Future sub-daily extreme precipitation: can a stochastic method based on temperature shifts agree with explicit simulations from an ensemble of convection-permitting models?

Extreme sub-daily precipitation can trigger natural disasters such as flash floods, urban floods, and debris flows, causing significant damage to infrastructure, homes, and livelihoods. With rising global temperatures, the atmosphere’s increased moisture-holding capacity enhances the potential for more intense and frequent extreme precipitation events. Sub-daily precipitation extremes are already increasing in magnitude, and the associated recurrence intervals are decreasing. A key component of climate change adaptation and resilience is quantifying the likelihood that future sub-daily extreme precipitation will exceed historical levels under different climate scenarios. Convection-permitting models (CPMs) are capable of resolving the physical processes driving precipitation extremes at high spatial and temporal resolutions. However, CPM simulations are computationally expensive and are available for a limited number of future scenarios. A recently proposed stochastic framework (TENAX) leverages temperature-precipitation scaling relationships and projected changes in daily temperature during wet days to estimate changes in extreme sub-daily precipitation. Can such a stochastic approach based on climate model simulations of temperature during wet days deliver projections of sub-daily extreme precipitation comparable to explicit simulations from CPMs?

This study evaluates the performance of TENAX in comparison to an ensemble of CPM simulations from the CORDEX-FPS Convection project over north-eastern Italy. Using historical (1996–2005) and far-future (2090–2099) CPM simulations under the RCP8.5 scenario and in-situ measurements of precipitation and temperature, we compare the return levels estimated using TENAX with the ones estimated with an extreme value method (SMEV) from the CPM ensemble. We assess two approaches for the application of TENAX: first, we train the model using CPM hourly precipitation and temperature for the historical period; then we train it using in-situ observations of the same quantities. In both cases, we project future return levels based on the changes in mean and variance of the daily temperature during the wet days as projected by the CPMs.

This analysis examines the potential of TENAX as a computationally efficient alternative to CPMs, as one of its key advantages is the ability to project sub-daily precipitation extremes even in the absence of CPM simulations, expanding its applicability to regions or scenarios where CPMs are not yet available.