Is the reverse orographic effect on hourly extreme precipitation well represented by a convection-permitting climate model?

A significant increase in the intensity of extreme precipitation of short duration is expected due to the global warming, and is already reported in recent literature. Estimating the changes in hourly extremes in climate models is of fundamental importance for improving risk management and adaptation to changing climate, because they cause numerous hydrological and geomorphological hazard. Convection-permitting models (CPMs) can resolve the scales at which convective processes occur, and thus provide higher confidence in the future estimates of hourly precipitation than coarser resolution models. However, the existing CPM simulations cover relatively short time periods (10–20 years), and this prevents the use of conventional extreme value methods for the estimation of extremes. Novel methods based on the concept of ordinary event have shown the potential of deriving accurate frequency analyses from short data records, and they can be successfully applied to CPMs.

Recent studies reported distinct orographic impacts on precipitation extremes. In particular, a “reverse orographic effect” was reported for hourly durations, meaning that the intensity of hourly extremes tends to decrease with elevation, as opposed to the orographic enhancement of precipitation for long durations. This reverse orographic effect was tentatively associated to orography-induced turbulence. As these processes could be sub-grid even for CPMs, it is crucial to understand whether and how CPMs are able to represent these effects before using these simulations to project future extremes in mountainous areas.

Here, we focus on a complex-orography area in North-eastern Italy and we use hourly precipitation records from ~150 5-min resolution rain gauges (our benchmark) and CPM simulations from COSMO model, run at 2.2 km resolution. The model is driven with ERA Interim reanalyses for the period 2000-2009. By applying a storm-based statistical method to both observed and simulated time series, we model the ordinary events tails using a Weibull distribution. We compute the distribution parameters and the extreme quantiles up to 20-year return period for 9 durations between 1 and 24 hours, and we evaluate: their dependence on elevation, the bias between the observation and the CPM, the dependence of the biases with elevation.

We find evident spatial patterns in the CPM biases on the annual maxima and the modelled quantiles, especially for short durations. The bias significantly depends on elevation, with increasing overestimation of the 1-hour quantiles with elevation. This seems to confirm our hypothesis that CPMs cannot well represent the “reversed orographic effect”. These findings can improve our understanding of the changes in the meteorological processes underlying the changes in the precipitation extremes, and could help us develop adjustment approaches that can account for the role of orography at multiple durations.