Analytics in support of regional scale risk reduction

Abstract

Disaster risk analysis quantifies potential damages and losses using accurate hazard, exposure, and vulnerability information. Risk reduction managers rely on disaster risk analysis to make informed decisions towards promoting greater resilience. However, current approaches lack key features for proactive decision-making, such as accounting for uncertainty in hazard modeling, time-dependent processes affecting vulnerability, and highlighting successes and benefits of risk reduction. This dissertation makes three main contributions to the field of regional disaster risk analysis. First, I develop a framework for calibrating hazard parameters (inversion) and estimating the response of a hazard system (forward estimation) using limited and uncertain spatial data. This addresses gaps in current inversion-forward estimation approaches, which includes not accounting for varying uncertainties and spatial distribution in data, as well as the assumptions of the cost function. As a test case, I implement the framework on a tephra dispersion model using thickness observations of varying uncertainty to reconstruct past volcanic eruption characteristics and associated tephra fallout. Results show consistent improvements in model predictive performance for both inversion and forward models. Secondly, I develop a computational framework for modeling time-dependent physical vulnerability in regional risk analysis. This study addresses a gap in regional earthquake risk analysis by accounting for processes that increase vulnerability (e.g. structural deterioration) and policies that mitigate it (e.g. regional-scale structural retrofit). The method provides a flexible tool for decision-makers to investigate the outcomes of various mitigation decisions. Finally, I propose a counterfactual probabilistic risk analysis framework to quantify and highlight the benefits of effective disaster risk reduction interventions. This approach addresses a key challenge in risk management: that successful mitigations can go unnoticed (often because they are successful), making it difficult to incentivise proactive risk reduction decisions. Through counterfactual analysis (i.e. imagining the ‘what-if’ scenarios in disaster risk), the thesis shows how to calculate the probabilistic benefits of interventions, highlighting and celebrating their success.