Probabilistic Evaluation of Drought in CMIP6 Simulations

Area of Science:

• Climate Science
• Environmental Science
• Meteorology

Background:

• Droughts pose significant social and ecological risks, necessitating effective adaptation and mitigation strategies.
• Climate models are essential tools for understanding and projecting drought dynamics.

Purpose of the Study:

• To evaluate the performance of 285 Climate Model Intercomparison Project Phase 6 (CMIP6) historical simulations in reproducing observed drought duration and severity.
• To develop and apply a novel probabilistic framework for assessing climate model accuracy in drought characterization.

Main Methods:

• Utilized 285 CMIP6 simulations from 17 climate models and three observational datasets.
• Employed the Standardized Precipitation Index (SPI) to quantify drought characteristics.
• Applied summary statistics beyond mean and standard deviation, and a Hellinger distance-based probabilistic framework to compare simulations and observations.

Main Results:

• Many simulations exhibited low error ($$) in reproducing observed drought summary statistics.
• The null hypothesis that simulations and observations follow the same distribution could not be rejected for over $$ of the grids using the Hellinger distance framework.
• No single model consistently outperformed others globally; tropical regions showed higher variance in drought statistics among simulations.
• Models accurately captured dry spell characteristics but showed biases in simulating low precipitation values.

Conclusions:

• Probabilistic evaluation is essential for identifying climate model weaknesses and selecting suitable models for impact assessments.
• Despite good performance in SPI metrics, models may still exhibit biases in simulating extreme low precipitation events.
• A comprehensive assessment is needed to understand the reliability of climate models for drought-related projections.