Environ Geol (2006) 51: 73–82 DOI 10.1007/s00254-006-0305-2 ORIGINAL ARTICLE Cheng-Haw Lee Estimation of groundwater recharge using Wei-Ping Chen Ru-Huang Lee water balance coupled with base-flow-record estimation and stable-base-flow analysis Abstract In this paper, the long- complex hydrogeologic modeling or Received: 14 February 2006 Accepted: 12 April 2006 term mean annual groundwater re- detailed knowledge of soil charac- Published online: 11 May 2006 charge of Taiwan is estimated with teristics, vegetation cover, or land- Ó Springer-Verlag 2006 the help of a water-balance ap- use practices. Contours of the proach coupled with the base-flow- resulting long-term mean annual P, record estimation and stable-base- BFI, runoff, groundwater recharge, flow analysis. Long-term mean an- and recharge rates fields are well nual groundwater recharge was de- matched with the topographical rived by determining the product of distribution of Taiwan, which estimated long-term mean annual extends from mountain range runoff (the difference between pre- toward the alluvial plains of the C.-H. Lee (&) Æ W.-P. Chen Department of Resources Engineering, cipitation and evapotranspiration) island. The total groundwater National Cheng Kung University, and the base-flow index (BFI). The recharge of Taiwan obtained by the Tainan, Taiwan BFI was calculated from daily employed method is about 18 billion E-mail: [email protected] streamflow data obtained from tons per year. Tel.: +886-6-2757575 Fax: +886-6-2380421 streamflow gauging stations in Tai- wan. Mapping was achieved by Keywords Groundwater recharge Æ R.-H. Lee using geographic information sys- Water balance Æ Base-flow-record Hydrology Division, Water Resources Agency, Ministry of Economic Affairs, tems (GIS) and geostatistics. The estimation Æ Stable-base-flow Taipei, Taiwan presented approach does not require analysis Æ Base-flow index Introduction Currently, standard techniques of estimating regional recharge most often involve (1) applying a water-bal- Estimating groundwater recharge is an important issue in ance model, where the moisture content of the soil is hydrogeologic studies. In most cases, recharge is esti- tracked through time (Finch 1998; Simmons and Meyer mated by multiplying the magnitude of water-level fluc- 2000; Chen et al. 2005), or (2) parameter-value adjust- tuations in wells by the specific yield of the aquifer ment of groundwater flow models (Lee et al. 2000; material or by applying the water budget model or using Jyrkama et al. 2002; McDonald and Harbaugh 2003). the water-balance method. While other parts of the wa- Application of the first approach, while generally less ter-balance equation, such as precipitation and runoff, intensive computationally, requires knowledge of the are relatively easy to measure, recharge remains an elu- vegetation and soil types within the study area, in sive process to quantify. This is especially so because it addition to a number of basic meteorological variables depends not only on precipitation but also on meteoro- such as air temperature and precipitation. The second logical conditions, as well as on soil type, soil–moisture approach is more taxing of computer resources because status, vegetation cover and condition, slope, cultivation a potentially complex groundwater flow model may practices, and most of all, on evapotranspiration, which have to be run repeatedly in search of a multidimen- is a function of the previously noted factors. sional parameter-value optimum. 74 With the purpose of inspecting recharge, estimating oration is negligible, i.e., for areas where the water table the groundwater component of streamflow has been a is not so close to the surface that the vegetation can use research focus for more than a century. Following the it through its root system. The approach combines the work of Boussinesq (1877), numerous studies (Bevans water-balance model, base-flow-record estimation, and 1986; Moore 1992; Rutledge 1992; Rutledge and Daniel stable-base-flow analysis. It is computationally simple, 1994; Mau and Winter 1997; Chen and Lee 2003) have requires minimal optimization, and does not need investigated the recession of streamflow, particularly information on vegetation and soil types. The technique baseflow, and have estimated the contribution of is mainly a collection of existing methods which, to the groundwater to streamflow. In some cases, the value of best knowledge of the authors, have not yet been com- baseflow is assumed to be equal to groundwater re- bined in a similar fashion for recharge estimation. It is charge. The primary purpose of most researches is to expected to be most practical for regional-scale studies determine the groundwater component of streamflow. where the long-term mean annual value of the spatially Nevertheless, only a handful of researchers, including variable recharge is of interest. The approach was Meyboom (1961), Rorabaugh (1964), and Rutledge applied using data from Taiwan