Radar-based precipitation type analysis in the Baltic area

A method to classify precipitation events based on their spatial extent and texture has been developed and applied to 3 yr of BALTEX Radar Data Center weather radar composites over the Baltic region. The method is capable of distinguishing large-scale precipitation features typically associated with frontal systems from more small-scale features, which are usually found in convective systems. Data used for this Study are 2-D radar images. The classification is performed in three steps. First, contiguous precipitation areas are identified in the radar data. In the second step, each of these areas is Subjected to an analysis where different texture parameters are calculated. In a third step, these texture parameters are evaluated by means of a neural network, and a type of precipitation is assigned to each class. The neural network has been trained using a large set of visually classified radar scenes. A validation of the results has been performed by (1) comparing regions where U.K. Met Office analysis shows large contiguous frontal areas and (2) by using surface observations where the surface observer reported precipitation events that could clearly be associated either with intermittent convective precipitation or with frontal systems. The results of the different comparisons are generally in good agreement with each other, and the false classification rate ranges from 10% to 20%. The application to 3 yr of radar data has resulted in estimation of the frontal fraction of precipitation in the Baltic Sea area. About two-thirds of overall precipitation events are determined by frontal passages with high seasonal and diurnal variations.