Premium
The detection of change in spatial processes
Author(s) -
Martin E. B.,
Scott E. M.
Publication year - 1995
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.3170060205
Subject(s) - range (aeronautics) , representation (politics) , scalar (mathematics) , kernel density estimation , statistics , computer science , spatial distribution , statistical physics , econometrics , mathematics , estimator , materials science , geometry , physics , politics , political science , law , composite material
The question of whether the spatial distribution of a characteristic of interest has changed over time is of interest within the environmental sciences, since spatial evolution may be informative about the underlying environmental processes affecting the characteristic. Techniques currently available for investigating change in spatially defined processes are in the main either subjective in nature or else ignore the spatial component. A hypothesis testing procedure based on a contour representation of the spatial distribution of the characteristic of interest has been developed and will be described in this paper. The approach does not require that observations be recorded at the same location, and thus is generally applicable. The degree of change between the contoured surfaces has been quantified in terms of various statistics which describe the rotational, translational and scalar aspects of change. The propoerties of the various statistics have been investigated through an extensive simulation study, which examined a range of factors, including the surface fitting procedure (Kernel density estimation) the number and choice of contour levels and the effect of random noise present within the system due to the measurement process. Finally the method is illustrated on the historical climate record of the U.S.A. The temperature field of the contiguous U.S.A. is studied for change over a 50 year period.