Abstract:
An understanding of the spatial distribution of informal settlements in an urban area is vital for ongoing development planning and the management of service infrastructure. It allows planners and decision-makers to formulate improved strategies to tackle environmental issues arising from informal settlement growth. The current work examined the City of Riyadh in Saudi Arabia. The objective of the research was to assess the performance of mapping techniques incorporating medium resolution satellite data (30 year time-series Landsat images) and specialist, expert knowledge (EK) unique to the informal areas located within the city boundary. Local knowledge was translated into a ruleset to identify and map the areas at object, settlement, and environ levels; both spatially and temporally. This process was combined with an object-based machine learning (ML) classification method (random forest (RF)) to produce a mapping scheme consisting of formal and informal settlement, road network, vacant block, and vegetation classes. It was found that the contribution of settlement-level indicators to the informal settlement classification was greatest in the ML model, followed by object-level and temporal indicators. Environ-level indicators contributed the least to the classification. The study shows that a process using a combination of RF and EK has the potential to provide planners and decision-makers with a very good change detection and landcover assessment technique to incorporate into ongoing urban planning activities. This technique would prove ideal for identifying informal settlements in Riyadh.
Keywords: OBIA, Random forest , expert knowledge, medium resolution satellite imagery, change detection, informal settlements