There are numerous algorithmic classification methods that attempt to address the connections between different scales of the atmosphere, such as EOFs, clustering, and neural nets. However, their relative strength lies in the description of the mean conditions, whereas extremes are poorly covered by them. A novel approach towards the identification of linkages between large-scale atmospheric fields and local extremes of meteorological parameters is presented in this paper. The principle is that a small number of objectively selected fields can be used to circumscribe a local meteorological parameter by way of regression. For each day, the regression coefficients form a kind of pattern which is used for a classification based on similarity. As it turns out, several classes are generated which contain days that constitute extreme atmospheric conditions and from which local meteorological parameters can be computed, yielding an indirect way of determining these local extremes just from large-scale information. The range of applications is large. (i) Not only local meteorological parameters can be subjected to such a regressionbased classification procedure. It can be extended to extreme indicators, such as threshold exceedances, yielding on the one hand the relevant atmospheric fields to describe those indicators, and on the other hand grouping days with “favourable atmospheric conditions”. This approach can be further extended by investigating networks of measurement stations from a region and describing, e.g., the probability for threshold exceedances at a given percentage of the network. (ii) The method can not only be used as a filtering tool to supply days in the current climate with extreme conditions, identified in an objective way. The method can be applied to climate model projections, using the previously found parameter-specific combinations of atmospheric fields. From those fields, as they constitute the modelled future climate, local time series can be generated which are then analysed with respect to the frequency and magnitude of future extremes. The method has sensitivities (i) due to the degree to which there are connections between large-scale fields and local meteorological parameters (measured, e.g., by the correlation) and (ii) due to the varying quality of the different fields (geopotential, temperature, humidity etc.) projected by the climate model.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.