Quantifying the uncertainty in future runoff projection over SSP scenarios, GCMs and hydrological model parameters

Abstract

Abstract This study evaluated the inherent uncertainty of future runoff prediction using eleven Coupled Model Intercomparison Project 6 (CMIP6) global climate models (GCMs) and a hydrological model (HM). The soil and water assessment tool (SWAT) model was used as a hydrologic model, and SWAT-CUP was used for parameter calibration. The future runoff projection was simulated utilizing two shared socioeconomic pathways (SSPs) scenarios, SSP2-4.5 and SSP5-8.5, for near (2021–2060) and far (2061–2100) futures. Jensen-Shannon divergence (JS-D) was used to quantify the uncertainties between the past and future probability distributions considering different GCMs and calibrated parameter sets of HM. The JS-D value for each GCM and calibrated HM parameter set was calculated at the range of 0.026–0.075 and 0.035–0.058, respectively. As a result, the uncertainty in the selection of GCMs was found to be greater than in the determination of values for HM parameters. Bayesian model averaging (BMA), which is a statistical approach that can combine estimations from multiple models and produce reliable probabilistic predictions, was applied to quantify the uncertainty by each GCM and HM parameters. When estimating the future runoff, INM-CM4-8 caused the greatest uncertainty, and the calibrated set of HM parameters using the year of high runoff caused the greatest uncertainty. This approach can help the uncertainty analysis in the future runoff estimation.