Table of Contents

Introduction

Part I

Appendix A

Appendix B

Appendix C

Part II A and B

Part II C and D

Part III

Part IV

References

Summary

Part II. Recharge and Water Budgets of the Kansas High Plains and Associated Aquifers

Preface to Part II

This part attempts to summarize most major studies that have quantified ground-water recharge in the Kansas High Plains and associated aquifers. (A note on aquifer nomenclature is presented below.) Those studies are divided into 1) regional climatic soil-water balance studies; 2) regional ground-water modeling or analysis studies; 3) Kansas basin- to county-scale ground-water studies; and 4) field-based experimental studies. For each of those studies, the methodology employed was briefly outlined and the water budget of the study region was summarized in a uniform style of inches per year of the hydrologic quantity of interest over the study area. This compilation and synthesis includes some original research as well, in that information not explicitly stated by the authors was derived from their data, and additional information in studies involving this author is presented in this compilation as well. The assumptions and limitations of the model or analyses used are emphasized; whenever estimation errors or uncertainties were quantified, those are explicitly stated in this report. (A note on uncertainty measures is presented below.) Emphasis also was placed on environmental and land-use factors affecting recharge estimates. Thus whenever possible, recharge estimates from predevelopment and development conditions were distinguished, and the recharge impacts of irrigation development are presented.

Note on Aquifer Nomenclature

The High Plains aquifer is a regional aquifer system underlying parts of eight states in the Great Plains from South Dakota to Texas. In Kansas, the High Plains aquifer lies beneath approximately 33,500 square miles of western and central Kansas and is composed of several units that are geologically similar and hydraulically connected. The most extensive unit of the High Plains aquifer is the Tertiary-age (Miocene-Pliocene) Ogallala Formation, popularly known as the Ogallala aquifer. The eastern extension of the High Plains aquifer in Kansas is composed of younger, generally Pleistocene sediments, similar to the Ogallala, that include the Great Bend Prairie and Equus Beds aquifers. Also lying above the Ogallala Formation are other Pleistocene and Holocene deposits that also fill the valleys of modern streams. Where these fluvial and eolian deposits and stream valleys are hydraulically connected to the High Plains aquifer, these are considered part of the High Plains aquifer.

Note on Measures of Uncertainty

To measure how accurate the estimate of the mean value of a variable (such as recharge) is, we can compute its standard deviation from the mean, most often referred to as standard error (Glantz, 1981). The term standard error is a statistical term for the degree of uncertainty inherent in estimating a mean value. The standard error quantifies the reliability of the estimate of the population (true) mean from a sample drawn randomly from the population. Because the certainty with which the mean can be estimated increases as the sample size increases, the standard error of the mean decreases as the sample size increases. Conversely, the more variable the original population, the more variable the possible mean values of samples. Because the population of all sample means follows a normal distribution at least approximately, the true mean of the original population lies within two standard errors of the sample mean about 95% of the time. With this information we can construct an interval that represents the range of values over which the mean can be expected to vary.

A. Climatic Soil-water Balance Studies on Regional Scales

Two major regional climatic soil-water balance studies have been recently conducted, one for the state of Kansas (Hansen, 1991) and the other for the entire High Plains aquifer (Dugan and Zelt, 2000). Because of their importance, and of the fact that the Hansen (1991) study was mostly adopted by the Division of Water Resources of the Kansas Department of Agriculture in the cases where more specific recharge data were not already available, those studies are analyzed in some detail in this report, especially in view of the fact that the Hansen (1991) study was somewhat short on details on methodology and data used.

A1. USGS Study on Natural Recharge for Principal Aquifers in Kansas (Hansen, 1991)

Hansen (1991) estimated “potential natural recharge” for the entire state of Kansas by extending the results of the soil-water budget model and methodology employed by Dugan and Peckenpaugh (1985) for the Central Midwest Regional Aquifer Systems Analysis (CMRASA) in parts of Arkansas, Colorado, Kansas, Missouri, Nebraska, New Mexico, South Dakota, and Texas, to the High Plains aquifer of Kansas. Potential natural recharge refers to the deep percolation rate of soil water (made available from precipitation) below the root zone, where the water is presumed to be below the zone of influence of evapotranspiration processes, and thus potentially available to move downwards towards the water table and thereby eventually recharge the aquifer.

Because, as mentioned previously, the results of that study have been generally adopted by the Division of Water Resources of the Kansas Department of Agriculture, but the data, methodology, and limitations have not been reported in detail in the Hansen (1991) report, a brief summary of these aspects is presented below based on the reports of Dugan and Peckenpaugh (1985) and Dugan and Zelt (2000), and personal communication with Hansen (July 2002).

The soil-water balance simulation procedure employed requires four types of input: 1) monthly precipitation (P), 2) computed monthly potential evapotranspiration (PET) values, 3) hydrologic properties of soils, and 4) vegetation types. The PET calculation was based on the Jensen-Haise method (Jensen, 1974), which requires monthly temperature and solar-radiation data. If direct solar-radiation measurements are not available, they can be estimated from the following data (Dugan and Peckenpaugh, 1985): percent of possible sunshine or cloud-cover data, angle of sun’s inclination at zenith (noon), hours of possible sunshine, and altitude above sea level.

The soil-water balance approach accounts for soil water entering, leaving, and remaining within the root zone, and can be summarized in the following simple equation:

R = ASW + P - SRO - AET - SC, . . . . . . . . . .(II-1)

where R = potential recharge (deep percolation), ASW = antecedent soil water within the root zone, P = precipitation, SRO = surface runoff, AET = actual evapotranspiration, and SC = total available soil-water storage capacity of the root zone. All above components are expressed in inches per month.

The soil-water balance approach emphasizes the physical factors (climate, soils, and vegetation) that determine the availability of water for recharge and consumptive water use. Observed monthly climatic data from numerous weather stations in Kansas were compiled for the period of study (1951-1980). Land-use data, by county, which provided vegetation patterns, were derived from 1978 statistics collected for the Census of Agriculture (U.S. Department of Commerce, 1980). Changes in vegetation over the period of study were considered small and thus neglected. The soils information was derived from a report by Dugan (1985) that consists of quantitative descriptions and areal distributions of the soils in Kansas and surrounding states based on their hydrologic characteristics, and from other generalized soils information in Kansas.

Six general vegetation types were considered in the state, each with distinctive seasonal consumptive water requirements, rooting depths, and infiltration-runoff relationships that create significant different demands on available moisture (Dugan and Peckenpaugh, 1985): 1) row crops, principally corn, soybeans, and grain sorghum; 2) tame hay, principally alfalfa; 3) small grain, principally winter wheat; 4) native grassland or pasture; 5) fallow or idle land; and 6) woodlands (urban area included). Each vegetation type is characterized by its consumptive water requirement (CWR) value, which is the quantity of water that vegetation type will consume if the availability of soil water is not a limiting factor. The CWR for each vegetation type was derived by multiplying the monthly value of potential evapotranspiration, PET, by a monthly crop coefficient (expressed as a simple ratio of CWR to PET; Dugan and Peckenpaugh, 1985). The difference between the amount of water required to meet the CWR and the water available within the plant root zone is the soil-water deficit (SWD).

The numerous soil groups within the state were reduced to 10 groups for computational purposes within the soil-water balance program (Dugan and Peckenpaugh, 1985). The availability of water for consumptive water use is influenced by three physical characteristics of the soil: hydraulic conductivity, available-water capacity, and slope, by regulating both infiltration and the ability of the soil profile to store water. Infiltration is largely a function of hydraulic conductivity and slope, while the water-storage capacity is determined by the product of the available-water capacity (AWC) of the soil and the root-zone depth. The relationship between precipitation and infiltration was incorporated in four empirical infiltration curves for varying soils, topography (slope), and land-use conditions (vegetation types) (Dugan and Peckenpaugh, 1985). Amounts of surface runoff and infiltration are computed from those infiltration curves. Runoff is that part of precipitation that does not enter the soil and is not accounted for in the soil-water balance. All infiltration is accounted for as either evapotranspiration or recharge. AET is a function of both the CWR of the vegetation type of interest and the soil water available within the root zone (Dugan and Peckenpaugh, 1985; Dugan and Zelt, 2000).

The soil-water balance program calculates, on a monthly interval, a variety of outputs such as infiltration, surface runoff, soil water stored in the root zone, consumptive water use, soil-water deficits, actual evapotranspiration, and deep percolation or potential recharge. That program computes the results for each climatic station for the various possible combinations of the soils and vegetation types in the study area. This output is then areally distributed through a “water-use program” (Dugan and Peckenpaugh, 1985) that weighs the outputs on the basis of percentage of occurrence of the various vegetation types and soils within the grid elements in the study area. The interpolation procedure is based on the distance of the two or three nearest climatic stations to weigh or adjust the soil-water balance program’s output to the centerpoint of each grid element (Dugan and Peckenpaugh, 1985). A simple flow chart of the input and output of the various soil-water simulation components is shown in fig. II-1, and the resultant distribution of potential natural recharge is shown in fig. II-2.

Figure II-1—Flowchart of soil-water simulation (adapted from Dugan and Zelt, 2000).

FIGURE II-2—Mean annual potential natural recharge (in inches per year) and extent of High Plains aquifer in Kansas (adapted from Hansen, 1991).

From the above summary of the methodology and data employed in estimating potential natural recharge, it is obvious that numerous simplifications and approximations were made in such estimation, resulting in relatively large uncertainties in the results. Equating deep percolation to aquifer recharge is premised on the assumptions that the immediate underlying aquifer is unconfined and true water-table conditions exist, and that all water that passes through the root zone into the unsaturated zone below ultimately reaches the underlying aquifer. In addition, such potential recharge is determined from factors independent of the properties of the aquifer. It also should be noted that return flow from irrigation and underflow from areas outside Kansas were not considered in estimating potential recharge in the Hansen (1991) study. Also the results of that study have not been calibrated or compared to actual measurements or more detailed estimates. Therefore, caution should be exercised in directly applying the results of that study to specific areas because of the general nature (generalized CWR/PET relationships, land use, and soil characteristics) of that study.

