Geoscience Reference
In-Depth Information
within an image (Garrigues et al., 2008a). Modeling the variogram of NDVI images
with high spatial resolution is an effi cient approach for characterizing and quantify-
ing heterogeneous spatial components (spatial variability and spatial structure) of a
landscape and the spatial heterogeneity of vegetation cover at the landscape level
(Garrigues et al., 2006, 2008a).
Reliable data analysis of spatially distributed data requires the use of appropri-
ate statistical tools and a sound data sampling strategy (Fortin and Edwards, 2001).
Spatial sampling schemes have been developed to determine the sampling locations
that cover the variation in environmental properties in a given area (Minasny and
McBratney, 2007). Moreover, data samples are transformed via a series of interpre-
tation steps to obtain complete descriptions of phenomena of interest (Edwards and
Fortin, 2001). Different sampling schemes are, say, random, systematic, stratifi ed, or
nested schemes (Edwards and Fortin, 2001; Thompson, 1992). The LHS is a stratifi ed
random procedure that is an effi cient way of sampling variables from their multivari-
ate distributions (Minasny and McBratney, 2006). Initially developed for Monte-Carlo
simulation, LHS effi ciently selects input variables for computer models (Iman and
Conover, 1980; McKay et al., 1979). Kriging, a geostatistical method, is a linear in-
terpolation approach that provides a best linear unbiased estimator (BLUE) for quan-
tities that vary spatially (Lin et al., 2008d). However, kriging interpolate algorithms
generate maps of best local estimate and generally smooth out the local details of the
spatial variation of an attribute (Goovaters, 1997). For sampled data, a geostatistical
conditional simulation technique, such as SGS, can be applied to generate multiple re-
alizations, including an error component, which is absent from classical interpolation
approaches (Lin et al., 2008d). In such conditional simulations, all generated realiza-
tions reproduce available data at measurement locations, and, on average, reproduce a
data histogram and a model of spatial correlations (i.e., variogram) between observa-
tions (Kyriakidis, 2001). In SGS, Gaussian transformation of available measurements
is simulated, such that each simulated value is conditional on original data and all
previously simulated values (Deutsch and Journel, 1992; Goovaters, 1997; Kyriakidis,
2001; Lin et al., 2008d). Geostatistical conditional simulations have been widely ap-
plied to simulate the spatial variability and spatial distribution of interest in many
fi elds. Moreover, geostatistical simulation techniques with LHS have been applied to
simulate Gaussian random fi elds (Kyriakidis, 2001; Pebesma and Heuvelink, 1999;
Xu et al., 2005; Zhang and Pinder, 2004).
This study applied variogram analysis to delineate spatial variations of NDVI im-
ages before and after large physical disturbances in central Taiwan. The NDVI data
derived from SPOT images before and after the Chi-Chi earthquake (ML = 7.3 on the
Richter scale) in the Chenyulan basin, Taiwan, as well as images before and after four
large typhoons (Xangsane, Toraji, Dujuan, and Mindulle) were analyzed to identify the
spatial patterns of landscapes caused by these major disturbances. Landscape spatial
patterns of different disturbance regimes were discussed. Moreover, cLHS schemes
with NDVI images were used to select spatial samples from actual NDVI images to
detect landscape changes induced by a series of large disturbances. The best cLHS
