Page 39 - IJPS-11-2

International Journal of
Population Studies Satellite data analysis of South Africa population grid

Eurostat (2024) also contends that generating grid Hoover’s concept posited that population distribution was
statistics typically necessitates georeferenced point images even if a state comprising 10% of the nation’s land area also
and datasets with a prominent level of spatial precision. contained 10% of the nation’s population (Hoover, 1941).
This involves, for instance, geocoding building, business, The index would reach zero if each area had an equal share
and address registers with geographic coordinates, enabling of the nation’s population and land, and it would approach
statistical information to be associated with specific a hundred if everyone resided in a single locality. Based on
locations. These point-based datasets can subsequently be the GPW population grids for South Africa, as displayed
aggregated to various areas as needed, including square in Figure 2, the HI was estimated at 90.4 in 2000 and
grid cells. 90.6 in 2020. This gives a statistical number to the GPW

One such gridded population dataset is the gridded population grids, confirming the view of a significant and
population of the world (GPW), which is produced by the increasing population concentration in South Africa. The
Socioeconomic Data and Applications Centre (SEDAC) HI for South Africa for 2000 and 2020 is visually presented
(about SEDAC, see https://sedac.ciesin.columbia.edu/ in Figure 3, which showcases the evident concentration of
about). SEDAC operates as a data center within the NASA’s the population.
Earth Observing System Data and Information System. To identify the clusters of population concentration,
SEDAC gathers data based on light and infrared emissions as proposed in Figures 2 and 3, the global spatial
from Earth, captured by satellites initially launched by autocorrelation (Moran I) can be used. The Moran I statistic
the United States Department of Defence. The GPW is a is a correlation coefficient that calculates the universal
minimally modeled collection of gridded population data. spatial autocorrelation of a data set, that is, the statistic
In order to facilitate data integration, the GPW aims to measures the similarity of one object to others surrounding
provide estimates of population counts in raster format it (Coetzee & Kleynhans 2018). The Moran I statistics
that are consistent with national censuses and population indicate values of 0.67 in 2000 and 0.68 in 2020, displayed
registers. in the first column of Figure 4. The absence of specified
The GPW (version 4) provides population distribution z-values for the two periods implies a strong acceptance
surfaces in both population counts (individuals per pixel) of the alternative hypothesis of spatial clustering. The
and population density (individuals per square kilometer) second column in the same figure, representing the
for the years 2000, 2005, 2010, 2015, and 2020 (Bustos cluster map, illustrates that high-concentration population
et al., 2020). These surfaces have a spatial resolution of clusters (depicted in red as high-high clusters) are sparse
30 arc s (approximately 1 km at the equator); however, and widely dispersed. In contrast, low-concentration
aggregated datasets are also accessible at resolutions of population clusters (depicted in blue as low-low clusters)
2.5 arc min, 15 arc min, 30 arc min, and one degree for encompass the majority of the South African landscape,
expedited processing. The GPW population surfaces are accounting for 78% in 2000 and 79% in 2020. The results,
created using areal weighting. depicted in the third column as a significance map,
indicate that the locations with significant clusters (shown
The SEDAC GPW for South Africa can be displayed in green) have remained largely unchanged over the two
through various GIS. The grid covers 86,948 spatial and/ periods, suggesting minor change in high/low population
or point locations for 2000, 2005, 2010, 2015, and 2020. concentration clusters.
The population estimates (estimated number of people)
for each location (30 arc seconds) for 2000 and 2020 are To further support the aforementioned findings, it
displayed in Figure 2. The distribution of the South African is essential to evaluate the complete spatial randomness
population appears to be uneven, with a concentration of (CSR) of the South African population’s spatial distribution.
people living in a limited number of locations, resulting in Complete randomness suggests that the observations
high spatial density in these areas and low spatial density in are scattered indiscriminately over the region. In no area
many others. This represents the geographic concentration would it then be possible to predict whether there would
of South Africa’s population. be more observations than in others. The chance to find
some observation should also be independent and not be
3. Key findings caused or altered by the presence of others (Dixon, 2018).
A scatter plot can be assessed to determine whether any
3.1. Spatial concentration, autocorrelation, and pattern exists among the observation data to determine
randomness the CSR. When there is normal restraint or rivalry between
The hoover index (HI), introduced by Edgar Hoover in the points, resulting patterns are regular, but if gravitation
1941, assesses the evenness or unevenness of population or transmission occurs between them, clustering of
distribution among U.S. states (Long & Nucci, 1997). observations is possible (Kyriakidis, 2009). Clustering

Volume 11 Issue 2 (2025) 33 https://doi.org/10.36922/ijps.3297