OTC 21532-- A Feature Oriented Regional Modeling System for the North Brazil Current Rings Migration after Retroflection

摘要

Southeast of the Trinidad-Venezuela region, the North Brazil Current (NBC) retroflects and forms about 5-8 rings annually. The ensemble of trajectories of rings extends offshore of the 500 m isobath, with a mean translation speed of approximately 14 km/day and a mean length scale of about 100 km (Goni and Johns, 2001, 2003). At least two distinct ring types exist: surface-intensified and thermocline-intensified, with differences evident in both azimuthal velocity and water masses. This paper presents a recent implementation of an operational modeling system for this region. The key to this modeling effort is to implement the feature oriented regional modeling methodology for the NBC rings with an advanced initialization scheme to incorporate the varying ring structure, shape, and associated currents made possible by regular surveillance and deployment of instruments into the ring. Multiple observations provide input and guidance to the ring initial conditions for the feature model system. Based on previous studies and data, a water-mass based feature model system for two distinct NBC rings is developed. The parametric models for temperature and salinity are built to capture the main features observed in vertical structure of those rings. For example, in the case of the thermocline-intensified ring, different empirical-analytical functions with tunable parameters are used to represent (i) the thermocline shoaling up in the intermediate depth of the ring, (ii) the dipping down in the inshore and offshore edges, and (iii) the presence of the maximum salinity water at 50-100 m. These feature models are first calibrated with available sea surface temperature (SST) data and then melded with background climatology in a featureoriented multiscale objective analysis to develop a three-dimensional description of the regional ocean. The feature oriented scheme is used to initialize an operational forecasting system using the Harvard Ocean Prediction System (HOPS) framework. Implementation, calibration, and validation of this system were carried out for multiple case studies during 2006 and 2007 when a number of drifter data sets were available. A hindcast study for 27 January 2010 is used for verifying the forecast system in a semi-operational mode. A fully operational system was launched in July 2010.