Within-type variability of 700 hPa winter circulation patterns over the Lake Superior basin

Abstract

A correlation-based map-pattern classification technique was applied to 25 winters of daily 700 hPa maps to group them into discrete types, which were then examined to determine their association with surface air temperatures over the Lake Superior basin. The focus of this study was to identify and to remove sources of within-type variability of temperature. Within-type variability was considerable. Its source turned out to be a combination of climate change and small-scale circulation features that were present in the data but were not captured by the map pattern classification. These small-scale features, identified using composite difference maps for cold and warm days of each map type, consisted of height anomalies centered to the east of Lake Superior. Viewing these differences as anomalous components of the mean 700 hPa height pattern-which when added to or subtracted from the pattern create subtle but important modifications-provided insight into the differences in the shape of the 700 hPa steering current for different surface weather systems. To incorporate these features into the classification, daily relative vorticity patterns were computed and discriminant analysis was used to create cold, average, and warm subtypes for each of the first five map types. Replacing the first five map types by their subtypes in the regression models increased the explained variance by 20- 30%. Collapsing warm and cold types did not increase the explained variance as much as it stabilized the models. The creation of cold and warm subtypes also removed a warming trend (significant for two of the map types) by translating the warming trend into a change in the frequency of the subtypes. The analysis used four different floating grids or windows. Classifications for windows centered east of the lake generated better results than those to the west.