Increased Dry Spells in Response to Explicitly Resolved Convection in High-Resolution Earth System Models

A warming climate is increasing both the severity and extent of drought conditions globally. The economic, agricultural, and environmental impacts are far ranging with recent examples of European forest health deterioration and falling hydroelectric output in China. Recent observed trends reveal longer dry spell lengths by 1-2 days per decade across northeast South America, southern North American, southern Africa. Further increases in temperature and atmospheric moisture are projected to exacerbate hydrological extremes through enhanced soil desiccation and less precipitation spatial evenness.

While most climate model predict increases in drought frequency and duration in response to rising greenhouse gases, there is still much uncertainty in how CMIP5/CMIP6 models simulate sub-daily precipitation patterns and how that effects future dry spell projections. The relatively coarse resolution, lack of ocean-atmosphere coupling, and parameterization of convection leads to the simulation of precipitation that is overly frequent, yet weaker in intensity, thus leading to shorter simulated dry spells. However, simply increasing model resolution when at the kilometer-scale does not necessary ensure better accuracy in convective organization and precipitation intensity.

On a regional scale, increasing model resolution and explicitly resolving convection normally leads to an improvement in convective precipitation patterns and dry spells, yet this is still unproven at a global scale. Here, the Next Generation Earth Modelling Systems (nextGEMS) project aims to address these issues with the development of convection-permitting, fully-coupled, Earth-system models. Using the ECMWF Integrated Forecast System (IFS) and Icosahedral Nonhydrostatic Weather and Climate Model (ICON), we examine the spatial distribution on hourly and daily precipitation and how this influences the simulation of the longest annual dry spells across the global mid-latitudes, experimenting with various kilometer scale resolutions and convection schemes.

Using ICON and IFS at resolutions ranging from 2.8–9 km over a 30 year historical (1990-2020) and a 30 year future (2020-2050) period, we find that explicitly resolving convection leads to a greater spatial concentration of weak (0.1 mm/hr), hourly precipitation occurrences when compared with IMERG observations, particularly over land. Within IFS, increasing resolution has no effect on spatial precipitation coverage, but turning off convection parametrization at 2.8 km leads to the most accurate representation. In the mid-21^st century simulations, IFS and ICON predict a greater increase in precipitation concentration compared to CESM2 simulations. This translates to a greater increase in projected longest annual dry spell trends globally, with hotspots in northeast South America, southern North American, southern Africa, and southern Europe having increased dry spell trends of 10-20 days per decade compared to 0-5 days in CESM2. While the single run nextGEMS simulations are unable to capture natural variability, these results indicate a potential underestimation in future drought projections that warrants further investigation.