Sie sind auf Seite 1von 14
Quantifying Ground Water Recharge at Multiple Scales Using PRMS and GIS by Douglas S. Cherkauer’ Abstract Management of ground water resources requires a method to calculate demonstrably accurate recharge rates at local to regional scales using readily available information bases, Many methods are available to calculate recharge, ‘but most are unable to satisfy all these conditions. A distributed parameter mode is shown to meet the stated needs. Such models are input intensive, however, so a procedure to define most inputs from GIS and hydrogeological sources is presented. It simplifies the PRMS calibration observed streamflow hydrographs by reducing degrees of freedom from dozens to four For seven watersheals (60 (0 500 km), the GIS-aided calibrations have average errors of 5% on recharge and 2% fon coal streamflow, verifying the accuracy of the process, Recharge is also calculated for 63 local-scale subwater sheds (average size 37 km’), For the study area, calculated recharges average 11 cmyr. Soil and rock conductivity, porosity, and depth to the water table are shown to be the physical properties which dominate the spatial variability of recharge. The model has been extended to uncalibrated watersheds where GIS and elimatic information are known, reproduces total annual discharge and recharge to within 9% and 10%, respectively, indicating the process ean also be used to calculate recharge in ungauged watersheds. It has not been (ested outside the study area, however. Introduction Q ‘As the demand for ground water supply grows with the Q,, ‘human population, the need for proactive management of ‘ground water supplies increases as well, Proper manage- ASM ment of the quantity of ground water resoutces requires an accurate understanding of the ground water budget. For a discharges to surface water bodies abstractions from ground — water through wells = change in ‘ground water system pee unit time Storage inthe portion of an aquifer system, the budget can be simply stated as R+GW-Q-Q,-GWay = ASH (1) where R= recharge tothe aquifer! GW,, = influxes from upgradient portions of ground water system* GW,.= effluxes. to downgradient portions of ground water system* Department of Geosciences, University of Wisconsi-Mil- waukee, PO. Box 413, Milwaukee, WI 53201; (814) 229-4563; fax (414) 229-5482; aquadacduwmeds Received February 2002, accepted Anil 2008. Copyright © 2008 by the National Ground Water Association, ‘Anificil recharge and leakage through aqutards are assumed negligible. ‘Recharge is heing defined as the net quantity entering the saturated zone (surface infitation less losses above the water table) ‘At the ground watershed scale (Table 1), Equation 1 simplifies to @ because there are no lateral flows across ground water divides, Under steady-state conditions (approximated in natural systems over long time periods), the budget equa- tion becomes. R=Q,4Q, Vol 42, No. 1~GROUND WATER-January-February 2004 (gages 97-110) 97 Recharge is the most difficult term in Equations 1, 2, and 3 t0 measure, because itis diffuse and varies in both space and time. Yet itis the single largest input to ground water systems and being able 10 quantity it spatially is essential for resource management. The temporal variation of recharge is beyond the scope of this paper. Good ground water management must be done at 2 scale that uses natural hydrologic, rather than political or ‘geologic, boundaries, The ground water budget therefore needs to be developed on a regional or large scale, for an entire aquifer oF geographic region, Communities within ‘that region can then assess their smaller scale portion of the resource in coordination with the larger area, Because recharge is the dominant source of ground water at either scale, 2 method is needed to determine its rates accurately across a regional scale. At the same time, that method must recognize spatial variation within the region and be able to resolve them to the small scale (Table 1). Many methods, including well and stream hydrograph analysis, water and geochemical budgets, and various model types, have been devised to estimate recharge. As a rule, however, these pro- codures fail to meet one or more ofthe piven criteria. While they may produce a good measure of recharge, they com- monly either are very scale- or site-specific or produce recharge values with large or undetermined uncertainty. ‘The purpose of this paper is to present a procedure that overcomes these limits. ‘The following cursory review of representative recharge methods is intended to point oat any limits in achieving the needs of resource management defined ear- Tier. Limitations in achieving these needs should not be inferred to suggest inaccuracy in a method, however. Scan- Ion et al. (2002) provide a more comprehensive review of available methods Previous Work Water level rises in well hydrographs have been used to measure recharges (Avery et al. 1999; Goes 1999; Delin etal. 2000; Ketchum et al. 2000) as AHI “ recharge rate (volume per unit surface area) AHV/t = rate of seasonal rise in water level in a well n effective porosity ofthe aquifer ‘This method i site specific and its validity depends on an accurate measurement of effective porosity, so itis gener- Table 1 Definition of Scale Terminology Used in Text ‘Seale Meaning Site specific Point measurement (< 100.0?) Local Subwatershed («25 km?) Large Watershed 25 10400 kn?) Regional Multiple watersheds (> 400 km) 98 DS. Cherkauer GROUND WATER 42, no. 1:97-110 ally difficult to extend to even the local scale. It is, how- ever, the only direct measure of recharge available ‘A number of methods have been used to estimate recharge indirectly. One common approach is to assume well abstraction is negligible in Equation 3, Then recharge can be determined by measuring base flow discharge to streams or springs (Meyboom 1961; Walton 1965; Gerhart and Lazorchick 1988; Hoos 1990; Mau and Winter 1997; Avery et al, 1999). The streamflow hydrograph from & sauging station can be separated into surface runoff and base flow components if effects from lake oF reservoir stor- age or human relocation of water do not affect total flow. ‘A wide variety of methods have been developed to separate base flow from hydrographs (Barnes 1939; Rorabaugh 1964; Linsley etal. 1958; Rutledge and Daniel 1994; Sloto and Crouse 1996; Arnold and Allen 1999) or 10 calculate recharge for specific events (Meyboom 1961; Bevans 1986; Rutledge 1998), Mau and Winter (1997) and Ansari (1999) have demonstrated there is a degree of con- sistency among the recharge estimates from manual and the rmore recent automated separation methods. Halford and Meyer (2000) raised concerns about the accuracy of base fow-derived recharge rates, but their concerns largely result from attempting to compare stream and well hydro graph recessions without incorporating the effects of poros- ity onthe later. Base flow separation has the advantage that ivimegrates ground water discharge aeross an entire water- shed, and the recharge values derived ae tied toa measured ‘ux (total streamflow). The method, however, can only be used in gauged watersheds, and the recharges cannot be resolved to smaller scales than that of the drainage basin, Recharge can also be obtained by measuring the other terms in Equation 1 of 2, and solving it for net rechange as the residual. Walton (1965), Cherkauer and Bacon (1978), Liu and Zhang (1993), Avon and Durban (1994), Amold and Allen (1999), Ketchum et al. (2000), Louie et a (2000), and Otto (2001, among