1961-1990 high-resolution monthly precipitation climatologies for Italy

High-resolution monthly precipitation climatologies for Italy are presented. They are based on 1961-1990 precipitation normals obtained from a quality-controlled dataset of about 6200 stations covering the Italian surface and part of the Northern neighbouring regions. The study area corresponds to the Italian administrative boundaries and includes the trans-national portions of Po basin which is the major water resource for Northern Italy. High-resolution climatologies are calculated by means of a local weighted linear regression (LWLR) of precipitation versus elevation: for each cell of a smoothed 30-arc second resolution Digital Elevation Model (DEM) the regression is performed considering the 15 stations with the highest weight. The station weights are expressed as the product of several weighting factors in the form of Gaussian functions in which the distances and the level of similarity between the station cells and the considered DEM grid cell in terms of orographic features are taken into account. In order to properly apply this procedure to the complex Italian domain, the coefficients regulating the decrease of the weighting factors are locally optimised by an iterative method. At each point of a 1\ub0x1\ub0 resolution grid covering the study area, the normals of the stations in the range of 200 km are recursively computed and the optimal values of the coefficients are those which minimise the error estimators. Optimised coefficients are then estimated for the high-resolution grid interpolating the 1\ub0x1\ub0 grid results by inverse distance weighting (IDW) and they are used to produce the climatologies. The performance of the model is evaluated by comparing, with a leave-one-out approach, the precipitation normals computed for each station to the corresponding observed values in terms of mean error (BIAS), mean absolute error (MAE) and root mean square error (RMSE). The results are then compared with those provided by other interpolation approaches, such as IDW and regression kriging (RK) with elevation as predictor. IDW exhibits the largest MAEs and RMSEs in any month (9.5 and 14.2 mm as monthly average), LWLR and RK are quite comparable in summer, while LWLR performs significantly better in winter. Mean monthly MAEs and RMSEs are 8 and 11.8 mm for LWLR, and 8.3 and 12 mm for RK. LWLR monthly precipitation maps show the finest spatial detail, nevertheless IDW and RK maps could be useful supports to detect and correct possible outliers in LWLR results