Table 1.
Summary information on input population count data used in the construction of the AsiaPop datasets.
Figure 1.
Köppen-Geiger climate classification [34] and associated land cover population density weightings for the Southeast Asia region.
Figure 2.The.
modelled spatial distribution of population in Southeast Asia, 2015.
Output datasets are produced in geographic projection with square grid squares that are 8.33×10−4 degrees on a side (∼100 meters at the equator).
Figure 3.
Visual comparisons are made of the AsiaPop, GRUMP v1, and GPW v3 gridded population datasets for Phnom Penh, Cambodia and Hanoi, Vietnam.
These datasets were generated using census data one level coarser than the finest detail available.
Figure 4.
Errors mapped for population distribution datasets produced using the AsiaPop, GRUMP v1, and GPW v3 methodologies for a) Cambodia and b) Phnom Penh, Cambodia.
Populations underestimated are highlight in blue and overestimated values are shown in red. The white outline highlights the boundary of Phnom Penh (source: GRUMP urban extent dataset).
Figure 5.
Scatter plots of observed, level 3 census data plotted against estimated modelled data from 2008.
Data for Cambodia are shown for the a) AsiaPop, b) GRUMP, and c) GPW methods. In the d) bean plot, the distribution of mean absolute error for Cambodia population estimates is plotted for all three model types. The median is indicated by the dark black line and the y-axis is log transformed. Each horizontal line represents an individual observation and the underlying black histogram indicates the spread of observations for each model type.
Table 2.
Accuracy assessment results for the AsiaPop, GRUMP and GPW modelling methods for a) Cambodia and b) Vietnam.