Principal component analysis of biological and physical variability in a Gulf Stream meander crest

Abstract

In September and October 1988 a series of physical and biological observations were collected by the R. V. Endeavor and R. V. Cape Hatteras in a Gulf Stream meander crest. The hydrographic data, vertical chlorophyll a profiles derived from CTD/fluorescence profiles (calibrated with discrete pigment samples), and zooplankton biomass data (20–120m, estimated from chlorophyll a exhibited increased variability in the upper 120 m with a zero-crossing at 55m. The first mode of zooplankton biomass is near-surface intensified and decreases between 40 and 100 m. For each variable, the individual profiles are regressed onto the corresponding modes to generate a spatial/temporal series of the first two principal components. These series are analyzed in a local curvilinear coordinate system, defined by the apex of the meander crest on a daily basis. The spatial structures of the principal components for each of the physical variables in the curvilinear coordinate system are very similar. The temporal variability in this data set is overwhelmed by the spatial variability for all variables except zooplankton biomass, which exhibits a clear diurnal signal because of diel migration. For all the variables, cross-stream variability dominates, but there is some indication of along-stream variability, particularly in the Slope Water. The large-scale Gulf Stream front produces 50–80% of the total data variance in all variables, except the zooplankton. Mesoscale phenomena, including meander-induced vertical motion with associated detrainment/entrainment on the western and eastern flanks of the crest, and a warm-core ring stream interaction, account for 15–30% of the total data variance for all variables. Only 18% of the zooplankton variability is due the large-scale Gulf Stream front, while 50% of the variability can be attributed to a combination of mesoscale phenomena and diel migration. On the order of 10% of the variability in the data is due to submesoscale phenomena ( L<20 km). The estimated variability associated with biological processes, other than diel migration, is 20%. The error for the terms in the variance decomposition is on the order of 5%.