Investigating orographic gradients and rainfall extremes through local regression models: an application over Italy

In areas with complex morphology the spatial variability of rainfall depths is known to be affected by the elevation, with orographic gradients that can vary depending on the duration and the location. In this work, more than 3800 time series of sub-daily annual maximum rainfall depths with at least 10 years of data are used to estimate the spatial variability of the average rainfall extremes over Italy. Most of the variability depends on elevation and this dependence is considered through a local regression approach applied to all the cells available in a grid with resolution of 1 km. The model is developed by considering various aspects, such as the low data density of some areas, the minimum difference in elevation of the sample, and the difficulties in extrapolating to high and low elevations. This model allows us to reconstruct maps of average extremes that consider the local effects of the topography. Our results indicate that, for the 1 h extremes, a negative gradient is present in large mountainous areas. In contrast, the 24 h mean rainfall extremes usually have positive orographic gradients, except for a few mountainous regions. In most cases, our findings confirm previous investigations available over areas of limited extension. The proposed approach allows us to identify areas where the reverse orographic effect occurs, providing a basis for future research on the underlying physical mechanisms. In addition, the obtained maps can be used in studies that focus on the spatial variability of rainfall quantiles, needed for hydrological design.