PhD Defence Ms Monika Kuffer

Department of Urban and Regional Planning and Geo-information Management


Title of defence

Mapping spatial patterns of deprivation in cities of the global south employing very high resolution imagery 


Most cities in the global South are unable to provide adequate housing and basic services to their low income inhabitants, causing that around one quarter of the global urban population is living in physically deprived areas, often referred to as slums. Spatial data on the location, extent and site-specific variations are considered essential to support planning, policy formulation and upgrading schemes. Such data, however, are often not available or outdated due to rapid development dynamics, the size of cities and the high number of deprived areas. Mapping these through ground surveys is rather labour intensive, ineffective (since developments are fast) and often not feasible with locally available resources and capacities. In this thesis, the use of space-born very high resolution (VHR) imageries is seen a potential alternative to provide rapidly base data on deprived areas at site, city or metropolitan scale for different moments in time. In the past years, many remote sensing methods have been developed that have the potential to identify and map such areas in VHR imageries. However, up till now, there is no consensus internationally on the most suitable method for deriving base data on deprived areas. To address this gap, this thesis aimed to develop and evaluate methods for mapping spatial patterns of deprivation in cities of the global South. For this purpose, four sub-objectives have been addressed in individual chapters, and their main results are summarized below.

To obtain an overview of available methods and image features to map deprived areas a review on the state-of-the-art on mapping deprivation by means of VHR remote sensing imagery was conducted. The review showed that our knowledge on mapping deprived areas in VHR imagery is limited to relatively few regions on the globe and that such areas display commonalities in terms of their spatial characteristics, besides context-specific differences. In general, they are densely built-up, have organic layout patterns, small roof areas and are often found at hazardous locations. Furthermore, the diversity of deprived areas was addressed in several publications but not mapped using image classification methods. Concerning methods to map deprivation, the review showed that machine learning and texture/morphological based methods are the most computationally efficient and achieved highest classification accuracies.

In order to capture the spatial characteristics, the utility of spatial metrics for mapping unplanned areas was analysed for the city of Dar es Salaam (Tanzania) and Delhi (India). An object level segmentation was used to quantify three morphological dimensions of unplanned areas (i.e., size, density, and pattern) with a selected set of spatial metrics. The results have been aggregated at the level of homogeneous urban patches (HUPs), representing homogeneous neighbourhoods. Results showed that aspects of the morphological differences between unplanned and planned areas could be captured with the employed set of spatial metrics (achieving an accuracy of around 70%). However, the results also showed computational constraints and transferability problems within cities, and in particular across the two rather different cities in terms of average building sizes and densities.

In search of a more transferable and computationally efficient approach the utility of image texture for mapping slums was analysed. Employing the grey-level co-occurrence matrices (GLCM) variance with a machine learning algorithm to differentiate slums and formal HUPs performed well in terms of accuracy levels (around 90%). Furthermore, the transferability showed consistently good results in three rather different cities, i.e., Mumbai (India) with many large and very diverse slum areas, the semi-arid city of Ahmedabad (India) with in average smaller slum areas and the city of Kigali (Rwanda) with much lower built-up densities and a complex topography. Parameter settings for defining suitable scale parameters for HUPs and window sizes of the GLCM variance were evaluated and required some adaptions for different urban environments, which reduced the transferability. However, the GLCM variance showed to be a rather robust image feature for mapping slums, avoiding high dimensionality of large features sets, which easily causes computational constraints.

Departing from the idea that deprived areas are homogeneous within a city, this thesis developed a methodology to map the diversity of deprived areas via image based features. For this purpose, the city of Mumbai (India) was used to develop a local typology of deprivation based on qualitative ground knowledge. The typology of four types of deprived areas and one formal built-up type was conceptualized into four physical dimensions of deprivation, i.e., environment, texture pattern, density, and geometry. All four dimensions were quantified with image based features, extracted via various image analysis methods such as GLCM, machine learning or spatial metrics. The significance of the set of these features was evaluated using logistic regression modelling, which also allowed to calculate the probability of a specific HUP to belong to one of the four deprived types. The final classification of the HUPs was determined by probability values within a vector domain, which is computationally much more efficient than working with pixel values in a raster domain. The result showed a moderate classification accuracy (around 79%). Comparing the results with a census based analysis of multiple deprivation showed great internal spatial diversity of deprivation within the large administrative units. This indicates that remote sensing based information on the diversity of deprivation is covering much better the spatial patterns and clustering of deprivation compared to deprivation mapping at administrative units. Thus remote sensing based information provides spatially detailed information on the extent of deprivation seen essential to inform urban planning and decision making.

In conclusion, deprived areas across Indian and East African cities used in this research showed commonalities in their morphology in terms of built-up densities, building sizes and layout patterns (being clearly different from formal built-up areas), which could be captured with the GLCM variance for mapping purposes. The aggregation to HUPs facilitated the analysis and provided more policy relevant information compared to a much noisier pixel level output. The use of GLCM, machine learning and spatial metrics allowed the mapping of the diversity of deprivation employing logistic regression modelling as a road forward to leave the idea that deprived areas are homogenous. Thus VHR imageries allowed to map deprivation as well as their diversity where statistical and machine learning methods were most efficient in terms of required computation resources and transferability.


I have graduated as a Human Geographer from the TU Munich (Germany) in 2001. I did my second MSc in Geographic Information Science at the University of London (UK) in 2010. Between 2001 and 2003, I was working as researcher at the Austrian Academy of Sciences (Institute for Urban and Regional Research) in Vienna (Austria). Since 2003, I am working as lecturer at the Faculty of Geo-Information Science and Earth Observation (ITC) of the University of Twente (Enschede, NL), Department of Urban and Regional Planning and Geo-information Management. My main research foci are urban remotes sensing, in particular monitoring deprived (slum) areas as well as analysing the urban form, structure and dynamics with remote sensing and spatial analysis and modelling.

Kuffer, M., van Maarseveen, M.F.A.M. (promoter) , Sliuzas, R.V. (co-promoter) and Pfeffer, K. (co-promoter)  (2017) Spatial patterns of deprivation in cities of the global south in very high resolution imagery. Enschede, University of Twente Faculty of Geo-Information and Earth Observation (ITC), 2017. ITC Dissertation 304, ISBN: 978-90-365-4369-9.

  Full text


Event starts: Wednesday 19 July 2017 at 14:30
Venue: UT, Waaier 4
City where event takes place: Enschede

previous page
more events

  1. Home »
  2. Organisation »