Premium
Modification of a correlation‐based circulation pattern classification to reduce within‐type variability of temperature and precipitation
Author(s) -
Brinkmann W.A.R.
Publication year - 2000
Publication title -
international journal of climatology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.58
H-Index - 166
eISSN - 1097-0088
pISSN - 0899-8418
DOI - 10.1002/1097-0088(20000630)20:8<839::aid-joc500>3.0.co;2-x
Subject(s) - anomaly (physics) , climatology , precipitation , environmental science , atmospheric circulation , circulation (fluid dynamics) , geology , atmospheric sciences , geography , meteorology , physics , condensed matter physics , thermodynamics
When a correlation‐based circulation pattern classification was applied to daily 700 hPa heights for 25 summer seasons, and the circulation types were related to surface temperature and precipitation for the Lake Superior basin, within‐type variability was found to be high. Regarding temperature, the difference between warm and cold days of a type is a positive height anomaly centred to the east of the lake, which is very similar to the height anomaly for winter temperatures identified in an earlier study. The difference between wet and dry days of a type is a pattern quite unlike that for temperature. It is a dipole pattern consisting of a negative anomaly to the northwest of the lake and a positive anomaly to the east and southeast. These small‐scale circulation components were missed by the correlation‐based classification because they are much smaller than the differences in height between circulation types. Regression models, relating monthly mean temperature anomalies to circulation type frequencies, produced results comparable to those obtained for winter temperature. For precipitation, the variance explained by the regression models is only half that obtained for temperature. Two approaches were used to reduce within‐type variability and to improve the regression results: the division of the first few circulation types into warm/cold and wet/dry subtypes on the basis of differences in 700 hPa vorticity (an approach introduced in the earlier paper), and a modified correlation‐based circulation pattern classification. The modification consisted of raising the correlation threshold for a few selected grid‐points to capture the small‐scale circulation component and also, of lowering the threshold for the comparison between maps and map sectors in order to minimize the number of unclassified days. Of the two approaches used to reduce within‐type variability, the modified classification scheme produced superior results. In particular, the variance in precipitation explained by the regression models is about 30% higher than that achieved using vorticity to create subtypes. Copyright © 2000 Royal Meteorological Society