However, such analysis provides valuable insights into the hydrologic system in an area. The spatial patterns of potential recharge and consumptive water use, systematically derived from measurable climatic, soil, and vegetation characteristics, are of great use to water-resources managers and planners. Such overall patterns of resultant recharge (fig. II-2) indicate that the controlling elements are the climatic factors themselves, particularly precipitation. The generalization can be made that as precipitation declines, both the magnitude and proportion of precipitation contributed to recharge declines (Dugan and Peckenpaugh, 1985). Also, areas of high cool-season precipitation tend to receive higher amounts of recharge (Dugan and Peckenpaugh, 1985). Smaller variations within local areas, however, are related to differences in soils and vegetation. The effect of soils, for example, is apparent in the westward extension of potential recharge contours in the Equus Beds and Great Bend Prairie aquifers as well as the Arkansas River sand dune areas in southwestern Kansas that coincide with sandy soils (fig. II-2). The role of vegetation is less apparent because regional vegetation changes usually are gradual (Dugan and Zelt, 2000). However, under similar precipitation and soils, potential recharge tends to be larger for cropland than natural vegetation.

A2. USGS Study of Soil-water Conditions in the Great Plains (Dugan and Zelt. 2000)

Dugan and Zelt (2000) expanded their Central Midwest Regional Aquifer Systems study (CMRASA; Dugan and Peckenpaugh, 1985) to the Great Plains and adjacent areas, including the entire Kansas High Plains aquifer. A major difference of this study to the previous regional climatic soil-water balance studies of Dugan and Peckenpaugh (1985) and Hansen (1991) was that the impacts of irrigation on recharge were considered. Thus, the soil-water balance, computed monthly over the period 1951-1980, is summarized by the following simplified equation (compare to eq. 1):

R = ASW + P + I - SRO - AET - SC . . . . . . .(II-2)

where I = irrigation water required in inches per month, and all other terms are identical to the ones shown for eq. II-1. The simple flow chart of the input and output of the various soil-water simulation components, shown in fig. II-1, explains the structure of the soil-water simulation employed, taking advantage of GIS technology. With the exception of incorporating irrigation return flow to recharge estimation, this study is subject to the same assumptions and limitations as in the previously mentioned USGS soil-water balance studies (Dugan and Peckenpaugh, 1985; Hansen, 1991). In all these studies, the following group of parameters were employed to define the initial physical boundaries of the calculation of the soil-water balance (Zelt and Dugan, 1993): 1) initial soil-water content as a proportion of the available water capacity (AWC) of the soil (long-term simulations are usually insensitive to this parameter); 2) infiltration-curve coefficients that define the equations that determine the amount of monthly precipitation that infiltrates the soil and the amount that becomes overland runoff; 3) consumptive water requirement (CWR) coefficients that determine the rate at which each vegetation type could potentially consume water, as a proportion of monthly PET; 4) monthly effective root-zone depth; 5) AWC of the soil. In addition, the following parameters also were specified for that study: 1) the irrigation threshold as a proportion of the AWC of the soil below which irrigation is required, and 2) irrigation season, as the months to which irrigation-water application will be restricted. For that study, the irrigation threshold was set at 50% of AWC, and irrigation season was specified according to the principal irrigated crops in the area of each simulation study (Zelt and Dugan, 1993).

The Dugan and Zelt (2000) study is one of the very few that provides some insight into the impact of irrigation on deep percolation and recharge on a regional scale {see also the Sophocleous and McAllister (1987, 1990) study summarized in Part I, section 8.1}. Figure II-3 shows a comparison of potential recharge for Kansas under nonirrigated conditions (A), irrigated conditions [(B), weighted towards high-water demand row crops—irrigated wheat excluded], and combined nonirrigated and irrigated conditions [(C), a weighted combination of (A) and (B)] extracted from the results of that study. As can be seen in that figure, deep percolation under irrigated conditions is higher than under nonirrigated conditions except where deep percolation under dryland conditions is large as a result of extensive areas of fallow conditions (Dugan and Zelt, 2000). Figure II-3, part (C), probably represents deep percolation more realistically than part (A) because the effects of irrigation are included, although the deep percolation patterns are similar. However, the fact that the closed 2-inches/yr contour near Garden City in southwestern Kansas is missing from the weighted combination part (C) of fig. II-3, probably represents a display error in the original reference.

The general increase in deep drainage under irrigated conditions does not result from excess irrigation but from an increased available-water capacity in the root zone at the end of the irrigation season (Dugan and Zelt, 2000) that maintains the soil profile wetter than otherwise possible, thus making infiltrating precipitation more effective in creating soil-water surpluses available for deep percolation. A comparison of deep percolation under irrigated and nonirrigated conditions for similar soils and crop types at two selected sites is shown in table II-1 (Dugan and Zelt, 2000). Although the absolute variations between deep percolation under nonirrigated and irrigated conditions are relatively small, the percentage differences can be considerable, particularly when deep percolation is small. The average percentage difference between deep percolation under irrigated and nonirrigated conditions for all soil- and crop-type combinations considered was 13% at Kearney, Nebraska, and 24% at Holyoke, northeastern Colorado (Dugan and Zelt, 2000). The generalization can be made that areas where cultivated crops are prevalent have larger potential recharge than areas in grassland with similar climatic and soil conditions. It should be noted, however, that the potential recharge increase under irrigated conditions does not necessarily coincide with a net gain by the underlying aquifer under ground-water-irrigated conditions (Dugan and Zelt, 2000); this gain, derived from an increase of water in the root zone, is at the expense of ground water in aquifer storage.

TABLE II-1—Comparison of deep percolation for irrigated conditions (DPI) with deep percolation for nonirrigated conditions (DPD) for selected soils and crop types at Kearney, Nebraska, and Holyoke, Colorado, 1951-1980 (values in inches/yr).

Table II-2 compares deep percolation for actual crops at four sites in North Dakota, Nebraska, Kansas, and Texas that are in areas predominantly in cultivation, with deep percolation for natural grassland only at those sites for Clay-Silty Clay Loam soil. All sites show that substantially greater potential recharge occurs under actual cultivated conditions than under grassland conditions alone (table II-2). Fallow conditions tend to increase the difference between the two potential recharge conditions (table II-2; Dugan and Zelt, 2000). The low potential recharge for both conditions at Sharon Springs, Kansas, and Muleshoe, Texas, as compared to those at Grand Forks, North Dakota, and Clay Center, Nebraska, is a result of the larger PET and, consequently, CWR at those sites (Dugan and Zelt, 2000). It is concluded that areas where cultivated crops are prevalent have greater potential recharge than areas in grassland with similar climatic and soil conditions.

TABLE II-2—Comparison of deep percolation for combined nonirrigated and irrigated conditions for actual vegetation with deep percolation for nonirrigated conditions for grassland for Clay-Silty Clay Loam soil at selected sites, 1951-1980.

[DPW, deep percolation for combined nonirrigated and irrigated conditions; DPD, deep percolation for nonirrigated conditions; PET, potential evapotranspiration; SG, small grain; HRC, high-water-demand row crops (corn, cotton)]

FIGURE II-3—Potential annual recharge for the High Plains aquifer in Kansas under nonirrigated conditions (A), irrigated conditions weighted towards high-water demand row crops (B), and combined nonirrigated and irrigated conditions (C)—a weighted combination of (A) and (B)— over the period 1951-1980. Contours in inches per year (adapted from Dugan and Zelt, 2000).

B. Large-area Ground-water Modeling or Analysis

Several regional ground-water modeling and other studies have been conducted in Kansas or in portions of the High Plains aquifer that include parts of Kansas, and these studies are analyzed here.

B1. USGS RASA Study of the High Plains Aquifer (Luckey et al., 1986)

Luckey et al. (1986) divided the High Plains (HP) aquifer into three segments, the southern HP, the central HP, and the northern HP, and employed a two-dimensional ground-water flow model in each such segment. The HP aquifer of northwest Kansas is included in the northern HP simulation model, whereas the rest of the Kansas HP aquifer (encompassed within GMD1, GMD3, GMD5, and GMD2 districts) is included in the central HP simulation model. The numerical model employed was the USGS Trescott et al. (1976) finite difference model using uniform 10-mile x 10-mile (100-mi²) grid cells. The model was run and calibrated in two dimensions under both predevelopment, steady-state conditions as well as transient conditions up to 1980. The results of that study will be presented for the central HP (which had 513 active grid nodes covering an area of 51,300 square miles) and the northern HP (which had 943 active grid nodes covering an area of 94,300 square miles).

a) Central High Plains Aquifer

The recharge distribution that resulted in the predevelopment (pre-1950) calibration is shown in fig. II-4. The sand dune areas in southwestern Kansas have the maximum recharge rate with a long-term average of 0.84 inch/yr, whereas the rest of southwest and west-central Kansas, consisting of clay-, silt-, and sandy-loam soils, had the minimum recharge value of 0.056 inch/yr. The Great Bend Prairie and southwestern Equus Beds regions, characterized by mostly sandy loam soils, were assigned a recharge rate of 0.28 inch/yr. Overall, the mean, long-term predevelopment recharge rate for the central HP was estimated to be 0.14 inch/yr (Luckey et al., 1986). Another 0.0056 inch/yr flowed into the central HP from the southern and northern HP. Recharge from streamflow losses was difficult to detect because of the coarse grid used in the model; that recharge was included as part of the recharge from precipitation assumed in the sand dune areas (Luckey et al, 1986). The steady-state central HP model calibration resulted in a mean difference between observed and simulated water levels at the 513 active model nodes of - 0.28 ft, with a standard deviation of 38.5 ft. At 98% of the nodes, the simulated water level was within 100 ft of the observed water level. Most of the discrepancies were in areas of sparse water-level data (Luckey et al., 1986).

FIGURE II-4—Estimated predevelopment, long-term average recharge rates (A) and generalized soil types (B) for the central High Plains aquifer (adapted form Luckey et al., 1986).

