceedings, 139-

Engineers 48(3):1079–1089.

Agricultural Engineers 46(4):1061–1067.

147. Delft, The Netherlands.

CES (Cooperative Extension Service). 1987. Iowa Drainage

He, X, M.J. Vepraskas, R.W. Skaggs, and D.L. Lindbo. 2002.

Moriasi, D.N., J.G. Arnold, M.W. Van Liew, R.L. Bingner,

Guide. Special Report (Revised). Ames, IA: Iowa State

Adapting a drainage model to simulate water table levels

R.D. Harmel, and T.L. Veith. 2007b. Model evaluation

University, Cooperative Extension Service.

in Coastal Plain soils. Soil Science Society of America

guidelines for systematic quantification of accuracy in

Journal 66:1722–1731.

watershed simulations. Transactions of the American

Dinnes, D.L., D.L. Karlen, D.B. Jaynes, T.C. Kaspar, J.L. Hatfield, T.S. Colvin, C.A. Cambardella. 2002. Nitrogen

Hooghoudt, S.B. 1940. Bijdragen tot de kenins van enige

management strategies to reduce nitrate leaching

natuurkundige grootheden van de ground. No. 7. Versl.

in tile-drained midwestern soils. Agronomy Journal

Landbouwk. Onderz. (Contributions to the knowledge

94(1):153–171.

of some physical constants of the soil. No.7. Report

Du, B., J.G. Arnold, A. Saleh, and D.B. Jaynes. 2005.

Agricultural Research) 46: 515-707.

Development and application of SWAT to Landscapes

Jin, C.X., and G.R. Sands. 2003. Long-term Field-scale

with tiles and potholes. Transactions of the American

Hydrology of Subsurface Drainage Systems in a Cold


Notes: Precip = precipitation. Sflow = streamflow. SRO = surface runoff. Bflow = baseflow. Tflow = tileflow. ET = evapotranspiration. SW = soil water. * Signifies incomplete year.

Society of Agricultural and Biological Engineers 50(3):885-900. NCDC (National Climatic Data Center). Climate Data Online. www.ncdc.noaa.gov. Nash, J.E., and J.V. Sutcliffe. 1970. River flow forecasting through conceptual models: Part 1. A discussion of principles. Journal of Hydrology 10(3):282-290.

Society of Agricultural Engineers 48(3):1121-1133.



523

Neitsch, S.L., J.G. Arnold, J.R. Kiniry, J.R. Wiliams, and K.W. King. 2002. Soil and Water Assessment Tool theoretical

Water Management 92(3):162-170.

documentation version 2000. GSWRL Report 02-01,

Singh, R., M.J. Helmers, and Q. Zhiming. 2006. Calibration

BRC Report 02-05, TR-191. College Station, TX:

and validation of DRAINMOD to design subsurface

Texas Water Resources Institute.

drainage systems for Iowa's tile landscapes. Agricultural

NCSS (National Cooperative Soil Survey). 1985. Soil Survey

Water Management 85(3):221–232.

of Hardin County, Iowa. USDA Soil Conservation

Skaggs, R.W. 1978. A water management model for shallow

Service and Iowa State University Cooperative

water table soils. Report No. 134. Chapel Hill, NC:

Extension Service. Washington, DC: US Government

Water Resources Research Institute of the University

Printing Office.

of North Carolina.

NCSS (National Cooperative Soil Survey). 1986. Soil Survey

Skaggs, R.W. 1980. Methods for design and evaluation of

of Hardin County, Iowa. USDA Soil Conservation

drainage-water management systems for soils with high

Service and Iowa State University Cooperative

water tables. Drainmod Reference Report. Fort Worth,

Extension Service. Washington, DC: US Government

TX: USDA Soil Conservation Society South National

Printing Office.

Technical Center. Skaggs R.W., and G. Fernandez. 1998. DRAINMOD ver.

surfaces. Transactions of the American Society of

5 for Windows 98/NT. Raleigh, NC: North Carolina State University.

Randall, G.W., D.R. Huggins, M.P. Russelle, D.J. Fuchs,W.W.

