Assessment of climate change impacts on floods with a hybrid data-driven and conceptual model across a data-scarce region

Abstract

In an era marked by unprecedented environmental shifts, comprehensively assessing the repercussions of climate change has become a paramount concern. This study investigates the impact of climate change on floods in the Kashkan watershed, located in western Iran, for the near future (2030–2059) and far future (2060–2089). First, a HEC-HMS model was developed for the study area, with a calibration period from January 1997 to December 2012 and a validation period from January 2013 to August 2019. Subsequently, to enhance the precision of our simulation, we employed long short-term memory (LSTM) as a methodological improvement. LSTM improved the ability of HEC-HMS to simulate maximum flows, reducing the annual average error peak flow (AEPFy) from 23.62–9.49% during the testing period. Then, 8 general circulation models (GCMs) were selected using a Taylor diagram for three climatic variables: cumulative daily precipitation, maximum daily temperature, and minimum daily temperature. These selected models were bias corrected using the quantile mapping method. The annual maximum cumulative 5-day precipitation was calculated for the SSP126 and SSP585 scenarios in the near and far future, revealing the potential for substantial increases. The result of the streamflow simulation with the hybrid model showed a significant increase in annual maximum discharge under both the SSP126 and SSP585 scenarios for the near and far future. Maximum discharge (mean ensemble of selected GCMs) is projected to increase by 45.08% and 37.59% in the near and far future for SSP126 and by 54.34% and 73.27% for SSP585. Most years will experience maximum flows exceeding the average baseline values. This increase, based on SSP126, will occur in most months, especially autumn, while SSP585 has similar patterns but with higher magnitudes. A 3-way ANOVA was employed to assess uncertainty in both the near and far future. The results suggest that individual factors such as Model, GCM, and SSP have limited influence, with the primary driver of uncertainty stemming from the interactions among these factors. The outcomes of this research will aid policymakers in integrating necessary measures to mitigate financial and human losses caused by the effects of climate change.