Climate and land use change threaten irrigation water quality in Bangladesh: machine learning and global climate models reveal future risks

Abstract

Climate change and land use transformations pose unprecedented threats to irrigation water quality in subtropical agricultural regions. This study evaluates the interactive effects of projected climate change and land use/land cover (LULC) dynamics on irrigation water quality parameters in the Old Brahmaputra River, north-central Bangladesh. We employed an integrated methodology combining three general circulation models (BCC-CSM1-1-M, HadGEM2-ES, and MIROC5) under RCP 4.5 and RCP 8.5 scenarios, random forest machine learning algorithms in Google Earth Engine for LULC modeling, and comprehensive water quality analysis including electrical conductivity (EC), major ions, sodium adsorption ratio (SAR), and Irrigation Water Quality Index (IWQI). Field sampling across 11 locations during contrasting seasons revealed significant spatial heterogeneity, with 36.4% of stations exhibiting high restriction levels (IWQI < 55) during dry periods. Climate projections indicate substantial precipitation variability, including 33.56% monsoon increases and 33.39% pre-monsoon decreases under RCP 8.5 by the 2070s, with temperature increases up to 19.14%. LULC analysis reveals dramatic urbanization (927.8% expansion by 2040) and concerning forest loss (42.95% reduction). Strong correlations between EC and sodium (r = 0.89) during dry seasons, coupled with projected environmental changes, suggest intensification of water quality challenges, highlighting urgent needs for adaptive irrigation management strategies in climate-vulnerable agricultural regions.