Estimating the Suitability of Circulation Classifications for Croatia

ABSTRACT Croatian territory is diverse and specific, composed of coastal, mountainous and plain parts. Thus, the differences between Croatian regions in climate and weather conditions during specific atmospheric circulation situations are significant. Since 1965, circulation types (CTs) in Croatia have been determined according to the subjective Poje's classification. Given various restrictions of this classification regarding its application in a synoptic‐climatological context, the application and evaluation of automated circulation classification (CC) procedures for the Croatian domain is sought. Hence, a representative selection of classification methods available from the catalogue established within the COST733 Action (‘Harmonisation and Applications of Weather Types Classifications for European Regions’) is applied and analysed regarding their synoptic skill performance with a focus on precipitation using four different datasets of surface observations for evaluation. To obtain a comprehensive assessment of the most appropriate CT approach for precipitation in Croatia, as a supplement to basic classification methods, specific settings (including location and size of the spatial domain, temporal domain, number of types and variables considered for classification) are taken into account. Thereby, various quantitative metrics to measure the synoptic skill are utilised and, in addition, CCs are used in an exemplary application for investigating the synoptic climatology of drought events in Croatia. Although no single superior CC becomes apparent from these systematic evaluation steps, valuable indications of advantageous features of CCs do emerge. Annual‐based CCs from the threshold‐based and optimisation methods with 18 CTs and for a Croatia‐centered domain tend towards the highest synoptic skill overall. A further improvement seems to be achieved by explicitly considering a vorticity index in the classification. This provides information on suitable existing CCs and important starting points for the intended development of an optimised CC for assessing spatiotemporal precipitation variability in Croatia.