Premium
The effect of station drift on the estimation of shifts in meteorological records due to site changes
Author(s) -
Rhoades D. A.,
Neill W. A.
Publication year - 1995
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/joc.3370150206
Subject(s) - series (stratigraphy) , autocorrelation , autoregressive model , environmental science , meteorology , variance (accounting) , random walk , climatology , statistics , mathematics , geology , geography , paleontology , accounting , business
The presence of station drift can seriously affect tests of significance of the shift due to a site change when neighbouring stations are used to adjust a meteorological series. The series of differences between a station and its neighbours (on which such tests are based) are often moderately autocorrelated; this is evidence that station drift effects commonly exist. Two types of station drift, both of which appear to be present in real meteorological series, are considered here: transient drift, modelled as a first‐order autoregressive process, and enduring drift, modelled as a random walk. A method is given to estimate the size of each of these effects in real data sets, using data from the periods between site changes, and hence to correct the variance of the estimated shift for station drift. The optimal length of the adjustment period can also be determined; it depends mainly on the size of the enduring station drift.