Remote Sensing and Geospatial Analysis for Landscape Pattern Characterization - Landscape Ecology for Sustainable Environment and Culture - page 206

centerpiece to address the spatial interaction between landscape patterns and
ecological processes that can help understand the intrinsic causality and underlying
landscape dynamics. On one hand, landscape patterns are strongly influenced by
ecological processes that include all aspects of biological, chemical, physical,
hydrological, and human-dimensional processes of the ecosystem. On the other
hand, landscape patterns can significantly affect ecological processes and land-
scape dynamics across spatial and temporal scales. To assess the two ways of
relationship, one needs to integrate the ecologically-oriented, vertical approach
with the geographically-oriented, horizontal approach that incorporates aggre-
gated, integrative environmental parameters (Bastian 2001 ). While a rich pool of
landscape ecological literatures have discussed specific pattern-process relation-
ships (e.g. Turner 1989 ; McGarigal and McComb 1995 ;Wu 2006 ), here we direct
our attention on some generic methodological issues for integration and synthesis
in a GIS environment.
11.3.1 Linking Landscape Patterns with Processes
Relating landscape spatial patterns with ecological processes involves the inte-
gration and synthesis of theories and technologies across spatial and temporal
scales, which can be pursued through either the qualitative or quantitative
approach. Given the scope of this chapter, we herewith limit our further discussion
to some technical issues when using the quantitative approach, especially multi-
variate statistical analysis, to address the pattern-process linkage at the broad scale
level. Multivariate statistical methods including linear and nonlinear multiple
regression can be used to examine the pattern-process relationship. When con-
structing a multivariate regression model to assess how ecological processes would
influence landscape patterns, each of the landscape metrics should be treated as a
dependent variable, while various biological, chemical, physical, hydrological, and
human-dimensional variables as independent variables (e.g. Lo and Yang 2002 ).
On the other hand, when examining how landscape patterns would affect eco-
logical processes, landscape metrics should be treated as independent variables,
while each of specific ecological indicators as a dependent variable (e.g. Yang
2012 ). Several statistical methods, such as ordinary least squares (OLS) regression
and logistic regression, can be used to determine the pattern-process relationship.
There are several issues one should pay close attention to when using the above
empirical method to study the pattern-process relationship. Firstly, since all
dependent and candidate independent variables are usually aggregated by areal
units, how these units are defined in terms of scale and zoning systems can affect the
results of parameter estimates in multivariate statistical analysis. This is actually the
modifiable areal unit problem (MAUP) that has been extensively discussed in
geography and spatial science literature (e.g. Openshaw 1984 ; Fotheringham and
Wong 1991 ; Jelinski and Wu 1996 ; Dark and Bram 2007 ). The modifiable areal unit