A Combined Stochastic–Analytical Method for the Assessment of Climate Change Impact on Spring Discharge

Abstract

This study describes a novel methodology for the prediction of spring hydrographs based on regional climate model (RCM) projections, with the goal of evaluating climate-change impact on karstic-spring discharge. A combined stochastic–analytical modeling methodology to predict spring discharge was developed and demonstrated on the Bukovica spring catchment at the Durmitor National Park, Montenegro. As a first step, climate model projections of the EURO-CORDEX ensemble were selected; and then bias correction was applied based on historical climate data. The regression function between rainfall and peak discharge was established by using historical data. Baseflow recession was described by using a double-component exponential model, where hydrograph decomposition and parameter fitting were performed on the Master Recession Curve. Rainfall time series from two selected RCM scenarios were applied to predict future spring-discharge time series. Bias correction of simulated hydrographs was performed, and bias-corrected combined stochastic–analytical models were applied to predict spring hydrographs based on RCM-simulated rainfall data. Both simulated climate scenarios predict increasing peak discharges and decreasing baseflow discharges throughout the 21st century. The model results suggest that climate change is likely to exaggerate the extremities both in terms of climate parameters and spring discharge by the end of the century both for moderate (RCP 45) and pessimistic (RCP 85) CO2 emission scenarios. To investigate the temporal distribution of extremities throughout the simulated time periods, the annual numbers of flood and drought days were calculated. Annual predicted flood days show an increasing trend during the first simulation period (2021–2050) and a slightly decreasing trend during the second simulation period (2071–2100), according to the RCP45 climate scenario. The same parameter shows a stagnant trend for the RCP 85 climate scenario. Annual predicted drought days show a decreasing trend both for the RCP 45 and RCP 85 climate scenarios. However, the annual number of drought days shows a large variation over time. There is a periodicity of extremely dry years with a frequency between 5 and 7 years. The number of drought days seems to increase over time during these extreme years. The study confirmed that the applied methodology can successfully be applied for spring-discharge prediction and that it offers a new prospect for its wider application in studying karst aquifers and their behavior under different climate-change scenarios.