Statistical Approach to Selecting Global Circulation Models (GCMs) – A Study Based in the Kalu Ganga Catchment of Sri Lanka

This study proposes a two-stage Global Circulation Model (GCM) selection approach that evaluates the performance of CMIP5 GCMs in simulating past rainfall for the Kalu Ganga catchment in Sri Lanka. In the initial selection, GCMs were screened based on their ability to replicate monsoonal rainfall patterns using data from the Global Precipitation Climatology Project (GPCP). The fine selection involved bias-corrected GCM simulations at the catchment scale across three rainfall categories—extreme events, normal rainfall, and no-rain days—validated against the Asian Precipitation Highly Resolved Observational Data Integration Towards Evaluation (APHRODITE) dataset. Several statistical performance indicators (PIs) were used to assess GCM performance at both stages, and Multi-Criteria Decision-Making (MCDM) techniques were applied to rank the GCMs accordingly. The results indicated that CanESM2 is most suitable for simulating extreme rainfall events, CNRM-CM5 for normal rainfall, and either CanESM2 or CNRM-CM5 for no-rain days. This highlights that no single GCM performs optimally across all rainfall categories, and researchers should select GCMs tailored to their specific focus on future rainfall extremes, flood risk, or drought analysis.