others, have used ground water budgets to calculate recharge rates. Sophocleous and MeAllister (1987), Swanson (1996), and Detin et al. (2000) have also used a soil water budget to estimate recharge. ‘Water budgets have the advantage that they ean be applied to almost any scale system for which fluxes ean be mea sured. They do not have to be constrained to gauged water- sheds. although lack of streamflow data can make quantifying ground water discharges to surface water box! ies difficult. The biggest concern, however, is the propaga tion oF eror through the calculations. Fach term in the budget has uncertainties, and there is no direct way to ver~ ify the accuracy of the resulting recharge. Furthermore the data needed to calculate all the other budget terms are fre= ‘quently not available, Geochemical methods can also be used to obiain recharge rates. Mass budgets of conservative fons or stable isotopes can be solved for recharge flux (Dettinger 1989: Woo! and Sanford 1995; Ting etal. 1998; Winograd etal 1998; Rosen etal. 1999; Cresswell etal. 1999: Jones etal 2000; Abd E] Samie and Sadek 2001) and/or to delineate the sources of recharge. Te chemical mass budgets are subject to the same uncertainties as water budgets. For isotopes ‘ther than "0 and 2H, additonal assumptions may need! 10 bbe made about the composition of recharge, or the absence ‘of sources or sinks in the geologic system. These methods fare best suited for estimating long-term recharge rates in large or regional scale areas. I is difficult to resolve local scale spatial variation with them. It is equally difficult to verily the accuracy of the recharges produced because many assumptions needed (about sourees, sinks, and reactions along flow) remain unstable Radioisotopes (Daniels et al. 1991; Solomon and Sudicky 1991; Zhu 2000) of artificial compounds with me-dependent sources (Delin etal. 2000) can provide information on water transit times, from which recharge rates can be calculated, These methods provide site-spe- cif, long-term average recharges, which can be mapped to show local scale variations. As long as the hydrogeology of the ste (eective porosity in particular) is well defined, the recharges produced are inherently accurate, Ground water transport pattems are subject to human interference, how- ever, so these dating methods aze best applied to undis- Curbed systems. Using them to define recharge for a region occupied by humans requires a complete understanding of how purmping and contamination may have altered the sys- tem through time ‘The recharge methods previously described share scale of application problems for their use in ground water man- agement, They either measure atthe ste or les scale, and ate dfficule to extrapolate to larger area, or they measure at large or regional scales, but are hard to resolve 10 the local scale. Many researchers have approached this problem by combining site or locally calulated recharges with some form of conceptual or mathematical model to extrapolate them o a larger scale. Many have used an empirical para metric approach in which recharges at disrete locales are comelated 10 readily measurable physical quantities that should function as controlling factors. The empirical relae sions can then be used to calculate recharge in areas where itfhas not been measured. Commonly, soil propeties, topog- raphy, and land eover (including vegetation) have been used to extrapolate recharge for regions or even entire Sates (Walton 1965; Charles et a. 1993; Holtsehlag 1997; ‘Cherkauer, in review), Arnold and Friedel (2000) and Eaton and Zaporozee (2002) have used a similar procedure 10 define the susceptibility to contamination of shallow aquifers, Those methods demonstrate the spatial distribution of relative recharge rates. Offen, however, i is sifficul 10 test the accuracy ofthe recharge values. The advent of GIS, however, makes this process more attractive than ever Ground water flow models have also been used to obtain estimates of recharge. The recharge flux can be extracted from the calibration process, or the model can be used to calculate te recharges that are necessary to match target heads under the modeled conditions (Stoertr and Bradbury 1989; Cherkauer and McKereghan 1991; Swan- son 1996). cis widely recognized, however, tht sich mod- els do not produce unique solutions, ‘The recharges obtained this way are entirely dependent on the accuracy of. the hydrogeological conceptual model and the resolution of the hydraulic propeties used. For example, asthe number of surface water bodies depicted in aground water model increases, the recharge rate needed 10 achieve mass balance will also rise; hence the recharge rate is controlled atleast parly by the design of the model One final modeling approach, distributed parameter models, holds considerable promise to meet the needs of regional ground water management, These mexdels use a holistic approach to water distribution. Precipitation on the modeled area is distributed among evapotranspiration, runoff, and infiltration according to prescribed formulas, CREAMS and HSPF have been used by Ostercamp etal (1994) and Dinicola (1997), respectively, 1 generate recharge rates. Both these models, however, are focused on surface hydrologic conditions. Sophocleous and Perkins (2000) have successfully Hinked SWATT (a semidistributed watershed model) (Amold et al. 1995) with MODFLOW, the widely used, fully distributed, finite difference, ground ‘water flow model (McDonald and Harbaugh 1988). SWAT. produces recharge inputs for MODFLOW, and the «wo models together allow simulation of all the ground water/surface water exchanges relevant to management at the watershed scale. This combination is an excellent approach, but does require development and calibration of ‘ground water model for each area to be examined, More recently, Steuer and Hunt (2001) used the pre cipitation runoi? modeling system (PRMS) to quantify recharge. It is a distributed parameter watershed model which allows a more rigorous simulation ofthe subsurface system than many surface water models, but does not require development of a ground water flow model. It can be calibrated to an observed streamflow bydrograph, but it allows the watershed to be discretized into local scale sub> divisions. Hunt et al. (2001) successfully used the recharges calculated by PRMS as input to a MODFLOW model tor 47 km? watershed, Because of its flexibility of seale and the fact that its ‘output ean be compared to abservable streamflows, PRMS: thas been selected as the method to quantify recharge inthis study. Its designed to incorporate multiple sees and can bbe tested against a readily available measure of flux. (streamflow). PRMS is data intensive, and calibration can invalve many dozens of degrees of freedom. To date, i has only been used (0 determine recharge at a relatively small scale (Steuer and Hunt 2001), In addition, as originally designed, the model is intended for use only on gauged ‘watersheds, Objectives ‘The purpose ofthis paper isto address three questions regarding the use of PRMS in quantifying recharge. 1. Can the PRMS approach be extended to areas larger than SO km?, thus allowing accurate determination of ground water recharge at the local through regional scales? Can the calibration process be constrained by the incorporation of parameters available from GIS data bases to streamline the model calibration process, but still produce accurate results? 