AbstractThe process of urbanization experienced world-wide has increased rapidly in recent
decades, with this trend set to continue. Urbanization is more pronounced in cities in the
Global South, and this brings with it significant social and environmental problems such as
uncontrolled urban sprawl and uneven resource distribution. While much urbanization in
the Global South is unplanned, there have been some rare attempts at strategic, large-scale
urban planning. One such example is Abuja, the capital of Nigeria, which is a new planned
city with its origins in a Master Plan devised in the 1970’s.
This research uses multi-temporal remote sensing to investigate urbanization in Abuja over
the last 40 years to critique the original Abuja Master Plan, showing the extent to which
urban development has kept with, or diverged from, the original Master Plan. The study
also investigated the potential of using remote sensing methods to distinguish unplanned
and planned urban settlements in Abuja, Nigeria.
First a time-series of multispectral Landsat images was acquired; cloud-free images from
1975, 1986, 1990, 1999, 2002, 2008 and 2014 were used, with some years specifically
selected to correspond with important dates in Nigeria’s socio-political development, and
to match major milestone targets as prescribed by the Master Plan.
The research also combined Landsat Thematic Mapper (TM)/Enhanced Thematic Mapper
plus (ETM+) image classifications of urban built-up land cover with Defence Meteorological
Satellite Program-Operational Linescan System (DMSP-OLS) stable nighttime lights imagery
to investigate, distinguish and map unplanned and planned urban areas. DMSP-OLS stable
nighttime lights imagery from 1999, 2002 and 2008 were selected. Thresholding techniques
with ancillary information were successfully applied to distinguish areas of unplanned and
Finally, the research focused on developing and applying deep learning and random forest
classification techniques on Very High Resolution (VHR) imagery to characterise and map
unplanned and planned built-up land at a finer spatial scale. This approach was able to
address some of the obvious limitations resulting from using coarse (DSMP-OLS) and
medium (Landsat) resolution imagery encountered in the earlier part of the research in
attempting to distinguish unplanned and planned built-up settlements. The results of the
study have shown deep learning can be successfully adapted to map unplanned and planned settlements in a city of the Global South, while random forest performed poorly
in distinguishing planned and unplanned settlements.
|Date of Award||24 Oct 2019|
|Supervisor||PAUL APLIN (Director of Studies)|