Simulating extreme total water levels using a time-dependent, extreme value approach

Abstract

Coastal flood hazard zones and the design of coastal defenses are often devised using the maximum recorded water level or a "design" event such as the 100 year return level, usually projected from observed extremes. Despite technological advances driving more consistent instrumental records of waves and water levels, the observational record may be short, punctuated with intermittent gaps, and vary in quality. These issues in the record often preclude accurate and robust estimates of extreme return level events. Here we present a total water level full simulation model (TWL-FSM) that simulates the various components of TWLs (waves, tides, and nontidal residuals) in a Monte Carlo sense, taking into account conditional dependencies that exist between the various components. Extreme events are modeled using nonstationary extreme value distributions that include seasonality and climate variability. The resulting synthetic TWLs allow for empirical extraction of return level events and the ability to more robustly estimate and assess present-day flood and erosion hazards. The approach is demonstrated along a northern Oregon, USA littoral cell but is applicable to beaches anywhere wave and water level records or hindcasts are available. Simulations result in extreme 100 year TWL return levels as much as 90 cm higher than those extrapolated from the "observational" record. At the Oregon site, this would result in 30% more coastal flooding than the "observational" 100 year TWL return level projections. More robust estimates of extreme TWLs and tighter confidence bounds on return level events can aid coastal engineers, managers, and emergency planners in evaluating exposure to hazards. Key Points Developed a probabilistic, full simulation total water level model (TWL-FSM) TWL-FSM captures seasonal and interannual climatic variability in extreme events TWL-FSM return levels are higher than those from the "observational" record