An extreme rainstorm can trigger multiple types of hazards simultaneously or in sequence, posing significant risks to human lives and urban systems. Under climate change, extreme rainfall is becoming more intense and frequent in many regions, including Hong Kong. Due to the dense population and urban systems in Hong Kong, hazards triggered by future extreme rainstorms can be unprecedented and catastrophic. Effective multi-hazard risk management techniques are needed for better disaster preparedness and crisis management. Structural mitigation measures are widely used for long-term risk mitigation to limit the systematic risk to an acceptable level. However, additional risks induced by hazard interactions, increasing vulnerability and element dependence, are often overlooked in long-term...[ Read more ]

An extreme rainstorm can trigger multiple types of hazards simultaneously or in sequence, posing significant risks to human lives and urban systems. Under climate change, extreme rainfall is becoming more intense and frequent in many regions, including Hong Kong. Due to the dense population and urban systems in Hong Kong, hazards triggered by future extreme rainstorms can be unprecedented and catastrophic. Effective multi-hazard risk management techniques are needed for better disaster preparedness and crisis management. Structural mitigation measures are widely used for long-term risk mitigation to limit the systematic risk to an acceptable level. However, additional risks induced by hazard interactions, increasing vulnerability and element dependence, are often overlooked in long-term risk management. To what extent the risk can be reduced with the existing structural mitigation measures and where should be prioritized for hazard mitigation in a multi-hazard context remain unknown. For short-term crisis management, measures including early warning, evacuation, resource allocation, and rescue are effective for minimizing the residual risk. However, existing hazard warning systems evaluate hazard intensities but lack critical information on likely consequences, hindering the effective implementation of crisis management measures.

In this thesis, prompt risk assessment methods are developed for rain-induced landslides and pluvial flooding to support risk-informed crisis management; a quantitative multi-hazard risk assessment method based on numerical simulation is proposed to investigate the risk amplification effect in a multi-hazard context and to support long-term risk management.

A novel prompt quantitative risk assessment method for rain-induced landslides is proposed in this thesis. Uncertainties in landslide occurrence, volume, and runout are quantified based on landslide records, and the propagation of uncertainties is quantified in a probabilistic framework. The proposed method was tested using 83 major rainstorms during 1995-2016 in Hong Kong. The method accurately predicts the number of affected buildings and potential fatalities and identifies rainstorms that can trigger fatal landslides. The proposed method automatically generates a one-page risk assessment report within minutes to support effective risk communication, resource allocation, and emergency response. The prompt landslide risk assessment method will contribute to the advancement of landslide emergency management from hazard-informed to risk-informed, significantly enhancing societal resilience and facilitating climate change adaptation.

For flooding hazards, a dynamic risk assessment method is proposed to evaluate urban system flooding upon extreme rainstorms in Hong Kong. The proposed dynamic risk assessment method quantifies the degree and duration of flood impact on urban systems based on hazard processes, which can be simulated by a hydrodynamic model. The proposed method is applied to assess the flood impacts on urban transportation and building systems in Kowloon, the most densely populated area in Hong Kong. Detailed urban settings such as engineering drainage facilities and buildings are incorporated into the flood simulation model, and the model is validated qualitatively and quantitatively by field observations. Flood hazard scenarios are investigated under designated rainstorms of different return periods, and the impacts on transportation and building systems are evaluated with the new process-based impact evaluation method. The dynamic risk assessment method enhances societal flood preparedness, warning, and response.

The implementation of crisis management requires timely information on flooding extent, severity, and duration. The existing flooding analysis methods are either inefficient for real-time application or unable to simulate the flooding process. This thesis further develops a rapid pluvial flooding process simulation method with a real-time level of simulation efficiency. The rapid pluvial flooding process simulation method adopts a coarse-grid hydrodynamic model to generate time-series flood maps in a low resolution. Then the low-resolution flood maps are converted to high-resolution maps by a deep learning model. The deep learning model can be trained using a limited number of flood scenarios under uniform and constant rainfall, which significantly reduces the effort in data preparation. The rapid pluvial flooding process simulation method is applied to a complex terrain of 352.2 km2 with both mountains and plains. Results show that the proposed rapid pluvial flooding process simulation method can simulate the spatiotemporal variation of flood depth with RMSE of 0.019-0.082 m. The simulation time for a 48-h rainfall event in the study area is less than 1 minute, reaching an efficiency improvement of 3000 times compared with the fine-grid hydrodynamic model. The proposed rapid pluvial flooding process simulation method shows great potential in emergency management for pluvial floods.

Challenges in multi-hazard risk assessment lie in the characterization of hazard interactions and element dependencies, and quantification of the changing vulnerability upon sequential hazards. This thesis presents a quantitative multi-hazard risk assessment method that addresses these challenges for rainstorm-induced hazards. The proposed multi-hazard risk assessment method integrates recent advancements in multi-hazard simulation and multi-hazard vulnerability analyses. A physically based multi-hazard model is adopted to simulate the interactions among rain-induced landslides and flooding. The dependencies between elements are considered using Bayesian networks. The changing vulnerability of buildings is evaluated based on reliability-based vulnerability functions recently developed for multi-hazard problems. The proposed multi-hazard risk assessment method is applied to an extreme rainstorm in Hong Kong. The proposed multi-hazard risk assessment method effectively identifies high-risk areas and assesses the potential losses in terms of fatalities and damaged buildings. The method is demonstrated to be an effective tool for managing rain-induced hazards.