Accurate and up-to-date population data is a prerequisite for many different applications, including risk
and vulnerability management. There is, however, a shortage of data with a high spatial resolution,
particularly in developing countries. Population densities are determined either by population census,
typically conducted every ten years, or from global grid-based (raster) population datasets with
relatively low resolution. Global population datasets are designed for global modeling studies,
including climate change research, and their resolution is generally too low for local or community
purposes. This paper presents a methodology for transforming population census data into grid-based
(raster) population data with a relatively high resolution (100 m). Population census data, land cover,
rural settlement data, and other geospatial datasets were utilized for a study area in the Khulna district
of Bangladesh. Local experts validated the geographic information system (GIS)-derived population
dataset as realistic and reasonably accurate. Our derived gridded population data was compared with
the available LandScan global dataset. The overall difference between the population for 2010, which
was projected from the 2001 census data, and our gridded population data was about 2.4% whereas the
LandScan data overestimated the population in the study area by 49%.