Estimating The Population Distribution
in a County Area in China Based on
Impervious Surfaces
Honglei Zhu, Ying Li, Zhaoli Liu, and Bolin Fu
Abstract
Information on the county level population distribution is
essential for supporting a wide variety of local planning pro-
cesses. Previous researchers found that impervious surfaces
had a strong linear relationship with population. But few
studies have reported estimating population in a county area
of China with impervious surfaces. In this paper, a population
distribution model for a county area in China was established
using demographic information at the county level (i.e., total
population, non-agricultural population, and agricultural
population) and impervious surfaces derived from Land-
sat TM images. The model was conducted to distribute the
population of Fujin County, Heilongjiang Province, China.
Population data of 123 villages and 11 towns of Fujin County
were used to evaluate the model. The results indicated that
the model performs better than area-based approaches, par-
ticularly in the estimation of town populations (17.86 percent
mean relative error). The results of village population estima-
tion were acceptable, with a mean relative error of 25.3 per-
cent. As a result, impervious surfaces cannot provide height
information and identify non-residential areas in developed
areas of zhens (urban cores of towns). The model suggests
the difficulty of non-agricultural population estimation.
Introduction
Information on the population distribution is essential for
decision makers in local development and for a better un-
derstanding and response to many potential social, political,
economic, and environmental problems (Lu and Weng, 2006).
In China, the census population data available to the public
are at the county level (i.e., total population of a county).
Therefore, timely and accurate estimation of population
distribution, particularly in county areas, is critical for public
planning and research.
As an alternative to the traditional census statistic method,
remote-sensing data have an advantage due to the capability
for collecting data over large geographic areas and the high
temporal frequency. Since the 1970s, remote sensing has been
used to estimate population distributions more frequently.
Four categories were proposed (Lo, 1986; Lu
et al.
, 2006; Wang
and Wu, 2010): (a) counts of dwelling units (Lo 1986, 1995), (b)
relations between population and built-up areas (Sutton
et al.
,
1997; Lo, 2001; Lu,
et al.
, 2010; Lu
et al.
, 2011; Azar
et al.
, 2013),
(c) land use classification (Langford
et al.
, 1991; Lo, 2003), and
(d) spectral features of individual pixels (Iisaka and Hegedua,
1982; Webster, 1996; Harvey, 2002; Jensen
et al
., 2010).
At the regional level, many studies have focused on the
second and third approaches, using medium-spatial-resolution
remote-sensing data such as the Landsat Thematic Mapper
(
TM
) (Li and Weng, 2005; Lu
et al
., 2006). For example, meth-
ods have considered the areas of cities (Lo, 1986, 2001; Sutton
et al
., 2001), areas of land use (Langford, 2006) and, partic-
ularly, areas of impervious surfaces (Lu
et al
., 2006; Wu and
Murray, 2005; Azar
et al
., 2010). The method linking popula-
tion and impervious surfaces is robust and attractive because
(a) impervious surfaces are stable and almost independent of
seasonal change and atmospheric conditions (Wu and Yuan,
2007), and (b) it requires no
a priori
knowledge of land cover.
Few studies have reported estimating population in a
county area of China with impervious surfaces. In this paper,
we develop a population distribution model with demo-
graphic information at the county level (i.e., total population,
non-agricultural population and agricultural population) and
impervious surfaces derived from Landsat
TM
images. The
manuscript is organized as follows. The next section de-
scribes the study area and data used in this study, followed by
the methods used for impervious surface estimation, populat-
ed places extraction, and the development of the population
distribution model. Then, the next section shows the accura-
cy assessment of the impervious surfaces, populated places
extraction, and population estimation, followed by the results
and the conclusion of this study.
Study Area and Datasets
Study Area
The county of Fujin (excluding Jiansanjiang farms), Heilong-
jiang Province, was selected to implement this study. Fujin
County has a geographic area of 5,172 km
2
, located between
131°25' and 133°26' E, and 46°51' and 47°31' N (Figure 1).
Fujin County is the key grain and sugar beet base in Heilong-
jiang Province and one of the country᾿s 100 major grain-pro-
ducing counties. With an annual grain output of 3.07 billion
kilograms, Fujin is known as the “Chinese soybean town” and
“China's northeast rice town.” Therefore, timely and accurate
population information is valuable for agricultural production
and land use planning.
According to the Fujin 2009 Statistical Yearbook, the county
has 11 towns and 266 villages. As the fourth-level administrative
Honglei Zhu and Bolin Fu are with the University of Chinese
Academy of Sciences, No.19A Yuquan Road, Beijing 100049,
China.
Ying Li and Zhaoli Liu are with the Northeast Institute of
Geography and Agroecology, Chinese Academy of Science,
4888 Shengbei Street, Changchun 130102, P. R. China (liying.
).
Photogrammetric Engineering & Remote Sensing
Vol. 81, No. 2, February 2015, pp. 155–163.
0099-1112/15/812–155
© 2014 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.81.2.155
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
February 2015
155