The long-term ground-water contribution to rivers is generally difficult to evaluate, but estimates are available for some rivers in Kansas. Fader and Stullken (1978) estimated base flow to the North and South Fork Ninnescah River as 38 cfs and 94 cfs, respectively; the model simulated 38 and 41 cfs, respectively. Gutentag et al. (1981) measured the baseflow of the Cimarron River as about 60 cfs at the Kansas-Oklahoma state line. The flow computed by the model for the same place was 80 cfs. Winter flow records for 1896-1908 for the Arkansas River indicate a ground-water contribution to the river between Garden City and Hutchinson, Kansas, of about 80 cfs. The model simulated 71 cfs for the same reach. The simulated outflow to rivers and model boundaries totaled 0.146 inch/yr (see table 3 of Luckey et al., 1986), which approximately balanced the equivalent total inflow into the model.

The development period for the central HP region was considered to be the 1950-1980 period. Additional stresses on the aquifer during the development period consisted of pumpage, return flow to the aquifer from irrigation, and additional recharge caused by human activities.

Pumpage was calculated using the method outlined by Heimes and Luckey (1982) by multiplying the product of irrigated acreage (derived from county-level census data) and composite crop demand for cells of dimension 10 minutes of latitude by 10 minutes of longitude (10-minute cells) by an irrigation efficiency factor (ranging from 45 to 70%, with efficiency improving with time) at 5-year intervals from 1949 to 1978 (Luckey et al., 1986). Those pumpage data were redistributed to 100-mi² model nodes, thus spreading the pumpage throughout a somewhat broader area than actually occurred (Luckey et al., 1986). Because return flow was assumed to reach the aquifer within the 5-yr pumping period, changing return flow during model calibration was exactly equivalent to changing net withdrawal (total pumpage minus return flow).

For the 1950-1980 simulation, the return flow (to the aquifer from irrigation when net withdrawal was assumed equal to irrigation requirement) ranged from 55% of total pumpage early in the development period to 30% later, and averaged 43%. This return flow appears large, but only the difference between total pumpage and return flow was important and both may be considerably overestimated whereas the difference remains correct (Luckey et al., 1986). The total pumpage for irrigation from the central High Plains during the 30-yr period 1950-1980 was estimated at 18,354,000 ac-ft (refer to table 4 of Luckey et al., 1986). Assuming that an average 43% of this total pumpage returned to the aquifer, the recharge from irrigation return flow would be 7,892,220 ac-ft over the 1950-1980 period or 263,074 ac-ft/yr or 1.0 inch/yr over the model area. This would be an additional recharge to the estimated predevelopment recharge.

The total volume of aquifer material dewatered was chosen as a calibration target for the development period simulation. Thus, the simulated water-level declines were 9% less than the observed ones on a volumetric basis. On an areal basis, the simulated water-level declines were 6% greater than the observed declines. The simulated change in ground-water storage was 54.9x10⁶ ac-ft, whereas the observed change in storage was 50.3x10⁶ ac-ft (Luckey et al., 1986). The differences between the observed and simulated water-level changes are believed to be primarily due to errors in the distribution of pumpage (Luckey et al., 1986).

b) Northern High Plains Aquifer

The northern High Plains was the last area of the High Plains to be developed for irrigation, with development generally starting after 1960. The calibrated predevelopment (pre-1960), long-term average recharge rate for the northwest Kansas High Plains, characterized by silt loam and clay loam soils, was 0.076 inch/yr.

The mean difference between the observed and simulated predevelopment water level at the 943 active model nodes of the northern High Plains was +0.30 ft, with a standard deviation of 55.2 ft. At 92% of the nodes, the simulated water-level altitude was within 100 ft of the observed altitude (Luckey et al., 1986). A comparison of estimated and simulated baseflows show that the simulated baseflow was less than the estimated baseflow for all river systems except the Republican River. (The total simulated baseflow was slightly more than 60% of the estimated baseflow.) According to Luckey et al. (1986), the smaller simulated baseflow probably is because some of the baseflow is contributed from local and intermediate aquifer systems, which were excluded in large-scale regional models such as the northern High Plains model.

The development period for the northern High Plains was simulated to be the 1960-1980 period. In that simulation, the return flow from irrigation ranged from 46% of total pumpage early in the period to 30% later in the period, and averaged 36%. The recharge from precipitation determined during the predevelopment-period calibration was assumed to have continued during the development period. In the development-period calibration, the difference between total pumpage and return flow for all areas was close to the estimated irrigation requirement. This also was similar to the central High Plains model simulation.

Recharge to the aquifer in the northern High Plains has been significantly increased by human activities such as leakage from canals and reservoirs, dryland farming, and other factors (Luckey et al., 1986). Cultivation practices associated with dryland farming can increase recharge from precipitation compared to the rate of recharge on rangeland (Ogilbee, 1962). For the northern High Plains, additional recharge due to cultivation was estimated to be 0.5 inch/yr throughout the dryland area.

The development-period composite recharge is shown in fig. II-5. This recharge is assumed to be the sum of five separate components (Luckey et al., 1986): 1) predevelopment period calibration recharge; 2) canal and reservoir leakage; 3) return flow from surface-water irrigation; 4) increased recharge due to dryland cultivation; and 5) additional recharge east of 100° longitude.

FIGURE II-5—Composite 1960-1980 recharge for the development-period model of the northern High Plains aquifer (adapted from Luckey et al., 1986).

Luckey et al. (1986) enumerated several additional factors to the above-listed ones that could account for the recharge increase simulated during the development period, such as the following: 1) decrease in runoff to streams due to cultivation; 2) decrease in baseflow to streams due to ground-water withdrawals; 3) change in downward leakage to underlying aquifers; 4) reduction of evapotranspiration due to lower water levels; 5) increase in downward leakage from a saturated zone above the water table; 6) greater specific yield of the aquifer than previously estimated; and 7) smaller total pumpage than estimated. None of these factors individually could account for the apparent increase in recharge, but collectively they might account for such increases.

The simulated net decrease in storage was 15x10⁶ ac-ft, whereas the observed decrease in storage was 6x10⁶ ac-ft. This simulation was accepted as a reasonable representation of the development-period operation of the aquifer in the northern High Plains (Luckey et al., 1986).

B2. USGS Study of the High Plains Aquifer in Oklahoma and Adjacent Areas, including the High Plains Aquifer of Southwestern Kansas (Luckey and Becker, 1999)

Luckey and Becker (1999) applied and calibrated a more spatially detailed finite difference numerical model for the central High Plains (exclusive of the High Plains area south of the Canadian River) than in the USGS RASA study (Luckey et al., 1986). They employed a single-layer, two-dimensional (2-D) MODFLOW model (McDonald and Harbaugh, 1988) using a uniform grid cell size of 6000-ft x 6000-ft (1.3-mi²). The flow model had an active cell area of 27,212 square miles. They calibrated the model for both predevelopment (pre-1946) or steady state, and development (1946-1997) or transient conditions.

Predevelopment-period simulation: To estimate recharge from precipitation, the model area was divided into zones of greater or lesser recharge (fig. II-6). The zones of greater recharge represented either sand dunes or areas of extremely sandy soils, whereas the zones of lesser recharge represented the remainder of the area (Luckey and Becker, 1999). The calibrated mean recharge value for the zones of higher recharge was 0.69 inch/yr, which represents 4% of the mean 1961-1990 precipitation in the area of 16.5 inches/yr. The mean recharge for the lesser recharge areas was 0.068 inch/yr, which represents about 0.37% of mean precipitation in the area. The predevelopment, overall mean simulated recharge from precipitation was 0.155 inch/yr.

Simulated predevelopment water levels were compared to observed water levels for the entire model area. The mean residual (i.e. differences between simulated and observed water levels at 21,073 active model nodes) is -2.9x10^-5 ft, with a standard deviation of 43.2 ft. At 97.7% of the nodes, the simulated water level was within 100 ft of the observed water level. Simulated predevelopment discharge to the Cimarron River near Liberal, Kansas, Forgan, Oklahoma, and Mocane, Oklahoma, was 5 to 10 cfs more than estimated discharge, whereas the estimated and simulated discharge to Crooked Creek were nearly the same (Luckey and Becker, 1999).

Development-period (1946-1997) simulation: The development period was simulated using five stress periods of 10 years each from 1946 to 1995, and one stress period of two years (1995-97). Pumpage was assumed constant within each stress period and was the average of estimated pumpage for all years within the period. Water use for irrigation was estimated based on Census of Agriculture data (U.S. Department of Commerce, 1949-1992) for the period 1946-1997 using the method outlined by Heimes and Luckey (1982), that consists of estimating irrigation use as the product of calculated irrigation demand (based on a modified Blaney-Criddle method), reported irrigation acreage, and assumed irrigation efficiency. The estimated pumpage was based on mean 1961-1990 monthly precipitation and temperature. According to Luckey and Becker (1999), the use of mean pumpage, temperature, and precipitation was assumed to have introduced “negligible error” by the end of the 52-year simulation. Time steps used in the simulation were 36.5 days long. Pumpage also was assumed to occur throughout the year, although most pumpage actually occurred during the summer. However, according to Luckey and Becker (1999), that assumption also should have introduced negligible error by the end of the 52-yr simulation.

FIGURE II-6—Simulated greater-recharge areas for the central High Plains aquifer (adapted from Luckey and Becker, 1999).

Simulated recharge due to irrigation averaged 24% of pumpage for the 1940’s and 1950’s, averaged 14% for the 1960’s, averaged 7% for the 1970’s, averaged 4% for the 1980’s, and averaged 2% for the 1990’s (Luckey and Becker, 1999). Recharge due to irrigation was subtracted from total pumpage before the simulation was made, so only net pumpage was input into the model. This operation means that recharge due to irrigation was assumed to occur within the same stress period as the pumpage. According to Luckey and Becker (1999), given the long period simulated and the low recharge near the end of the simulation, that assumption probably did not cause substantial error in the model.

Recharge due to dryland cultivation was estimated to be 3.9% of mean 1961-1990 precipitation (which is 16.5 inches/yr) or 0.64 inch/yr over the area in dryland cultivation (which also included irrigated wheat). Over the entire study area, dryland cultivation recharge amounted to 0.24 inch/yr.

Thus total simulated recharge in the development-period model consists of 0.15 inch/yr background recharge from precipitation, 0.24 inch/yr due to dryland cultivation, and 0.022 inch/yr due to irrigation, totaling 0.41 inch/yr recharge from all sources.

