Including prior information on temperature-dependent sub-daily extreme precipitation in a Bayesian framework

Improving our estimates of extreme precipitation magnitudes with low exceedance probability under climate change scenarios is crucial for disaster preparedness. The task is particularly challenging for sub-daily extremes, as they are hardly resolved by current climate models and they are expected to change at faster rates than longer-duration extremes. A statistical approach to predict future sub-daily extremes using a physically-based dependence on temperature was proposed (TENAX). The approach establishes a functional dependence between the parameters of the statistical model and near-surface air temperature. A temperature model is then used to represent the probability of having a precipitation event at a given temperature. While an exponential relation between scale parameter and temperature can be physically obtained from the Clausius–Clapeyron relation, the dependence of the shape parameter (related to tail heaviness) on temperature is less trivial and may significantly affect the model’s accuracy. Here, we implement a Bayesian framework to investigate this issue and to include prior knowledge on the parameter in the statistical inference. We test both linear and exponential dependencies of the shape parameter on temperature, as well as different temperature models. Preliminary results on several stations in Germany, Japan, the UK, and the USA show consistency of the past return levels with the previous TENAX model (based on maximum likelihood estimation with only the scale parameter dependent on temperature), and with benchmark estimates from a non-asymptotic method (SMEV), in both its classic and time-dependent implementations.