Environmental Engineering Reference
In-Depth Information
the time) half-century-old US Geological Survey (USGS) topo-
graphic sheet and, more importantly, his personal knowledge
of Cape Cod in order to make inferences about which portions
of Cape Cod were populated and to what degree of density.
He assigned population densities,
apriori
, to sparsely populated
township subdivisions and then computed the population den-
sities for the remaining more densely-populated subdivisions
in magnitudes that would ensure that the overall population
densities for each township remained the same as before dasy-
metric recalculation. Wright's method can be described more
formally as:
computational devices and the dearth of suitable ancillary data.
Beginning in the late 1980s and early 1990s, interest was renewed
in dasymetric mapping, as several researchers in the United States
and the United Kingdom investigated various methods of areal
interpolation (Goodchild and Lam, 1980; Lam, 1983; Goodchild,
Anselin and Deichmann, 1993; Flowerdew and Green, 1989;
Flowerdew, Green and Kehris, 1991; Langford, Maguire and
Unwin, 1991; Langford and Unwin, 1994). The development
and proliferation of geographic information systems (GIS) in the
1990s along with the advent of systematically collected satellite
remote sensor data presented new opportunities for the practical
application of dasymetric mapping techniques.
D
D
m
a
m
1
D
n
=
a
m
−
a
m
where
D
is the mean population density per area (in Wright's
case, the township),
D
m
is the estimated population density
for subarea
m
(in Wright's case, the larger and less densely-
population township subdivision,
D
n
is the estimated population
density for subarea
n
(in Wright's example, the smaller and
more densely-population township subdivision), and
a
m
is sub-
area
m
's proportion of the total area. Knowing
D
and
a
m
,and
treating
D
m
as a known quantity, Wright was able to solve for
D
n
, for each township. He then created a dasymetric map of
the resulting township subdivision population densities. Wright
admits that his methods of assigning population densities are
not exact, but represent a good approximation, based on ''con-
trolled guesswork
...
perhaps the best that could be accomplished
without intensive field work'' (Wright, 1936, p. 104).
Most follow-on dasymetric mapping researchers refer to this
as the first-published use of dasymetric mapping, although some
authors did acknowledge, including Wright, the Russian origin
of at least the word
dasymetric
: ''It is a map to which the Russians
have applied the term 'dasymetric' (density measuring)'' (Wright,
1936, p. 104). In a footnote, Wright also refers to a similar type
map that was printed in an English language publication three
years earlier by the American Geographical Society. There are
at least two additional examples of the use and description of
dasymetric mapping,
per se
, in the English-language literature
prior to Wright in 1936 (De Geer, 1926; and Semenov-Tian-
Shansky, 1928).
Fortunately, Petrov (2008) demystifies the origins of the
term ''dasymetric map'' and discusses early uses by Russians,
particularly Benjamin Semenov-Tian-Shansky. Petrov and oth-
ers (MacEachren, 1979; McCleary, 1969; Robinson, 1982), as
cited in Petrov, 2008; and Maantay, Maroko and Herrmann,
2007) describe perhaps the earliest map that appears to use
dasymetric mapping principles: George Julius Poulett Scrope's
world population density map of 1833 (Scrope, 1833). Four
years later, Henry Drury Harness created a population-density
map for Ireland (Robinson, 1955). Both Scrope's and Harness'
maps distinguished between areas based on population density;
however, as Petrov (2008) points out, neither described their
process for constructing the maps. The first description of the
dasymetric methodology in the English-speaking literature was
by De Geer (1926) who reviewed the series of population density
maps for Europe that were compiled by Semenov-Tian-Shansky.
Semenov-Tian-Shansky subsequently described his dasymetric
maps two years later in the same journal (Semenov-Tian-Shansky,
1928).
Further development and use of dasymetric mapping stag-
nated for several decades, perhaps largely due to the analytic
complexity of the methods along with the absence of automated
1
−
−
14.2.4
Dasymetric mapping
variations
14.2.4.1
Binary dasymetric
The most basic technique for adding intelligence to the process of
creating a dasymetric map is to differentiate between populated
and unpopulated areas. Population or other sociodemographic
data are reapportioned to those areas that are considered to be
populated. The mapping units themselves are spatially refined,
and the resulting population densities are changed due to the
change in areal support; in a binary approach, this results in
higher population densities than in the original population map.
It is necessary to use ancillary data to make the distinc-
tion between populated and unpopulated areas. For the binary
approach, it is assumed that population can only occur, and
for dasymetric mapping should only be depicted, in residential
areas. Thus, the ancillary data or information should accurately
estimate those areas that can be considered residential in nature.
These may include land-use/landcover categories (LULC) such as
Urban or Built-up Land, which may involve subcategories such
as Residential, Mixed Urban or Build-up Land, Other Urban or
Built-up Land (Anderson
et al
., 1976), as well as rural areas in
which people may reside, such as forested and agricultural land
(e.g., Anderson's Level I categories 2 and 4). The choice depends
upontheassumptionsoftheresearchers,thescaleatwhichLULC
data are available, and the particular characteristics of the study
area. In cases where multiple LULC classes are used, these classes
are collapsed into one overall class that represents populated
areas; by default the remainder of the study area is considered
unpopulated; thus, the approach is considered ''binary''. This
methodology for dasymetric mapping was described in detail
by Langford and Unwin (1994) and subsequently referred to as
the ''binary dasymetric method'' by Eicher and Brewer (2001).
This technique also has been referred to at least once as ''filtered
areal weighting'' (Maantay, Maroko and Herrmann, 2007). Many
examples of its use exist (Langford and Unwin, 1994; Holt, Lo snd
Hodler, 2004; Lo, 2004; Poulsen and Kennedy, 2004; Chen
et al
.,
2004; Langford and Higgs, 2006; Langford, 2007) and Eicher and
Brewer (2001) have compared this method to other dasymetric
mapping techniques. We provide a current example of binary
dasymetric mapping and describe the methodological steps in
further detail later.
Fisher and Langford found that binary dasymetric mapping
was very robust to classification error in LULC derivation from
satellite imagery. Classification errors up to 40% could be expe-
rienced before the resulting binary dasymetric map accuracy
Search WWH ::
Custom Search