Soil Science Society of America Journal 66:1722-1731 (2002)
© 2002 Soil Science Society of America
DIVISION S-10—WETLAND SOILS
Adapting a Drainage Model to Simulate Water Table Levels in Coastal Plain Soils
X. Hea,
M. J. Vepraskas*,a,
R. W. Skaggsb and
D. L. Lindboa
a Dep. of Soil Science, Box 7619
b Dep. of Biological and Agricultural Engineering, Box 7625, North Carolina State Univ., Raleigh, NC 27695
* Corresponding author (Michael_Vepraskas{at}NCSU.edu)
 |
ABSTRACT
|
|---|
Seasonal saturation in soils is expensive and time consuming to document, but the information is needed for land use assessments. Hydrologic models can be used to assess saturation occurrence quickly if the models are calibrated for individual sites. This study determined whether a drainage model (DRAINMOD) could predict water table levels in soils with and without a perimeter ditch. Water table levels were monitored for up to 3 yr at two toposequences that contained a total of 21 soil plots (3 m by 3 m). Soils included Typic Paleudults, Aquic Paleudults, and Umbric Paleaquults. Each plot was instrumented with a recording well to monitor daily water table levels. DRAINMOD was calibrated for each soil plot using measurements of in situ saturated hydraulic conductivity, soil water characteristic, depth to impermeable layer, depth of rooting, and rainfall. A plot's water table fluctuation was simulated by a system of virtual drains whose distance and depth were adjusted to produce simulated water table fluctuations in line with those actually measured. Further calibration adjusted drainable porosity in the upper 20 cm of the soil, depressional storage, evapotranspiration rate, and depth to impermeable layer. Adjustments were made by iteration to minimize the absolute average deviation between simulated and measured water table levels. Calibration had to be done by plot. Average absolute deviations were generally <20 cm for periods ranging from 1 to 3 yr. The results showed that DRAINMOD could be adapted to simulate water table levels in landscapes that do not contain a network of parallel drains.
Abbreviations: PET, daily potential evapotranspiration • WTD, water table depth
|
INTRODUCTION
|
|---|
FREQUENCY AND DURATION of soil saturation determines a soil's suitability for a variety of uses such as on-site waste water disposal, and also determines whether the site is in a jurisdictional wetland (Mitsch and Gosselink, 1993). Soil scientists predict the shallowest depth that is seasonally saturated by looking for the depth to low chroma or gray colors in a soil profile (Schoeneberger et al., 1998; Vepraskas, 1999). Using soil color to predict depth to seasonal saturation is generally reliable in areas where hydrology has not been altered, but it does not tell us the duration or frequency of saturation at a given depth. Land-use regulations are increasingly requiring that saturation frequency and duration be evaluated (Environmental Laboratory, 1987).
Long-term daily measurements of water table levels are relatively simple to make and provide reliable data on saturation frequency and duration. Due to the variability of the weather, however, a long period of monitoring is required (e.g., >15 yr) to make sure that the monitored water-table levels represent average conditions. Such long-term measurements of water table levels are time-consuming and too expensive to do at all sites where the information is needed. An alternative approach for assessing seasonal saturation is to use hydrologic models to estimate daily water table levels across long (e.g., 40-yr) periods. The water table data can be correlated to soil color patterns, and the colors then used to extrapolate the water data to other sites containing similar soils. Such an approach was used by Boersma et al. (1972) and Simonson and Boersma (1972) with limited success.
Hydrologic modeling can be used to predict long-term historic water table fluctuations on a day-to-day basis for a soil (Skaggs, 1999). The required input data for model calibration can be acquired in a short period (e.g., 6 mo) and the long-term simulations using historic rainfall data can be done on a desktop computer rapidly. Once the long-term daily water table data are obtained, probability values for a specific duration of saturation can be computed for any soil depth. The major advantage of using simulation models is that the effects of annual and seasonal variability of weather can be considered in the analysis.
The hydrologic model DRAINMOD has been extensively used in the USA to analyze the long-term effects of drainage on water table fluctuations (Skaggs, 1980; Fouss et al., 1987; Konyha et al., 1992; Skaggs et al., 1994). The DRAINMOD model was originally developed for poorly drained agricultural fields. It is normally used to simulate the performance of drainage structures and related water table management systems across a long period of climatological record (e.g., 20–50 yr). The model can calculate how often the soil is saturated within a given depth for a specific duration during a certain period in a year. In recent years it has been modified to work at watershed scale to describe the hydrology of drained forested land (McCarthy et al., 1992; Amatya et al., 1997). Abdel-Dayem and Skaggs (1990) extended the application of DRAINMOD to arid regions. Additional versions of the model have also been developed to predict the effects of drainage on N losses (Breve et al., 1997; Zhao et al., 2000) and on soil salinity (Kandil et al., 1993). Reliability of the DRAINMOD model has been verified in extensive field experiments on a wide range of soils, crops, and climatological conditions (Skaggs, 1982; Gayle et al., 1985; Fouss et al., 1987).
