The interplay between uncertainties, transferability & policymaking in remote sensing-based slum mapping

The global attention on improving slums and providing adequate living conditions, aligning with the Sustainable Development Goals (SDGs) target 11.1, which aims to achieve universal access to secure and affordable housing by 2030 (United Nations, 2014) underscores the importance of developing reliable and transferable approaches to measuring the advancement of the global slum improvement initiative.
The provision of updated and reliable data on the geographical locations, size, structure, and socio-economic conditions within slums plays a pivotal role for decision-makers in the strategic development, execution, and monitoring of the effectiveness of initiatives aimed at enhancing slum conditions in diverse settings and levels of governance such as neighbourhoods, cities, and even at a national level. Remote sensing (RS) technology, such as satellite imagery and aerial photography, plays a pivotal role in providing intricate spatial information concerning urban areas, including detailed data on the physical attributes of slum settlements. This technology captures the dynamics of slums, particularly in monitoring rapid urbanisation and the evolving dynamics of slums and identifying specific morphological characteristics of slum areas.
While RS technology offers many advantages in monitoring and supporting policy formulation related to slums, it also has limitations. Slums vary widely in their physical characteristics, construction materials, and density. RS may struggle to distinguish between different types of slums accurately. In addition, commonly used visual indicators (in images) to produce reference data in measuring the performance of RS-based slum detections are affected by interpreters' different local and professional experiences, the choice of indicators used to conceptualise slums, and the fact that interpreters may differ in detail when delineating objects, which introduces uncertainty in producing reference data regarding slums.
These uncertainties can significantly affect decision-making processes. However, no study focuses on the interplay between uncertainties in RS-based slum detection, the transferability of slum detection methods, and its impact on the slum policymaking process, thus becoming the research objective. This dissertation aims to uncover this interplay by employing Object-Based Image Analysis, user-generated delineations, interviews with slum policymakers, and a literature review. The research discusses the effects of respondents' backgrounds on slum conceptualisations. It also discusses the impact of uncertainty on the comparability of RS-based mapping across different locations (spatial transferability) and the effect of uncertainties in measuring the slum dynamics (temporal transferability). Lastly, the research discusses the interplay between these uncertainties and slum policymaking.
The dissertation underscores the critical importance of accurately communicating uncertainties in remote sensing (RS)-based slum detection to policymakers, as these uncertainties significantly impact the reliability of data used in policy formulation. It shows that respondents' backgrounds, especially their local knowledge and GIS skills, greatly influence slum conceptualisation and delineation, which affects the accuracy of automated methods like Object-Based Image Analysis (OBIA). Detection accuracy is improved by integrating local knowledge and non-observable indicators such as zoning maps. However, without transparent reporting of uncertainties, policymakers risk relying on flawed or oversimplified data. This can lead to misguided decisions, significantly when such data influences interventions and budget allocations for slum management and urban poverty reduction. The study highlights the need for context-specific approaches, combining automated RS methods with ground-reference data, ensuring policymakers have a nuanced understanding of slum dynamics. This approach enables policymakers to develop more informed, adaptable, and effective strategies tailored to the complexities of slum areas, ultimately leading to better-targeted and sustainable interventions.