Mapping informal settlements in a Middle Eastern environment using remote sensing techniques

Informal settlements and slums have become the new reality for much of the world’s urban poor. Informal settlements result from the urgent need for shelter, high urban growth rates and a shortage of suitable affordable housing. Mapping informal settlements from satellite imagery is challenging in terms of definitions, data availability and methods. Spectral and spatial indicators have been used successfully to analyse informal settlements, using very high spatial resolution imagery (VHR), but require substantial effort in terms of tuning thresholds to local conditions. This research was designed to categorise and explore informal settlements attribution in a Middle Eastern context, using Jeddah, Saudi Arabia as a case study. The thesis poses three research questions and concludes by adapting an ontological framework for mapping informal settlements in this environment. The first research question asks, “How can object-based image analysis (OBIA) be used for mapping informal settlement indicators using VHR imagery in a Middle Eastern context?” It documents the application of OBIA to map informal settlements, drawing on an existing ontology Kohli et al. (2012) and the indicators of Owen and Wong (2013). Three informal settlements with different land-use histories were selected to represent old and new informal settlements in the city of Jeddah, Saudi Arabia. Vegetation extent was the most useful indicator, followed by road network with 100% and 84% producer accuracies, respectively. Informal and formal settlements were mapped with an overall accuracy of 83%. It concludes that OBIA is a useful method for mapping informal settlement indicators in Middle Eastern cities. However, a generic ruleset for mapping informal settlements remains elusive, as each indicator requires significant localised ‘tuning’. The second research question asks, “Can OBIA and machine learning (ML) becombined to incorporate additional information at the object, settlement and environ levels to assist with the identification of informal settlements and their characteristics?“ Fourteen indicators of informal settlement characteristics were derived and mapped using an object-based ML approach using VHR imagery. These indicators were categorised according to three different spatial levels: environ, settlement and object. The most useful indicators for prediction were found to be density and texture measures, (with random forest (RF) with relative importance measures of over 25% and 23% respectively). The success of this approach was evaluated using a small, fully independent validation dataset. Informal areas were mapped with an overall accuracy of 91%. Objectbased ML, as a processing chain approach, performed better (8%) than objectbased image analysis alone due to its ability to encompass additional geospatial datasets. The third research question asks, “Can a time-series approach help to identify informal settlements and their characteristics at a city scale?” It evaluates the utility of time series analysis (TSA) with machine-learning (ML) methods to map informal settlements. Both spatial and temporal indicators were used to perform the analysis. These indicators were categorized according to four different spatial and temporal domains: environ, settlement, object and time. This work proposed a novel approach combining an object-based ML with TSA for informal settlement mapping. The overall accuracy achieved was 95%. Among the spatial and temporal levels examined, the contribution of the settlement level indicators was most significant to ML model followed by object level indicators. Adaptation of this method allows for the combination of a wide ranging and diverse group of indicators in a comprehensive ML framework. The fourth research chapter re-examines an existing ontological framework,the generic slum ontology (GSO). It asks, “Can an existing ontological framework be adapted for mapping informal settlements based on quantitative approaches including OBIA, ML and TSA?” This work examines the utility of the GSO proposed by Kohli et al. (2012) for mapping informal settlements. Generally, the framework of the GSO performed well in the Middle Eastern context but some refinements are suggested. The adaptation of GSO and an OBIA approach in the first case study Fallatah et al. (2018) resulted in better informal settlement classification accuracy (83%) compared to similar studies conducted at the same scale (5𝐾𝑀2). In the second case study, ML was used to integrate other datasets at different GSO levels, specifically the settlement and environ levels. This work enhanced mapping informal settlements by (8%) compared to the previous work using case study the city scale level (70𝐾𝑀2). A TSA approach was used to introduce the temporal domain as a new paradigm to GSO in case study three. A slight improvement (1%) was noted when utilizing TSA for mapping informal settlements. Commentary is provided on the utility of the GSO in providing a generic and flexible framework to allow integration of future datasets including 3D information and ancillary datasets into informal area mapping.