Tide predictions have been around for centuries. They rely on a relatively long time series of tide measurements collected at a location from which the components of dozens of astronomical influences are calculated and added together; these are called harmonic constituents
. NOAA creates these tide predictions for over 200 tide gauge locations in the United States, and the vertical reference coordinate, or datum, is updated approximately every 20 years. The current national tidal datum epoch
is 1983-2001, so tide predictions are inherently biased low because of sea level rise (which is roughly 0.25 inches/year in southeast Florida).
Tide predictions cannot and are not intended to account for current conditions in the atmosphere and ocean, which can often have a significant impact on the water level -- they are astronomical and climatological by design. The timing of high and low tides is captured very accurately by the predictions, but the amplitude can be off by as much as one to two feet. In low-lying coastal areas, tidal flooding is becoming a more frequent and more severe problem, and being able to forecast when high water events are expected is of great value to municipalities and water managers.
The model described here was initially conceived and developed in 2020-2022 as the centerpiece of Nathan Taminger's Departmental Honor's Thesis at the University of Miami (see references below for details). It uses the NOAA tide predictions as the baseline, then adds a series of three dynamic refinements to arrive at tide forecasts out to ten days. Sea level rise is accounted for by a linear trend through the data then adding its value in the current year to the tide predictions. Then, fluctuations of the regional-average atmospheric surface pressure provide a daily adjustment (1 cm per hPa via the inverted barometer effect
). Lastly, a multiple linear regression model is run on four other regional enviromental parameters to provide the final tide forecast. The model is run daily and all stations are complete by approximately 1100 UTC (7am EDT); forecast data come from the 0600 UTC cycle of the GFS global model. Beginning in December 2022, this work is continuing thanks to funding from and collaboration with the South Florida Water Management District.