Climate change influences the variability of hydrologic responses (i.e.
, precipitation amounts and frequencies, groundwater recharge and discharge, evapotranspiration, runoff processes, etc.
) across the land surface [1
]. Many studies have reported the steadily increasing variability of hydrologic responses to have negative impacts on water resources management, such as increased severity and frequency of natural disasters [3
]. According to the fourth report of the Intergovernmental Panel on Climate Change (IPCC), it seems that climate change will aggravate water stress throughout the world, together with population growth, urbanization and changes in land use [5
]. Especially the variability of water availability due to climate change has made it more difficult to manage water resources efficiently at the field scale in regions, such as the Republic of Korea (ROK), where flow-duration coefficients (indicating the rate of runoff) are relatively higher. Thus, available groundwater resources provide a natural means to alleviate the effects of the highly variable availability of water in ROK. Furthermore, climate change has more long-term and extensive impacts on groundwater than direct runoff, because the spatio-temporal variability of direct runoff is strongly affected by precipitation. Therefore, sustainable groundwater management is needed for adaptation to climate change.
One of the highest priorities for sustainable groundwater management is to evaluate and appropriate the share of available groundwater that can be feasibly extracted. For this purpose, we need to estimate accurate amounts of groundwater recharge at the watershed scale. In particular, understanding the characteristics of baseflow could be the first step toward a better estimation of groundwater recharge. Great efforts in baseflow estimation using historical streamflow records are based on several approaches, such as the recession curve displacement (RCD) method [6
], the curve-fitting method [14
] and the water-table fluctuation method [16
]. Furthermore, the Hydrograph Separation Program (HYSEP) [20
], PART [10
], RORA [10
], BFLOW [11
], the Web-GIS-based Hydrograph Analysis Tool [21
] (WHAT) systems and tracer-based hydrograph separation methods [23
] have been employed to separate baseflow from streamflow. For example, Kim [25
] estimated groundwater recharge using the baseflow-recession curve and a rainfall-runoff model and made a comparison of the results from these two methods. Kim et al.
] and Bae and Kim [27
] estimated baseflow quantity and the groundwater recharge rate using the Natural Resources Conservation Service (NRCS)-Curve Number (CN) method and baseflow separation, respectively. Yang and Chi [28
] obtained high correlation coefficients between baseflow rates and groundwater-table elevations in the analysis using WHAT. Among these baseflow separation models, WHAT is the web-based tool that can separate baseflow from direct runoff by using the Local Minimum method, BFLOW filter [29
] and Eckhardt filter [30
]. Eckhardt [28
] proposed representative values of baseflow index (BFImax
) for various aquifers, but the use of BFImax
, the maximum of the long-term ratio of baseflow to total streamflow) values specific to regional situations is recommended instead. For this reason, a genetic algorithm-based BFImax analyzer was developed for providing the optimal BFImax parameter to obtain local watersheds and aquifer characteristics for the long-term separation of baseflow from streamflow [22
]. With antecedent soil moisture, groundwater storage, precipitation rates and amounts, the analysis of the recession curve plays an important role in understanding the baseflow characteristics at the watershed scale. For quantifying baseflow from hydrograph separation, the WHAT system [21
] adopting the BFLOW parameter [11
] has mainly been used. However, the digital filter [33
] is limited in its consideration of hydrologic watershed features, because the digital filter algorithm simply separates low-frequency signatures from the high-frequency signatures through the signal-analysis process.
The United States Geological Survey (USGS) RECESS model [10
], one of the most widely used baseflow-separation programs, can perform more accurate baseflow separations compared to HYSEP or BFLOW, by constructing a master-recession curve (MRC) from streamflow. Furthermore, the recession index (K
) derived from RECESS, which is the time required for the streamflow recession by one log cycle [10
], can be used to identify the recession characteristics of watersheds in many parts of the world. Many watersheds within ROK are usually small and steep, because about 70% of ROK is mountainous. These geological characteristics can contribute to a shortened travel time of flow from one point of a watershed to another. The RECESS model has been suggested for analyzing long-term recession characteristics that had runoff histories [10
]. The RECESS model is a long-standing method used to analyze recession in many other studies across the world, but its accessibility should be improved, while its applicability needs to be extended to include other operating systems and not only MS-DOS. One method to improve the applicability is a web-based system, which builds upon the field-scale RECESS model’s advantage for policymakers. Specifically, the web-based HYSEP [34
] and WHAT systems [21
] were developed and widely applied in various field-scale studies. The RECESS model provides the recession index (K
), which can be used to estimate the alpha factor, a parameter also used in the Soil and Water Assessment Tool (SWAT). Therefore, the applicability of RECESS also needs to be extended to web-based systems.
Typically, the RECESS baseflow separation requires observed streamflow data at gauge stations; the SWAT model [35
] can simulate land-atmosphere processes (i.e.
, streamflow, baseflow, etc.
