About these forecasts
High tide flooding, minor coastal flooding that occurs at high tide even without a storm, disrupts roads, infrastructure and coastal communities. NOAA's National Ocean Service Center for Operational Oceanographic Products and Services (CO-OPS) issues a Monthly High Tide Flooding Outlook, giving daily flood likelihoods up to a year ahead that communities use to plan road closures, prevent saltwater damage to electrical grids and infrastructure, and mitigate costly impacts to local economies. The forecasts on this page come from a next-generation version of that outlook. It was developed by NOAA's Physical Sciences Laboratory (PSL) and the Cooperative Institute for Research in Environmental Sciences (CIRES) in partnership with CO-OPS, and is being evaluated for transition into the operational product. It is designed to give more accurate flood probabilities, particularly for extreme water levels and at stations where climate variability matters most.
How the forecasts are made.
Water level at a gauge is treated as the sum of the predicted astronomical tide, the long-term linear trend, the seasonal cycle, the monthly mean sea level anomaly, and faster hourly-to-weekly fluctuations. The monthly anomaly is forecast by a calibrated multi-model ensemble, currently based on the ECCC operational "GEM5.2-NEMO" seasonal forecast model and an empirical model ("CS-LIM"). The faster fluctuations, which are skewed toward high water and tend to persist from one day to the next, are represented by a statistical model fitted to each gauge's hourly record. Together these generate a large ensemble of possible hourly water levels; the probability of flooding on a given day is the fraction of ensemble members that exceed the threshold in any hour of that day.
What is new
Compared with NOAA's current high tide flooding outlook, this next-generation system (1) predicts the monthly anomaly with a calibrated ensemble of dynamical and empirical seasonal forecasts rather than damped persistence, (2) uses a skewed rather than Gaussian distribution for hourly variability, and (3) computes daily flood risk directly from hourly water level realizations.
Forecast skill
Hindcasts for 1993-2020 were evaluated out of sample (10-fold cross-validation, trend removed) for a high, station-specific flood threshold. At every lead from one to six months, the new daily probabilities are more skillful than the current outlook at most stations, and beat a seasonally varying climatology at about 70 of 116 gauges (about 63 for the current outlook). Skill is highest on the West Coast, Alaska and the Pacific Islands, where monthly sea level anomalies are most predictable; on the Atlantic and Gulf coasts, most of the skill comes from the tide. Experiments in which the monthly anomaly is known perfectly show that substantial improvement remains possible, especially on the Atlantic and Gulf coasts.
