SYNOPTIC WEATHER-MAP CLASSIFICATION: CORRELATION VERSUS SUMS-OF-SQUARES*

Authors

  • Cort J. Willmott

    1. CORT I. WILLMO'TT is Associate Professor of Geography and Marine Studies at the University of Delaware, Newark, DE 19716 He is also a member of the University's Center for Climatic Research and Chair of the AAC Climate Specialty Croup. His research interests include the relationships between land-surface processes and cdirnate, the statistical analysis oi large-scale climate fields and the evaluation of model performance.
    Search for more papers by this author

  • *

    This note grew out of discussions following a climate paper session held at the 1986 AAG annual meeting. Robert C. Balling (Arizona State), James E. Burt (Wisconsin-Madison), Robert G. Crane (Penn State), Laurence S. Kalkstein (Delaware), Brenton M. Yarnal (Penn State) and David R. Legates (Delaware) thoughtfully evaluated an earlier draft of this note and I most gratefully acknowledge their assistance.

Abstract

A sums-of-squares similarity measure (S), attributed to Kirchhofer, has been characterized by a number of synoptic climatologists as superior to correlation for quantifying the similarity between two gridded weather maps. The sums-of-squares metric and product-moment correlation, however, are virtually identical so S cannot be regarded as better. It is also suggested that correlation, in any form, is not generally a satisfactory measure of the similarity between weather maps of comparable units.

Ancillary