) and has been used to predict hydrologic responses at ungauged watersheds. SWAT has an advantage when applied to large-scale ungauged watersheds by adapting auto-calibration tools, such as Parameter Solution (ParaSol) [36
], Sources of Uncertainty Global Assessment using Split Samples (SUNGLASSES) [37
], Shuffled Complex Evolution-University of Arizona (SCE-UA) [36
] and SWAT-calibration and uncertainty procedures (CUP) [39
]. For example, Park et al.
] and Lee et al.
] tested the applicability of SWAT to small watersheds and performed the estimation of runoff curve coefficients, respectively. Furthermore, Lee et al.
] and Jung et al.
] suggested parameter regionalization using SWAT at ungauged watersheds. If calibrated by observations of streamflow in a watershed at the gauged downstream, streamflow at the ungauged upstream can be predicted by the SWAT model. Currently, SWAT has adopted several auto-calibration tools for the training of SWAT parameters (e.g., curve number (CN), threshold depth of water in the shallow aquifer required for return flow to occur (GWQMN), baseflow recession factor (ALPHA_BF)). These calibration tools are designed to find the parameter sets satisfying specific given thresholds, from numerous parameter combinations, which may be physically meaningless. For this reason, it is difficult to consider the calibration characteristics of each SWAT parameter with the auto-calibration tools. Among the SWAT parameters, the alpha factor, which is strongly dependent on stream recession, can be distorted by the calibration processes of the various other parameters related to streamflow, including peak flow and low flow. In such cases, it is difficult to obtain accurate streamflow recession during SWAT calibrations. In this regard, recession information from the observed streamflow can contribute to the improvement of streamflow prediction using SWAT for baseflow separation in an ungauged watershed.
Accordingly, the objectives of this study are three-fold (Figure 1
): (1) to develop the web-based RECESS model to build up the advantages of the RECESS program; (2) to evaluate the recession in streamflow prediction using SWAT at a gauged watershed; and (3) to estimate baseflow at ungauged watersheds.
Primary objectives of this study. SWAT, Soil and Water Assessment Tool; WHAT, Web-GIS-based Hydrograph Analysis Tool.
Primary objectives of this study. SWAT, Soil and Water Assessment Tool; WHAT, Web-GIS-based Hydrograph Analysis Tool.
Groundwater is an important water resource to build up the capacity to adapt to climate change, and scientific information on groundwater characteristics can be used to implement sustainable river management. To promote the capacity of the RECESS program to analyze groundwater characteristics in practice, the web-based RECESS model was developed in this study. Like other gauge stations that provide climatic and hydrologic data, this model can provide information on the alpha factor (α). This study evaluated the alpha factor obtained from the web-based RECESS model as an input, not as a parameter in SWAT applications. To this purpose, the alpha factor derived from the web-based RECESS model was applied to the SWAT model for the separation of baseflow from streamflow. Specifically, we assessed the impacts of streamflow recession on the baseflow characteristics by treating the alpha factor derived from the web-based RECESS model as an input to the SWAT model.
The main conclusions obtained from this study can be summarized as follows.
The web-based RECESS model and SWAT-CUP produced alpha factors of 0.108 and 0.663, respectively. These alpha factors were estimated by using two different methods, which was reflected by the considerable differences between them, by the influences on accuracy with which streamflow could be predicted. This might indicate that SWAT-CUP has a limited ability to correctly simulate the characteristics of streamflow recession due to the weighted auto-calibration on the entire streamflow, insufficient observation and (consequently) the lack of a spatially representative distribution of streamflow data.
The alpha factor obtained from the web-based RECESS model was applied to the calibration of SWAT for streamflow recession periods. As a result, the web-based RECESS model produced good calibration results (NSE
: 0.82; PBIAS
: −12.2%). The application of the web-based RECESS alpha factor to the SWAT calibrations for streamflow recession periods (Scenario II) resulted in better predictions of streamflow recession (Figure 7
). Comparing the individual simulations using Scenarios I and II, Scenario II predicted the recession of low flow more accurately than Scenario II. Based on these findings, this study revealed a significant effect of recession on baseflow. This conclusion is consistent with previous studies that have found the recession to play a major role in baseflow separation [67
]. However, for better calibration, spatially distributed alpha factors and other parameters associated with groundwater that could affect SWAT simulations should be considered.
Initially, it was expected that baseflow would be mainly affected by rainfall and streamflow. However, our findings showed different results compared to those expected (see Section 3.3
). For two watersheds that were different in terms of land use, soil texture and topography, similar precipitations produced significant differences in baseflow.
This study showed that the ratio of baseflow to streamflow (B/S) affected the temporal baseflow distribution in ungauged watersheds. As B/S is higher, the fluctuation of the temporal baseflow distribution becomes lower.
Based on these findings, the application of the web-based RECESS alpha factor to auto-calibrations for the estimation of recession periods could improve streamflow prediction. Furthermore, the web-based RECESS model can provide easy access to recession information (alpha factor), contribute to extending the applicability of the original RECESS model, help increase its accessibility and increase the convenience with which hydrological modeling may be performed. We expect that the web-based RECESS model will be useful in the identification of the roles of streamflow and baseflow in integrated river basin management and sustainable watershed development.