Evaluation of Climate Models Using Bergen Metrics as a Multi-Metrics Comparison over Indonesia

Abstract Recent climate events have become increasingly unpredictable, emphasizing the need for climate model projections with reliable performance. Future climate projections derived from the Shared Socioeconomic Pathways of Coupled Model Intercomparison Project Phase 6 require careful evaluation against historical observations to ensure their applicability. This study evaluates six CMIP6 historical simulations namely ACCESS-CM2, CNRM-CM6-HR, EC-Earth3, EC-Earth3-Veg, HadGEM3-GC31-MM, and MPI-ESM1-2-HR using cases in Indonesia. Model performance was assessed using the SA-OBS dataset with a horizontal resolution of 0.25 degrees. Five evaluation metrics were applied namely Taylor Skill Score, Interannual Variability Score, normalized Mean Absolute Error, Kling Gupta Efficiency, and Willmott Index. These metrics were synthesized using the Bergen Metrics framework to provide an integrated measure of model performance. In addition, a rank based weighting approach was employed as a comparative assessment. Results indicate that all models show limited skill in simulating precipitation, while minimum, mean, and maximum temperatures are reproduced more accurately. Across all variables, EC-Earth3-Veg consistently achieved the best overall performance, followed closely by EC-Earth3. The remaining models exhibited comparable performance but ranked lower. These findings are consistent with previous regional evaluations over Southeast Asia. The relatively strong performance of EC-Earth3-based models is likely associated with their finer spatial resolution. Overall, this study demonstrates that Bergen Metrics provide a robust and flexible framework for climate model evaluation, while rank based weighting offers complementary insight into relative model consistency.