Comparisons of simulated and observed predevelopment to 1998 water-level changes were done only for 162 observation wells in Oklahoma, resulting in a mean residual of -2.9x10-3 ft, with root mean square difference of 17.9 ft. 98.1% of simulated values were within 50 ft of observed values.

The simulated discharge to the Cimarron River at the end of 1997 was 51.2 cfs, a decrease of 10.9 cfs from simulated predevelopment discharge. According to Luckey and Becker (1999), both the discharge and simulated change in discharge appear reasonable.

Simulated water-level changes for 1998-2020 using mean 1996-97 pumpage indicate more than 100-ft declines in Finney and Haskell counties (fig. II-7); an area of simulated decline of 50-100 ft covers most of Grant and Haskell counties, and substantial parts of Stanton, Stevens, Seward, Meade, Gray, Finney, and Kearny counties (Luckey and Becker, 1999). Very little additional decline is simulated to occur in southeastern Gray County, but this is an area of little saturated thickness. A summary of simulated predevelopment, 1998, and 2020 water budgets for the entire model area is shown in fig. II-8. Ground-water storage over the central High Plains study area is simulated to decrease by 49 x 10⁶ ac-ft (or 33.76 inches) from the end of 1997 to the beginning of 2020 (Luckey and Becker, 1999). The increase in stream discharge at the beginning of 2020 compared to the one at the end of 1997, evident in fig. II-8, occurs mostly in the eastern portion of the simulated area, mainly along Beaver River below Optima Lake and along the Canadian River (fig. II-6). The cause seems to be the extra recharge that was simulated on dryland wheat (R. R. Luckey, written communication, March 2003).

FIGURE II-7—Simulated water-level changes in the central High Plains aquifer for 1998-2020 using mean 1996-97 pumpage. Bold dots with numbers 1, 2, and 3 represent experimental recharge sites described in text subsection 4: Field-based experimental recharge studies at the local level (adapted from Luckey and Becker, 1999). Areas where High Plains aquifer is absent are shown in black.

FIGURE II-8—Simulated predelopment, end of 1997, and end of 2019 water budgets for the central High Plains studied by Luckey and Becker (adapted from Luckey and Becker, 1999).

B3. USGS Study of the High Plains Aquifer in Western Kansas (Stullken et al., 1985)

Stullken et al. (1985) employed the two-dimensional finite difference USGS model developed by Trescott et al. (1976) to simulate the steady-state, predevelopment ground-water budgets of northwestern Kansas and southwestern Kansas using uniform 2.84-mi x 2.84-mi (8-mi²) grid cells.

Steady-state (pre-1950) Simulation of the High Plains Aquifer in Northwestern Kansas:

Following model calibration, the predevelopment recharge varied from 0 to 0.79 inch/yr and averaged 0.20 inch/yr, which is consistent with Jenkins and Pabst’s (1975) recharge estimate (see section B10). The mean difference between observed and simulated hydraulic head values (mean residual) was -2.0 ft (buildup), with a standard deviation of 8.12 ft. Simulated hydraulic-head difference in 88% of the nodes was within 10 ft of observed values. According to Stullken et al. (1985), estimated streamflow gains also were simulated “closely” by the model. The predevelopment water budget (table II-3) indicates that approximately 80% of total recharge came from precipitation, and that approximately 75% of discharge was by leakage to streams.

TABLE II-3—Water budget from steady-state simulation of High Plains aquifer, northwest Kansas (values in inches/year).

Steady-state (pre-1950) Simulation of the High Plains Aquifer in Southwestern Kansas

Recharge from precipitation indicated by the calibrated model ranged from 0 to 2.0 inches/yr and averaged 0.24 inch/yr. The greatest recharge occurred in the area south of the Cimarron River and in the area between the Cimarron River and Crooked Creek where a large percentage of the surface is dune sand. The calibrated recharge from precipitation for the Arkansas River valley and dune-sand areas to the south was 0.25 inch/yr (Stullken at al., 1985). The simulated recharge to the aquifer owing to leakage from streams and creeks in the western part of southwestern Kansas (Big Bear, Little Bear, and Sand Arroyo creeks, the North Fork of Cimarron River, and the western reach of the Cimarron River) totaled 0.036 inch/yr, whereas recharge due to stream leakage from the Arkansas River, Crooked Creek, and the eastern reach of the Cimarron River was 0.045 inch/yr, although overall Arkansas River, Cimarron River, and Crooked Creek were gaining streams in the area (Stullken et al., 1985). Estimates of recharge from streams in southwest Kansas using channel geometry methods (Hedman and Osterkamp, 1982) ranged from 0.08 to 0.14 inch/yr in the reaches of the losing streams.

Movement of water from the High Plains aquifer to the underlying Lower Cretaceous (Dakota) sandstone aquifer was simulated in parts of Grant, Haskell, Stevens, Seward, and Meade counties, and totaled 0.032 inch/yr (table II-4).

The mean residual for all 1,028 active nodes was -1.08 ft (a buildup), with a standard deviation of 10.5 ft.

The simulated steady-state water budget is given in table II-4 and shows that about 60% of the total recharge came from precipitation, approximately 19% came from boundary inflow, and 21% came from leakage of streams. Also, 38% of the total discharge from the aquifer went to boundary outflow, 8% went to the underlying Lower Cretaceous sandstone aquifer, and 54% went to streams.

TABLE II-4—Water budget from steady-state simulation of the High Plains aquifer, southwest Kansas (values in inches/yr).

B4. USGS Study of the Dakota and High Plains Aquifers of Southwestern Kansas (Watts, 1989)

Watts (1989) developed a layered model of the High Plains, Dakota, and Cheyenne Sandstone aquifers separated by the Niobrara-Graneros and Kiowa confining units for southwestern Kansas centered around a study area consisting of Kearny, Finney, Gray, Hodgeman, and Ford counties. The 3-D MODFLOW model was employed for this purpose using a 6-mi x 6-mi square grid for the above-mentioned study area. The model was calibrated to simulate the measured water-level changes in the High Plains and the Dakota aquifers for the period 1975-1982. Each year of the calibration period was divided into three stress periods representing 1) the relatively static water-level period of January to April, 2) the high ground-water-withdrawal period of May through September, and 3) the recovery period of October through December, respectively. Annual irrigation pumpage from the High Plains aquifer was estimated for 1974 and 1978 irrigated acreages multiplied by crop-water demand (Heimes and Luckey, 1982) and divided by an irrigation efficiency of 0.8. Straight-line projection and extrapolation were used to estimate annual pumpage between 1974 and 1978 and 1979-1982, respectively. Estimates of the withdrawal from the Dakota aquifer were based on water-use reports from the Division of Water Resources, Kansas Department of Agriculture.

The model calibration resulted in an average difference between January 1982 measured and simulated water levels in the High Plains aquifer of 1.04 ft, and a standard deviation of about 17 ft. The differences between measured and simulated potentiometric surfaces for the same time period for the Dakota aquifer resulted in an average difference of -0.09 ft and a standard deviation of 55 ft. The simulated 1982 water budget for the High Plains and Dakota aquifers is shown in table II-5.

As can be seen from table II-5, outflow from and decrease in storage in the High Plains aquifer dominate the water budget of the study area. Average recharge to the High Plains aquifer during 1982 was estimated at 0.6 inch/yr. Although downward leakage was only a minor component of the High Plains aquifer water budget (0.09 inch/yr), it was a major source of inflow to the Dakota aquifer. Leakage across the base of the High Plains aquifer in the model area is simulated as predominantly downward in areas where the Niobrara-Graneros confining unit is present and upward where the Dakota aquifer subcrops below the High Plains aquifer, such as occurs in parts of southern Finney and Kearny counties in the model study area as well as in Stanton, Grant, Haskell, and other counties in the general model area, as shown in fig. II-9.

TABLE II-5—Simulated 1982 water budgets for the High Plains and Dakota aquifers (values in inches/yr).

FIGURE II-9—Finite-difference grid of model and study areas in southwestern Kansas with simulated leakage between High Plains and underlying aquifers, 1982 (adapted from Watts, 1989).

B5. USGS Study of the High Plains Aquifer in Oklahoma, including Some Kansas Counties (Havens and Christenson, 1984)

Havens and Christenson (1984) simulated the High Plains aquifer in Oklahoma, including the southern tier of Kansas counties (Morton, Stevens, Seward, and Meade) and a portion of Baca County, Colorado, as the northern boundary of the model using the USGS two-dimensional finite difference model (Trescott et al., 1976). The Canadian River was used as the approximate southern boundary of the study area. The model area was discretized using a regular 5-mi x 5-mi grid network that included 888 active nodes covering an area of 22,200 square miles. The model was calibrated for both predevelopment (1941), steady-state, and transient conditions from 1941 to 1980. Eight five-year pumping periods with one-year time steps were employed in the transient model. Pumpage used in that study was calculated as a percentage of the total crop demand as determined by Heimes and Luckey (1982) and crop- distribution data published by the U.S. Department of Commerce as a Census of Agriculture. The predevelopment (1941) and the transient, 1980 simulation water budgets are presented in table II-6.

TABLE II-6—Steady-state (1941) and end of simulation (1980) water budgets for the High Plains regional aquifer in northwestern Oklahoma (values in inches/year).

Recharge from precipitation was estimated as 0.34 inch/yr. However, the eastern half of the model area (that includes Meade, Seward, and the eastern portion of Stevens counties in Kansas), that also had higher precipitation, had a higher value of recharge (0.45 inch/yr), whereas the western half of the model area (including Morton and most of Stevens counties in Kansas) had half of that recharge value (0.23 inch/yr). The steady-state simulation resulted in a mean difference between computed and measured heads of -0.044 ft in the 356 nodes that were in the Oklahoma portion of the model, and a mean absolute value of the differences of 50.1 ft. For the transient simulation, the mean difference between computed and measured heads was -0.011 ft, and the mean of the absolute values of the differences was 48.0 ft. Although no streamflow data were available for predevelopment-conditions calibration, the end-of-simulation (1980) leakage to streams (118.2 cfs) was very close to the total estimate of 117.8 cfs of low-flow discharge of streams draining the High Plains aquifer of northwestern Oklahoma for the period 1970-79.

