Premium
Subjectivity in A computer‐assisted synoptic climatology I: Classification results
Author(s) -
Yarnal Brent,
White Dale A.
Publication year - 1987
Publication title -
journal of climatology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.58
H-Index - 166
eISSN - 1097-0088
pISSN - 0196-1748
DOI - 10.1002/joc.3370070203
Subject(s) - metric (unit) , sample (material) , computer science , correlation , classification scheme , statistics , population , artificial intelligence , data mining , mathematics , machine learning , engineering , operations management , chemistry , geometry , demography , chromatography , sociology
Numerous computer‐assisted synoptic climatological classifications were performed to determine the validity of a commonly used ‘objective’ typing scheme. In this study, the correlation‐based Kirchhofer technique was used to classify a large population of daily mean sea level pressure grids. Unlike previous methodological studies, metric parameters were held constant while the form of the data used in the classifications was varied. It is demonstrated that sample size, number of grid points and the particular random sample used will affect the classification results, thereby confirming previous work that has suggested that such typing schemes are extremely subjective. It is recommended that users of correlation‐based techniques obtain a number of relatively large random samples of size n ≥ 1000 and perform several classification runs in order to reach an ‘optimum’ solution.