Premium
Combining data from multiple years or areas to improve variogram estimation
Author(s) -
Walter John F.,
Christman Mary C.,
Hoenig John M.,
Mann Roger
Publication year - 2007
Publication title -
environmetrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.68
H-Index - 58
eISSN - 1099-095X
pISSN - 1180-4009
DOI - 10.1002/env.825
Subject(s) - variogram , spatial analysis , autocorrelation , geostatistics , statistics , kriging , estimation , data set , computer science , spatial variability , environmental science , mathematics , engineering , systems engineering
A requirement for geostatistical prediction is estimation of the variogram from the data. Often low sample size is a major impediment to elucidating a variogram even for a highly autocorrelated spatial process. This paper presents a methodology for improving variogram estimation when samples exist from multiple years or regions sharing a similar process for generating spatial autocorrelation. Such samples may come from annual monitoring programs for natural resources or from multiple geologic regions. As each set of samples contains some information on the spatial autocorrelation, combining information through the construction of a combined variogram cloud and binning to obtain a common variogram improves the estimation of the variogram. In both simulations and in real datasets of oyster abundance the method proposed here reduces the likelihood of failing to obtain a variogram from a set of samples and improves the efficiency of variogram estimation. In practice, the benefits obtained by estimating an otherwise elusive variogram generally outweigh the costs of using a slightly incorrect variogram model if different sampling stanzas are combined when they do not share the same spatial autocorrelation. Copyright © 2007 John Wiley & Sons, Ltd.