B6. KGS Study of Geohydrology of Southwestern Kansas (Gutentag et al., 1981)

Gutentag et al. (1981) studied the ground-water resources of southwestern Kansas (11 counties) and compiled a (partial) water budget for the area as of 1975 (table II-7).

TABLE II-7—1975 water budget (partial) for southwestern Kansas (values in inches/yr).

Recharge was considered to be 10% of precipitation during the growing season (April-October) on irrigated land, and 1% of precipitation on nonirrigated land. The amount of annual recharge from runoff into surface depressions and streams is assumed to be included with the total estimate of recharge from precipitation. Records for 1975 showed that about 1,400,000 acres were irrigated; the remaining 2,824,000 acres were not irrigated. Annual recharge to the aquifer, based on about 15 inches of precipitation during the growing season, is computed to be 0.6 in over the total area (1.50 inches over the irrigated area and 0.15 inch over the nonirrigated area). Streamflow gains (ground-water contribution to streamflow or baseflow) represent the total average baseflows of the Cimarron River at Mocane, Oklahoma, Crooked Creek near Nye, Kansas, and Arkansas River leaving the study area. It was estimated that pumpage for irrigation (of 1,400,000 acres) during 1975 was between 2.1 and 2.8 million acre-feet (6.0-8.0 inches over the total area). Figures derived from Meyer et al. (1970) for irrigated land in Finney County (see section C5) showed about 20% of the water applied to irrigated land was returned to the aquifer by percolation below the root zone. Thus, calculations suggest that about 420,000 to 560,000 ac-ft/yr (1.2 to 1.6 inches/yr) returns to the ground-water reservoir as recharge from irrigation return flow in southwestern Kansas (Gutentag et al., 1981). The number of irrigation wells in January 1975 was estimated at about 7,000 in southwestern Kansas. Ground-water evapotranspiration was considered negligible in southwestern Kansas because the water table in most of the area is too far below the land surface to be affected by evaporation and transpiration (Gutentag et al., 1981). In order to balance this total outflow (well pumpage, baseflow to streams, and boundary outflows—table II-7), ground-water storage was being depleted by 4.4 to 5.8 inches/yr (as of 1975), and this depletion is reflected in the observed long-term declines in ground-water levels and saturated thickness.

B7. Kansas Governor’s Task Force on Water Resources Interim Report (1977)

The Governor’s Task Force on Water Resources in its December 1977 Interim Report included a section, entitled “A Case Study of the Ogallala” to provide some basic background information for understanding the physical, economic, legal, and management problems involved on ground-water depletion in western Kansas. The Task Force presented the following information on recharge and withdrawals from the Ogallala aquifer in western Kansas (table II-8), and summed up the situation in the following way:

The estimated amount of groundwater withdrawn in Region No. 1 [GMD1] of western Kansas in 1975 was almost 15 times greater than the estimated recharge; in Region No. 3 [GMD3], withdrawals approximated 18 times the recharge; and in Region No. 4 [GMD4] withdrawals were about seven times the recharge. For western Kansas as a whole, withdrawals are estimated to average 14 times the recharge rate. With this great a disparity between withdrawals and recharge and with current indications pointing to even greater withdrawals for irrigation in the future, it seems apparent that the water table has nowhere to go but down if current conditions and projections prevail.

Theoretical solutions to increasing recharge were considered by the Task Force, which concluded that “. . . there are no simple physical solutions and we must look beyond the physical to economic, legal and management factors for enlightenment on how to resolve groundwater depletion problems.”

TABLE II-8—Hydrologic budget component estimates for the Ogallala aquifer in western Kansas as of 1975.

B8. KGS Study of the Ogallala Aquifer in Kansas (O’Connor and McClain, 1982)

O’Connor and McClain (1982) studied the Ogallala aquifer in Kansas (approximately 28,560 square miles in a 32-county area of western Kansas divided into northwest, west-central, and southwest subregions). Based on previous recharge estimates, they approximated the recharge from irrigation return flow and from precipitation on the area overlying the Ogallala and peripheral aquifers as shown in table II-9. Overall recharge for the Ogallala aquifer in western Kansas was estimated as 0.57 inch/yr. Nonirrigated or dryland average recharge overlying the Ogallala aquifer was estimated as 0.3 inch/yr. This includes recharge from ephemeral streams, depressions, and sand-dune tracts that have higher than average recharge. Recharge on irrigated land above the Ogallala (including that from precipitation and irrigation return flow) was estimated to be 10% of an average irrigation application of 18 inches annually, or 1.8 inches/yr. The recharge figures in table II-9 do not include subsurface inflow or recharge from the Arkansas or Cimarron rivers. Because of declining water levels along those stream valleys, much of the streamflow was lost by influent seepage to the ground-water reservoir and not by flow out the east and south sides of the study area.

TABLE II-9—Estimated recharge from precipitation and irrigation return flow to the Ogallala aquifer in western Kansas, as of 1977.

B9. Great Bend Prairie Aquifer Regional Recharge Estimates (Fader and Stullken, 1978; Cobb et al., 1983)

Fader and Stullken (1978) evaluated the ground-water resources of the Great Bend Prairie in south-central Kansas. They estimated ground-water recharge for the combined drainage area above the stream-gaging stations of Raymond (Rattlesnake Creek), Arlington (North Fork Ninnescah River), and Murdock (South Fork Ninnescah River), where average annual precipitation was estimated to be 25 inches/yr. The ground-water drainage area above the three stations was estimated to be 2,280 square miles based on a December 1973 potentiometric surface map of the region. Fader and Stullken’s recharge by precipitation estimate to the above ground-water drainage area was 240,000 ac-ft/yr or 2 inches/yr. Based on streamflow- duration curve analysis, the combined ground-water contribution to streamflow at these stations was estimated as about 110,000 acre-ft/yr or 0.9 inch/yr (Fader and Stullken, 1978).

Recharge to the ground-water reservoir is principally by direct infiltration of precipitation and irrigation on the land surface throughout the area plus underflow laterally from the west, and leakage upward from the bedrock. Recharge to the area by underflow occurs only across the western Kiowa County line and was estimated to be 500 to 1,000 acre-ft/yr. The inflow from the bedrock was estimated to be 5,000 to 10,000 acre-ft/yr (based on the assumptions that the Cedar Hills Sandstone is the major contributor, the hydraulic gradient in that formation is virtually equal to and in the same direction as in the overlying unconsolidated deposits, and the hydraulic conductivity of the Cedar Hills Sandstone is about 25 ft/day).

Fader and Stullken (1978) estimated that 900,000 acre-ft of water was withdrawn by wells through the Great Bend Prairie during 1952-1971, of which 680,000 acre-ft was for irrigation and 220,000 acre-ft was for municipal and industrial use. Sixty-two percent of the wells recorded in May 1974 were within the ground-water drainage area above the stream-gaging stations near Raymond, Arlington, and Murdock.

Cobb et al. (1983), previously of the Kansas Geological Survey, calibrated (by trial and error) the Trescott et al. (1976) USGS two-dimensional finite-difference flow model using a grid spacing of 15,000-ft x 15,000-ft throughout the Great Bend Prairie region. The resulting average recharge was 0.75 inch/yr.

As previously mentioned, Luckey et al. (1986), employed the Trescott et al. (1976) USGS two-dimensional finite-difference flow model using a grid spacing of 10-mi x 10-mi and calibrated it for the central High Plains, which incorporates the Great Bend Prairie region (please also refer to sections C10-C12 and D2-D3 below). The estimated predevelopment long-term average recharge rate for the Great Bend Prairie was 0.28 inch/yr.

B10. Other Regional Studies Involving Kansas High Plains Recharge Estimates (Jenkins and Pabst, 1975; Landon, 2001, 2002; Hecox, 2003)

Jenkins and Pabst (1975) in a study of northwest Kansas (nine northwest counties covering an area of 8,050 mi2) estimated annual recharge from precipitation to be 0.25 inch.

Landon (2001, 2002), in a preliminary MODFLOW modeling study of the High Plains aquifer in the Republican River basin in Nebraska, Kansas, and Colorado, estimated average predevelopment recharge rate of 0.26 inch/yr across the entire 30,224-mi² active grid-node Republican River basin model area.

Hecox (2003), in a recent MODFLOW modeling study of a 19,629-mi² region of the High Plains that includes all of the High Plains aquifer of northwestern Kansas, estimated steady-state recharge to be approximately 0.68 inch/yr.

C. Basin-scale to County-scale Ground-water Studies

The basin- to county-scale ground-water studies highlighted here consist of studies that addressed ground-water recharge in some quantitative fashion either through numerical modeling or hydrologic-budget analysis. Additional studies in this category are summarized in Part IV.

C1. Wichita and Scott Counties

Dunlap et al. (1980) applied the USGS two-dimensional finite difference ground-water flow model (Trescott et al., 1976) to a 480-square-mile area centered northeast of the town of Leoti, in Wichita County. The model was calibrated for both predevelopment (1950-51) or steady-state, and transient (1951-1977) conditions. The calibrated, uniform, steady-state recharge was 0.28 inch/yr. A soil-zone model (Lappala, 1978) was used to estimate crop-water demand and irrigation needed to maintain the available soil moisture at 50%. The annual pumpage employed in the transient model was calculated based on annual changes in crop acreage and crop-irrigation demand (derived from the soil-zone model). Irrigation return flow was estimated to be minimal. The numerical model was more sensitive to changes in pumpage than to hydrogeologic parameters and recharge, and thus the latter remained unchanged during the transient simulation.

C2. Lane and Scott Counties, West-central Kansas

Gutentag and Stullken (1976) studied the ground-water resources of Lane and Scott counties and presented a detailed water budget for Scott County for 1971 (a dry year) and 1972 (a wet year) as shown in table II-10.

