Application of machine learning models in assessing the hydrological changes under climate change in the transboundary 3S River Basin

Abstract

Abstract This paper aims to evaluate two machine learning (ML) algorithms, namely, convolutional neural network (CNN) and long short-term memories (LSTM) deep learning algorithms, to predict the hydrological regime of the 3S River Basin under various climate change scenarios. Climate models CMCC-CMS, HadGEM-AO2, and MIROC5 were used to predict future climate and streamflow for three future periods: near-future (2020–2050), mid-future (2050–2080), and far-future (2080–2100) under two Representative Concentration Pathways (RCPs) 4.5 and 8.5. The future projection shows an increase in mean annual temperature from 0.08 to 4.3 °C by CMCC-CMS, from 0.13 to 4.4 °C by HadGEM-AO2, and −0.07 to 4.2 °C MIROC5 models. Similarly, the annual precipitation is projected to fluctuate from 13.3 to 62.5% by CMCC-CMS, from −12.4 to 26.1% by HadGEM-AO2, and from 6.9 to 49% by the MIROC5 model. The 3S River Basin expects an increasing trend in streamflow in the Srepok and Sesan Rivers, while the Sekong is projected to have reduced streamflow. ML models predicted the increasing flood risk in the Sekong and Sesan catchments with the increase of the Q5 index in the future but a decrease in the Srepok.