Towards Risk-Informed Development: Improving Political Disaster Risk Modeling in the Nile Basin Region Using Big Data and Machine Learning

The purpose of this dissertation is to analyze disaster risk components and how they impact intrastate and interstate conditions in the context of resilient development. There are two main factors that affect disaster risk: exposure to specific natural hazards and vulnerability of a given region, community, or state regarding susceptibility factors, coping abilities, and adaptive capabilities. This dissertation also builds a custom database given the name of the Structural, Survey & Events (SSED) from 95 data sources for disaster risk components in the Nile Basin Initiative (NBI) states between 2000 and 2020. The modeling takes place using machine learning algorithms running on High-Performance Computing (HPC) resources as part of the Extreme Science and Engineering Discovery Environment (XSEDE) program. Furthermore, the dissertation illustrates the fact that political risk is closely linked to lower susceptibility in NBI states, whereas bilateral cooperation is dependent on exposure and coping capacities. On the other hand, the risk of inequality relies on the adaptive capabilities of the Nile Basin region. Deep Learning models have shown promising results, indicating that disaster exposure elements do indeed fit greatly across different explained disaster risks or impacts. Thus, building an End-to-End Machine Learning Pipeline for data processing and modeling using HPC helps reach the best-fitting model, confirming that granular spatiotemporal yields a better fit. In this dissertation, machine learning prediction is used to rank NBI states across political risk, bilateral diplomatic, cooperative interstate aptitude, and risk of inequality showing varying results across NBI states.