3. Can the results be extended to ungauged areas? Study Area “The study has boen conducted in southeastern Wiscon- sin in conjunction with the development of regional scale, DS.Cherkauer GROUND WATER 42,n0.1:97-110 99 ‘ground water flow model, PRMS was used to check esti- ‘mates of recharge in seven gauged watersheds (Figure 1) ‘The region receives 75 10 90 cmlyr of precipitation. The centire area has been glaciated, with drift ranging from 0 t0 ‘more than 60m in thickness. ‘The drift is predominantly clay-tich, low-conduetivity tll along Lake Michigan in the east. To the west, tills become sandier and outwash ‘becomes more common, so the conductivity ofthe deposits A series of Paleozoic sedimentary units Tie beneath the ilacial deposits and dip eastward toward the Michigan Basin, In most of the study area, the topmost rock is a dolomite whieh, together with the glacial deposits, forms 2 shallow, unconfined aquifer canging up to 200 m in thick- ness. Below the dolomite isa shale, which is a continuous aquitard to the east of its suberop limit (Figure 1). Beneath the shale fie deeper, confined aquifers consisting of sand- stones and dolomites ranging from 200 to more than 600 thick. To the west of the shale suberop, this dominantly sandstone aquifer is unconfined and in direct hydraulic ‘commonication with the surficial glacial sediments ‘The seven study watersheds were selected to cover the full range of hydrogeotogie, topographic, and land cover ‘conditions in the region, Each contains a USGS gauging station for a drainage area of ~$00 km? or less. Most have continuous gauging records for the period 1963 10 1995, Five lie entirely east of the shale suberop, where the aquifer receiving recharge isthe glacial/dolomite combination, The other «wo extend to the west of the suberop. In these areas, recharge goes to the glacial sediments and then into the sandstone aquifer, Figure 1. Location map of the study watersheds. The eal brated watersheds are the seven southernmost, and the two test ones are in the north, Both sets include their HRU subdi visions. Suberop of Maquoketa shale i solid pattern, but only shown outside the designated watersheds. Heavy bar is 20 km long. 100 DS. Chetkauer GROUND WATER 42, no. 1:97-110 Watershed Model PRMS (Leavesley et al. 1983) is a deterministic, dis tributed parameter model designed to evaluate the effects of precipitation, temperature, and land cover on the distribu- tion of water among runoff, evapotranspiration, and infil- tration (Figure 2). It calculates the water fuxes of each ‘component of the hydrologic eyele at a user-defined time interval, Within PRMS, a watershed is conceptualized as a series of interconnected, cascading reservoirs including, interception storage on vegetation, impervious-surface stor- ‘age atthe ground surface, and storages in the soil, both the ‘unsaturated and saturated ones, The storage in each reser- voir, and the movements between them, are calculated at ceach time interval ‘A watershed is subdivided into up 1 50 hydrologic response units (HRUs) of relatively homogeneous proper ties such as slope, aspect, vegetation, sol, oF land cover. The ‘model calculates water budgets for each HRU indepen: dontly and then sums their outputs to produce a composite shed response, PRMS accepts inputs and produces ‘outputs as time series, which are stored in a WDM (water- shed data management file. The programs [OWDM (Lumb Cal. 1990) and ANNIE (Flynn etal, 1995) are used to pre pare, transfer, and observe the time series. ‘As used in this study, PRMS requires the input of dozens of variables for each HRU, or the entire watershed. Many are commonly available climatic (temperature, precip- itation) or ground surface (slope, aspect, vegetative cover) measures. There are, however, 16 that are more difficult to ‘obtain, Rather than leave them as free to vary in calibration, wherever possible these parameters have been defined in this ‘study as functions of properties commonly available from GIS. Predominantly, they fall into the ground surface and subsurface categories described later in this paper. Data Needs. ‘Climatic and Streamflow Time Series PRMS was used in daily mode, which requires inputs for each HRU of latitude, longitude, aspect, daily values of precipitation, and maximum and minimum temperature Because along. term average recharge was sought, 30 years ‘of records were used to calculate daily average precipite tions and temperature at NOAA climatic observation sites throughout the region. For each study watershed, daily time series for these parameters were then calculated as ‘Thiessen-polygon weighted averages of the nearest climate reconis, The resultant annual precipitations range from 80 to 87 emir, Average daily maximum and minimum tem peratures have a range of less than 2°C among the water- sheds. Typically, the input temperatures vary smoothly throughout the year, but single extreme events remain apparent for precipitation (Figure 3). A composite hydrograph was generated for each gaug- ing site. Ibis the average daily flow for the 30-year study period plotted as a function of time during the water year. As with precipitation, extreme flows on a single day in the record persist despite the long-term averaging. The average daily flow per unit drainage area varies from 21 co 32 em/yr resulting from differences among runoff, recharge, and Figure 2 Schematic diagram of the precipitation runoff ‘modeling system (PRM). Modified from Leavesley et al, 11983). cevapotranspiration. Nowe that the gauging site is often located upstream from the mouth of a geographic water- shed. PRMS can only be calibrated for the portion of the ‘watershed draining to the gauge. Land Surface Conditions Conceptually, PRMS distibutes precipitation among interception, runoff, evapotranspitation, an infiltration. Init tration is calculated trom precipitation and interception coet- ficients. Values of the latter from Steuer and Hunt (2001) were used because of that site’s proximity andl similar glacial history. Surface sunofT is generated from both the impervious surfaces and the soil zone reservoirs (Figure 2) using ROjqy = (Ppt ~ Retip) x Impery 6) RO y= Pin x [Sen + [(Sex-Sen) x Rechn/Rems)]} tm tao 000300, Deysin Water Yeor (Ey i — Fo Figure 3. Climatic inputs for the Root River watershed, where Impery = the percentage of ground surface that is| impervious Ppt total precipitation Pin et precipitation (Ppt-interception) Rechr torage in the soil zone Remx ximum storage capacity inthe soil zone Retip = maximum retention capacity of impervious surfaces ‘Sen and Sex = minimum and maximum runoff contribut- ng areas tn Equations 5 and 6, Ppt is input, Pin and Rechr are caleu- lated daily by PRMS, Retip was set universally as 0.5 and the other four terms were defined based on GIS-avail- able values (Table 2) and soil porosity, which was calibrated Evapotranspiation js calculated based on daily tem: peratures and the maximum possible hours of sunshine (Hamoa 1961), It can be modified by application of a ‘monthly coefficient, whieh has been calibrated in this study. Finally, net infitation ( the subsurface is ealeulated by PRMS as the residual water after interception, sunoff, and evapotranspiration have been removed. Conditions Below the SoilZone PRMS allows water to exfiltrte downward from the (6) soil reservoirto the subsurface and ground water reservoirs, ‘Table 2 Formulation of PRMS Coefficients for Distribution of Water at Ground Surface S% Ag = apealval and ever ina RU GW = open wae aes a pon nan HRD 2” Selective preity Parameter PRMS Label GIS-Based Formula’ Runoff coefficient for impervious surfaces nperv 0.1 x (Dev) Sex 