Assimilation of HF radar-derived surface currents on tidal-timescales

Abstract

A surface current observation system based on high-frequency (HF) radar has been developed for Raritan Bay and the coastal waters of New York and New Jersey. An HF radar network provides synoptic surface current maps in near-real-time that can be optimally combined with ocean circulation models using data assimilation (DA) framework to obtain the best possible estimate of a three-dimensional ocean state. A nudging or Newtonian damping scheme has been developed to assimilate HF radar data into an estuarine and coastal ocean circulation model. This model, with an extensive embedded real-time observational network, is called the New York Harbour Observing and Prediction System (NYHOPS). A nudging parameter is introduced into the equations of motion which affects the model dynamics. The data is imparted to neighbouring (three-dimensional) grid points via model dynamics. The impact of HF radar DA is analysed by computing the DA skill score (DAskill) based on the mean-square-error (mse). The DAskill is computed by comparing non-assimilated and assimilated model solutions with in-situ observations of three-dimensional currents, temperature and salinity, which have not been included in the assimilation. A positive DAskill (0∼1) represents an improvement in the model performance by assimilation. HF radar data covering Raritan Bay and the New York Bight (NYB) Apex were assimilated into the NYHOPS model in the model hindcast cycle (–24h to 0h) on a daily forecast basis for a period of 40 days. The DAskill is assessed with respect to the NYHOPS model hindcasts (daily model solutions from –24h to 0h) as well as the first day forecasts (daily model solutions from 0h to 24h). HF radar DA improved the NYHOPS model performance during both the hindcast and forecast periods. The model skill metrics for the near-surface layers in the inner-NJ shelf region shows a hindcast DAskill of 24% (14%) and forecast DAskill of 18% (7%) for horizontal velocities u: east-west component (v: north-south component), and a hindcast DAskill of 33% (38%) and forecast DAskill of 25% (30%) for temperature (salinity). The nudging scheme is robust and efficient for the HF radar DA into the NYHOPS operational forecast model. The NYHOPS-HF radar DA system is capable of importing in the observations and produce useful hindcasts/forecasts with minimum computational expense.