DRAINMOD was developed specifically for a soil containing a network of parallel drainage ditches or subsurface drains. Many sites for which hydrologic simulations could be used for land-use assessments are not in agricultural fields and do not have parallel drain tubes or ditches. Hydrology of sites containing a single perimeter ditch, rather than a network of parallel drains, also cannot be simulated with DRAINMOD without adjustment. Despite these limitations, DRAINMOD is a simple model to use because it requires easily measured soil properties. Its output shows water table fluctuations across time, and such data are needed in soil studies. The objective of this study was to test whether DRAINMOD could accurately simulate water table fluctuations in soils with and without a perimeter drain. This work was the first step of a broader investigation that related the results of the hydrologic simulations to soil color patterns.
|
MATERIALS AND METHODS
|
|---|
Experimental Sites
Research was conducted in the North Carolina Coastal Plain at two sites containing toposequences of soils that ranged from well drained to very poorly drained. The Greenville site (Fig. 1A) was located in Pitt County, NC, 
5.1 km southwest of Greenville at 35°34'10'' N, and 77°26'26'' W. A ditch along the perimeter of a portion of the site was 1 to 2 m wide and 0.6 to 1.5 m deep. The average slope at the site was 2%. Vegetation consisted of loblolly pine (Pinus taeda L.), red maple (Acer rubrum L.), and white oak (Quercus alba L.). Most of trees were between 10 and 50 yr old. Additional data on soil morphology at the site was previously reported by Hayes and Vepraskas (2000).
The soil toposequence at the site consisted of soils in the following series: Goldsboro (fine-loamy, siliceous, subactive, thermic Aquic Paleudults), Lynchburg (fine-loamy, siliceous, semiactive, thermic Aeric Paleaquults), Rains (fine-loamy, siliceous, semiactive, thermic Typic Paleaquults), and Pantego (fine-loamy, siliceous, semiactive, thermic Umbric Paleaquults). Soil boundaries were determined by observations made on-site. Experiments were conducted along four transects which contained a total of 13 plots. Each plot measured 3 m by 3 m. The transects were established at increasing distances from the ditch. Each transect consisted of three or four plots along the soil toposequence that were placed to include Goldsboro (moderately well drained), Lynchburg (somewhat poorly drained) and Rains (poorly drained) soils. Each soil contained one plot along each transect. Transect 4 contained an additional plot in the (very poorly drained) Pantego soil. These plots are labeled in Fig. 1A as: 1G, 1L, 1R... 4G, 4L, 4R and 4P to represent transect and series name.
The second site, the Bertie site (Fig. 1B), was located in Bertie County, NC, at 76°48'00'' N, and 36°5'30'' W. Vegetation at the site consisted of loblolly pine, red maple, sweet bay (Magnolia virginiana L.), white oak, red oak [Q. borealis F. Michx. (= Q. rubra L.)], and black cherry (Prunus serotina subsp. serotina). No drainage ditches were present at this site. The soil toposequence consisted of soils in the following series: Noboco (fine-loamy, siliceous, subactive, thermic Typic Paleudults), Goldsboro, Lenoir (fine, mixed, semiactive, thermic Aeric Paleaquults), and Leaf (fine, mixed, active, thermic Typic Albaquults). The experiment was conducted along two transects, labeled North (N) and South (S), and each transect contained five plots. There was only one plot in the Noboco soil. Plots were labeled as 1, 2N, 3N, 4N, 5N, 2S, 3S, 4S, and 5S to represent the soil series (1-Noboco, 2-Goldsboro, 3-Lenoir, 4,5-Leaf) and transect.
Water table levels were monitored daily to depths of 2 m at 0000 h (midnight) in each of the 22 plots at the two experimental sites using RDS automatic monitoring wells (Remote Data Systems, Inc., Wilmington, NC). The water table data were collected from November 1996 until March 1999 at the Greenville site and from December 1996 to October 2000 at the Bertie site.
Wells were installed by boring a 10-cm diam. hole to a depth of 2.25 m, inserting the well, and filling in the space between the well screen and soil with sand. The well screen extended the length of the well beginning at 15 cm below the surface. A 3-cm thick layer of dry bentonite pellets was then placed on the top of the sand to seal the well from surface water inflow. A conical mound of soil was placed over the bentonite to direct surface water away from the well.
To ensure that the recording wells were monitoring daily water levels accurately, a manual check well was installed at each plot to a depth of 127 cm below the mineral soil surface. Every 2 to 3 wk the check wells were measured to compare with the water table data from the recording wells. Rainfall was also measured daily at each site using recording gauges (Onset Computer Corp., Bourne, MA).