035 x (% Dev} +02 X (4 Ag) -O4 X (% OW) Son BAS x Dev) Maximum storage in soit, ems OLN X Dy, ‘Dove devioped nd over race, clad ond oan BRD 1. = average depth o water ae ahswerge hike of use 098) ne coetiens in hee Fomnlashave hee oti yaa em, With the exception of oy ll se pt aise sa in IS 0S. Cherkauer GROUND WATER 42, no. 1297-110 101 Table 3 Formulation of PRMS Components Below the Ground Surface! ‘Budget Component PRMS Label PRMS Calculation? Gits-Based Formula? Subsurface Reservoir Storage in unsaturated 20ne Res Calculated with water budget Anil Valuc of Res tun assure nl value 208-WH) x ax 12 Discharge ostream Res RaseRes x Re? Ref = {Ks x $ x Le x 24M(Ad X Discharge to saturated zone Gat Gil = Rsep % (ResMResms)*= Rep = Ket 00 Resmx = |; Rexp= 005 Ground Water Reservoir Direct inflow rom sei Sep Assigned value Sep= Ks/100 Storage in saturated zone ow Called from water budget al Value of and asigned nl vale =0NT=B) x0 x 12 Discharge to steam Bas Bas=Reb X GW! Reb= [Kg x $x 2 x LewAdX>)] Discharge wo deeper aquifer Sok Sok = Gonk x GW \Kb x Ww x Ad x 12VGWI 2am nthe sone oan we dete a Tle, “Tos em ns arn say oni inn which are directly analogous to the unsaturated and satu- rated zones in an unconfined ground water system, Each of these accepts inflow from reservoirs above it in the con- cceplual system, releases water to either still deeper ground ‘water zones or the stream, and can also store water (Fig- ure 2), A water budget is continually calculated for both reservoirs. PRMS uses conservation of mass equations to calculate inflow to and storage in each and an equation with ‘ user-supplied flux coefficient to calculate the outflows, In the outflow equations are of the form lux coefficient x height of water stored in the reservoir a The user may make the outflow equations either linear or nonlinear, and the linear option has been selected herein. Numerous input parameters are needed to make these subsurface calculations, but site-specific values are olten ‘not known, One could assign uniform values based on regional measures, but this would not allow for any spatial variation within the subsurface. To overcome this problem, site-specific values have been defined from available GIS and other data sources. First, the flux equations have been transformed to a Darvian analog: Flow = conductivity x gradient x aren (8) where the area through which the flow occurs isthe coluran height of the stored water muttiplied by the effective width of the flow. Hence, the PRMS flux coefficient becomes PRMS coefficient = conductivity x gradient x width (9) ‘Table 4 Definition and Data Source of Terms in Table 3 Source Abbreviation Definition Available GIS Other Ad Deainage are of HRU Yes B Average elevation of base of shallow aquifer in HIRU Somet HG! maps w ‘Vertical hydraulic grafient at base of system HG reports Kb Vertical hydraulie conductivity of material below saturated zone HG repoes Ka, Effective horizontal hydeulic conductivity of saturated zone HG repons Ks fective horizontal hydraulic conductivity of soil Yes Le Length of primary channel io HRU Yes 1 Efective porosity in HRU Calibrated 5 ‘Average ground surface slope in HRU Yes wr ‘Average water table elevation in HRU. Yes 102 _DS.Cherkauer GROUND WATER 42, no. 1:97-110 ‘As an example, for the saturated zone flows to the stream (base flow or BAS) and deeper systems (SNK) are calculated as BAS K, x1, (GWin) 2L, (10) SNK =K, x1, x Ad ay where K, = horizontal hydraulic conductivity of the saturated geologic column K, = vertical hydraulie conductivity of the basal unit below the aquifer 1, = horizontal hydraulic gradient in the saturated zone 1). = vertical gradient at the base the height of stored water in the saturated zone clfeetive porosity length of main channel in HRU surface area of HRU GW (storage) is calculated by PRMS as the height of the stored water columa, [t must be divided by porosity to make it the equivalent of « head in the saturated zone, The horizontal gradient has been arbitrarily approximated as half the GIS-measured ground surface slope, The coetli- Cients used to calculate the fluxes below the ground surface are summarized in Tables 3 and 4. The relations presented define the inputs needed for the subsurface zone caleula- tions. Porosity cannot be directly linked to any available GIS measures, so it must be obtained from calibration. It has been assumed to be uniform throughout a watershed. in the application of PRMS. | Bach HRU has been defined as 2 subwatershed that has 4 perennial stream in it, and has relatively homoge- ‘neous soil, slope or land cover. 2. The model was run exclusively in daily mode so all HRU outputs are made to arrive instantaneously at the ‘gauging station. 3. A single set of climatic inputs was provided for a ‘watershed. All HRUs within it were assumed to have the same precipitation and temperature. 4. Several parameters for which good information on regional-scale distributions is not available were set as constants. ‘These included interception, soil- and root- zone depths, snow albedos, and water contents, among others 5. Because total relief in the study watersheds never exceeds 150 m, the adiabatic variation of temperature and precipitation is assumed negligible within each watershed. 6. An annual temperature increase between 1° and 2°C hhas been applied to urbanized HRUs to account for the turban heat island effect. Model Calibration ‘A physically based estimation of 12 subsurface or ‘ground surface input parameters has been provided. After inclusion of the available climatic and runoff information, only four parameters that must be determined through calibration (porosity, precipitation correction fac: tors, temperature correction, and evapotranspiration coeffi cients). This 75% reduction in degrees of freedom ‘simplifies the calibration process. ‘The process followed isto calibrate the four remaining ‘unknowns via a sequence of four sets of iterations, Step 1. The effective porosity of the geologic column is modified to produce the correct shape of the annual hhydrograph. In this study. it is incorporated in the storage terms of the soil, and unsaturated and saturated zones (Table 3), Low porosity produces a lower storage capacity and flashier runoff, and makes the peak spring flow much Ihigher tan fall and winter low ows, Step 2. The precipitation input is raised or lowered by correction factors until the simulated total stream discharge approaches the observed value, Separate corrections ean be used for snow and rain, primarily t overcome inaccuracies that may result from using weather stations outside the watershed, Step 3. Temperatures for December through February are raised slightly, and March may be reduced, to better ‘match the observed hydrograph. Both insolation on artifi- cial surfaces and road salting tend to make snow melt faster in the winter than would be dictated by ambient air tem- perature. In addition, discontinuity of snow cover or frozen ground can lead to earlier melting or greater infiltration (Chetkauer 2001) than PRMS can simulate. Without this adjustment, PRMS tends 10 generate too Title base flow and. runoff in the winter, and too much in the spring as the excess snow melts, Step 4. Evapotranspiration coefficients are modified (primarily in the fall and in March) to fine-tune the simu- lated hydrograph, The viability of the hygrograph's cal bration has been assessed by visual comparison and by the ‘goodness-of-fit measure for time series defined by Nash and Sutcliffe (1970). PRMS has been calibrated on the seven study water- sheds