Evaluating CMIP6 precipitation simulations across different rainfall regimes in the Amhara Region, Ethiopia

Precipitation patterns play a crucial role in sustaining rainfed agriculture, particularly in regions like Ethiopia, where millions depend on it for their livelihoods. Understanding the ability of climate models to simulate precipitation accurately is essential for predicting the potential impacts of climate change, especially in regions with complex topography like Ethiopia. This study assesses the performance of 16 CMIP6 climate models in simulating precipitation patterns across different rainfall regimes in the Amhara region. By comparing model outputs with observed data, the research aims to identify which models best capture the seasonal and annual variability in precipitation, thereby informing climate adaptation strategies.

To conduct the analysis, observational data from the Enhanced National Climate Services and Climate Hazard Infrared Precipitation with Stations datasets were utilized. The study focused on evaluating the models' performance over the period 1985–2014 using statistical metrics such as the coefficient of determination (R²), root mean square error (RMSE), mean bias (MB), and Student's t-test. In the Western Amhara region, the models were further assessed with Taylor Skill Score (TSS), Inter-annual Variability Score (IVS), and Comprehensive Rating Method (CRM). The top-performing models for simulating the annual rainfall cycle in this region were GFDL-ESM4 (CRM: 0.88), EC-Earth3-Veg (CRM: 0.85), and CESM2 (CRM: 0.82). Other notable models included MRI-ESM2.0, MPI-ESM1-2-HR, and IPSL-CM6A-LR. In the Eastern Amhara region, which follows a bi-modal rainfall pattern, EC-Earth3-Veg (CRM: 0.85, R: 0.83, RMSE: 52.4 ​mm/month) showed the best performance, followed by MRI-ESM2.0 (CRM: 0.76, R: 0.76, RMSE: 55.6 ​mm/month) and CESM2 (CRM: 0.82, R: 0.88, RMSE: 65.46 ​mm/month). GFDL-ESM4 also demonstrated strong performance (CRM: 0.88, R: 0.77, RMSE: 55.6 ​mm/month). CMCC-ESM2, although lower-ranked in Western Amhara (CRM: ∼0.41), performed better in the Eastern region. However, most models exhibited a dry bias and substantial inter-model variability, particularly in simulating rainfall patterns in highland areas. Only 31 ​% and 44 ​% of models successfully captured the seasonality of rainfall in the Western and Eastern Amhara regions, respectively. Even the top models showed significant discrepancies compared to observed rainfall, indicating ongoing challenges in model accuracy. These findings provide valuable insights for developing targeted climate adaptation strategies in the Amhara region and highlight the need for improving climate models to better assess future climate change impacts in Ethiopia.