View article

Our study reveals some problems related to the digital mapping of mean annual precipitations using (geo) statistical methods in GIS environment. The applications are carried out for a 4950 km2 region situated in Vrancea County, using a sample of 34 rain gauges. We first address the data uncertainty issue looking for georeference errors and data errors. As a result we chose to eliminate 2 rain gauges from our analysis significantly evading the general spatial precipitations pattern probably due to the shorter data recording intervals. We show how easily such outliers can mislead us by inducing a false precipitations–latitude correlation. We then proceed by deriving mean annual precipitation spatial models using classical statistical approaches (ordinary kriging, cokriging) and a more elaborated approach (residual kriging). Comparison of the results proves the superiority of the latter. Still, the uncertainty of the output has to be considered, especially when it comes to the extrapolation of the residual kriging model outside the calibration area.