in an effort to define the likely range of ground water recharges in the region, The watersheds incorporate the full range of hydrogeologic, climatic, topographic, and land ‘cover conditions present (Tables 2 and 3). The calibration was done on a one-year time series. In this study, PRMS rnceded to be run for about three years of simulation before the calibration year to allow stabilization of its responses, Ground water recharge has been defined 2s the inflow to the saturated zone, and bas been calculated as the sum of| the discharges from the saturated zone to the stream (base flow) and t0 deeper aquifers (Figure 2), Base Flow Separation Asa check on the validity of the base flows calculated by PRMS, manual base flow separation was conducted on both the composite hydrograph for the study period and individual year hydrographs for each watershed, The pro- cedure uscd is slightly modified from Linsley etal. (1958). (On a semilog plot ofthe hydrograph, a straight line is iden- tified passing through the discharge minima along the spring flow recession. The slope of this recession line is assumed constant throughout the entire year. The line is extrapolated backward to a critical time, t, after the pre- DS.Cherkauer GROUND WATER 42,n0.1:97-110 103 aaa] Discharge cm) 120180 240300340 ‘Dayetn Water Year 6 . “ A i at . if 3 5 bo & ee eel Es bP 44 i - ily ge 2 a ae | 2 ° Lt | Flt 6 SSE ee ° © mm me am 300 ° @ mm wm a0 ayn ne eer aye nite eae —Moaed = Observed —Motied = Obed ceding flow peak, where (= (drainage area)? (Linsley et Rorabaugh’s (1964) master recession curve and, in the al. 1958; Rutledge 1998). Similarly, itis extrapolated for- author's view, a separation that is Jess sensitive to user- ward to t, after the succeeding flow peak, The recession defined controls than some of the available automated sepa: segments are then connected at each critical time by verti- tation techniques (Sloto and Crouse 1996; Rutledge 1998), cal lines. For the year, the volume of water beneath the con- tinuous recession line is base flow. ‘The vertical connecting line is less subjective and pro duces a more conservative (lower) base flow than the pro- cedure originally described by Linsley et al. (1958). This is Watershed Scale similar to Bevans (1986), but uses a shorter cvtical time. Ineach watershed, PRMS could be calibrated to repro- The overall procedure produces a lower base flow than duce the composite hydrograph reasonably well (Figure 4). Results Table 5 Water Budget Components for the Study Watersheds! T ‘Hydrograph Otservatons RMS Caran _| seperation Ad [ Poh Qube | Qaim Ean? wr RO te RA | Com ak sz | mse [253 oom mr as Cetet asf msm easy a2 sks ge | nya Potency || rise | ac [esis | ile 0K) ee sal ora | Oe een 2 ell Oni {as M9 i et Se ror | ate | sta im 0 most aso Tonle so fsa or oy os) 07? 8s | 3a Mon | or | mas dst os tos Ass | al Qa = cero pith septs atc nnn yer " ve, = [Qaim ota ~21Quin = Qobo? V/s = Qube} whete Cha nd Qos ae te ily saa a nerve dsharg ad Quis he anna avergs 104 D'S Cherkaver GROUND WATER 42, no. 1297-110 © mw 2308380 Days n Water Year — Madoled > Observed Figure 8. Simulation of water year 1982 for the Bark River. Parameters from the calibration to the composite hydro: graph were used unchanged. ‘The coefficient of model-fit efficiency ranged between 0.57 and 0.78 forall but the Menomonee River (Table 5). ini eating the model accounts for ~60% (0 80% of the daily variation in observed discharge. The relative weaknesses in the Menomonee watershed are because itis urban and the model’s response during the spring thaw is extremely flashy. The reasonable success of the calibrations immedi- ately demonstrates two things. First, the mechanism for defining surface and subsurface input terms presented in Tables 2 and 3 works. Second, using PRMS at a scale where the HRUs average 37 km? is viable. PRMS tends to underestimate winter discharge and then overestimate peak spring flows when compared tothe composite hydrographs (Figure 4) due to the smoothing inherent in the composite hydrograph. The spring flood ‘occurs at different times cach individual year, but these are averaged into a single event which is longer but with a lower peak discharge than any given year. When the model-fit efficiency coefficient is recalculated without the months of January rough March, the values rise (Table $), indicating. the simulations (including the Menomonee) account for 55% to 89% of the observed variation, consid- cred to be an excellent ft ‘The Bark River simulation is probably the weakest of the study watersheds during the winter/spring transition (Figure 4). When the model is rerun for an individual year without any recalibration, the hydrograph fit during this period is visibly much better (Figure 5). In addition, the ‘model calculates a base flow of 16.3 em, for water year 1982, and manual hydrograph separation results in 16.5 em, a difference of 1.2%. This test demonstrates that the use of hydrographs for individual years would improve the el ciency coefficients; however, because the objective of this study was to simulate the spatial distribution of recharge (not temporal), the composite hydrograph was used. Its smoothing effet was simulated by slightly increasing tem- peratures in the winter and lowering them in March, as explained earlier. Note that for the individual year (Figure 5), simulated flow peaks precede the observed by about five days (Figure 5). This isthe result of the model delivering all flow to the. ‘outlet instantaneously. Using the composite hydrograph also smooths out this disparity Recharges calculated atthe watershed scale range fom 3.5 wo 182 em/r (Table 5). In the modeled watersheds, recharge averages ~11.79% of precipitation, while 10.2% runs off directly, 49.2% evapotranspires, and another 10.3% infiltrates only to the unsaturated zone before dis- charging as interlow (Table 5). The remainder is inter- cepted and returned to the atmosphere before distribution ‘within the land portion of the hydrologic eyele To test the validity of the recharges obiained from PRMS, manual base flow separations were done on both the composite and individual year hydrographs. PRMS recharges are generally a litle smaller than the base lows from individual years (Figure 6, Table 5), but the average difference is only 14%. PRMS always underestimated the recharge derived from the separation of the composite hhydrograph (Table 5). The latter's smoothing effect blurs the distinction between interflow and base flow compo- nents, For individual years when the spring peak is small, the interflow component gets included in the composite separation, but when the peak flow is high it ges excluded Hence, a better test is fo compare the PRMS base flow and half of the interflow to the base flow separated from the ‘composite graph When this is done, the two methods show ‘ery similar values Figure 6), with an average difference of only 5%. In short, PRMS and the calibration process used in ths study produce accurate estimates of ground water recharge at the watershed scale. Generally, a8 the simulated recharge in a watershed imereases (tothe lft on Figure 7), both interflow and runoft decrease. Figure 7 also shows the relation of watershed scale recharges to the primary controlling hydrogeotogic factors in the PRMS inputs develope previously. I has & ‘rong direct relation with sol and geologic conductivities and porosity. Ground surface slope, land cover, and depth to the water table show weaker relations tothe simulated recharges. Subwatershed Scale ‘One advantage