Regional frequency analysis, bias correction and downscaling of ERA5-Land snow related variables

This paper proposes the regional frequency analysis, bias correction and downscaling of ERA5-Land snow depth and increase of snow depth over three consecutive days, with the aim of quantifying design loads at ungauged sites and accurately assessing avalanche hazard. Local statistical analyses are typically performed by extracting extremes for snow depth time series. The latter are often incomplete, especially during extreme snowfall events. Furthermore, in most cases the stations are not found at elevations where avalanches typically release. In contrast, ERA5-Land provides broader spatial and temporal coverage, though it is affected by biases that lead to an under- or overestimation of extreme values. To address these limitations, a regional frequency analysis is proposed for both observed and ERA5-Land snow related variables. The adopted method is the index value regional method, which lead to the identification of homogeneous areas  according to common snowfall and avalanche occurrence patterns. A bias correction and downscaling approach was then applied, based on a comparison between the cumulative distribution functions of regionalised observed and ERA5-Land variables for the area of Central Italy, taken as case study. Calibration values derived from this comparison were then used to adjust ERA5-Land quantiles, thus providing a simple procedure to extrapolate design loads at the local scale.