Study On The Impact Of Baseflow On Watershed Water Balance

Computation of watershed water balance is the main component of water resourcesmanagement.Baseflow plays an important role in maintaining streamflow and watershedwater balance. Exploring the impact of baseflow on watershed water balance has significantpractical value for understanding interactions between surface water and groundwater.Seventeen gauging stations across Michigan were chosen for this study. Seventeen watershedswere delineated to summarize various watershed characteristics using ArcGIS. Baseflow wasestimated from daily streamflow records (1967–2011) using the two-parameter recursivedigital filter method (i.e., Eckhardt method) forbaseflow separation of the Web-basedHydrograph Analysis Tool (WHAT) program. The statistical analysis software (SAS) wasused to develop multiple regression models for baseflow and baseflow index (BFI).Meanannual and annual water balances of the17watersheds were evaluated by comparingobserved streamflow with simulated streamflow estimated using Fu’s equation, which isbased on the Budyko Hypothesis. The Budyko Hypothesis describes mean annual waterbalance as a function of available water and energy. Impacts of baseflow on mean annual andannual water balances were also investigated with Fu’s equation.Main results are shown asfollows:(1) Observed average annual baseflow ranged from162to345mm, and BFI varied from0.45to0.80during1967–2011period. The average BFI value during the study period was0.71, suggesting that about70%of long-term streamflow in the study watersheds is likelyderived from baseflow contribution. Three equations (two for annual baseflow and one forBFI estimation) were developed and validated. The regression models estimated baseflow andBFI with relative error (RE) varying from29%to48%and from14%to19%, respectively.In absence of reliable information to determine groundwater discharge in streams and rivers,these equations can be used to estimate annual baseflow and BFI in Michigan.(2) Observed streamflow ranged from237to529mm/yr with an average of363mm inthe study watersheds during1967–2011. The performance of Fu’s equation in estimatingmean annual streamflow resulted in a Nash-Sutcliffe efficiency (ENS) value of0.07and RootMean Square Error (RMSE) value of64.1mm/yr.The17study watersheds were categorized into3groups based on different w(w is a parameter in Fu’s equation) values. Correspondingwvalues were2.40,1.83and2.05. Streamflow coefficient during1967–2011periodranged from0.29to0.34,0.36to0.38and0.41to0.51, respectively. Average streamflow coefficient was0.42, indicating that about40%of long-term precipitation in the study watersheds wasconverted into streamflow on average.(3) Climate sensitivity of mean annual streamflow showed that a10%increase in meanannual precipitation increased mean annual streamflow by16.7%, while a10%increase inmean annual potential evapotranspiration (ETp) decreased streamflow by6.9%on average.This suggested that mean annual streamflow was sensitive to changes in mean annualstreamflowand less sensitive to changes in mean annual ETpin all17watersheds. Except forTrap Rock River (04043050) and MacatawaRiver (04108800) watersheds, the correlationcoefficient between streamflow coefficient and BFI in the remaining15watersheds was0.46.This positive correlation suggested thatstreamflow coefficient increasedwith the increase ofBFIs. Watersheds with similar BFI values would have similar ETa/P(the ratio of actualevapotranspiration (ETa) to precipitation (P))ratios. Inversely, if BFI values were different,ETa/Pratios would be different.(4) Estimated annual streamflow using Fu’s equation varied between105mm and716mm with ENSand RMSE values ranging from2.0to0.98and48.5mm/yr to91.9mm/yr,respectively. While Fu’s equation provided appropriate estimates of annual streamflow inmost study watersheds, it showed relatively poor performance in simulating annualstreamflow in watersheds with high BFI values.The sensitivity coefficient of annualstreamflow responses to annual precipitation and ETpranged from0.25to0.71and0.27to0.24, respectively. This indicates that annual streamflow was also sensitive to changes inannual precipitation and less sensitive to changes in annual ETpin the watersheds.Correlationcoefficients between annual ETa/P and aridity index (defined as the ratio of ETPto P, ETp/P)ranged between0.65to0.40. ETa/P ratioswere negative correlated with ETp/Pratios forwatersheds with BFI values varying from0.78to0.80, which was deviated from Budykohypothesis.However, with the decrease of BFI values, correlation coefficients between ETa/PratiosandETp/Pratios generally transfered from negative into positive values that followedwith Budyko hypothesis. Therefore, different contributions of baseflow to streamflowchanged the impact of climate controls on annual water balance in the baseflow-dominatedwatersheds.(5) Coefficients of variation and extremes ratio for annual streamflow during1967–2011period varied from0.08to0.30and1.5to4.4, respectively. Interannual variability ofstreamflowwas obviously low forwatersheds with high BFI values, indicating that interannual variability of streamflow tended to be stable in baseflow-dominated watersheds. Thecorrelation between the relative error of streamflow deviation ratio (i.e., the ratio of annualstreamflow deviation and that of annual precipitation, SDR) and BFI is0.62. BFI as anexplanatory variable was used to develop model for estimating SDR. The revised modelimproved the predictive ability at watersheds with high BFI values compare with the originalKoster and Suarez’s analytical framework based on Fu’s equation. Specifically, R2increasedfrom0.28to0.62.RMSEand Mean Absolute Difference (MAE) increased from17.67mm/yr to14.07mm/yr and19.51mm/yr to14.54mm/yr, respectively. ENSreduced from0.28to0.12.However, the predictive ability of the revised model would improve significantly if onewatershed (MacatawaRiver watershed(04108800)) was excluded. CorrespondingENS,R2,RMSEand MAEwas0.45,0.52,13.03mm/yrand11.62mm/yr. Overall, the predictiveaccuracy of interannual variability of streamflow could be improvedby considering the effectof baseflow on watershed water balance.Regression models between baseflow (BFI) and watershed characteristicswere developedin this study. Based on the evaluations of watershed water balance, streamflow sensitivity toclimate variation in P and ETpwas quantified and the impact of baseflow on watershed waterbalance was qualified. Research results were helpful for expanding the research idea forstudying on watershed water balance and water cycle (focus on the impact of baseflow onwatershed water balance). It has important reference values for water resources developmentand utilization in the study areaas well as carrying out hydrological cycle process and variablerule in the ungauged sites in China, especially for analyzing the impact of climate change onhydrology and water resources variations at watershed scales.