to PRMS is thatit calculates u recharge for each of the 63 HRUs in the study watersheds. The over~ all distribution (Figure 8) ranges from 1.9 to 20.7 emvyr With a mean of 10.7. In general, recharges are highest in the ‘western Watersheds (Bark, Fox, Turtle) and lowest in those along Lake Michigan (Oak, Root). Glacial sediments in the former watersheds include considerable outwash and ice contact deposits, so soil conductivities are high. Clay-tich ‘deposits with much lower conductivities underlie the east- em watersheds. These latter areas occupy the low end of Figure 8 and exhibit much higher interflows (Table 5). Regression statistics were calculated forthe correlation of recharges in the HRUs to various controlling parameters (Table 6). As atthe watershed scale, the strongest relation is a direct one between recharge and soil conductivity Aquifer conductivity and porosity have weaker direct rela- tions, because their variation among the HRUs in a water- shed has been limited in the calibration process. Depth to the water table has the second strongest relation, As it increases, the opportunity for both evapotranspiration and interflow discharge from the unsaturated zone increases, DS. Cherkauer GROUND WATER 42,n0,1:97-110 105 S20 Ew | ge | Eu g? a fe Be | i capamaralees 3 it =] pile Pras recharge mun) = conse Inde rarest: PDI a Figure 6. Comparison of recharges calculated by PRMS and by hydrograph separation. Separations have been done on Doth the Tong-term composite hydrographs and those for Individual water years within the study period. with a resultant decrease in recharge to the saturated zone. ‘The poor correlation to both the length of overland flow and slope is because they have only limited variation at the THRU scale, ise | | ‘Ma Bar Tur Cad Men Fox Roo Oak [Porosity [llsione ZA Depthtowr! Response of Model to Calibration Steps For the Root River, the steps of calibration were repeated by sequentially inputting the final values for each ‘change to PRMS, and then observing the overall effect it had on the hydrograph. The starting point has individual values for topography, land cover, and hydrogeologic para meters within each HRU, except that porosity was set at & regional average of 8%, No adjustments had been made to raw precipitation or temperature inputs, and the evapotran- spiration multiplier was held constant throughout the year ‘The result was a poor simulation of the hydrograph (Fig- ure 9 and Table 7), withthe peak discharge much too high and a flow recession slope that is much too gentle for the remainder of the year. Calibration of porosity to 2% vastly alters the shape of the hydrograph, producing a recession slope very close to that observed (Figure 9), This step had the single greatest ‘effect on the model fit when either the whole year or just the nonwinter period was considered (Table 7). Steps 3 through 5 in the calibration process involve sequentially altering the climatic inputs to the model. Their effects were generally individually indistinguishable, but important col- Teetively (Table 7). Refining the spatial precipitation ‘weighting, by altering the annual precipitation multipliers Bi Avvicuitura Developed Figure 7. Relation of watershed scale recharge to other water budget components and independent input parameters. In each graph, watersheds are arranged from highest recharge on the left to lowest on the right. Parameters with strong control on. recharge show clear change in helght from left to right. 106 DS. Cherkaver GROUND WATER 42, 0, 1:97-110 ‘Table 6 Correlation of Recharge in Inc Correlation Parameter Coefficient (r) oil conductivity (KS) ost Depth to water table (D,.) 0.62 Conductivity of geology (Ke) ss Porosity (n} oss Natural land cover (Nat) oa Developed land cover (Dev) 037 Length overtand flow (Lot) 09 Surface slope ($) 0.08 from 1.0 10 0:95, and 1,02 for rain and snow, respectively, resulted ina better water budget match, Temperatures were anificially raised about 1°C in December to February and reduced about the same in March to better match winter low flows and the spring melt Finally, the monthly evapotran- spiration coefficients were lowered in March and raised luring the growing season to reflect seasonal variation in plant activity. Together these climatic variable adjustments ‘dramatically improved the calibration in the January through March period, increasing the model fit coefficient by five times for the whole year, but only slightly in the April through December period (Table 7). ‘The final simulated hydrograph is still not a perfect fit with the observed (Figure 9). It does, however, match the post spring recession very well. This is the period during which the bulk of recharge occurs, Extrapolation of Results to Other Watersheds One limitation of PRMS is that it can only be eali- brated on gauged watersheds, but once several calibrations have been done, how effectively could their results be used to generate simulations on other watersheds? AS a test, two additional gauged watersheds were simulated, the north and west branches of the Milwaukee River, at the northern extreme of Figure 1 period of 13 years, and drain 379 and 388 ky They were actively gauged for a respec Frequency (%) was Recharge Rate mye) Figure 8 Frequency distribution of HRU scale recharge rates calculated by PRMS. charg for ET & Observed fem) charge for Start & Port cms) Figure 9. Progressive changes in the simulated hydrugraph of Root River during the callbration process. Steps and 2 {start and porosity) use the left axis, The others use the right tively. PRMS was provided withthe actual preipitations and temperature for the watersheds and the GIS-derived variables forthe HRUs. All the calibrated climatic parame- ters were input as the averages resulting from the calibra tion of the seven study watersheds. A porosity of 9% was input, because the local geology is elosest 10 the Fox Watershed, PRMS was then run just once on each water shed to generate & hydeograph. Figure 10 shows the com. patison of those uncalibrated Simulations to the observed hydrographs. The extrapolation reproduces the total observed streamflow on both watersheds 10 within 9% and recharge to within 10% ofthat obtained from separation ofthe com- posite hydrograph (Table 8). Both of these emors ate slightly larger than the averages forthe calibrated water: sheds (1.9% und 5%, respectively), but are deemed accept able. The extrapolation process thus appears 10 be a legitimate way 10 calculate recharges in ungauged water sheds. Two cautions are in order. Firs al the study and test watersheds are smaller than 500 km? and were simulated With HRUs averaging 35 co 45 kn’. No testing on larger watersheds has been done, PRMS is limited to 3 maxiraum of 50 HRUSs,s0 itis anticipated there could be problems in watersheds larger than ~2000 km?, Second, no testing on extrapolation outside the region where the calibrations ‘were done has been conducted Table 7 Sensitivity of Hydrograph Fit to Individual Calibration Steps in Root River Watershed ‘Api erg Sep (RU perio GIs) DN Skp2icfisane porty) ONES LA Sep 3 orphan pio) 9206, Sep tinimertenpertus) O20 12 OM DS.Chereaver GROUND WATER 42,n0.1:97-110 107 in . | om im oH ‘Days Water Vear — Modeled = Observed Figure 10, Comparison of simulated and observed hydro- graphs on two uncalibrated watersheds. These to lie at the far northern edge ofthe study area in Figure 1, Conclusions Management of ground water resoures for human use requires knowledge of recharge rates They must be known across aquifer or regional scale