This is the text from the
People and Place poster
People and PLace
by Steven R Holloway, James Schumacher and Roland L Redmond
To better understand patterns of human settlement, migration, and related
economic activities, social scientists traditionally have relied on data gathered
directly from individuals and their families. Such "census" data can provide
information about the socioeconomic characteristics of people living in different
geographic areas at different times. In the United States, complete census data
are collected every 10 years from people residing in geographic units defined as
census "blocks". These collection units are of variable size and shape, but each
are delineated to include about 100 people. Census blocks are nested
hierarchically within block groups, census tracts, counties, and states.
Choropleth maps of population densities are made by simply dividing the number of
occurrences found in the enumeration unit (i.e. block group) by either the size
of the unit or the significant population unit. Such maps are used, for example,
to show distributions of people per square mile, housing units by age or value
per unit area, or even births per 1000 women.
Even though census data are collected at the "block" level, it would be
difficult to analyze or display all the data for these smallest collection units,
especially for an entire state or region of the country. Consequently, and for
reasons of privacy, population number is the only variable available for
individual blocks. At higher levels of organization, like the block group, census
tract, or county, more detailed information about people and households becomes
available. This information can be used to provide valuable socioeconomic profiles
of population groups.
The hierarchical relationship among census blocks, block groups, and
tracts is shown to the right for an area around Missoula, Montana. The 74 block
groups contain 2,238 blocks and fall within 18 census tracts and one county.
Again, because individual blocks contain approximately the same number of people,
it should be no surprise that blocks and block groups increase in size with
increasing distance from the urban center. To the far right is a choropleth map
showing population density for the area.
Choropleth maps are the easiest and most common way to display
tabular census data. Despite their simplicity, these maps have limited utility
for detailed spatial analysis of socioeconomic data, in part because by definition
the boundaries of both the enumeration units and the mapping units are the same
(i.e., the census block groups). In western North America, human populations are
concentrated in relatively few towns and cities, found at lower elevations and
along major river corridors. Relatively large expanses of land are essentially
uninhabited, especially block groups in tracks that lie farther away from urban
areas. When population density, or any other socioeconomic variable, is mapped by
traditional choropleth techniques, the results often tell us more about the size
and shape of the block group, or other enumeration unit, than about the people
actually living and working within them.
Dasymetric maps can circumvent many of the above limitations because
the boundaries of their mapping and enumeration units are independent. Described
below is the process of creating a dasymetric map of population density based on
land cover (and use), land ownership, and topography. Note that the enumeration
units are still the census block groups, but the mapping units become land cover
polygons within each block group. The first step (A) removes portions of all block
groups uninhabited by people. These areas were identified by selecting: 1) all
census blocks with zero population; 2) all lands owned by the state or federal
government; and 3) all corporate timberlands. Once uninhabited areas are removed,
populations in each block group can be reassigned to areas based on land cover,
land use, and slope (B). Four general land cover/land use classes were selected:
Urban, Agriculture, Forested, and Open. All Urban and Agriculture polygons were
assumed to be populated, whereas only those areas of Forested or Open land with a
slope of less than or equal to 15% were assigned people (C). People are then
assigned to land cover/land use polygons on a per unit area basis (D) - urban
polygons were assumed to have higher densities and were given a relative
weighting of 80 people per 100 population, whereas open polygons were weighted at
10, and agriculture and forested polygons at 5 each. The number of people
assigned to each mapping unit (land cover/land use type) within the enumeration
polygon (block group) was based on the formula shown below.
The dasymetric population map shows spatial patterns much more precisely than
is possible with traditional choropleth maps. The same process can be applied to
other census variables, such as income, gender, race, religion, occupation,
housing units, etc., and then analyzed across regions that extend over multiple
states and encompass numerous interdependent economies. This approach will allow
much more accurate and revealing analyses of spatial relationships between humans
and the natural landscape. For example, we can examine settlement patterns over
time and in relation to distances from water or forest or wilderness, and thereby
better understand why people live where they do in the Rocky Mountain West.
Finally, when coupled with other broad scale natural resource data that are
becoming available in digital form, the database can serve as a predictive tool
to identify areas most vulnerable to any negative consequences of future
development or land use change.
Missoula County occupies an area of 2,618 square miles in
northwestern Montana with a population in 1990 of 78,687. The bar charts below
each map legend show how much area in the entire county is occupied by each class
(see the companion poster). The choropleth map shows five classes of population
density; the lowest class, "less than 100", covering 2,531 sq. mi. (97%). The
dasymetric map, on the other hand, identifies a sixth class, unoccupied lands or
"none", which covers 2,354 sq. mi (90%). The dasymetric map's "less than 100"
class is only 185 sq. mi. (7%), a far more accurate picture of population density
in the county than the 97% indicated by the choropleth map. Changes in class size
in the higher classes appear relatively small at this scale, indicating that most
of the diVerences in the two mapping techniques are due to the exclusion of
unoccupied lands (A) and to a lesser extent slope restrictions (B). This suggests
that the database for excluded lands be developed with some care. The use of land
cover/land use (C) and relative densities (D) filters are CPU intensive and
careful consideration of their effectiveness should be made.
Go Back to Posters Page
Please mail input to Webmaster