Pits were dug in each plot to a depth of 1 m and major soil horizons were described. Undisturbed soil cores (7.6 cm diam. by 7.6 cm height) were collected in each soil horizon except the O horizon (organic layer) by using a Uhland core sampler (Uhland, 1950). Soil water characteristics, which relate soil water contents to specific soil water potentials, were determined for each undisturbed core using the standard pressure plate method (Klute, 1986; Richards and Weaver, 1943). Vertical saturated hydraulic conductivity was measured on each undisturbed core using the constant head method (Klute and Dirksen, 1986) after the measurements of soil water characteristic were completed for pressure levels > -500 cm. Lateral saturated hydraulic conductivity was measured in the field for each layer at each plot using the Compact Constant Head Permeameter (Amoozegar, 1992; Amoozegar and Wilson, 1999). Soil samples were collected from each plot in 15-cm-depth increments to a depth of 225 cm. The samples were air-dried and ground to pass through a 2-mm mesh sieve. Particle size distribution was determined using the hydrometer method (Gee and Bauder, 1986). The position and elevation of each plot in the two experimental sites was determined with a surveyor's transit.
DRAINMOD Description
DRAINMOD is a hydrologic model that simulates water table levels in a soil plot across time from input data consisting of precipitation, evapotranspiration, infiltration, runoff, and subsurface drainage (Skaggs, 1999). DRAINMOD was developed specifically for shallow water table soils with parallel drains that occur on nearly level landscapes. The model computes a water balance on a soil pedon of unit cross-sectional area. A water balance is determined on a day-by-day, hour-by-hour basis, and a water table depth (WTD) is computed for each time step.
The water balance for a time increment 
t can be written as (Skaggs, 1999):
 | [1] |
where Va is change of water free pore space or air volume (cm), D is drainage (or subirrigation) from the soil profile (cm), ET is evapotranspiration (cm), DS is deep seepage (cm), and I is infiltration entering the soil profile (cm). A water balance is also computed at the soil surface for each time increment using:
 | [2] |
where P is precipitation (cm), S is the change in volume of water stored on the surface (cm), and RO is the surface runoff (cm).
The components for Eq. [1] and [2] are illustrated in Fig. 2
for a soil plot like those studied here. The plot has a unit cross-sectional area and extends from the soil surface to the top of a restrictive layer. Precipitation falling on the plot surface collects in shallow depressions and then infiltrates the soil. Once the storage capacity of the depressions is filled, remaining precipitation leaves the plot area by surface runoff. Water that infiltrates moves through an unsaturated zone to the water table. Below the water table, the water drains from the plot by lateral movement or vertically downward through a restrictive layer. The volume of water-free pore space above the water table, Va (cm3 cm-2 surface area), is related to WTD and is a function of the soil water characteristic and thickness of individual soil horizons (Skaggs, 1999). Each plot is considered in isolation, and water draining from a plot is assumed to be lost from the landscape and does not directly affect another plot. The components shown in Fig. 2 apply to any soil plot whether it is artificially drained or not. The figure can be applied to soils having either episaturation or endosaturation (Soil Survey Staff, 1999). Episaturation occurs when the soil below the restrictive layer is unsaturated within 2 m of the surface. Endosaturation is found when the soil is saturated from the top of the water table to below a depth of 2 m.
View larger version (72K):
[in this window]
[in a new window]
|
Fig. 2. Schematic diagram showing the principal components of input and outputs of water used in the DRAINMOD model. ET, evapotranspiration.
|
|
The subsurface drainage rate of water that leaves the pedon below the water table is computed by DRAINMOD in one of two ways depending on whether the surface is ponded with water or not. For the more typical nonponded conditions, the subsurface drainage rate is computed using the steady-state Hooghoudt equation (Bouwer and van Schilfgaarde, 1963). It is assumed that the water table is elliptically shaped between parallel drains and that most drainage occurs below the water table by water moving laterally toward the drains (Fig. 3)
. The Hooghoudt equation may be written as:
 | [3] |
where q is drainage rate (cm h-1), de is the effective depth of the restrictive layer below the drain (cm), m is the water table height above the drain (cm) in the soil plot which is assumed to be located at a point midway between ditches, K is effective lateral saturated hydraulic conductivity (cm h-1), and L is distance between drains (cm).
View larger version (65K):
[in this window]
[in a new window]
|
Fig. 3. Principal components used for the Houghoudt equation in DRAINMOD, which include: de, effective depth of restrictive layer below drain; L, distance between drains; m, height of water table above drain midway between drains; and q, drainage rate.
|
|
During large storms or long continuous wet periods, the water table may rise to the surface. In such cases, the water table does not have the elliptical shape assumed in Hooghoudt equation, and an equation developed by Kirkham (1957) is used in DRAINMOD to quantify drainage.
Climatic inputs for the model include hourly values for precipitation and daily potential evapotranspiration (PET). Direct PET input data were not available at either of the two study sites. The Thornthwaite (1948) method was used to calculate PET in this study. Daily maximum and minimum temperature data were obtained from the nearest available weather station. The monthly correction factors of PET used in the model were obtained from Workman et al. (1994). Because the vegetation at the two experimental sites was forest, rooting depth was the only plant factor considered. Rooting depth was determined from soil pits dug in each plot.