to demonstrate the (1992), have focused on groundwater recharge through utility of the technique. analyzing the streamflow data. Rutledge (2005) further summarizes constraints involved with the application of the Rorabaugh model for estimating groundwater Methodology recharge. Mau and Winter (1997) have provided the instantaneous recharge method and the constant The water balance of a geographic region can, in gen- recharge method of hydrograph analysis to estimate eral, be written as recharge. Although several methods have been used to estimate P ¼ ET þ qs þ qb þ qN þ DS; ð1Þ the groundwater discharge and recharge from stream- where P is the precipitation (LT)1); ET is the evapo- flow records, the most commonly used are the tech- )1 )1 niques of baseflow separation. These methods aim at transpiration (LT ); qs is the surface runoff (LT ); qb is the groundwater contribution to runoff (LT)1), which is estimating a continuous or daily record of baseflow )1 under the streamflow hydrograph. In other words, it the definition of baseflow; qN is the net flux (LT )of requires an extended period of recording efforts in esti- any water entering or leaving the region other than mating the long-term groundwater discharge, as well as precipitation (e.g., water diversions, groundwater flux across the basin boundaries, and irrigation); and DS is the exercise of a variety of manual methods (Horton )1 1933; Barnes 1939; Olmsted and Hely 1962; Dzhamalov the change in stored water (LT ) within the area. 1973; Zektser 1977) or a rapid analysis and that intro- Generally, evapotranspiration is by far the largest loss duces some elements of subjectivity in the research for term in Eq. 1, amounting to 70% of precipitation the base-flow-record estimation (Rutledge 1992; Mau (including evaporation from open water surfaces) on a and Winter 1997). One study employed a water-balance global basis (Brutsaert 1982). Long-term ET measure- approach and digital filter method to estimate base ments are practically nonexistent, and the available ET recharge to groundwater in Nebraska (Szilagyi et al. estimation methods may differ by as much as 10–20% 2003). on an annual basis (Vorosmarty et al. 1998). In light of To increase the speed of analysis and reduce the these uncertainties, the general assumption that DS is subjectivity inherent in manual analysis, Rutledge (1993) negligible in most cases on a long-term basis may be well proposes several computer programs: RECESS, RORA, justified. For our purposes, this assumption is employed, and PART, and newer versions have been proposed acknowledging that for some watersheds where (Rutledge 1998, 2000). The research of this paper is hydraulic heads have changed significantly in the past, it accomplished using an automated analysis procedure by may lead to biased recharge estimates. It is further as- the programs described above. sumed that qN in Eq. 1 can be neglected as well, at least To prevent overestimation caused by rainstorm on a regional scale. events, several studies (Rutledge 1993, 1998, 2000; With regard to the stated assumptions, Eq. 1 sim- Zektser 2002; Chen and Lee 2003) indicate that the plifies to baseflow in the dry season should be chosen to be the P À E ¼ qs þ qb ð2Þ average value of the year. For this purpose, the stable- base-flow analysis is developed in this study to obtain a which states that the difference between precipitation more reliable result. and ET emerges as surface runoff and baseflow. If the Based on our previous research (Chen and Lee 2003), change in the stored water volume is negligible, as was the proposed approach in this paper offers an estimate assumed, then on a long-term basis, baseflow must of total recharge for regions where groundwater evap- represent a lower bound to groundwater recharge within 75 a given watershed. By quantifying qb, one obtains an Construct three parallel 1-dimensional arrays: estimate of recharge, provided that the portion of the 1. Streamflow areal ET originating from the groundwater is negligible 2. Base flow when compared to the total ET of the watershed. 3. "ALLGW" Read a data file of daily mean streamflow, and assign values Flow as completely groundwater discharge (while the to Streamflow. Assign all values of ALLGW=0. surface runoff is negligible) can be based on the ante- cedent recession. Linsley et al. (1982) proposed the empirical relation that Locate all days that fit the antecedent recession requirement (see EXPLANATION). On these days, reassign ALLGW=* 0:2 N ¼ A : ð3Þ and assign Base flow=Streamflow. This relation gives the time base of surface runoff (N [d]) as a function of the drainage area (A) upstream from Locate each day when ALLGW=*. If it is followed by a daily decline a streamflow-gauging station, in square miles. The time of the log of streamflow exceeding 0.1, then reset ALLGW=0. base of surface runoff is the number of days after a peak in the hydrograph of streamflow while the component of Locate all day when ALLGW=0.