Use of a neural network for a classification of wind regimes in the Central Victoria Land and their relative coupling with the synoptic description provided by the ECMWF model

A 15-years database of meteorological measurements gathered by the Italian Automatic Weather Stations deployed in the Central Victoria Land has been used to determine a classification of the surface wind field patterns through a SOM neural network. The problem of the stability and of the meteorological meaning of such a classification has been investigated with particular attention to the patterns representing localized and generalized katabatic wind events and upslope glacier-winds conditions.

The evolution of the instantaneous wind field in the pattern-space determined by the neural network is also considered.

For each wind field pattern the correspondence with the synoptic and meso-scale configuration in the Victoria Land and Ross Sea area is derived basing on satellite imagery and on the ECMWF model description.