Geoscience Reference
In-Depth Information
the whole study area, so it is not easy to use them to explore local variations in the
impact of urbanization on water quality.
In recent years, a simple but powerful statistical method named geograph-
ically weighted regression (GWR) has been developed to explore the spatial
non-stationarity among spatial variables (Brunsdon et al.
1998
; Fotheringham et al.
2002
). Spatial non-stationarity means that the relationships between independent
and dependent variables are not constant over space (Fotheringham et al.
2002
).
Unlike OLS that generates a set of regression results applied to all the regression
points, GWR can calculate a set of local regression results for each regression point.
In other words, the model parameters (e.g. regression coefficients and
R
2
)from
GWR might be different at different sampling points. The local model parameters
can be mapped using visualization (e.g. GIS) so that local spatial variation in the
regression relationships can be investigated. GWR is an increasing widely used spa-
tial statistical method that is an important geospatial technology along with GIS
and remote sensing (Fotheringham et al.
2002
). Therefore, GWR provides a useful
geospatial tool to explore the spatially varying relationships among spatial variables,
such as population density, land uses, water quality, and climate variables, which
usually change over space. This technique has been applied in some fields, such
as ecology (Shi et al.
2006
), social studies (Fotheringham et al.
2001
), and urban
studies (Yu
2006
), which are all good examples of the application of geospatial
technology. However, GWR is seldom used in water resources research.
This study applied the GWR technique to examine the spatially varying relation-
ships between urbanization and water quality indicators in eastern Massachusetts,
USA. GWR is an exploratory spatial data analysis (ESDA) technique. Thus, the
objective of this study is not so much to study mechanism of water pollution, but
to explore how the relationships between urbanization and water quality indicators
change across space and to compare the influence of urbanization on water quality
among watersheds with different degrees of urbanization distributed from central
city to suburbs.
9.2 Study Area
An area covering fifteen river basins in eastern Massachusetts as defined by the U.S.
Geological Survey (USGS) and the Massachusetts Department of Environmental
Managements (MADEM) has been chosen for this study. The study area contains
metropolitan Boston and its surrounding areas up to 80 km away from Boston in
eastern Massachusetts (Fig.
9.1
). The area is about 10,000 km
2
. The population
in year 2000 is about 5.2 million. This area is more densely populated, urbanized
and industrialized than most parts of New England. Most of lands in metropoli-
tan Boston and the City of Worcester are highly developed, including residential,
commercial, and industrial lands. The rest of the study area is mainly covered
by forest. A small amount of agricultural land also exists in some places outside
