Geoscience Reference
In-Depth Information
Massachusetts Geographic Information System (URL
http://www.mass.gov/mgis
/).
The land use data were originally interpreted from 1:25,000 aerial photographs
by the Resource Mapping Project at the University of Massachusetts, Amherst
(MassGIS
2005
).
The drainage area (watershed) for each sampling site was delineated from digi-
tal elevation data provided by the USGS National Elevation Dataset (NED) 1 Arc
Second (about 30 m resolution, URL
http://seamless.usgs.gov
/ website/Seamless/)
using ArcGIS spatial analysis tools, and then used for the calculation of urbanization
indicators.
PDLU and PD for each sampling site were calculated in ArcGIS by overlapping
land use and population layers to the drainage area layer. PDLU for a sampling site
was calculated by dividing the amount of the developed lands within the watershed
of the site by the total area of the watershed. PD for a sampling site was calculated
by dividing the total population within the watershed of the site by the total area of
the watershed. It is always a challenge for geospatial technologies application to link
a variable limited by administrative boundaries (e.g. population density) to another
that derived from natural features (e.g. land cover, soil, climate, etc.). To estimate
the total population within a watershed, the proportion of each census block that fell
within the watershed was calculated, and then multiplied by the population in the
entire census block. Obviously, this will introduce some error in population density
of watersheds because the calculation assumes that population is evenly distributed
over a census block. However, census blocks are the smallest spatial units for the
population data available in the study area. Compared with other spatial units for
population data (e.g. municipalities and counties), the error was relatively small.
This method has been widely used to derive population variables in many previous
studies (Ahearn et al.
2005
; Alberti et al.
2007
; Xian et al.
2007
). Through these
steps, a linkage of water quality for points to urbanization indicators for areas was
established. They were used for further statistical and spatial analyses.
The temporal discrepancy between the water quality data (1990-2005), and the
land use data (1999), and population data (2000) might raise some doubt about the
comparability of the water quality and urbanization indicators. Temporal agreement
was not vital with respect to the land use and population data since they did not
normally change dramatically over a short time period. As for the water quality
indicators, they might change daily or hourly. However, for this study that takes
advantage of geospatial technologies to analyze data from dozens of water quality
sampling sites on a regional scale at once, it is infeasible to set up many sampling
sites with multiple water quality indicators over this study area during a short time
period (e.g. 1999 or 2000). In addition, the purpose of this study is to test the general
pattern in the spatially varying relationship between water quality and urbanization,
not to explore their temporal relationship (i.e. how water quality changes over time
affected by urbanization). The water quality data from 1990 to 2005 is used to rep-
resent the average situation of water quality at different sampling sites associated
with different urbanization levels. The spatial relationship between water quality
and urbanization can have implication for their temporal relationship. Therefore, as