Soil property inputs include soil water characteristic of the A horizon and saturated hydraulic conductivity of soil horizons from the surface to the top of the restrictive layer. Infiltration is calculated by the Green-Ampt equation (Green and Ampt, 1911):
 | [4] |
where f is infiltration rate (cm h-1), F is cumulative infiltration (cm), and A and B are parameters that depend on soil properties. The relationship of parameters A and B with WTD were calculated in this application using the soil preparation program within DRAINMOD.
Surface storage capacity was characterized by the average depth of depression storage that must be satisfied before runoff can begin. Depressional storage parameters are generally estimated visually in the field according to the topography. In most cases, it is assumed that depressional storage is evenly distributed over the field. Depressional storage parameters <0.5 cm indicate little ponding of water occurs because the surface is relatively smooth. Areas with some depressions where water ponds after rains have depression storage values between 1 and 1.5 cm, while areas with many depressions causing widespread ponding of water after heavy rains have depression storage values >2 cm.
Model Calibration
The DRAINMOD model was calibrated separately for each experimental plot using a short-term record of observed weather data and water table measurements recorded across a 1- to 3-yr period. Predicted and monitored water table fluctuations were compared and then model parameters were adjusted individually to bring predicted values in line with measured ones. The agreement between monitored and predicted daily water table depths was quantified by the absolute deviation (
) for the observed period, defined as follows:
 | [5] |
where n was number of days in observed period, Ym was monitored WTD at midnight of each day (cm), and Yp was corresponding predicted WTD (cm).
Assumptions
This research was based on three assumptions that were made for the experimental sites and tested as part of the work: (i) Water table levels can be predicted for individual soil plots in a landscape by treating each plot in isolation, and each plot is calibrated separately from the other plots; (ii) Subsurface drainage rates can be approximated in landscapes by using the Hooghoudt equation which uses drain spacing and depth as well as depth to a restrictive layer to compute the drainage flux (the drainage system used for calibration in this case is a virtual one); and (iii) Deep seepage losses are virtually zero, or so small that they can be included with losses by subsurface drainage.
These assumptions apply to many Coastal Plain soils, but may not apply to soils in other landscapes. Plots were treated in isolation because we wanted to simulate water table fluctuations at each well in a soil plot. DRAINMOD is appropriate for such simulations because it does not use a groundwater inflow variable to simulate water table levels into a soil plot. Groundwater flow across the landscape was not of interest here. When groundwater flow lines need to be computed across a landscape, then models other than DRAINMOD should be used.
|
RESULTS AND DISCUSSION
|
|---|
Soil Properties
The particle size distribution data for selected plots at both sites, Table 1, show the range of textures found in the soils. Soils at the Greenville site tended to be sandier than those at the Bertie site. The saturated water content ranged from 0.31 to 0.43 cm3 cm-3 at the Greenville site, and from 0.32 to 0.46 cm3 cm-3 at the Bertie site (data not shown). Lateral saturated hydraulic conductivity (K) values (Table 2) for representative plots at both sites decreased with depth and were similar among the sites for comparable depths. The data used to initialize the DRAINMOD program for each plot included the K values and soil water characteristic. Crop inputs included the rooting depth for each plot observed in soil pits.
View this table:
[in this window]
[in a new window]
|
Table 1. Particle size distribution data for two plots that contained Goldsboro soils at the Greenville and Bertie sites.
|
|
Calibrating DRAINMOD to Account for Drainage Rate
Water table levels were monitored at the Greenville site from November 1996 to March 1999, and from December 1996 to June 2000 at the Bertie site. At the Bertie site, rainfall data prior to December 1999 were incomplete due to malfunctions of the gauge during a hurricane and to a bird nesting in the gauge during another period. The monitored rainfall and well data from January 2000 to June 2000 were selected to calibrate the model because reliable data were available for the period.
After inserting the values for measured soil properties and daily rainfall into the model, DRAINMOD needed additional calibration for each plot in order to simulate measured daily water-table levels accurately. The calibration process was found to produce the best results when variables were adjusted in the order of: drain spacing, drain depth, depth to a restrictive layer, depressional storage, drainable porosity of O horizons, and ET.
The calibration process began by adjusting subsurface drainage, because this variable has a major influence on the water-table levels and is very sensitive to drain spacing. As drain spacing increases, the drainage rate decreases for a given WTD. For the plots considered here, the drainage rate varied with a plot's position in the landscape. Plots at higher elevations had deeper water tables and drained faster than those at lower elevations that had higher water table levels. Rather than rewriting the drainage algorithm in DRAINMOD, we used the Hooghoudt equation with drain spacing and depth as calibration parameters. The drain spacing was adjusted by trial and error for each plot to minimize the average absolute deviations between measured and predicted water table levels. For these adjustments, a drain depth of 65 cm was used initially because this was the average depth of the perimeter ditch at the Greenville site. We did not consider any other aspect of the perimeter ditch in this calibration process because there is no place in DRAINMOD for perimeter ditches. The model was developed for a network of parallel ditches, not simply one.
