Data mining of extreme value modelling European precipitation data

Abstract. This paper shows how can the peaks over threshold model of gridded European precipitation data be combined with various data mining tools. The motivation is that even the 0.5 grade-grid of 63 years of the European Climate Assessment daily precipitation data is a massive data set, where there is little hope to find valuable results without reasonable preprocessing. This step is based on the peaks over threshold approach, which is a sound model for the extremes. We have applied a moving window methodology in order to catch the changes in the pattern of the high precipitations. Our results show that indeed there are spatially different tendencies observable.