areas, but the method of obtaining them must also be able to resolve smaller scale spatial variation, and be testable for accuracy. I also needs 10 use data that are readily available (eg, soils, topogra- phy) rather than those that frequently ae not (such as iso topic compoxition or water budget uxes). To date, the best ‘method availuble to meet all daese criteria has been the use ‘of the distributed parameter model PRMS (Stever and Hunt 2001). This model is data intensive, requiring site local scale input for dozens of system properties and had only been applied atthe local scale, In this paper, the model has ‘been applied at scales from local to regional. A procedure has been presented! to define all topographic, hydrogeo- logic, climatic, and land cover inputs. Precipitation, tem- perature, and streamflow have been input as daily averages ‘over 30 years of record, Ground surface and subsurface parameters have been defined as functions of commonly available GIS and hydrogeological data, Four input vari- ables (porosity, evapotranspiration coefficients, and minor adjustments to precipitation and temperature) remain unde- fined and are obtained by calibrating PRMS to the time- averaged annual streamflow hydrograph. ‘The procedure has vastly streamlined the PRMS cali bration process by fixing the values of 12 parameters pre- viously free to vary during calibration. It has been calibrated for seven study watersheds < 500 km? in size. Each has been subdivided into subwatersheds with rela tively homogeneous soil conductivity, land cover, and/or ‘around slope. The calibration process simulated. hydro- ‘eraphs that matched observed annual streamflow to within 1.9% and which account for 55% to 89% of the temporal variation of discharge. The ground water recharges calcu~ luted by PRMS reproduce those obtained by hydrograph separation for each full watershed to within 59% to 15%, meaning they are accurate. At the subwatershed scale, recharge averages 11 em/yr and ranges between 2 and 21 cemvyr. Soil conductivity, porosity, and depth to the water lable are the parameters with the strongest control on the calculated recharge The calibrated parameters for the study watersheds have been used as input to PRMS to simulate other sites. Without any further calibration, PRMS reproduced total streamflow and recharge to within 9% and 10%, respec- Lively. This indicates that PRMS could be used to accu- rately determine recharge rates in ungauged watersheds within the same region as the calibrated ones, This study hhas not, however, tested extension of the model's calibra- tion beyond the study area or to watersheds > 500 kin? (throne easing nod ae Table 8 Results of Extrapolating PRMS to Uncalibrated Watersheds “Total Streamflow PRMS Results Recharge Measures ‘Compos Bo + Test Qobs Qaim Difference | Io Bo HG sep Mlo’2 Difference Watershed | (em/year) (emiyear) —(4)* |(envyear) (emiyear)® Ban Ept | (emiyearyYemiyenr)* (2) N.Brach | 228 241-453 | G76, -0265. O73) 1541S =S8 W.Branch | 214233489 | 782 925 40208 07%] 1 132100 {CompontGtp te fam tise ogee degen “Rertg ie nk ine coarse thn omy yoga, 108 D'S Cherkaver GROUND WATER 42, no, 1:97-110 ‘The three questions this study sought to answer ean all be answered unequivocally yes. PRMS can provide a demonstrably accurate measure of recharge in watersheds between 60 and 500 km? and ean calculate them at multi ple scales. Input from GIS substantially simplifies its cai bration by uniquely defining the majority of input parameters atthe ground surface and inthe subsurface. The calibration results and GIS-derived inputs have been used to extend the model to uncalibrated watersheds, which ‘means extrapolation (0 ungauged watersheds. will also work. The success of the calibrations confirms the validity of the formulations presented calculate 12 surface and subsurface input variables (Tables 2 and 3). They have worked throughout the nine watersheds examined ‘The method presented to determine recharge thus meets the specific needs of grounel water resource planners Although many other methods to estimate recharge exist it has been the premise of this paper that those other methods fofien cannot meet planners’ needs for demonstrably accu rate recharge values resolved to varied scales and using readily available information sources. ‘Acknowledgments ‘This work was partially funded by both the University ‘of Wisconsin System Ground Water Research Program and the Wisconsin Department of Natural Resources Source ‘Water assessment Program. | want to thank D. Feinstein, J Krohelski, C, LaCosse, P. Lenaiker, J. Steuer, and H, Yantz {or their contributions to the conceptualizaton, implemen tation, and interpretation of this study. Laura Toran and an ‘anonymous reviewer made suggestions that substantially improved the manuscript References ‘Adb El Samie,S., and M. Sodek, 2001, Ground water recharge and flow inthe lower Cretaceous Nubian sandstone aquifer inthe Sinai Peninsula, using isotopic techniques a BYdto~ chemistry. Hydrogeology Journal 9, no, 4: 78-389, Ansari, S.A, 1999, Determination ofthe magnitude of and com: trols ground-water recharge. Unpublished M.S. thes Department of Geosciences, University of Wisconsine Milwaukee, Amold, J, and P. Allen, 1999. Automated methods for estimat ing baseflow and ground water recharge from streamflow records, Journal of the American Water Resources Assucta tio 35, 90.2: 411-424 Amold, J.P. Ailea, R. Muth, and G. Bernhard, 1995, Auto mated base flow separation and recession analysis oe niques. Ground Water 33, 90.6: 1010-1018, Amold,T., and M. Friedel 2000, fects of land use on recharge potenial of surficial and shallow bedrock aquifers in the Upper Mlinois Basin. U.S. Geological Survey Water Reseurees Investigations Report. 00-4027, Avery, W., J. Donovan, and N. Ketchom, 1999, Recharge esti mation by stage-diseharge interpretation trom exoss-coere- lated well measurements, Ground Water 37, no. 3: 332-337. ‘Avon, Land T. Durbin. 1954. Evaluation of the Maxe ‘ethod for estimating recharge to ground water basins in Nevada. Water Resources Bulletin 30,n0, 1: 99-111 Bares, B.S. 1929. The structure of discharge recession curves, In Transactions of the American Geophysical Union 20, m5, Bevan, H. 1986, Estimating stea areas of eastern Kansas auiter interactions in col we sivearaflow records, in Selected Papers in the Hydrologic Sclences, ed. S. Suita, 51-64. US. Geological Survey Water Supply Paper 2290, s, B.C. Behroozi 1. Schooley, and J, Hoffman, 1993, A, ‘method for evaluating ground-water recharge areas in New Jersey. New Jersey Geological Survey Report GSR-32, (Cherkauer, D. In review. Estimating round water recharge using opography. hydrogeology, land cover, and precipitation Ground Waver Cherkaver, D., and V. Bacon, 1978, Is there a ground-water shortage in southeastern Wisconsin? SE Wisconsin Regional Planning Commission Tectocal Record 4, no. 1: 19-30. (Cherkaver, D., and P. MeKereghan, 1991, Ground water dis- charge to lakes: Focusing in embayments. Ground Water 29, no. 1: 72-80. CCherkauer, K. 2001. Understanding the hydrologic ec of frozen soil Unpublished, Ph.D. dissension, Departmen of Civil and Environmental Engineering, Univesity of Washington, Cresswell. RJ. Wischuson, G: Jacobson, and K. Fifield. 1999, “Assessment of recharge (0 ground water systems inthe arid southwest part of Notthem Teritory, Austalia, using chlo rine-36. Hdrogeology Journal 7, 6.4: 393-404 Daniels, D. 8. Fritz, and D. Leap, 1991. Estimating recharge rates through unsaturated glacial till by itive tracing, Ground Water 9, 90, |: 26-34 Deli, G.,R, Healy. M. Landon, and J. Bobike. 