Changing drain spacing had a considerable impact on the simulation of the water-table levels (Fig. 4)
. Figure 4 shows the effect that changes to drain spacing (L) had on the simulated water table levels compared with the measured values for the period November 1996 to March 1999 for plot 2G at the Greenville site. The optimum value that produced the lowest value for L in this plot was 110 cm. As L increased, the drainage rate decreased, and the predicted water table level rose closer to the surface. Plots in Lynchburg and Rains soils were at lower elevations than the Goldsboro plots, and had smaller drainage rates. Therefore, larger values of L were required to optimize drainage rate and the final values were 1700 cm for Lynchburg plots and 4000 cm for Rains plots.
The final drain spacing parameters for all plots are given in Table 3. There were differences in drain spacing for some plots found in the same soil series because small elevation differences and location on the landscape among plots affected drainage rates. The effect of elevation differences on spacing parameter values can be seen by comparing elevation and L values of plot 2R (0.01 m above datum) to those of plot 1R (0.37 m above datum) at the Greenville site. Plot 2R required a larger drain spacing than plot 1R because its lower elevation caused its natural drainage rate to be lower. While the ditches used in this calibration process were virtual ones, increasing ditch spacing has the same effect on the water table fluctuations as reducing the hydraulic gradient in Darcy's law. Deep, closely spaced ditches result in large hydraulic gradients with relatively high drainage rates as would be found in a moderately well-drained soil. Conversely, widely spaced ditches result in small hydraulic gradients with slow drainage rates as would be found in a poorly drained soil. The effect of slope on drainage rate can be considered in the model (Fipps and Skaggs, 1989), but was not considered in this investigation.
View this table:
[in this window]
[in a new window]
|
Table 3. Calibrated drainage input parameters for the Greenville and Bertie sites. These parameters were obtained by calibration to represent the best relationship between net drainage flux and water table depth for each plot.
|
|
Once the L parameter was determined, the drain depth parameter (de) was adjusted. The effect of different values for the drain depth parameter on the predicted water table levels is shown in Fig. 5
. The predicted water table was raised with decreased drain depth. The drain depth of 65 cm, used for the initial calibration, was found to be optimum for all plots at the Greenville site. As shown in Table 3, optimum drain depths were either 50 or 100 cm at the Bertie site.
View larger version (29K):
[in this window]
[in a new window]
|
Fig. 5. Effect of different drain depths (Dd) on water table fluctuations in Plot 2G at the Greenville site.
|
|
The de parameter, representing the equivalent depth to the restrictive layer in the Hooghoudt equation, also needed to be adjusted. This parameter affected the predicted water table level during the dry season (Fig. 6)
. The shallower the restrictive layer, the higher (i.e., closer to the surface) the predicted water table will be during the dry season. The K measurements suggested the restrictive layer would occur between 1 to 2 m for most plots but no clear depth could be identified in any plot where the Ksat values abruptly decreased. However, Eq. [3] requires that a specific depth be used in the model where a restrictive layer effectively begins. The actual values of this parameter for all the plots were determined by trial and error and ranged between 120 and 250 cm for both sites (Table 3). The restrictive layer occurred in the mid- to lower portions of Bt horizons whose saturated K values were between 0.01 and 0.02 cm h-1. There was no abrupt increase in clay percentage at the point where the restrictive layer began (Hayes, 1998), so its precise depth could not be determined in the field. We did not evaluate whether saturation occurred below the restrictive layer and therefore do not know whether the soil plots were episaturated or endosaturated.
View larger version (25K):
[in this window]
[in a new window]
|
Fig. 6. Effect of different depths to restrictive layer (de) on water table fluctuations in Plot 2G at the Greenville site.
|
|
Calibrating DRAINMOD for Depressional Storage, Drainable Porosity, and Evapotranspiration
Depressional storage is important for sites where the water table frequently rises to the soil surface. This parameter represents the average depths of pocket-like depressions on the surface that hold water. It has a minor effect for sites having deeper water tables. Runoff occurs after depressional storage sites are filled. Values of the surface depressional storage were selected to optimize agreement between predicted and measured water table depths. As shown in Table 3, depressional storage values were lowest in the upland areas and increased as soils progressively became poorly drained. The highest depressional storage value of 5 cm was found for the Pantego plot (4P) at the Greenville site. The plot was in a depression that had a hummocky surface which may have been created as tree-throw mounds.
During high water table seasons (from November to March), simulated and measured water tables in some plots of Goldsboro and Lynchburg soils at the Greenville site showed good agreement (data not shown). However, other wetter plots showed much greater fluctuations in the predicted water table levels than in the measured values, particularly where the measured water table was within 20 cm of the surface, as illustrated in Fig. 7A
for Plot 2R of the Greenville site. The cause of this great variability was improper selection of drainable porosity values for the O horizons.