Recharge was considered to be 10% of the precipitation on irrigated land during the growing season and 1% of the precipitation on nonirrigated land during the growing season. Infiltration of streamflow (streamflow losses) is considered to be included with the recharge estimate from precipitation (Gutentag and Stullken, 1976). Thus recharge from precipitation during 1971 (a dry year) was estimated to have been 0.36 inch, and during 1972 (a wet year) 0.62 inch for Scott County. The corresponding values of estimated recharge for Lane County were 0.08 inch (1971) and 0.16 inch (1972), respectively. An additional source of recharge is return flow from irrigation. Gutentag and Stullken (1976), using figures experimentally derived by Meyer et al. (1953) for irrigated land in Finney County, assumed that 20% of withdrawal by wells subsequently returns to the ground-water reservoir (see section C5). The total amount of water pumped for irrigation in Scott County was 150,000 acre-feet (3.88 inches) in 1971 and 95,000 acre-feet (2.46 inches) in 1972. The part of the total water pumped that returns to the reservoir in Scott County was thus estimated to be 30,000 acre-ft (0.78 inch) in 1971 and 19,000 ac-ft (0.49 inch) in 1972. If the seepage of irrigation water is added to the recharge inflow and boundary inflows given in the budget for Scott County, the total recharge would be 1.37 inches in 1971 and 1.35 inches in 1972, indicating that total recharge remains relatively constant from year to year when there is little change in irrigated acreage (Gutentag and Stullken, 1976). Evapotranspiration from ground water was considered negligible because the water table was well below the root zone nearly everywhere in the area (Gutentag and Stullken, 1976).

TABLE II-10—1971 and 1972 water budgets for Scott County (values in inches/yr).

C3. Arkansas River Valley in Hamilton and Kearny Counties

Barker et al. (1983) applied a USGS two-dimensional finite element ground-water flow model (written by J. V. Tracy and documented in Dunlap et al., 1984) to nearly 110,000 acres of the Arkansas River valley between the Colorado-Kansas state line and the Bear Creek Fault Zone in southwestern Kansas. The model was calibrated for both steady-state (pre-1970, considered as averaged 1951-1969 conditions) and transient conditions (1970-79). Monthly pumpage data were estimated from energy-consumption records. The simulated hydrologic fluxes for two periods, 1970-74 and 1975-79 are presented in the water budget table II-11 below.

Recharge to the aquifer from incident water (precipitation plus irrigation) amounted to 22 to 26% of that total water, although approximately 15% of that recharged water was lost though ground-water evapotranspiration.

TABLE II-11—Simulated water budget for Arkansas River alluvium between Colorado-Kansas state line and Bear Creek Fault Zone, Kearny and Hamilton counties, Kansas.

C4. Unconsolidated Aquifer System of Kearny and Finney Counties

Dunlap et al. (1985) applied the three-dimensional USGS finite difference model (Trescott, 1975; Trescott and Larson, 1976) to simulate the High Plains, Arkansas River valley, and sandhills aquifer system in Kearny and Finney counties. Three vertical layers were used in the model: the top layer represented the valley and upper aquifers; the middle layer, the confining zone; and the bottom layer, the lower aquifer, which is the principal water-source aquifer in the area. The model was calibrated for both predevelopment (1940) or steady-state conditions, and transient (1974-1980) conditions. The pumpage from 1974 to 1980 was estimated over 4-month periods from 1) a soil-zone model (Lappala, 1978), calculating crop-water demand for the major crops in the area, and 2) irrigated-acreage data.

Predevelopment recharge for the Arkansas River valley and sandhills area from precipitation and irrigation was estimated to be 11% of the normal rainfall at Garden City, or 2.08 inches/yr (the same value also was used for transient conditions), and was estimated at less than 0.5 inch/yr for the High Plains aquifer (that was considered negligible in the model application). In the Dunlap et al. (1985) study, irrigation return flow was found insignificant in comparison to other recharge (such as precipitation recharge and river and canal seepage) and pumpage. Seepage losses from the Arkansas River are a major source of recharge to the aquifer system. The simulated river and canal seepage and boundary inflow averaged 1.13 inches/yr (or 36,200 ac-ft/yr) over the 1974-1980 simulation period.

The simulated water budget for 1980 is shown in table II-12, which shows that the lower aquifer is recharged by 1) leakage from the confining zone; 2) lateral, subsurface inflow; and 3) canal seepage. The major source of water for pumpage (approximately 58%) is from downward leakage of water from the overlying upper and valley aquifers. Approximately 42% of the ground water pumped from the lower aquifer came from storage in the lower aquifer (table II-12).

TABLE II-12—Simulated water budget for unconsolidated aquifer system of Kearny and Finney counties, 1980 (values in inches/yr).

C5. Upper Arkansas River Corridor in Southwestern Kansas

Whittemore et al. (2001) applied and calibrated a two-layer MODFLOW model for the entire corridor of the Upper Arkansas River from the Colorado-Kansas state line to the Crooked Creek-Fowler Fault Zone in eastern Ford County. A rectangular area of 29-mi width and 126-mi length was employed to incorporate the extent of the alluvial trough in Hamilton and western Kearney counties, all of the ditch-irrigation service areas, and the High Plains aquifer to the north and south of the river valley. A uniform grid cell size of 0.5-mi x 0.5-mi (0.25 mi2) was employed, resulting in an active-cell model area of approximately 2,328 mi2. Two layers were employed in the model, one for the alluvial aquifer along the Arkansas River valley, and the other for the High Plains aquifer and the older alluvial aquifer underlying the sand dunes south of the river floodplain in Hamilton and western Kearny counties. A zone of low hydraulic-conductivity clays and silty clays underlies much of the Arkansas River coarse sand and gravel alluvium and slows the downward movement of shallow ground water into the underlying High Plains aquifer. The model was calibrated for predevelopment (1938-1942), steady-state conditions, and for 1991-2000 conditions using 10-year averaged water-level measurements. Because of various artifices introduced in the 1990’s model simulation conditions, only the predevelopment simulated water budget will be highlighted here. The model calibration involved trial-and-error, cell-by-cell adjustments of hydraulic conductivity and recharge values using the GIS ArcView environment to minimize computed head error. This procedure resulted in a simulated mean head error of less than 1 ft with a standard deviation of less than 3.28 ft (Whittemore et al., 2001). According to the authors, the computer model simulated the net gain in the Arkansas River from Garden City to Dodge City as well as that from Syracuse to Garden City (fig. II-7) well.

The model-estimated total recharge from precipitation, irrigation canal seepage losses, and irrigation return flows was 1.04 inches/yr over the total active model cell area of 2,328.25 mi2 (1,490,080 acres). The simulated steady-state water budget is given in table II-13 (D. O. Whittemore and S. P. Perkins, personal communication, January 2003), and shows that approximately 81% of the total recharge came from precipitation and irrigation return flows, approximately 15% came from river leakage, and less than 4% from boundary inflow. The major discharge from the alluvium and High Plains aquifer system went to Arkansas River baseflow (45%), with approximately 28.7% being artificially discharged by mostly irrigation wells, and approximately 26% being discharged as boundary outflow.

Whittemore et al. (2001) estimated (by taking into account gaging stations’ streamflow differences, streamflow diversions, estimated evaporation, and phreatophyte use) that the average net recharge from the Arkansas River to the alluvium followed by leakage into the underlying High Plains aquifer during the decade 1989-1998 was about 7.4 inches/yr over the entire area of the Arkansas River alluvial valley from Hartland to Dodge City (which is equivalent to 118,000 acres; most of the flow losses between the Kansas-Colorado state line and Garden City occur from the western edge of the High Plains aquifer underlying the river valley near the former town of Hartland). Whittemore et al. (2001) estimated that recharge rates to the river alluvium would be over 1 ft/yr during 1995-2000, when the river flows and recharge were greater.

TABLE II-13—Water budget from steady-state (1940’s) model simulation of Upper Arkansas River corridor in south-western Kansas (values in inches/yr).

C6. Finney County

Meyer et al. (1970) conducted a detailed study of recharge from streams, precipitation, return flow from irrigation, and inflow through the aquifers in Finney County, southwest Kansas. They developed a detailed water budget for all of Finney County minus the panhandle area (a total of 24 townships or 552,960 acres) using the following equation:

Total Recharge = Change in Storage + Total Discharge . . . . . . . (II-3)

where all elements of the right-hand side of the equation were observed or estimated, and Total Discharge = well pumpage + (lateral ground-water outflow - inflow) + streamflow seepage (positive for baseflow, negative for streamflow losses). Data from the 24-year period (1940-1964) were used in the calculations. Observed values of the components on the right-hand side of the water-balance equation (II-3) are summarized in table II-14.

TABLE II-14—Water-balance components for the Finney County study area (552,960 acres). Data for 1940-1964. Values in inches/yr.

Thus, the average recharge indicated by the water-balance equation (II-3) for the 24-year period (1940-1964) was 124,000 ac-ft/yr or 2.7 inches/yr over the area. If the ground-water system were considered to be under near-equilibrium conditions before pumping began, the components in the water-balance equation would be as shown under the column “Predevelopment conditions” of table II-14. However historic streamflow records for the period 1922-1930 indicate a seepage loss from the Arkansas River between Syracuse and Garden City of 21,000 ac-ft/yr or 0.46 inches/yr over the study area. Because the Arkansas River gains and loses flow in about equal proportions as it passes through Hamilton and Kearny counties, most of the 21,000 ac-ft/yr loss during 1922-1930, like the 2000 ac-ft/yr loss during 1940-1964, is believed to be directly contributed to the ground-water reservoir within Finney County. (This loss would be reduced by about a 300-ac-ft/yr gain in flow of the river east of Garden City.) The long-term average predevelop-ment recharge to the aquifer was apparently less than 0.5 inch/yr (based on 1922-1930 conditions) or even less than 0.05 inch/yr (based on the 1940-1964 conditions).

Thus, according to Meyer et al. (1970), the development-conditions recharge rate of 2.7 inches/yr probably reflects an additional recharge resulting from recycled ground water from irrigation and an accompanying increase in effective recharge from precipitation on the irrigated land. The change in land use from native grassland to cropland is a factor contributing to increased recharge. As the water table is lowered by pumping, the evapotranspiration losses also are reduced and the effective recharge to the aquifer is increased. Thus, Meyer et al. (1970) list four factors as probably having a role in increasing recharge during the 1940-1964 period: 1) increased precipitation (the period 1940-1951 was marked by precipitation rates significantly above average), 2) irrigation return flow, 3) change in land-use practices, and 4) evapotranspiration reduction due to lowered water tables.

