Levee vulnerability assessment using an integrated unsaturated transient seepage model, sensitivity analysis, and fragility curves

Abstract

The accelerated changes in climate resulting in more frequent and disruptive floods necessitate broader perspectives of levee vulnerability assessment. This paper aims to advance levee vulnerability curves using global sensitivity analysis to identify critical inputs for developing multi-variable fragilities. This analysis is efficient and effective for skewed data such as extreme precipitation. The study categorizes inputs into geometry, precipitation, and soil characteristics, generating 30,000 scenarios for analysis. A transient unsaturated seepage analysis is conducted to examine different failure modes such as piping, erosion, and overflow, as well as erosion initiation and enlargement time and locations for each scenario. Results show that, in addition to the initial upstream water level and precipitation characteristics, soil properties—such as gravel and clay content, along with water retention parameters—are crucial for developing fragility curves across different soil types. Additionally, comparing fragility curves for historical data and future precipitation projections highlights the importance of integrating these projections into levee risk analysis for the next 30 years. As a practical implication, these fragility curves are applied to calculate failure probabilities for a levee case study. This research would support the integration of levee vulnerability assessments with social factors and stakeholder perspectives which also increases the applicability of fragility functions in flood risk mitigation.