Evaluation of CMIP6 model performance in simulating climatology of rainfall and temperature across different agroecological zones in North Western Ethiopia

In order to better understand the future climate, and its impact, accurate crop yield prediction by using crop simulation models and develop effective adaptation and mitigation measures for climate change, the best global circulation model should be identified. This study was initiated to evaluate the performance of eleven CMIP6 global climate models in simulating ENACTs observed climatology of rainfall, maximum temperature, and minimum temperature across the Gojjam and South Gonder Administrative Zones and its five Agroecological Zones in North West Ethiopia. The comprehensive Rating Index (CRI), which is based on seven statistical metrics, correlation coefficient, mean absolute error, root mean square error, percent bias, Kling–Gupta Efficiency, Nash–Sutcliffe Efficiency, Wilmot index of Agreement and Taylor diagrams were used to evaluate the model performance across multiple temporal scales. CMCC-CMC-SR5, GFDL-CM4, GFDL-ESM4, BCC-CSM2-MR, and MPI-ESM1-2-HR, INM-CM4-8, consistently outperformed for rainfall across multiple spatial and temporal scales. In simulating maximum temperature, CMCC-CMC-SR5, GFDL-CM4, GFDL-ESM4, and INM-CM4-8 were the top-performing climate models. For minimum temperature, the top best-performing climate models were GFDL-CM4, INM-CM4-8, GFDL-ESM4, BCC-CSM2-MR, and MIROC6. The findings highlights that the importance of selecting global circulation models based on specific variables, seasons, and spatial contexts rather than a one-size-fits-all approach. The research findings underscore that CMIP6 models exhibit notable biases, particularly in the simulation of rainfall, largely due to scale mismatches associated with coarse model resolution and the complex topography of the study area. Despite these limitations, the identified best-performing models provide a reliable foundation for climate change impact assessments. Accordingly, future research should prioritize statistical or dynamical downscaling, robust bias-correction techniques, and multi-model ensemble approaches, as well as their integration with hydrological applications. Such efforts are essential for improving the accuracy of future climate and crop yield simulations and for supporting evidence-based policy formulation in Ethiopia’s highland regions, particularly in North-West Ethiopia and other areas with similar agroecological contexts.