Development of NOAA’s Next-generation Prediction System for High Tide Flooding Risk on Subseasonal to Annual Timescales

In the United States, the National Oceanic and Atmospheric Administration’s National Ocean Service (NOAA/NOS) has developed a statistical model to predict the daily risks of high tide flooding (HTF), for forecast leads of up to one year, at 98 tide gauge locations along the US coastline. NOAA/NOS predicts the daily probability of exceedance of hourly water levels above a specified flood threshold by combining the tide prediction with the (extrapolated) linear trend of mean sea level and a probabilistic prediction of the anomalous hourly non-tidal residual (NTR). In turn, the NTR anomaly prediction is made up of two components: (1) a prediction of monthly mean NTR, currently based upon the observed autocorrelation function of linearly detrended NTR, with uncertainty based upon the observed standard deviation of monthly NTR; and (2) a prediction of the probability distribution function (PDF) of hourly NTR anomalies, which uses observed historical dependence upon the total water level and is assumed to be Gaussian. These forecasts are available at https://tidesandcurrents.noaa.gov/high-tide-flooding/monthly-outlook.html.

 

In this presentation, we introduce an updated version of the NOAA/NOS HTF framework, with three key improvements: (1) the trend estimate is determined empirically and is allowed to be nonlinear; (2) monthly mean SLA is predicted by either an empirical or dynamical climate forecast model, and includes an ensemble spread; and (3) the PDF of hourly NTR anomalies is non-Gaussian and determined separately for each month from past observations using a “stochastically-generated skewed” (SGS) distribution. Skill of the updated version is compared to the original (currently operational) version at all tide gauge locations, and the impact of each of the improvements on skill is diagnosed. Further prospects for improvement of the HTF framework are also discussed.