DMSP/OLS IMAGERY TO ESTIMATE POPULATION 
Mikiyasu Nakayama 
Associate Professor, Faculty of Agriculture, Utsunomiya University 
Commission VII, Working Group 5 
KEYWORDS: Population, DMSP, OLS 
ABSTRACT: The Operational Linescan System (OLS) 
on the Defense Meteorological Satellite Program (DMSP) 
has a unique capabılity to detect low levels of visible - near 
infrared radiance at night. An attempt was thus made to 
estimate population using the "city lights" data. The 
Hokkaido islandin Japan was selected as the case study area. 
The coefficient of correlation between the digital number of 
the DMSP/OLS imagery and the logarithm of the 
population was 0.53. Multiple regression models suggest 
that neither normalized vegetation index nor altitude does 
not significantly contribute to improve the accuracy in 
estimating the population. Analysis was also made for 
population in square grids of 10 km by 10 km to 40 km by 
40 km. The correlation was calculated between the number 
of bright pixels in a grid and population in the same gnd 
These models give better results (with R — 0.60 to 0.86). It 
is thus suggested the number of "bright" pixels in a grid (in 
this case between 100 and 1600 sq. km.) obtained may serve 
as a better indicator of population. Another case study 
conducted on the east coast of U.S.A., with "synthesized" 
city lights data showed almost identical result. The case 
study in U.S.A. showed better results as compared with 
the same for Hokkaido. It was thus assumed that population 
of a region could be estimated fairly accurately with 
DMSP/OLS night time imagery. 
POPULATION AS CAUSE OF 
ENVIRONMENTAL DEGRADATION 
Presence of humanity is the major factor of environmental 
degradation. Obtaining population data is thus essential to 
identify the probability of environmental degradation to 
Occur in a certain area. However, in most part of the world 
particularly in the developing world, little reliable data on 
population distribution are available with reasonable spatial 
resolution (e.g. a few km.) Moreover, population data (if 
any) are generally given by administrative units (e.g. 
county, state or prefecture) and they are not given as grid 
(mesh) data. 
Despite the increasing use of satellite imagery in various 
environmental fields, few has attempted to estimate the 
population distribution with satellite data, except for a 
couple of papers dealing with such highly populated areas as 
Tokyo and Hong Kong. 
ATTEMPT TO EMPLOY DMSP/OLS DATA TO 
ESTIMATE POPULATION 
An attempt is now being made by the author to figure out 
the feasibility of estimating population with data by 
Operational Linescan System (OLS) on the Defense 
Meteorological Satellite Program (DMSP). The OLS hasa 
unique capability to detect low levels of visible to near 
infrared radiance at night, thanks to its low light sensing 
capabilities, which is more than four orders of magnitude 
lower than other sensors such as NOAA/AVHRR or 
LANDSAT/TM (Elvidge, et. al., 1997a). 
The "city lights" are clearly visible in night-time scenes. 
The magnitude of "city lights" in a given areais supposed to 
be proportional to the population of the area. Elvidge 
showed that the "lit" area of a nation has high correlation 
with the population of the nation (Elvidge, 1997b), though 
few analysis has been made on the feasibility of estimating 
population using the "city lights" data. 
  
Figure 1: Nighttime Imagery of Hokkaido 
456 International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998