View larger version (33K):
[in this window]
[in a new window]
|
Fig. 7. Comparison of the simulated and measured water table fluctuations at Plot 2R of the Greenville site illustrating the effects of drainable porosity on degree of fluctuations. (A) Simulated water levels show a high degree of fluctuation during the period of November 1996 to May 1997 because drainable porosity within 20 cm of the surface was assumed to be low. (B) Fluctuation has been reduced as drainable porosity was increased to account for a porous O horizon.
|
|
Drainable porosity is defined as the volume of water drained from the soil plot for each unit change of the WTD. The affect of drainable porosity on water table change can be illustrated with a simple example. If 1 cm of water falls on a mineral soil having a drainable porosity of 0.01 cm cm-1, the water table would rise 100 cm (i.e., 1 cm/0.01 cm cm-1). On the other hand, if 1 cm of rain falls on a soil having a drainable porosity of 0.2 cm cm-1, the water table would rise only 5 cm. Thus, drainable porosity values have a large influence on how much the water table will fluctuate as the soil gains or loses water.
The relationship between drained volume, or water-free pore space (cm), and WTD was originally calculated from the soil water characteristics for all the soil horizons except the O horizon. The drainable porosity of the upper 20 cm of mineral soil at Plot 2R ranged from 0.001 to 0.01 cm cm-1. However, the overlooked O horizon (organic soil material) in the plot was also 20-cm thick and consisted of partially decomposed leaves, twigs, and roots that made the horizon very porous. It had a much higher drainable porosity than the mineral soil beneath it. Thus, when the relationship between drainage volume and WTD was based solely on soil water characteristics of the mineral soil (ignoring the O horizon), the simulated water table fluctuated rapidly with the addition or removal of water, because the mineral soil had relatively little pore space to store water. In the drier plots (e.g., 1 L of the Greenville site), the water table rarely came within the O horizon and, consequently, the high drainable porosity of the organic layer did not affect simulated water table fluctuations. The drainable porosity of the horizon at a depth of 0 to 20 cm was adjusted to 0.2 cm cm-1 to take into account the properties of the O horizon. These adjustments minimized the fluctuations in the water table at this depth (Fig. 7B).
The last parameter adjusted was the ET factor. Differences between predicted and measured water tables were also found in a few months such as March and April (Fig. 7B). The correction factor for ET was adjusted to improve the simulation in these months (Fig. 8)
. The ET correction factors for selected plots are shown for all months in Table 4 to illustrate the range in values found.
View larger version (25K):
[in this window]
[in a new window]
|
Fig. 8. Water table fluctuations with adjusted evapotranspiration (ET) at Plot 2R. The effect of the adjustment can be seen by comparing this figure for the month of March with that of the Fig. 7B, which did not have ET adjusted.
|
|
View this table:
[in this window]
[in a new window]
|
Table 4. A summary of monthly correction factor for evapotranspiration calculation in DRAINMOD for selected plots at both sites.
|
|
Evaluation of the Adjusted DRAINMOD Models
A summary of the average deviations between measured and simulated water table levels for individual plots at the two experimental sites is presented separately in Table 5. The water table fluctuations within 100 cm of the surface were of greatest interest in this study. During the wet season, November to April, the predicted and observed water table depths for most plots at the Greenville site were in good agreement with the average deviations ranging from 3 cm to 25 cm. The agreement between predicted and observed values at all plots was particularly good during the period of November 1996 to April 1997 and the period of November 1997 to April 1998. The average deviations for most plots were <10 cm from November 1996 to April 1997. During the next wet season (1997 to 1998), the average deviations were <15 cm for most plots. Although agreement between predicted and observed results was not as close as at the other plots, it is still satisfactory considering the complexity of drainage processes and the variability of field conditions. Because the wells did not measure water tables below a depth of 2 m during the dry season, the average deviations for this period were not considered in the evaluation of simulation results.
View this table:
[in this window]
[in a new window]
|
Table 5. A summary of average absolute deviations () used for comparison of observed water table fluctuations with predictions by DRAINMOD for the Greenville site and Bertie sites.
|
|
At the Bertie site, rainfall data were not measured for as long as at the Greenville site and were also not as complete as at the Greenville site. Fortunately, we did have reliable data during the wet season from January to June of 2000. Average deviations varied from 7 cm to 12 cm. The simulated water table agreed very well with monitored values. The prediction models calibrated with the 2000 data were also compared with the measured data from January to June of 1999, which were not available at 2S and 5S. Agreement between observed and predicted results were considered satisfactory for this period.
|
SUMMARY AND CONCLUSIONS
|
|---|
DRAINMOD was tested for its ability to simulate water table levels at two experimental sites in the North Carolina Coastal Plain. The model was developed for fields that contain a network of parallel drains, but neither site in this study satisfied the requirement for the application of DRAINMOD, and a virtual drainage network was assumed to be present at both sites. In this study, six model parameters were adjusted to bring simulated water table levels in line with measured values: virtual drain spacing, virtual drain depth, depth to restrictive layer, depressional storage, drainable porosity of the organic horizons, and evapotranspiration. These parameters probably need to be adjusted in this order to obtain optimum results, because each causes a specific change in the simulation results.
