Abstract
A coastal ocean forecasting system was developed for the Long‐term Ecosystem Observatory (LEO) on New Jersey's inner shelf. The forecast system comprised an ocean model, the Regional Ocean Modeling System, forced by a high‐resolution atmospheric forecast, with assimilation of ocean data from ships and coastal radar systems. The forecasts were used to aid the deployment of real‐time adaptive sampling observing systems during the July 2001 Coastal Predictive Skill Experiment. Temperature and salinity assimilation data were prepared by optimal interpolation of shipboard towed‐body data. Surface current observations from coastal radar were projected vertically for assimilation using a statistically based extrapolation. The assimilation methods tested with the operational forecast system in July 2001 were continuous nudging and intermittent melding of the model forecast and gridded data. Observations from a validation array of current meter and thermistor moorings deployed on a cross‐shore line through the center of the LEO intensive observing area were used to formulate a set of quantitative model skill metrics that focused on aspects of the two‐layer wind‐driven upwelling and downwelling circulation that characterizes ocean dynamics during the stratified summer season along this coast. An ensemble of model and data assimilation configurations were tested, showing that the k profile parameterization for vertical turbulence closure, and assimilation by intermittent melding, comprised the forecast system with the more significant skill as measured by the mean squared error of the validation metric time series.