Tree-Structured Modeling of the Relationship Between Great Lakes Ice Cover and Atmospheric Circulation Patterns

摘要: Abstract Seasonal maximum ice concentration (percentage of lake surface covered by ice) for the entire Laurentian Great Lakes and each Lake separately is modeled using atmospheric teleconnection indices. Two methods, Linear Regression (LR) Classification Trees (CART), are used to develop empirical models interannual variations cover. Thirty-four winter seasons between 1963 1998 nine indices were in analysis. The cover characteristics different Lake. data lent itself better CART analysis, because does not require a priori assumptions about distributions perform well. stepwise LR needed more variables, general, did explain as much variance models. Multivariate ENSO index Tropical/Northern Hemisphere index, explained Composite circulation patterns threshold values these two found be associated with above-and below-normal Lakes. Thus, also provided insight into physical mechanisms (atmospheric characteristics) underlying statistical relationships identified