Assimilation of Observed Water Levels into Storm Surge Model Predictions

Storm surge models must compromise between model run time and accuracy. This frequently results in model simplifications to the real world, such as utilizing barotropic, depth-integrated equations and neglecting lesser fluvial inputs. Consequently, physical drivers such as steric effects, large-scale coastal currents, and rainfall have limited influence on model predictions. Similarly, advances in meteorological forecasts and hindcasts have substantially improved surge models, however imperfections in wind/pressure fields remain first-order drivers of surge model errors, particularly in event based forecasting applications. The combined simplifications in model physics and errors in model inputs form the basis for errors in storm surge prediction. Importantly, nearly all of these drivers are gradual, with temporal scales from ~1 day to ~6 months, and spatial scales on the order of ~10 to ~1000 kilometers. This means that while accounting for these phenomena explicitly through model physics is difficult, there is substantial opportunity to do so through other means.

In this talk, we present a methodology for assimilation of water level gage data into model predictions to correct for these errors through the introduction of a pseudo atmospheric pressure correction. The scheme is low-cost, flexible, and robust against data issues, making it particularly useful for storm surge forecast applications. We demonstrate the effectiveness of the approach and discuss issues associated with its application.