Integrated assessment of bias correction techniques, CMIP6 model rankings, and multi-model ensemble optimisation across diverse temporal scales for regional climate projection in Kerala, Southwestern India

Abstract

In the context of climate change, CMIP6 (Coupled model intercomparison project phase 6) General Circulation Models (GCMs) are indispensable for projecting global and regional climate impacts, including temperature rise, precipitation variability, and extreme weather events. These models serve as the basis for Intergovernmental Panel on Climate Change (IPCC) assessments and are crucial for informing mitigation and adaptation strategies. However, their coarse resolution and systematic biases limit their direct application in local-scale climate impact studies. This motivates the present study, which aims to enhance the reliability of CMIP6 precipitation projections over Kerala, a monsoon-dominated, topographically complex region susceptible to rainfall variability. This study employs the CRITIC–TOPSIS (Criteria Importance through intercriteria correlation and technique for order of preference by similarity to ideal solution) framework to comprehensively evaluate bias correction methods, GCM performance, and multi-model ensembling (MME) techniques across multiple temporal scales. Observed daily rainfall data from the India Meteorological Department (IMD) serve as the reference for model evaluation. This integrated, data-driven approach enables robust ranking and selection of optimal models and techniques for regional application. The findings reveal considerable variability in model performance across time scales. ACCESS-ESM1-5 performs consistently well, while MRI-ESM2-0 and HadGEM3-GC31-LL are more suited to long-term projections. IITM-ESM and CMCC-CM2-SR5 show strength in short- to medium-term applications. Advanced ensemble methods, such as Support Vector Machines, Gradient Boosting Machines, Random Forests, and LightGBM, outperform simpler methods in capturing rainfall variability. The study’s results provide practical guidance for selecting climate models and designing ensemble strategies, particularly for hydrological forecasting, infrastructure planning, and climate risk assessment in Kerala and similar monsoon-prone regions. Overall, this research contributes to advancing regional climate modelling practices and supports informed, climate-resilient decision-making at policy and planning levels. © The Author(s) under exclusive licence to Institute of Geophysics, Polish Academy of Sciences 2025.