Model simulations were accurate among the 21 plots evaluated at two sites. The successful application of DRAINMOD to these sites showed that our three assumptions were justified. DRAINMOD has to be calibrated by soil to accurately simulate water table fluctuations in a landscape. Subsurface drainage rates were approximated using the Hooghoudt equation, which uses drain spacing and depth as well as depth to a restrictive layer to compute a drainage flux. Deep seepage losses were either zero or included with losses by subsurface drainage. The drainage algorithm used (Hooghoudt equation) considered both hydraulic conductivity by horizon and hydraulic head gradient which decreased as the water table fell. Algorithms in DRAINMOD can be used to consider effects of slope on drainage rates, but were not applied in this analysis.
Porous organic layers were found to have large effects on water table fluctuations in the plots where water tables came to the surface. The large drainable porosity of such layers must be accounted for to minimize water table fluctuations near the surface. Because most agricultural fields do not have such layers, their effect has been overlooked in earlier applications of this model.
Calibration by adjusting input parameters is justified because measured inputs always include some errors. Workman and Skaggs (1989) stated that errors caused by the limitations of the accuracy of the required inputs in DRAINMOD appear to be more important than errors due to the approximations in the model. Calibrating the model exposes the problems in the measured data inputs and so improves the accuracy of the final prediction. Once the model is calibrated, long-term water table levels can be computed using historic rainfall data in order to determine the probability that a site will be saturated at selected depths.
|
ACKNOWLEDGMENTS
|
|---|
Funding for the project was obtained from the U.S. Environmental Protection Agency (contract no. CR 824735-01-0) and the Water Resources Research Institute of the University of North Carolina (WRRI Project no. 70175). Their assistance is greatly appreciated.
Received for publication August 10, 2001.
|
REFERENCES
|
|---|
- Abdel-Dayem, S., and R.W. Skaggs. 1990. Extension of DRAINMOD for simulating water management in arid regions. p. 201–212. In Proc. Symp. Land Drainage for Salinity Control in Arid and Semi-arid Regions, Vol. 2., Cairo, Egypt. 25 Feb.–2 Mar. 1990. Drainage Res. Inst., WRC, Cairo, Egypt.
- Amatya, D.M., R.W. Skaggs, and J.D. Gregory. 1997. Evaluation of a watershed scale forest hydrologic model. Agric. Water Manage. 32:239–258.
- Amoozegar, A. 1992. Compact constant head Permeameter: A convenient device for measuring hydraulic conductivity. p. 31–42. In G.C. Topp et al. (ed.) Advances in measurement of soil physical properties: Bringing theory into practice. SSSA Spec. Publ. 30. SSSA, Madison, WI.
- Amoozegar, A., and A.W. Wilson. 1999. Methods for measuring hydraulic conductivity and drainable porosity. p. 1149–1205. In R.W. Skaggs et al. (ed.) Agricultural drainage. Agron. Monogr. 38. ASA, CSSA, and SSSA, Madison, WI.
- Boersma, L., G.H. Simonson, and D.G. Watts. 1972. Soil morphology and water table relations: I. Annual water table fluctuations. Soil Sci. Soc. Am. Proc. 36:644–648.
- Bouwer, H., and J. van Schlifgaarde. 1963. Simplified method of predicting fall of water table in drained land. Trans. ASAE 6:288–291.
- Breve, M.A., R.W. Skaggs, J.E. Parsons, and J.W. Gilliam. 1997. DRAINMOD-N, A Nitrogen Model for Artificially Drained Soils. Trans. ASAE 40:1077–1085.
- Environmental Laboratory. 1987. Corps of Engineers Wetlands Delineation Manual. U.S. Army Corps of Eng. Tech. Rep. Y-87-1. U.S. Gov. Print. Office, Washington, DC.
- Fipps, G., and R.W. Skaggs. 1989. Influence of slope on drainage of hillsides. Water Resour. Res. 25:1717–1726.
- Fouss, J.L., R.W. Skaggs, and J.S. Rogers. 1987. Two-stage weir control for subsurface drainage in humid areas. Trans. ASAE 30:1713–1719.
- Gayle, G.A., R.W. Skaggs, and C.E. Carter. 1985. Evaluation of a water management model for a Louisiana sugar cane field. J. Am. Soc. Sugar Cane Tech. 4:18–28.
- Gee, G.W., and J.W. Bauder. 1986. Particle-size analysis. p. 383–411. In A. Klute (ed.) Methods of soil analysis. Part 1. 2nd ed. ASA and SSSA, Madison, WI.
- Green, W.H., and G. Ampt. 1911. Studies of soil physics, Part I: The flow of air and water through soil. J. Agric. Sci. 4:1–24.
- Hayes, W.A., Jr. 1998. Effect of ditching on soil morphology, saturation, and reduction in a catena of Coastal Plain soils. M.S. thesis. North Carolina State Univ., Raleigh, NC.
- Hayes, W.A., Jr., and M.J. Vepraskas. 2000. Morphological changes in soils produced when hydrology is altered by ditching. Soil Sci. Soc. Am. J. 64:1893–1904.[Abstract/Free Full Text]
- Kandil, H.M., R.W. Skaggs, S. Abdel-Dayem, and Y. Aiad. 1993. DRAINMOD-S: Water management model for irrigated lands, crop yield and application. p. 257–275. In E. Lorre (ed.) Subsurface drainage simulation models. Trans. Workshop, 15th Congress ICID. The Hague, The Netherlands.
