CR 17:303-324 (2001)  -  doi:10.3354/cr017303

ABSTRACT: By applying the dynamical cluster algorithm to large-scale circulation patterns at various tropospheric levels (Z700, Z500, and SLP [sea-level pressure]), we obtain the so-called weather regimes (WRs). WRs are the cluster central patterns; there is no subjectivity at all involved in the procedure. A red noise test allows one to select the best number of clusters at each level. A comparison is performed between the different level classifications, and highly significant correlations are found. While previous attempts to classify daily circulations in a fully objective way were concerned with mid-tropospheric levels and could not go further back than 50 yr, here we classify 120 yr of SLP patterns. We arrive at the important conclusion that the same 5 WRs are found for the 3 periods 1880-1918, 1919-1957 and 1958-1997. The linkage between these WRs and the local tangible weather is then investigated for both temperatures and precipitation. It is found that the instantaneous departure of local weather from average climate is highly correlated with the WRs, making this approach a challenging and coherent description of local climates. The atmosphere does not merely evolve around its mean state, but instead spends more time around a few peculiar (large-scale) states with specific consequences for local weather. As a result, the WRs may provide a reliable (and moreover fully objective) framework for building downscaling algorithms appropriate for local climate change studies.

KEY WORDS: Classification · Large-scale circulation · Weather regime · Local climate · European Alps · France