Impacts of Irrigation on Recharge

With regard to irrigation-related return flow to the aquifer in Finney County, Meyer et al. (1953) conducted detailed ditch-loss studies and irrigation efficiency experiments at the Garden City Experiment Station. The ditch-loss studies showed losses of 10% occurring in a quarter-mile length of a farm-supply ditch. Experiments on irrigation efficiencies in the area showed that deep percolation losses (those penetrating more than 6 ft) in a well-drained system are frequently 20% or more. With these facts in mind, Meyer et al. (1970) concluded that “. . . it seems reasonable to assume that 15 percent of the water applied to irrigated fields could return to the groundwater reservoir, and that another 10 percent would return from ditch leakage.”

According to Meyer et al. (1970), of the total 1940-March 1964 pumpage of 3,530,000 ac-ft, 25% or 882,000 ac-ft returned to the ground-water reservoir, and an additional 219,000 ac-ft returned from surface water. This would be a total contribution of about 1,100,000 ac-ft for the 24-year period, or an average annual irrigation return contribution of approximately 45,700 ac-ft, which is equivalent to 1 inch/yr over the area.

C7. Ford County

Spinazola and Dealy (1983) evaluated the hydrologic conditions in the Ogallala aquifer in Ford County during 1980 and 1981. They produced an approximate water budget for 1980 conditions for that part of Ford County principally underlain by the Ogallala aquifer, an approximately 700-mi² area that includes the four northwestern townships and all townships south of the Arkansas River (table II-15).

TABLE II-15—1980 water budget for Ford County underlain by the Ogallala aquifer (values in inches/yr).

Recharge to the aquifer, 0.6 in/yr, was considered as 3% of annual precipitation. Ground-water evapotranspiration along the 48-mi-long reach of the Arkansas River valley was considered to be the same as that calculated along the Arkansas River valley in Hamilton and Kearny counties between 1970-79 by Barker et al. (1983).

C8. Pawnee Valley (Sophocleous, 1980, 1981)

In a hydrogeologic study of Pawnee Valley, western Kansas, Sophocleous (1980, 1981) estimated regional ground-water recharge in that valley using two different methods:

i) Interpretation of streamflow records at the discharge end of the flow system and of pumpage data over the 1925-1945 period, where near-equilibrium conditions could be assumed. Thus the long-term average recharge to the alluvial aquifer was assumed to equal the long-term ground-water outflow during the 1925-1945 period. Such analysis resulted in a recharge rate of 0.6 inch/yr over the alluvial aquifer area.
ii) Analysis of the soil-moisture budget based on hydrometeorological and soil data of the composite Pawnee watershed. This analysis, based on 20 years of hydrometeorological data (1959-1978) using the Thornthwaite method for calculating potential evapotranspiration in conjunction with the modulated soil moisture technique of Holmes and Robertson (1959) to arrive at actual evapotranspiration and moisture surplus and deficit, resulted in a value for regional ground-water recharge of 0.4 inch/yr.

Thus, the average estimated regional ground-water recharge for the Pawnee Valley was about 0.5 inch/yr, which represents 2.2% of the 1959-1978 average annual precipitation of 22.7 inches. The Sophocleous (1980) study indicated that by 1978-79, the Pawnee Valley had been depleted by 37% compared to 1945-47. It also is interesting to note that during 1978-79, the ground-water appropriations in the Pawnee Valley alluvial aquifer (that reached at least 84,000 ac-ft) amounted to about 11 times the amount of estimated natural ground-water replenishment for the Pawnee Valley.

C9. Wet Walnut Creek Basin, West-central Kansas

Koelliker et al. (1999) Integrated Basin Modeling Study

Koelliker et al. (1999; see also Ramireddygari et al., 2000, and Sophocleous et al., 1998) developed an integrated watershed and ground-water model by combining the watershed model POTYLDR (Koelliker et al., 1982) with the two-dimensional USGS MODFLOW model, and applied and calibrated it to the Wet Walnut Creek basin in west-central Kansas. The basin was divided into 78 subbasins and the Wet Walnut Creek valley aquifer was discretized into 0.5-mile x 0.5-mile-square cells. Using 1960 initial conditions (assumed to be near-equilibrium conditions, the model was run in monthly steps for the period 1960-1996. The data for the period 1960-1990 were used to calibrate the model, whereas the data for the 1991-96 period were reserved for verifying the calibrated model. The mean residual during the 1960-1990 calibration period was 1.51 ft with a standard deviation of 6.59 ft. The mean residual for the 1991-96 period was 0.53 ft with a standard deviation of 8.22 ft. The mean and median 1960-1996 water-budget components for the Wet Walnut Creek valley aquifer are shown in table II-16. The relatively large contrast of the mean and median values are indicative of the high variability of water-budget components in the Wet Walnut valley.

TABLE II-16—Mean and median of the 1960-1996 water budget components for the Wet Walnut Creek valley aquifer (in inches/yr).

The average 1990-96 recharge to the aquifer (which also includes pond seepage from the watershed structures and irrigation return flow in addition to precipitation percolation) was estimated at 1.9 inches/yr; however, it ranged widely from less than 0.05 inch/yr to more than 16 inches/yr during the 1993 flood year. In comparison, the average pumpage from the aquifer over the same period was 4.0 inches/yr. The average net stream seepage to the aquifer (i.e. streamflow loss minus streamflow gain or baseflow) was estimated at 1.9 inches/yr (table II-16). Wet Walnut Creek was a net-gaining stream up to the mid- to late 1960’s, but since the late 1960’s it became a net-losing stream to the alluvial aquifer. By the early to mid-1980’s, the stream network system became the major source of recharge to the aquifer, even exceeding precipitation percolation- and pond-seepage-based recharge.

Other Wet Walnut Creek Valley Ground-water Studies

Gillespie and Slagle (1972), in a ground-water recharge study of Wet Walnut Creek valley from Bazine to Albert, estimated the average annual recharge to the aquifer for 1965-69 to be 13,000 acre-ft/yr. If we approximate the study area to be 64,000 acres (100 mi²), this recharge value converts to an estimate of 2.4 inches/yr.

Nuzman (1990) conducted a numerical modeling study of the Walnut Creek valley from near Ness City to Great Bend, an area of 124,160 acres (194 mi²), using the USGS MODFLOW program (McDonald and Harbaugh, 1988) and a grid of 1-mi x 1-mi. A trial-and-error calibration resulted in a recharge estimate of 10% of an average 22 inches of annual precipitation (i.e., 2.2 inches/yr).

Finally, as a result of the 1990-91 public hearings on the designation of an Intensive Groundwater Use Control Area (IGUCA) in Barton, Rush, and Ness counties, Kansas, the Chief Engineer concluded, regarding long-term ground-water recharge in the Walnut valley, “. . . that the long-term sustainable yield of the aquifer within the boundaries of the proposed control area as set forth in Conclusion No. 8 (i.e., an area of 348,800 acres) is no more than approximately 22,700 acre-ft per year” (Division of Water Resources, 1992, p. 96). This translates to 0.8 inch/yr. However, the declared IGUCA encompasses areas beyond the Walnut Creek alluvium. Considering that the Walnut Creek alluvial aquifer area from near Ness City to the confluence at Great Bend is approximately 128,000 acres (200 mi²), the sustainable yield figure of 22,700 acre-ft/yr translates to a long-term recharge estimate of 2.1 inches/yr.

C10. Solomon River Basin, Kansas

Jorgensen and Stullken (1981) studied the North Fork Solomon River valley from Kirwin Dam to Wakonda Lake (Glen Elder Dam) in north-central Kansas. They applied the USGS two-dimensional finite-difference model (Trescott et al., 1976) to the area using the 1946 water-level conditions as approximately steady-state conditions, and employing a rectangular 0.25-mi x 0.50-mi grid network. The estimated effective recharge from precipitation (i.e. aquifer gain from precipitation minus evapotranspiration) was 2.55 inches/yr. The simulated steady-state hydrologic budget is shown in table II-17 below.

TABLE II-17—North Fork Solomon River valley from Kirwin to Glen Elder dams used to simulate steady-state (1946) conditions (values in inches/yr).

According to Jorgensen and Stullken (1981), the simulated net leakage of approximately 2.5 inches/yr (i.e. leakage to the river minus leakage from the river, table II-16) was consistent with the estimated gain in baseflow of the river within the area modeled.

Burnett and Reed (1986) applied the same USGS model and grid network used by Jorgensen and Stullken (1981) to the South Fork Solomon River valley from Webster Reservoir to Waconda Lake (Glen Elder Dam) in north-central Kansas. They calibrated the model for transient 1970-79 conditions using two pumping patterns per year—one pumping period simulating the nonirrigation season (September through May) of each year, and another period simulating the irrigation season (June through August) of each year. The simulated hydraulic heads were within ± 5 ft of the measured hydraulic head at 49% of the sites, and were within ± 10 ft of the measured hydraulic head at 82% of the sites. Table II-18 shows the average simulated hydrologic budget for the area.

TABLE II-18—Average (1970-78) hydrologic budget for the Sourth Fork Solomon River valley from Webster to Glen Elder dams (values in inches/yr).

Average annual recharge to the aquifer from precipitation was determined to be 1.7 inches/yr for 1970-78. Average annual recharge from the Osborn irrigation canal was estimated to be 0.9 inch/yr over the same period. The 1970-78 average annual discharge from the aquifer to the river (baseflow) was estimated to be 2.3 inches/yr, whereas total well pumpage over the same period averaged about 1 inch/yr.

Sophocleous et al. (1990; see also McClain et al., 1995) estimated ground-water recharge for the Solomon River basin using two different methods: 1) interpretation of early streamflow records (1960-61) at the North Fork Solomon River gaging station near Glade (Phillips Co.), and 2) using a monthly soil-moisture budget for the Solomon River basin over the 25-yr period 1964-1988.

