Verification of winds and seas forecasts from the FNMOC ensemble forecast system

Winds and seas have always been of primary interest to the U.S. Navy. Launch and recovery of aircraft at sea, underway replenishment, assessing conditions for landing small boats, and even interdicting pirates are still important naval activities that are impacted by high winds and high seas. While forecasts have improved over the years, quantifying forecast uncertainty remains an important requirement in Navy meteorology. One method to estimate that uncertainty makes use of an ensemble of model runs, each with slightly different initial conditions. The mean of such an ensemble remains reliable for a longer lead time than a single (deterministic) forecast, and the spread of such an ensemble is related to forecast uncertainty. However, the skill in the ensemble mean forecast and the relation of ensemble spread to forecast skill must still be evaluated.

The FNMOC ensemble forecast system includes the NOGAPS atmospheric and WAVEWATCH III ocean wave models. Recently added to the operational system are a data post-processing system that computes ensemble means, forecast probabilities and estimates of forecast error (ensemble spread) as well as verification against observations for the ensemble forecasts.

This presentation will describe the FNMOC ensemble forecast system, including data post-processing and verification, with emphasis on the latter. The verification system computes metrics that include bias and mean-squared error, conditional error, conditional rank probability score, and Brier Skill Score. It also produces information for reliability and Receiver Operating Characteristic (ROC) diagrams. Examples and results relevant to Navy operational activities will be provided. While current results are from raw ensemble fields, plans for future operational bias correction will be discussed.