Empirical orthogonal function analysis of advanced very high resolution radiometer surface temperature patterns in Santa Barbara Channel

Abstract

Empirical orthogonal function (EOF) analyses of a yearlong, advanced very high resolution radiometer (AVHRR) sea surface temperature (SST) data set from the Santa Barbara Channel describe a spatial and temporal variability which is consistent with in situ measurements. The analyses combine an unconventional decomposition of the spatial variance with a more conventional analysis of temporal variance modes to identify features in the AVHRR data such as fronts, eddies, and annual temperature cycles. The methodology has practical applications in the EOF analysis of remote sensing data which are more densely sampled in space than in time. The first mode (60%) of the spatial variance shows a persistent front aligned northwest‐southeast in the west central channel, which separates warmer surface water to the east from colder waters off Point Conception to the west. The eigenvector time series shows a seasonal modulation, peaking in early summer, consistent with in situ measurements of temperature gradient and current. The second spatial variance mode (7.7%) suggests an eddylike structure. The conventional EOF decomposition of temporal variance yields a heavily dominant first mode (91.6% variance) representing the basin‐wide seasonal warming and cooling. This mode peaks in early autumn and lags by 90 days the first‐mode time series revealed by the spatial variance analysis. The first temporal variance mode has little spatial variation in the eigenvectors, while the second and third modes show more spatial structure. The second and third temporal variance modes are extremely low energy (2.5% and 1.1%) and do not reveal statistically significant information on the frontal and eddy patterns evident in the AVHRR and in situ data.