2000. Effects of topography and sol properties in techarge at two sites in an agricultural field. Journal ofthe American Water Resources Association 36, 00.6: 1401-1416, Dettinger, M. 1989. Reconnaissance estimates of natural recharges to desert basins in Nevada, USA, by using cho Fide baianee caleuations Journal of Hydrology 106, 38-78, Dinieols, R. 1997, Estimates of recharge from runoff atthe Han ford site, Washington. U.S. Geological Survey Water Resources Investigation Report 97-4038, Eaton, T, and. A. Zaporozee. 2002. Contamination potential in southeastern Wisconsin. In Ground Water Resources of Southeastern Wisconsin, SE: Wisconsin Regional Planing ‘Commission Technical Report 37. Flynn, K.P. Hummel, A, Lamb, and J Kittle, 1995, User's man- ual for ANNIE, version 2, a computer program for interac- tive hydrologic data management, U.S. Geological Survey ‘Water Resourees Investization Report 95-4085, Gerhart J., and G. Lazorchick. 1988. Evaluation of the ground ‘water resources of the lower Susquehanna River basin, Pennsylvania and Maryland. U.S. Geological Survey Water Supply Paper 2284 Goes, B, 1999, Estimate of shallow ground water recharge inthe Halejia-Ngura Wetlands, semi-arid northeast Niger, Hydrogeology Journal 7, no, 8: 294-303, Halford, K., and G. Meyer. 2000. Problems associated with esti- rating ground-water discharge and recharge from stream discharge records. Ground Water 39,90, 3: 831-32 Hamon, W. 1961. Estimating potential evapotrunspiration. In Proceedings of American Society of Ciil Engineers, Jour- nal of the Hydraulics Division 87,10, HY3: 107-120. Holisehkig. D. 1997. A generalized estimate of ground-water recharge rates in the Lower Peninsula of Michigan. U.S, Geological Survey Water Supply Paper 2437 Hoos, A. 1090, Recharge rates and aquifer hydraulic charters Ties for selected drainages in middle and eastern Tennessee. USS. Geological Survey Water Resource Investigation Report 91-4015, Hunt, RJ. Steve, J. Manson, and T, Bullen, 2001, Delineating a recharge area for a spring using numerical modeling, Monte Carlo techniques, and geochemical investigations. Ground Water 39, 90. : 102-112. Jones, 1. J- Banner, andl J. Humphrey. 2000, Estimating recharge ina tropical karst aguier. Water Resources Reseurch 36, 0. 5: 1289-1300. Ketchum, JJ. Donovan, and W. Avery, 2000, Recharge chart teristics of phrcatic aquifer as determined by storage aceu- ulation, Hydrogeology Journal 8,00, 6: 879-593, Cha DS. Cherkauer GROUND WATER 42,10. 1:97-110 108 Leavesley, G., R, Lichty, B. Troutman, and L. Saindon. 1983, Preciitaion-runoff modeling systen)—User's manual, US, Geological Survey Water Resources Investigation Report 83-4237. Linsley, R,, M, Kohler, and 1 Pauthus, 1958. Hydrology for Engi- neers. New York: MeGras-Hil, Liu, ¥., aad C, Zhang, 1993. A comparative study of calculation mthods for recharge of rainfall seepage to ground water in plain area (China). Ground Water 31, no. 1:1 Louie, M., P. Shelby, J. Smessud, L, Gatchell, and J. Selker. 2000 Field evaluation of pussve capillary samplers forest ‘mating ground water echarge. Water Resources Research 136, no 9: 2407-2416, Lumb, ‘A.J. Kite, and K. Flynn. 1990. Usee's manual for ‘ANNIE, a computer program for inteructive hydrologic analyses and data management, US. Geological Survey ‘Water Resources Investigations Report 89-4080, Mau, D, and T, Winter. 1997, Estimating ground water recharge ‘rom streamflow hydrographs for a small mountain water shed in a temperate humid climate, New Hampshire Ground Water 35, no. 2: 291-304. MeDonald, M, and A. Harbaugh. 1988, A modula, three-dimen- sional finite-difference ground-water flow model. US, Geo- Togeal Survey Techniques of Water Resources Investigations Book 6, Chaper A Meyboom, P. 1961, Estimating ground water recharge from stream hydrographs, Journal of Geophysical Research 66, 1203-1214, Nash, J. and J. Sutcliffe, 1970, River flow forecasting through | ‘conceptual models, Par I: A discussion of principles. Jour- zal of Hydrology 10, 282-200. COstercamp, W., L- Lane, and C. Savard, 1994. Recharge esti- ‘males using @ geomorphieldiseibuted-parameter simulation approach, Amangosa River Basin. Wazer Resources Bulletin 30, no. 3493-807. (Ono, R. 2001. Estimating ground water recharge rates in the ‘southeast Holstein region, northem Germany, Hydrogeol- ‘gy Jounal 9, no. :498-S11 Rorabaugh, M.l. 1964, Estimating changes in bank storage and ‘groundwater eoncibution to sireamflow,Fnternational ASSo- lation of Seiemfic Hydrology Publication 3, 432-441. Rosen, MJ. Bright, P. Carron, M. Stewart and R, Reeves. 1999. Estimating rainfall echarge and soil moisture residence times in Pukekahe, New Zealand, by combining geophysics, chemistry, and isotopic methods. Ground Water 37, no. 6: 536-84, Rutledge, A. 1998. Computer programs for describing the reces- sion of ground-water discharge for estimating mean ground ‘water recharge and discharge from streamflow reconls: Update, U:S. Geological Survey Water Resources Investi- ‘gation Report 98-418, 110 _DS.Cherkauer GROUND WATER 42, n0.1:97-110 Rutledge, A., and C. Daniel, 1994, Testing an automated method to estimate ground water eecharge from streamflow recon Ground Water 32, no. 2: 180-189. Scanlon, B.,R. Healy, aad P. Cook. 2002. Choosing appropriate techniques for quantifying ground-water recharge. Hiro geology Journal 10, no. 1: 18-39. Sloto, R, and M, Crouse. 1996, HYSEP: A computer program for steamflow hydrograph separation and analysis. US. Geological Survey Water Resources Investigation Report 96-4080. Solomon, D.,and E.Sudicky. 1991. Tritium and helium 3 isotope ratios for direct estimation of spatial variation in ground water recharge. Water Resources Research 27, 90, 9: 2309-2319. Sophocleous, M., and J. MeAllister. 1987. Basinwide water-bal- lance modeling with emphasis on spatial distribution of ‘round water recharge. Water Resources Bulletin 23, n0. 6 597-1010. Sophecleous, M., and S. Perkins. 2000. Methodology and appli- cation of combined watershed and ground-water models in Kansas. Joumal of Hydrology 236, 185-201 Stover, J, and R. Hunt. 2001. Use of a watershed-modeling ‘approach to assess hydrologic effects of urbanization, North Fork Pheasant Branch Basin near Middleton, Wisconsin, USS. Geological Survey Water Resources Investigation Report 01-113, Stoenz, M., and K. Bradbury. 1989, Mapping recharge areas using @ ground water flow model: A case study. Ground Water 27, no, 2: 220-229. ‘Swanson, $. 1996, A comparison of two methods used to esti ‘mate ground water recharge in Dane County, Wisconsin, Unpublished M.S. thesis, Department of Geology and Geo- physics, Univesity of Wisconsin-Madison, ‘Ting, C.,T- Keck, and C. Lig. 1998, Estimation of ground water ‘echarge using the chloride mass-balance method, Pingtung Plain, Taiwan. Hydrogeology Journal 6, no. 2: 282-292, ‘Walton, W. 1965. Ground water recharge and runoft in Iino, Mlinois State Water Survey Report of Investigation 48 ‘Winograd, A. Riggs, and. Coplan, 1998. The relative comti- ‘butions of summer and coo-season precipitation to ground water recharge, Spring Mountains, Nevada. Hydrogeology Journal 6, no. {2 77-98, ‘Wood, W., and W. Sanford. 1995. Chemical and isotopic meth- ‘ods for quantifying ground water recharge in a regional semiarid environment. Ground Water 33, no. 3: 458-468. ‘Zhu, C. 2000. Estimate of recharge from radiocarbon dating of ‘ground water and numerical flow and transient modeling Water Resources Research 36, no. 9: 2607-2620.

Das könnte Ihnen auch gefallen