Thomas, D.L., P.G. Hunt, and J.W. Gilliam. 1992. Water table

Nelson, and J.L. Anderson. 1997. Nitrate losses through

management for water quality improvement. Journal of

subsurface drainage in conservation reserve programs, alfalfa, and row crop systems. Journal of Environmental

Soil and Water Conservation 47(1):65–70. Tomer, M.D.,T.B. Moorman, D.E. James, G. Hadish, and C.G. Rossi. 2008b. Assessment of the Iowa River’s South Fork

Quality 26:1240–1247. Richardson, C.W., D.A. Bucks, and E.J. Sadler. 2008. The

watershed: Part 2. Conservation practices. Journal of Soil

Conservation Effects Assessment Project benchmark

and Water Conservation 63(6):371-379, doi:10.2489/

watersheds: Synthesis of preliminary findings. Journal of Soil and Water Conservation 63(6):590-604,

jswc.63.6.371. Tomer, M.D., T.B. Moorman, and C.G. Rossi. 2008a. Assessment

doi:10.2489/jswc.63.6.590. Salazar, O., I. Wesstro¨ma, and A. Joel. 2008. Evaluation of

of

the

Iowa

River’s

South

Fork

watershed: Part 1. Water quality. Journal of Soil and

DRAINMOD using saturated hydraulic conductivity

Water

estimated by a pedotransfer function model. Agricultural

jswc.63.6.360.

Conservation

63(6):360-370,

doi:10.2489/

USDA NRCS (Natural Resources Conservation Service).

Water Management 95(10):1135-1143. Saleh, A., and D.W. Fryrear. 1999. Soil roughness for revised

2003. Keys to Soil Taxonomy, 9th ed. Washington, DC:

wind erosion equation (RWEQ). Journal of Soil and

USDA Natural Resources Conservation Service, Soil

Water Conservation 54(2):473-476.

Survey.

Sands, G.R., I. Song, L.M. Busman, and B.J. Hansen. 2008.

http://soils.usda.gov/technical/classification/

tax_keys/keysweb.pdf.

The effects of sub-surface drainage depth and intensity

USGS (US Geological Survey). 2001. National elevation

on nitrate loads in the northern Cornbelt. Transactions

database. Reston,VA: US Geological Survey. http://ned.

of the American Society of Agricultural and Biological

usgs.gov. Watson, B.M., S. Selvalingam, and M. Ghafouri. 2003.

Engineers 51(3):937-946. Schilling, K.E., and M. Helmers. 2008. Effects of subsurface

Evaluation of SWAT for modelling the water balance

drainage tiles on streamflow in Iowa agricultural

of the Woady Yaloak River catchment. In Proceedings

watersheds:

of the Modelling and Simulation Society of Australia

Exploratory


Onstad, C.A. 1984. Depressional storage on tilled soil Agricultural Engineers 27(3):729-732.

hydrograph

analysis.

Hydrological Processes 22(23):4497–4506. Singh, R., and M.J. Helmers. 2008. Improving crop growth simulation in the hydrologic model DRAINMOD to

and New Zealand, 873-878, Townsville, Queensland, Australia: Modelling and Simulation Society of Australia and New Zealand.

simulate corn yields in subsurface drained landscapes. In

Wrighr, J., and G. Sands. 2001. University of Minnesota

2008 American Society of Agricultural and Biological

Extension Service. http://www.info.usda.gov/CED/

Engineers Annual International Meeting. American

ftp/CED/neh-16.htm.

Society of Agricultural and Biological Engineers

Zucker, L.A., and L.C. Brown, eds. 1998. Agricultural

Paper No. 083571. St. Joseph, MI: American Society of

Drainage: Water Quality Impacts and Subsurface

Agricultural and Biological Engineers.

Drainage Studies in the Midwest. Ohio State University

Singh, R., M.J. Helmers, W.G. Crumpton, and D.W. Lemke. 2007. Predicting effects of drainage water management

524

in Iowa’s subsurface drained landscapes. Agricultural


Extension Bulletin 871. Columbus, OH: Ohio State University Extension.
