Comparison of indicators to evaluate the performance of climate models

TL;DR

This paper evaluates six climate model performance indicators, introduces a new multi-component measure, and assesses 48 models' ability to replicate regional climate variables, highlighting the importance of indicator choice.

Contribution

It proposes a new multi-component performance measure and a model selection approach for better climate model evaluation.

Findings

The new measure effectively differentiates model performance.

Even top models poorly reproduce some regional climate variables.

The selection method improves model assessment accuracy.

Abstract

The evaluation of climate models is a crucial step in climate studies. It consists of quantifying the resemblance of model outputs to reference data to identify models with superior capacity to replicate specific climate variables. Clearly, the choice of the evaluation indicator significantly impacts the results, underscoring the importance of selecting an indicator that properly captures the characteristics of a "good model". This study examines the behavior of six indicators, considering spatial correlation, distribution mean, variance, and shape. A new multi-component measure was selected based on these criteria to assess the performance of 48 CMIP6 models in reproducing the annual seasonal cycle of precipitation, temperature, and teleconnection patterns in Central America. The top six models were determined using multi-criteria methods. It was found that even the best model reproduces one derived climatic variable poorly in this region. The proposed measure and selection method can contribute to enhancing the accuracy of climatological research based on climate models.