- Kirkham, D. 1957. Theory of land drainage. In J.N. Luthin (ed.) Drainage of agricultural lands. Agron. Monogr. 7. ASA, Madison, WI.
- Klute, A. 1986. Water retention: Laboratory methods. p. 635–662. In Methods of soil analysis. Part 1. 2nd ed. ASA and SSSA, Madison, WI.
- Klute, A., and C. Dirksen. 1986. Hydraulic conductivity and diffusivity: Laboratory methods. p. 687–734. In Methods of soil analysis, Part 1. 2nd ed. ASA and SSSA, Madison, WI.
- Konyha, K.D., R.W. Skaggs, and J.W. Gilliam. 1992. Effects of drainage and water management practices on hydrology. J. Irrig. Drain. 118:807–819.
- McCarthy, E.J., J.W. Flewelling, and R.W. Skaggs. 1992. A hydrologic model for a drained forested watershed. J. Irrig. Drain. Eng. 118:242–255.
- Mitsch, W.J., and J.G. Gosselink. 1993. Wetlands. 2nd ed. Van Nostrand Reinhold Co., New York.
- Richards, L.A., and L.R. Weaver. 1943. Fifteen atmosphere percentage as related to the permanent wilting percentage. Soil Sci. 56:331–340.
- Schoeneberger, P.J., D.A. Wysocki, E.C. Benham, and W.D. Broderson. 1998. Field book for describing and sampling soils. USDA-NRCS, National Soil Survey Center, Lincoln, NE.
- Simonson, G.H., and L. Boersma. 1972. Soil morphology and water table relationships. II. Correlation between annual water table fluctuations and profile features. Soil Sci. Soc. Am. Proc. 36:649–653.
- Skaggs, R.W. 1980. Combination surface-subsurface drainage systems for humid regions. J. Irrig. Drain. Div. Am. Soc. Civ. Eng. 106(IR4):265–283.
- Skaggs, R.W. 1982. Field evaluation of a water management model. Trans. ASAE 25:666–674.
- Skaggs, R.W. 1999. Drainage simulation models. p. 461–492. In R.W. Skaggs et al. (ed.) Agricultural drainage. Agron. Monogr. 38. ASA, CSSA, and SSSA, Madison, WI.
- Skaggs, R.W., D. Amatya, R.O. Evans, and J.E. Parsons. 1994. Characterization and evaluation of proposed hydrologic criteria for wetlands. J. Soil Water Conserv. 49:501–510.
- Soil Survey Staff. 1999. Soil taxonomy. USDA Agric. Handb. 436. USDA-NRCS, U.S. Gov. Print. Office, Washington, DC.
- Thornthwaite, C.W. 1948. An approach toward a rational classification of climate. Geog. Rev. 38:55–94.
- Uhland, R.E. 1950. Physical properties of soils modified by crops and management. Soil Sci. Soc. Am. Proc. 14:361–366.
- Vepraskas, M.J. 1999. Redoximorphic features for identifying aquic conditions. Tech. Bull. 301. NC Agric. Res. Serv., Raleigh, NC.
- Workman, S.R., J.E. Parsons, G.M. Chescheir, R.W. Skaggs, and J.F. Rice. 1994. DRAINMOD User's guide. USDA-NRCS, U.S. Gov. Print. Office, Washington, DC.
- Workman, S.R., and R.W. Skaggs. 1989. Comparison of two drainage simulation models using field data. Am. Soc. Agric. Eng. 32(6):1933–1938.
- Zhao, S.L., S.C. Gupta, D.R. Huggins, and J.F. Moncrief. 2000. Predicting subsurface drainage, corn yield and nitrate nitrogen losses with DRAINMOD-N. J. Environ. Qual. 29:817–825.[Abstract/Free Full Text]
This article has been cited by other articles:
|
|
|
|
 
C. P. Morgan and M. H. Stolt
Soil Morphology-Water Table Cumulative Duration Relationships in Southern New England
Soil Sci. Soc. Am. J.,
March 29, 2006;
70(3):
816 - 824.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. J. Vepraskas, X. He, D. L. Lindbo, and R. W. Skaggs
Calibrating Hydric Soil Field Indicators to Long-Term Wetland Hydrology
Soil Sci. Soc. Am. J.,
July 1, 2004;
68(4):
1461 - 1469.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
S. Fiedler and M. Sommer
Water and Redox Conditions in Wetland Soils--Their Influence on Pedogenic Oxides and Morphology
Soil Sci. Soc. Am. J.,
January 1, 2004;
68(1):
326 - 335.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
X. He, M. J. Vepraskas, D. L. Lindbo, and R. W. Skaggs
A Method to Predict Soil Saturation Frequency and Duration from Soil Color
Soil Sci. Soc. Am. J.,
May 1, 2003;
67(3):
961 - 969.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
