Frontiers in Marine Science (Dec 2022)

A novel statistical approach to predict seasonal high tide flooding

  • Gregory Dusek,
  • William V. Sweet,
  • Matthew J. Widlansky,
  • Philip R. Thompson,
  • Philip R. Thompson,
  • John J. Marra

DOI
https://doi.org/10.3389/fmars.2022.1073792
Journal volume & issue
Vol. 9

Abstract

Read online

Sea level rise is increasing the frequency of high tide flooding in coastal communities across the United States. Although the occurrence and severity of high-tide flooding will continue to increase, skillful prediction of high tide flooding on monthly-to-annual time horizons is lacking in most regions. Here, we present an approach to predict the daily likelihood of high tide flooding at coastal locations throughout the U.S. using a novel probabilistic modeling approach that relies on relative sea-level rise, tide predictions, and climatological non-tidal residuals as measured by NOAA tide gauges. A retrospective skill assessment using the climatological sea level information indicates that this approach is skillful at 61 out of 92 NOAA tide gauges where at least 10 high tide flood days occurred from 1997–2019. In this case, a flood day occurs when the observed water level exceeds the gauge-specific high tide flood threshold. For these 61 gauges, on average 35% of all floods are accurately predicted using this model, with over half of the floods accurately predicted at 18 gauges. The corresponding False-Alarm-Rate is less than 10% for all 61 gauges. Including mean sea level anomaly persistence at leads of 1 and 3 months further improves model skill in many locations, especially the U.S. Pacific Islands and West Coast. Model skill is shown to increase substantially with increasing sea level at nearly all locations as high tides more frequently exceed the high tide flooding threshold. Assuming an intermediate amount of relative sea level rise, the model will likely be skillful at 93 out of the 94 gauges projected to have regular flooding by 2040. These results demonstrate that this approach is viable to be incorporated into NOAA decision-support products to provide guidance on likely high tide flooding days. Further, the structure of the model will enable future incorporation of mean sea level anomaly predictions from numerical, statistical, andmachine learning forecast systems.

Keywords