Enhancing Reliability in Climate Projections: A Novel Approach for Selecting Global Climate Models

Reliable climate projections are essential for informed decision-making in the context of climate change. However, selecting suitable Global Climate Models (GCMs) for such projections remains challenging, especially when computational resources are limited. This study introduced an innovative approach to GCM selection, emphasizing the identification of models that exhibit consistency in projecting future climate changes and skill in representing current climate conditions, including average climate, seasonal patterns, and climatic variations. GCM performance in simulating these critical properties was evaluated for three major climatic variables: rainfall, maximum temperature, and minimum temperature. This assessment results in a structured 3×3 performance matrix for each GCM, encapsulating its ability to capture these essential climate features. The matrix distances, quantifying the disparities between each GCM's performance matrix and the ideal reference matrix, were used to collectively represent the overall model performance. Finally, GCMs were ranked based on these differences using the Jenks natural break classification method, a robust statistical technique, to aid in identifying the top-performing GCMs, making them ideal candidates for ensemble model construction. The newly developed method was tested by applying it to select GCMs for Nigeria from a pool of 19 CMIP6 GCMs. The results indicate that 15 of 19 GCMs consistently projected future climate within a 95% confidence interval. Further evaluation of matrix distances and natural classification reveals a subset of GCMs, ACCESS.ESM1.5, BCC.CSM2.MR, CMCC.ESM2 and MRI.ESM2.0 are the most suitable choice for simulating Nigeria's climate. The multimodel ensemble mean of the selected GCMs projected a notable increase in rainfall by 10 to 40% over most of the country and the maximum and minimum temperatures by 1.0 to 3.5°C and 0.5 to 4.0°C across the country. The method introduced in this study can be an effective tool for GCM selection to enhance climate projection reliability.