Comprehensive Performance Evaluation of CMIP6 Models in Simulating Precipitation and Temperature: A Multi-Scale and Altitude-Based Analysis in the Upper Yangtze River Basin

In recent years, the accelerated impacts of global climate change on sensitive regions have drawn increasing attention. This study focuses on the Upper Yangtze River Basin, evaluating the performance of CMIP6 models in simulating precipitation and temperature during the historical period (1980–2014), providing different insights for future climate studies.

Seventeen commonly used CMIP6 models were selected and systematically evaluated at both annual and monthly scales for their ability to simulate precipitation and temperature. Performance evaluation employed multiple metrics, including bias, standard deviation, root mean square error (RMSE), and correlation coefficients. The Comprehensive Rating Index (CRI) was introduced to quantify the overall performance of each model. Additionally, F-tests and T-tests were conducted to analyze the statistical significance of differences between model simulations and observational data: F-tests assessed the homogeneity of variances between model outputs and observations, while T-tests evaluated differences in means.

Building on this assessment, a single evaluation metric derived from the historical period (1980–2014) is utilized to compute model rankings and a Composite Rating Index (CRI). Subsequently, a rank-based weighting (RBW) method is applied to assign weights to each model at both annual and monthly scales. This approach considers skill differences among models and provides insights for weighted multi-model ensemble (MME) analysis.

The results indicate that most models tend to underestimate annual mean temperature, with CESM2 performing relatively better than other models (CRI = 0.94), while annual cumulative precipitation is generally slightly overestimated, with FGOALS_g3 showing better performance (CRI = 0.89). On a monthly scale, CESM2 performs better in more months for temperature simulation, and FGOALS_g3 similarly performs better in more months for precipitation simulation. However, differences between monthly and annual performance are observed: certain models, such as IPSL_CM6A_LR and INM_CM5_0, which perform less effectively on an annual scale, exhibit relatively better performance in specific months. These findings highlight the variability in model performance across temporal scales and the importance of assessing models on both annual and monthly basis. Additionally, different models exhibit varying simulation capabilities at low-altitude, mid-altitude, and high-altitude observations. This underscores the heterogeneity in model performance across temporal and spatial scales, emphasizing the necessity of rigorous evaluations at both annual and monthly resolutions, as well as across varying spatial scales.

This study provides a comprehensive analysis of the applicability of CMIP6 models in the Upper Yangtze River Basin using multi-scale and multi-altitude approaches, incorporating CRI and RBW methods. The findings emphasize the importance of multi-metric analyses, significance testing, and weighting approaches in optimizing model selection and advancing the understanding of climate change impacts.