Integrating remote sensing and GIS for environmental justice research - Urban Remote Sensing: Monitoring, Synthesis and Modeling in the Urban Environment - page 216

Anderton et al ., 1994; Bowen et al ., 1995). However, the vast
majority of academic environmental justice research supports
the notion that the burden of environmental risk in urban areas
from pollution and waste management falls disproportionately
on minorities and the poor.
A number of authors have offered critiques of environmen-
tal justice research methods (Zimmerman, 1994; Bowen and
Wells, 2002; Maantay, 2002), perhaps the most fundamental of
which concerns the issue of causation. Clearly, simply because
there is evidence that there is an association between the spatial
distribution of race and environmental hazard does not neces-
sarily imply that the association is the direct result of acts of
intentional discrimination by environmental decision-makers or
others. This argument was embodied in the debate over com-
peting explanations of ''race versus class'' for explaining patterns
of environmental inequity (Cutter, 1995; Downey, 1998). Here,
the issue concerns which socioeconomic characteristics may be
the actual drivers of the relationship of socioeconomic charac-
ter with environmental hazard, when collinearity among race,
ethnicity, and indicators of class (e.g. poverty, income, and edu-
cational attainment) occurs. Many authors have noted that it is
impossible to disentangle the effects of race and class, as these
characteristics have been so closely woven together in the his-
torical processes of industrial development and residential and
labor segregation that have produced the patterns of environ-
mental inequity that may be currently observed (Pulido, 2000).
Several longitudinal quantitative studies, as well as a handful of
qualitative studies employing archival research, have attempted
to address such processes, with mixed evidence of intentional
discrimination (Boone and Modarres, 1999).
Szasz and Meuser (1997) provide a helpful typology of expla-
nations for environmental inequity. The typology distinguishes
between a situation in which a hazardous facility location is cho-
sen because of demographics and a situation in which a facility
location is chosen for reasons other than demographics. In the
former case, a location for a facility may be chosen because
of the economic benefits associated with facility development,
because of political disempowerment of the targeted commu-
nity, or because of outright racial prejudice on the part of facility
developers or policy makers. In the case where a facility is sited
for reasons other than demographics, the choice of facility loca-
tion may be based on the availability of inexpensive, industrial
land that (coincidentally or not) coincides with socioeconomic
disadvantage. Additionally, a facility may be initially located in
a community that is not socioeconomically disadvantaged or
in a relatively uninhabited area, and socioeconomic disadvan-
tage proximate to the facility increases subsequent to the facility
being built.
Other critiques of environmental justice research concern
methodology. First, the measurement of environmental risk is
subjective, in that a researcher must decide what constitutes
risk and how is the magnitude of that risk calculated. Early
environmental justice studies simply considered the presence
versus the absence of a hazard, say, aTSDF, as a proxy for risk. This
approach was criticized for being too blunt to capture magnitude
of actual risk. More recently, environmental justice researchers
have considered more sophisticated measures of hazard, such
as the distance from a hazardous facility (Mennis, 2002), the
density of hazardous facilities (Downey, 2003), the toxicity and
volume of toxic chemicals released (Sadd et al ., 1999), measures
of ambient pollution gathered from sensor networks (Jerret et al .,
2001), and distributions of hazards derived from computational

simulations of diffusion of toxic materials through various media
(Fisher, Kelly and Romm, 2005).
Another methodological challenge concerns the quantifi-
cation of the relationship between environmental hazard and
socioeconomic character. Since environmental data and socioe-
conomic data typically take different forms, with census-based
socioeconomic data often only made available in aggregated
form, data integration is necessary to develop case-based flat
files of observations used for statistical analysis. This challenge
has emerged in the debate over the appropriate ''scale of analy-
sis'' for environmental justice research, which is perhaps better
formulated as a debate over the appropriate analytical spatial
resolution. Researchers have argued for instance, whether US
zip codes, US Census tracts, or US Census block groups pro-
vide more accurate measures for capturing the demographic
character of the neighborhood surrounding a hazardous facility
(Willliams, 1999).
The choice of scale-of-analysis is related to what is often
referred to as the modifiable areal unit problem (MAUP) (Open-
shaw, 1984), which concerns the fact that different ways of
spatially aggregating punctiformdata influences statistical results.
The MAUP takes two forms. The first may be considered an issue
of scale, where, for instance the statistical results of analyzing
population data at the unit of the US Census block group will
differ from the results of an analysis of the same raw data aggre-
gated to the US Census tract level (where many block groups
spatially nest within a single tract). The second form is an issue
of partitioning, where the same dataset, partitioned into different
regions for the purpose of spatial aggregation, may yield different
statistical results, even when each partitioning scheme uses the
same number of regions (i.e. is roughly the same scale of analysis)
(Fotheringham and Wong, 1991).
Other authors have criticized environmental justice research
for its use of statistical methods that do not account for the
spatial nature of most environmental justice data. Ordinary
least squares (OLS) regression, for example, produces biased
parameter estimates in the presence of spatial autocorrelation
in the model residuals. Because socioeconomic and environ-
mental characteristics tend to exhibit strong spatial dependency,
models of environmental equity often violate the assumptions
of OLS regression and other conventional statistics. A number
of environmental justice researchers have addressed this issue
by incorporating statistical methods adapted for spatial data,
such as spatial econometric modeling (Buzzelli et al ., 2003) and
geographically weighted regression (Mennis and Jordan, 2005).

16.3 Remote sensing for
environmental equity
analysis

The role of environmental remote sensing in environmental
justice research is two-fold. First, remote sensing may be used
to quantify environmental characteristics of the earth surface,
including environmental hazards and environmental amenities,
so that the locations of such hazards and amenities may be
compared to distributions of demographic characteristics. Typ-
ically, this application of remote sensing requires medium to
high resolution imagery (e.g. 1 m-30 m) that indicates land