In risk assessment of spatially distributed infrastructure, the probability of demand exceeding capacity is evaluated across the system. We describe and compare two levee system reliability analysis frameworks for seismic and high-water demands. The first approach is general, but computationally intensive, and uses Monte Carlo simulations to model capacity and demand for “segments” (i.e., elemental levee lengths) as spatially correlated random variables. We apply a capacity correlation model derived from seismic case histories in Japan. The seismic demand correlation model is based on global ground motion data, whereas the high-water correlation is taken as unity. The second approach achieves computational efficiency by grouping segments into physics-based “reaches” (i.e., length of levee having uniform statistical distributions of capacity and demand). Statistics and spatial correlation of the limit state function are computed using a procedure based on the firstorder reliability method. The probability of failure of the reach is then computed using level-crossing statistics. The application of level crossing statistics required an adjustment, introduced here, to previously utilized capacity correlation functions. We apply both methods for a levee system subjected to realistic demand and capacity distributions and show that characteristic lengths (defined as lengths of levee that can be considered as statistically independent) are comparable for high-water and seismic demands. This outcome is specific to the considered failure mechanisms and is driven by use of similar capacity correlation models, whereas differences in demand correlation models have limited impact.

Introduction

Levees are defined as man-made or natural embankments along rivers or water bodies. Their primary purpose is to provide protection against high-water events. The performance of levees when subjected to high-water or earthquakes is essential for the resilience of surrounding communities. Despite their critical function, many levees were not engineered at the time of their construction and are often founded on soft and weak soils. As a result, levees are frequently damaged during high-water events (water level rise in the river channel; e.g., [1–۳]) and following major earthquakes (e.g., [4–۸]). For levees that continuously impound water, a single failure anywhere along their length will produce flooding, and hence comprises system failure. For levees that intermittently impound water, the seismic failure probability is related to the combination of seismic deformation potential and probability of high-water during or shortly following the event, whereas the high-water failure probability is simply the single-segment failure probability during a high-water event. In either case (continuously or intermittently loaded), levees constitute spatially distributed series systems, which present particular challenges for reliability assessment. This paper describes two conceptually similar approaches for analysis of levee reliability, with an emphasis on the system probability of failure given knowledge of capacity and demand on a more local level. We defer to other documents for recommended analysis procedures for computing capacity at the segment, or crosssection level (Zimmaro et al. [9] for seismic, URS Corporation, Jack R. Benjamin & Associates Inc. [10] for high-water).