Environmental Engineering Reference
In-Depth Information
50 states and Puerto Rico (MRLC, 2009). LULC data (LandPro)
for 2001 were also obtained from the Atlanta Regional Com-
mission, the regional planning organization for metropolitan
Atlanta; these data were derived from manual interpretation and
GIS digitizing of true-color aerial photographs with 1 m spatial
resolution obtained in 2001. Parcel data for 2001 were obtained
from the Gwinnett County GIS Department. We used these three
sources of ancillary data to test for differences in dasymetric
population estimates based on different spatial resolutions and
postulated LULC accuracies. We also obtained data on the loca-
tions of parks, airports, interstate and other primary highways,
railroads, and landmarks (ESRI Data and Maps, 2008, ESRI, Red-
lands, CA); these data were used to create spatial masks for areas
that were assumed, apriori , to have no residential population.
Boundary data for counties, Census block groups, Census blocks,
and voting districts were downloaded from US Census 2008
TIGER/Line files for the 2000 Census boundaries (US Census
Bureau, 2008).
(a) Census 2000 total population by blocks
Total Persons
0
1 - 100
101 - 350
351 - 850
851 - 1800
1801 - 5735
County boundary
N
W
E
S
14.3.2 Dasymetric maps
0
5
10
Kilometers
20
(b) Census 2000 population density by blocks
We performed all the analysis using ArcGIS (ArcInfo) ver-
sion 9.3.1 (Environmental Systems Research Institute, Redlands,
CA) with the Spatial Analyst Extension. We used four pro-
cesses to estimate the population to the pixel (e.g., grid) level:
three implementations of the binary dasymetric method and
one implementation of the N-class dasymetric method. First, we
employed the binary dasymetric method (Langford and Unwin,
1994) using the NLCD 2001 LULC as the ancillary data layer.
We refined the spatial extent of the populated area by extracting
parks, airports, landmarks (excluding D10 military installations
and reservations; D31 hospitals, urgent care facilities, and clinics;
D37 federal penitentiaries, state prisons, or prison farms; por-
tion of D43 educational institutions - colleges and universities
for which the Census Bureau collects institutional population
counts), interstates, highways and railroads as a binary mask
after recoding a clipped raster layer as 1
Persons per
Square Kilometer
residential (NLCD
classes: Developed Low Intensity [22] and Developed Medium
Intensity [23], Developed Open Space [21] and Developed High
Intensity [24] may be appropriate additions to residential area,
based on the particular study area.) and 0
=
0.0
0.1 - 100.0
100.1 - 350.0
350.1 - 850.0
850.1 - 1800.0
1800.1 - 410380.6
County boundary
N
W
E
non-residential (all
other NLCD classes). We used the block group population as
our population source layer. We used the Tabulate Area tool
in ArcGIS for block groups and the reclassified NLCD 2001
layer to obtain residential population areas for each block group.
We joined the table back to the block group feature class and
calculated population densities for all block groups ( n = 208).
We converted block groups to a raster layer of population den-
sity. Finally, we used map algebra, multiplying the block group
population density raster by Reclassified NLCD 2001 (binary
raster layer), to derive the residential area population density
(Fig. 14.3a).
For the second and third implementations we used binary
dasymetric mapping with LandPro 2001 and the Gwinnett
County 2001 parcel data as the ancillary data layers. These
two ancillary layers provided increased spatial resolution and/or
attribute specificity for delineating residential areas. The GIS
processing steps for both implementations were identical to
those of the first implementation, which we described in the
preceding paragraph. The only exception was that we did not
=
S
0
5
10
Kilometers
20
FIGURE 14.2 Block-level population (a) and population
density (b) at the block level for Gwinnett County.
2000) at the block and block group levels for Gwinnett County,
Georgia (FIPS = 13135). Figure 14.2a and b depict the 2000 pop-
ulation and population density at the block level for Gwinnett
County. We obtained three types of ancillary data in order to
evaluate their impact on dasymetric mapping accuracy. LULC
data for 2001 (the most proximate data available for comparison
to the 2000 sociodemographic data) were obtained from the
National Land Cover Database (NLCD) produced by the Multi-
Resolution Land Characteristics Consortium (MRLC, 2009), a
group of nine US federal government agencies whose objective
is acquiring Landsat 5 and Landsat 7 imagery and generating a
national LULC database, at 30 meters spatial resolution, for the
Search WWH ::




Custom Search