Atmospheric circulation pattern classification for South-West Germany using hydrological variables

A fuzzy rule based classification of circulation patterns is used to investigate the link between large scale atmospheric features and regional scale hydrological variables (precipitation and discharge). The rules are assessed using an optimization procedure with different precipitation related objective functions. Due to the high spatial and temporal variability of precipitation, an optimization using precipitation measured at one or more selected locations might lead to classifications which are only optimal for the training period. Instead simpler objective functions defined on the basis of different indicators: such as interpolated areal precipitation, the precipitation coverage over one or more selected areas, the occurrence or magnitude of discharge increases and the nearly no changes in the discharge series are used. These indicators are strongly linked to precipitation but due to their integral character provide smooth objective functions with less local maxima and which lead to robust classifications. In order to compare different classifications and to find the optimum number of patterns, different classification quality measures based on the probability of exceedence of a given threshold and on conditional precipitation amounts are introduced. The comparison of classifications corresponding to different indicators and defined on different areas show that the patterns capture the behavior reasonably well over large areas and some improvement can be achieved for mesoscale areas (up to a a few 10,000 km2).The methodology is applied for South-West Germany for the state Baden-Württemberg with a total area of 35,000 km2. Daily precipitation amounts measured at 294 locations are used to demonstrate the methodology. CP dependent wetness indices and probabilities of exceedences are mapped using external drift kriging to show the geographical distribution of the link to CPs. Information measures are also mapped and allow the identification of regions where classification provides useful information for precipitation. Finally the link between discharge extremes and CPs is discussed.