Uncertainty and Nonstationarity in Streamflow Extremes under Climate Change Scenarios over a River Basin

Abstract

The present study analyzes the various uncertainties and nonstationarity in the streamflow projections over the Wainganga River Basin, India, under representative concentration pathways (RCPs) 4.5 and 8.5 using the 3-layer variable infiltration capacity (VIC-3L) model. The uncertainties associated with the multiple climate models, parameters, and return levels were modeled using reliability ensemble averaging (REA), Bayesian analysis, and the delta method, respectively. With the recent development of extreme value theory (EVT), the annual maximum flows for the past and future were modeled with nonstationary assumption and validated using the Akaike information criterion (AIC) value and likelihood ratio test. The results obtained from the study indicate that the stationary assumption is a good fit for the observed and stabilized radioactive forcing scenarios (RCP4.5); whereas, for the highest greenhouse gas emission scenarios (RCP8.5), nonstationary modeling is more suitable. The obtained future flood quantiles under RCP4.5 and 8.5 are not likely to be critical in the coming century for both stationary and nonstationary assumptions. However, the nonstationary estimate of the return levels under lower return periods will be more useful to design low-capacity hydraulic structures. Further analysis of nonstationary return levels revealed that the change detection in the return levels under a lower return period was much earlier than the higher return period. The uncertainty analysis of the return levels showed larger uncertainty bound in the case of RCP8.5 rather than the RCP4.5. Furthermore, the quantification of the uncertainty between the stationary and nonstationary assumptions using Bayesian analysis with Markov chain Monte Carlo (MCMC) simulation provided a high uncertainty range in the case of nonstationary assumption compared with stationary assumption.