Characterization of mesoscale eddies in the lee of the Hawaiian Islands from direct observations and numerical simulation

The interactions between the trade winds and the Hawaiian Islands determines favorable conditions for the formation of mesoscale eddies in the lee of the islands. The region is therefore ideal for the study of these features in the open ocean. Direct observations collected during the E-Flux III field experiment (March 2005) are used in this study to analyze the physical, biogeochemical and optical characteristics of Hawaiian eddies. A numerical simulations was generated to further characterize the eddy activity in the region. This analysis required the development of an automated eddy detection and tracking algorithm. Chapter One focuses on cyclone Opal, observed during E-Flux III. Opal was a fairly circular eddy with a radius of &ap 80 km. Its core was characterized by the intense doming of isopycnal surfaces and nutrient upwelling. This triggered a phytoplankton bloom, which resulted in enhanced chlorophyll concentrations, and a shift to a diatom dominated community. The analysis of the potential vorticity field suggests that radial exchanges of water might have occurred between the eddy core and its surrounding. A new conceptual model for the life cycle of an eddy, which does not imply a closed system, is thus proposed. In Chapter Two, optical measurements are used to test the sensitivity of two parameters (chlorophyll concentration to attenuation coefficient ratio, and backscattering ratio) to the observed changes in ecological community at Opal's core. The results show that the backscattering ratio can detect the shift from a small phytoplankton to a diatom dominated community. This suggests that optical properties could be successfully used to retrieve in-situ information on the ecological communities within open ocean eddies. Chapter Three presents the eddy detection and tracking algorithm. The method is based exclusively on the geometry of the velocity vectors. Four constraints characterizing the spatial distribution of the velocity vectors around eddy centers were derived from the general features associated with eddy velocity fields. The points in the domain for which these four constraints are satisfied are detected as eddy centers. Eddy sizes are computed from the streamfunction field, and eddy tracks are retrieved by comparing centers at successive time steps.