Statistical analysis of a satellite imagery time series: Implications for the study of coastal and estuarine circulation.

摘要

Empirical Orthogonal Functions analysis of a thirteen month AVHRR imagery time series was performed to assess the temporal and spatial variability of turbidity patterns in Delaware Bay and the coastal area to the south. The statistical results were interpreted in light of existing knowledge of the coastal and estuarine circulation in the study area, and compared to data sets of wind and river discharge for the concurrent period of time. Several analyses, including eigenvector rotation for the EOF modes and cluster analysis, were run on different subsets of the imagery data, normalized to tidal cycle or to wind strength as well as to location within and outside the estuary. The results produced spatial patterns of variability and plots of temporal modulation of such variability over subtidal time scales; tidal cycle normalization allowed a separate examination of the variability due to tidal forcing, the dominant circulation driving force, which was inadequately sampled in the original series.;The estuarine turbidity maximum variability was captured by the dominant EOF modes and corresponded to variability in the river discharge records. Analysis of the turbidity signal along the coastal zone to the south of Delaware Bay revealed a spatial pattern that is entirely consistent with the presence of a buoyant coastal current flowing out of the Delaware Estuary and hugging the coast in a southward flow, as reported recently by Garvine (1991) and Munchow (1992). Tidal variability analysis produced ambiguous results which suggest that further study is needed in this area. The main source of error was the occasional presence of localized clouds, but it did not affect the dominant first EOF mode, because of the random nature of such error.;The techniques developed for this study can be useful as a data reduction tool in the management of the large volumes of operational satellite-derived data, since the first few EOF modes summarize the variability of the data set.