Urbanization encompasses both physical and social dimensions. The former includes the
attributes of size, layout, land use pattern, and the pattern and rate of growth or expansion. The
social dimension on the other hand, embodies a panoply of subjects that include, but are not
limited to, the composition and distribution of the population, housing, the availability and
accessibility to services, and the livelihood strategies of the urban dwellers. While data on most of
these variables may be collected using scientific sampling techniques such as those used for the
National Science Foundation sponsored project, Creation of a GIS for six Cities in Arid
Environments: in Morocco, Senegal, Mali, Niger, Tanzania and Botswana, which aim for
representative sampling, research on some of these and other urban issues may also require
complementary techniques. This paper seeks to examine the adequacy of the remote sensing
imagery-based random sampling technique in collecting information on the social dimensions of
urbanization such as the urban migrant and to demonstrate the need for complementary techniques
such as migrants’ network analysis at both ends of the rural and urban continuum. This paper is
part of an on-going dissertation research on land tenure reforms and social transformation in
Botswana: implications for urbanization and development.
The project attempts “to document and assess the linkages between land cover change, urban
structure change and the work/livelihood history of the urban poor in six African cities located in
arid environments.” To accomplish this objective, the project processed and analyzed remote
sensing imagery to ascertain land use/land cover changes as well as the framework for selecting a
random sample of households for interview. The first part entails the analysis of the remote
sensing imagery to identify the various categories of urban land use at each time and over the
study period. This methodology also allows for the quantification of the observed land use changes
through the computation of pixels in the imagery. The second component of the project entails the
conduct of household interview to reconstruct the livelihood history of the urban residents
especially, the poor. The processed imagery with a grid overlay is also used for the spatial
stratification of the cities (based on the perceived land use classes) from which a random sample of
households are selected for interview. This process starts with a random selection of cells by a
computer software, with a bias for poor residential sections as well as adjustments for stages of
growth in the different sections of the cities. A list of households closest to the center of each cell
are then drawn up from which ten are randomly selected for interview. The claim that is made for
this sampling method is that it allows for a stratification of the random sample based on the land
use patterns perceived from the remote sensing imagery. In short, the sampling approach aims for
a scientific research that could be replicated.
While this sampling method may be appropriate in tackling a broad array of urban issues that
the NSF project is concerned with, it may not be adequate in investigating quite a number that
relate to the urban migrant and rural-urban migration in which the project is also interested. This
paper will use urbanization in Botswana, one of the six countries covered by the NSF project, as a
case study.