Ranked and Bias-corrected CMIP6 Projections Reveal a Seasonal Shift of Monsoon Rainfall and Accelerated Warming in a Humid Tropical River Basin of the Western Ghats, Kerala, India

Basin-scale water planning requires climate projections that have been screened and corrected against local observations. This study ranked 15 CMIP6 global climate models (GCMs), compared bias-correction methods and projected future climate for the Valapattanam River Basin (1,238.81 km²) in humid tropical Kerala, India, using daily rainfall, maximum temperature (Tmax) and minimum temperature (Tmin) from the India Meteorological Department observatory at Kannur (1990–2014). Trends were tested with the Mann–Kendall test, Sen's slope and the Pettitt test, and GCMs were ranked for each variable using Compromise Programming and TOPSIS. Observed annual rainfall averaged 3,414.6 mm, 77.3% of it in June–September, without a significant trend, while Tmax rose by 0.51 °C per decade. MIROC6 ranked best for rainfall and MRI-ESM2-0 for Tmax, and the two ranking methods agreed closely. Among four rainfall corrections, distribution mapping (DM) reproduced the observed daily distribution most closely (Kolmogorov–Smirnov D = 0.063); linear scaling gave the best monthly NSE but 38% too many wet days and a 99th percentile 35% too low. A two-fold split-sample test, repeated for the six GCMs ranked best for rainfall, confirmed this out of sample: DM gave the lowest D in 10 of 12 cases (mean 0.065 against 0.148 for linear scaling), although mean volume errors of 10–17% remained for all methods. DM-corrected projections for 2081–2100 under SSP1-2.6 to SSP5-8.5 indicate that Tmax rises by 1.3–3.9 °C, Tmin by 1.6–5.4 °C and hot days increasing from 16 to 73–217 per year. Annual rainfall changes little (−4% to +4%), but under SSP3-7.0 and SSP5-8.5 south-west monsoon rainfall falls by 12–16% and north-east monsoon rainfall rises by 55–71%, with fewer wet days and longer dry spells. As the projections are based on one station and one GCM per variable, they are indicative and should be bracketed by a multi-model ensemble. Distribution mapping is recommended for preparing CMIP6 inputs for hydrological modelling, and the projected seasonal shift calls for more storage of late-season rainfall and more efficient irrigation in the dry season.