Premium
A procedure for blending manual and correlation‐based synoptic classifications
Author(s) -
Frakes Brent,
Yarnal Brent
Publication year - 1997
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/(sici)1097-0088(19971115)17:13<1381::aid-joc204>3.0.co;2-q
Subject(s) - computer science , correlation , artificial intelligence , statistics , data mining , mathematics , geometry
Manual and correlation‐based (also known as Lund or Kirchhofer) classifications are important to synoptic climatology, but both have significant drawbacks. Manual classifications are inherently subjective and labour intensive, whereas correlation‐based classifications give the investigator little control over the map‐patterns generated by the computer. This paper develops a simple procedure that combines these two classification methods, thereby minimizing these weaknesses. The hybrid procedure utilizes a relatively short‐term manual classification to generate composite pressure surfaces, which are then used as seeds in a long‐term correlation‐based computer classification. Overall, the results show that the hybrid classification reproduces the manual classification while optimizing speed, objectivity and investigator control, thus suggesting that the hybrid procedure is superior to the manual or correlation classifications as they are currently used. More specifically, the results demonstrate little difference between the hybrid procedure and the original manual classification at monthly and longer time‐scales, with less internal variation in the hybrid types than in the subjective categories. However, the two classifications showed substantial differences at the daily level, not because of poor performance by the hybrid procedure, but because of errors introduced by the subjectivity of the manual classification. © 1997 Royal Meteorological Society. Int.J.Climatol., Vol.17, 1381‐1396 (No. of Figures: 9 No. of Tables: 1 No. of References: 30)