Environmental Engineering Reference
In-Depth Information
Gradient Analysis
Categorical Analysis
a
b
Quantitative Data
Spatial Statistics
Spatial Autocorrelation
Range
Anisotropy
Qualitative Data
Landscape Metrics
Composition
(patch type and proportion)
Configuration
(patch size, patch shape,
patch orientation, spatial
arrangement)
Fig. 7.3 Spatial analysis can be undertaken on different types of spatial data. (a) Raster-based
quantitative spatial data (e.g., forest height, basal area, NDVI) that can be analyzed using spatial
statistics to determine the intensity, spatial range, and directionality (anisotropy) of the spatial
pattern. (b) Categorical and qualitative forest data (e.g., species, stand age) require a different
analytical approach that typically includes landscape pattern metrics
vector format (points, lines, polygons) and the raster (i.e., pixel) format [
60
].
Representation of spatial structure in one form of data does not preclude use of
another form. For example, annual polygons of insect defoliation data can be
converted into raster form and analyses can be undertaken using time series of
raster values at a specific location [
61
]. Furthermore, binary rasters (presence/
absence) can be converted to continuous rasters by increasing the cell size and
counting the number of “presence” pixels surrounding a focal pixel using
4-neighbour, 8-neighbour, 16-neighbour rules and assigning this value to the to
the new, larger pixel.
Raster data can be used to represent any continuous variable. In contrast to
point data, in raster data types, the information fully covers the extent of the study
area. There is also a unique grain (or cell size) to each raster “pixel” that
determines the subarea of continuous space that is discretized by the raster cell.
The selection of raster grain can have important consequences to the results of
spatial analyses [
6
]. Raster data can be used to represent any number of spatial
variables relevant to disturbance ecology including, but not limited to, tree species
[
62
], stand age [
18
,
63
], basal area [
64
], insect damage [
61
],andnumberoffire
occurrences [
65
]. The continuous coverage of raster data makes it amenable to
many different analytical techniques such local quadrat variance, lacunarity, and
wavelets [
56
].
