Multi-year droughts in CMIP6 large ensemble models

Abstract Multi-year droughts (MYDs) are extreme drought events leading to long-lasting impact. Due to their limited number in observational records, global climate models with large ensembles can contribute to understanding by increasing the sample size of MYDs. However, the knowledge on their ability to simulate MYDs is limited on a global scale. In this study, we evaluate six different CMIP6 models in simulating MYDs by comparing them to ERA5. We assess frequency, time spent in MYDs versus shorter droughts, seasonality, and physical drivers. The multi-model mean (MMM) performs robustly across these metrics, with strong inter-model agreement in deviations from ERA5. Deviations from ERA5 and inter-model spread are larger for MYD drivers compared to normal drought drivers. This can result from either model biases, ERA5 biases, or a limited sample size of MYDs within ERA5. The differences between the MMM and ERA5 are explained primarily by internal variability, which underscores the value of large ensembles for studying rare extremes such as MYDs.