The long-term average recharge to the alluvial aquifer was assumed to equal the long-term average ground-water outflow during the early times of the Solomon watershed irrigation development. Such an equilibrium condition existed in the watershed until the early 1960’s (Weston, 1979). During 1960 and 1961, the average amount of ground water appropriated in the ~395,674-acre area drained by the North Fork Solomon above Glade was 13,860 acre-ft/yr, which amounts to 0.42 inch/yr over that subwatershed area (water appropriation data from Division of Water Resources, Kansas State Board of Agriculture). The average annual baseflow during the period 1960-61, as derived from the streamflow data at Glade, was ~10,200 acre-ft/yr, which amounts to 0.31 inch of water over the same subwatershed area. Thus the total ground-water outflow (baseflow plus pumpage) for 1960-61 was 0.73 inch/yr, which, under the assumption of equilibrium, represents the amount of ground-water recharge. Ground-water outflow through evapotranspiration was presumed negligible and therefore was not considered in the calculations.

The second method for estimating regional ground-water recharge in the Solomon River basin is the moisture-budget technique. If the 25 years of record (1964-1988) for all 19 weather stations covering portions of the Solomon River basin (Sophocleous et al., 1990) are considered representative of the average conditions in the Solomon River watershed, moisture budgets indicate that the total annual average-moisture surplus or average potential-annual ground-water replenishment in this watershed for predominant 12-inch soil-moisture-capacity soils varies from 0 inches to 3.8 inches. For the 1960-61 period, and based on the climatic data from the Kirwin Dam station (the closest station to the Glade streamgaging station and centrally located within the entire watershed) and the predominant soil-moisture capacity of 12 inches, precipitation totaled 27.84 inches, Thornthwaite potential evapotranspiration and actual evapotranspiration totaled 27.79 inches and 25.98 inches, respectively, and moisture surplus totaled 1.89 inches, which is above normal compared to the 25-year average of 1.16 inches for the same conditions.

During the 1960-61 period, the average total streamflow at Glade was 34,720 ac-ft/yr and the average baseflow was 10,200 ac-ft/yr, resulting in a direct surface runoff (the difference between total streamflow and baseflow) of 24,520 ac-ft/yr (0.74 inch/yr). The moisture surplus must, however, satisfy both the surface runoff and the ground-water recharge. This surface runoff figure, when subtracted from the average 1960-61 moisture surplus of 1.89 inches, based on the Kirwin station, results in a value for regional ground-water recharge of 1.15 inches. This value is of similar magnitude as the recharge value (0.73 inch/yr) calculated from baseflow and ground-water pumpage data.

Thus, assuming that the more than 395,000-acre subwatershed above Glade is typical of the entire Solomon watershed, the average estimated regional ground-water recharge for the Solomon watershed is 0.94 nch (based on the two previously mentioned recharge-estimation methods), which represents only 4% of the average annual precipitation (23.29 inches/yr). During 1980-81, the ground-water appropriations in the Glade subwatershed, which reached 146,182 ac-ft, compared to 13,860 ac-ft in 1960-61, amounted to more than 4.7 times the amount of estimated natural ground-water replenish-ment for that subwatershed.

South Fork Solomon River Basin above Webster Reservoir

Weston (1979) computed an annual ground-water budget for the South Fork Solomon River basin above Webster Reservoir for the period 1947-1976. By assuming equilibrium conditions prior to 1966, and by estimating long-term average outflow during that period (consisting of estimated baseflow, and total appropriated pumpage minus 20% to reflect net withdrawal because a portion of what is pumped is returned to the ground-water system), Weston (1979) estimated total inflow or recharge for the South Fork Solomon River basin above Webster Reservoir (an area of 1,044 mi2) as 0.56 inch/yr for the predevelopment, steady-state period 1947-1965. Average annual recharge for the entire 1947-1976 study period was estimated at 0.51 inch/yr.

C11. Arkansas River Valley from Kinsley to Great Bend

Sophocleous et al. (1993) modeled the Arkansas River valley from Kinsley to Great Bend using the MODFLOW model in two dimensions in conjunction with the parameter-estimation model MODINV (Doherty, 1990). They employed a regular, rectangular 1 mi x 1 mi grid, oriented in a southwest to northeast direction incorporating the Arkansas River from Kinsley to Great Bend, and calibrated the model for both predevelopment, steady-state (1955) and transient (1955-1990) conditions using annual time steps and pumpage data from DWR water-rights data.

The calibrated steady-state recharge was estimated as 1 inch/yr with a standard error of 0.02 inch/yr, and the average 1955-1990 recharge was estimated as 1.8 inches/yr with a standard error of 0.06 inch/yr. A summary of all inflows and outflows in the region is presented in the predevelopment and 1985-1990 development-period water budgets (table II-19). The ratios s/ h (square root of error variance over the difference between the highest and lowest value of head in the model region) are relatively small (0.009 and 0.015 for the 1955 and 1990 water budgets, respectively), indicating satisfactory model fit.

TABLE II-19—Simulated steady-state and transient ground-water budgets of Arkansas River valley from Kinsley to Great Bend (values in inches/yr).

The predevelopment budget shows that the major input to the aquifer is ground-water recharge and the major outputs from the system are lateral outflow from the model boundary near Great Bend and stream baseflows (streamflow gain). In contrast, the major inputs to the aquifer in recent times come from both streamflow losses and recharge, and the major outflow is well pumping, whereas streamflow gains from baseflows decreased significantly. Ground-water pumpage for irrigation is now balanced by an increase in recharge (mainly from irrigation return flow and conversion of grasslands to croplands), a decrease in natural discharge (mainly decreases in baseflow), and a net loss in aquifer storage (as manifested by long-term ground-water-level declines).

C12. Rattlesnake Creek Watershed, South-central Kansas

Sophocleous and Perkins (1993 a,b) studied the lower Rattlesnake Creek watershed from west of the Macksville stream-gaging station near the southwest Stafford County boundary to the confluence with the Arkansas River in Rice County, an area of approximately 560 mi². They applied and calibrated the USGS MODFLOW finite-difference model to the study area using 1-mi² grid cell network in two dimensions and employed parameter-estimation techniques to optimize model parameters for both steady-state (mid-1950’s conditions) and transient (1955-1991) conditions using 3-yr stress periods and annual time steps. The parameter-estimation program MODINV (Doherty, 1990) was employed to optimize model parameters. The predevelopment (c. 1955) recharge was estimated as 1.3 inches/yr with a standard error of 0.4 inch/yr, whereas the average development-period recharge (1955-1990) was estimated as 1.9 inches/yr with a standard error of 0.3 inch/yr. The predevelopment and development water budgets are displayed in table II-20. The ratio of the square root of the error variance in the parameter estimation model over the difference between the highest and lowest value of head in the region was 0.0094 to 0.0096, a relatively very small value for both the steady-state and transient simulations, indicating that errors in the model were considerably less than the model response as indicated by the maximum head loss of 335 ft.

TABLE II-20—Predevelopment (c. 1955) and development (1988-1990) water budgets for the lower Rattlesnake Creek watershed (values in inches/yr).

In contrast to the 1950’s water budget, where the largest outflows from the aquifer were baseflow to streams and evapotranspiration losses mainly to the Quivira marsh region, the present day (1990) dominant outflow is ground-water pumpage for irrigation. This superimposed discharge to the aquifer system is balanced by an increase in recharge (mainly through irrigation- return flow and conversion of grasslands and dryland farming to irrigated agriculture), a decrease in discharge (mainly through decreased baseflow contributions to streams and decreased evapotranspiration), and a loss from aquifer storage (as manifested by long-term ground-water-level declines).

Sophocleous et al. (1997, 1999) also developed and applied an integrated watershed and ground-water model to the entire Rattlesnake Creek basin (approximately 1,317 square miles) by combining the USDA Soil and Water Assessment Tool (SWAT) watershed model (Arnold et al., 1993) with the USGS MODFLOW model. The basin was divided into 35 topographic subbasins (fig. II-10) and the stream-aquifer system was modeled for both predevelopment or steady-state conditions (pre-1960), and development (1955-1994) or transient conditions. The calibration period spanned from predevelopment to 1980; the data for the post 1980 period were reserved for verifying/corroborating the calibrated model. The parameter-estimation program (PEST; Doherty et al., 1994) was employed to optimize the aquifer parameters. The predevelopment average recharge was estimated at 1.4 inches/yr, whereas the average 1990-94 period recharge was estimated at 2.2 inches/yr but ranged from 0.5 inch/yr during 1994 to 5.0 inches/yr during the flood year of 1993. Table II-21 displays the water budget for the Rattlesnake Creek basin during predevelopment and present-day (1991, 1992, 1993, and 1994) conditions, representing a transition from extreme dryness to extremely wet conditions.

The average calibrated recharge for the entire transient simulation period (1955-1994) in each subwatershed in the Rattlesnake Creek basin is shown in fig. II-10. Because the effective recharge to the aquifer is largely a function of the soil type and land use, given similar precipitation patterns, a wide variation in recharge can be observed. For example, subwatersheds 1 through 4 in the far upstream boundary of the watershed near the Kiowa-Edwards-Ford counties border (fig. II-10), consisting predominantly of the low hydraulic-conductivity Harney soil, have recharge averaging 0.4 to 0.8 inch/yr. On the other hand, subwatersheds 33 and 35 in southwest and northeast Stafford County, respectively (fig. II-10), that consist predominantly of the highly permeable Tivoli soils, show much higher recharge, averaging more than 4 inches/yr. (This more detailed and up-to-date study supercedes the recharge estimates based on the soil-water budget pilot study of the Rattlesnake Creek basin by Sophocleous and McAllister [1987, 1990]. However, the impact of soil, plant, and land-use factors on recharge from that study are presented in part I, section 8.1.) The overall area-weighted average recharge during the 1955-1994 simulation period was 2.1 inches/yr.

FIGURE II-10—Rattlesnake Creek subwatersheds and simulated average (1955-1994) recharge rate (inches/yr) in each sub-watershed. Bold numbers are recharge rates; smaller numbers are subwatershed identification numbers (adapted from Sophocleous et al., 1997).