Assessing the effect of climate change on drought and runoff using a machine learning models

Abstract

Nowadays, droughts and the impacts of climate change on water resources and the environment have had significant negative effects. Investigating the effects of climate change on drought indices and streamflow is crucial for water and environmental resource management. Therefore, the present study was conducted in two parts to examine the impact of climate change on drought indices and the amount of watershed streamflow. In the first part of this study, drought modeling was performed using the Standardized Precipitation Index (SPI) and emission scenarios (RCP4.5 and RCP8.5) at three temporal scales (3, 6, and 12 months) during the period of 1995–2055. Then, the climatic impacts on SPI for the period 2030–2055 under different climate scenarios were evaluated. The Karun basin in south west Iran, which is affected by droughts and the impacts of climate change, was selected as the study area. In the second part, the Adaptive Neuro-Fuzzy Inference System (ANFIS) was utilized to estimate watershed streamflow for a 20-year period. Subsequently, in this section, the Whale Optimization Algorithm (WOA) was employed to improve the results of ANFIS. Finally, streamflow prediction for the future period (2035–2055) was carried out using the hybrid model. The results indicated that analyzing precipitation through SPI under different climate scenarios could influence severe fluctuations in droughts within the study area. Frequency analysis of droughts under climate scenarios, RCP4.5 and RCP8.5, demonstrated an upward trend with diverse spatial prevalence patterns. On the other hand, the duration of droughts increased towards the RCP4.5 scenario and remained unchanged according to the RCP8.5 climate scenario. The northeastern, eastern, and southeastern regions will experience the longest and most frequent droughts compared to current conditions. Furthermore, the results of the second part showed that the developed ANFIS-WOA model provides better results (RMSE = 127, MAPE = 98.50, NSE = 0.73) compared to the ANFIS-based model with evaluation criteria of RMSE = 127, MAPE = 98.50, NSE = 0.73. Additionally, in the investigation of the impact of climate change on streamflow using ANFIS-WOA in the time range of 2030 to 2055, the flow rate in most months of the year will decrease by approximately 20 units compared to the baseline period, with a greater intensity of reduction in the RCP8.5 scenario than RCP4.5. However, there will be an increase in streamflow by approximately 20 (m³/s) only in October. The approach used in this study demonstrates the effects of climate change on the level of drought and watershed streamflow, serving as a warning for decision-makers and managers to better manage available water resources. Finally, this approach is recommended for implementation in other similar regions for water resource management and water supply assessment.