Time evolution of surface chlorophyll patterns from cross‐spectrum analysis of satellite color images

Abstract

Sequences of coastal zone color scanner (CZCS) images from the offshore region adjacent to Vancouver Island, Canada, have been analyzed to estimate the time rate of decorrelation of surface phytoplankton chlorophyll pigment patterns. In these high‐latitude, high‐pigment areas, CZCS‐derived pigment estimates were lower than those obtained from ship samples by about a factor of 3, their frequency distributions were skewed in opposite directions, and subareas of the images often showed a discontinuity in the frequency distribution at a concentration of 1.5 mg m−3, where the algorithm changes CZCS bands. We selected cloud‐free subareas that were common to several images separated in time by 1–17 days. Image pairs were subjected to two‐dimensional auto spectrum and cross‐spectrum analysis in an array processor, and spectra of squared coherence were formed. The squared coherence estimates for several wave bands were plotted against time separation, in analogy with a time‐lagged cross correlation function. Threshold levels for significant coherence were estimated from many realizations of squared coherence calculated for pairs of synthetic random uncorrelated fields with specified power law behavior κ−1.5, near the observed range κ−1.5–κ−2. For wavelengths of 50–150 km, significant coherence is lost after 7–10 days, and for wavelengths of 25–50 km, significant coherence is lost after 5–7 days; in both cases offshore regions maintain coherence longer than coastal regions. For wavelengths of 12.5–25 km, only the offshore regions maintained coherence after 1 day, but that was clearly lost after the next time separation of 6 days. The implication for the formation of monthly average large‐scale surface maps to estimate open ocean productivity (e.g., Esaias et al., 1986) is that all mesoscale patterns (<150‐km length scale) will not be resolved.