Predicting Snowfall from Synoptic Circulation: A Comparison of Linear Regression and Neural Network Methodologies

Abstract

Much of the current debate regarding global warming and the regional manifestation of climate change relates to the ability of General Circulation Models (GCMs) to adequately represent modern climate features. Although the GCMs represent large-scale features well, their ability to model regional or local climate remains highly suspect. Important developments for the analysis of regional climate-change include statistical methodologies that translate between the large-scale and local-scale extremes. Traditionally, such methodologies have been limited to linear relationships; however, neural networks provide a new way to accomplish the same goal, with the added benefit of addressing the non-linear relationships that are characteristic of some climatic fields.