Improved ocean chlorophyll estimate from remote sensed data: The modified blending technique
Gregg and Conkright (2001) who pioneered the use of the blending technique in an attempt to calibrate ocean chlorophyll, expressed the need for further work to be done in order to obtain improved results. One problem faced when using this technique with spatially sparse data, is distortion of the resulting blended field when approaching the coastal boundaries. In this paper, the causes of the distortion and alternative methods for solving it are discussed. One of these method herein termed the corrector factor method, appeared the most appropriate in correcting the problem. In it, the blending process is done twice. This method sees the reduction of the mean squared difference between the blended and satellite fields from 6.299 in the normal blending to 0.347 in the corrector factor blending. This figure is also below the tolerance margin (the mean squared difference between the satellite and in situ fields) for the real data which was 0.989. Furthermore, this method is backed by a standard statistical procedure which produces identical results to its own even though the two methods differ in structure. A mathematical proof as to why these results coincide is also outlined. Validation study carried out by the authors showed that at least 80% of the times these methods are used the corrector factor will provide a better estimate of chlorophyll concentration than the original blending method. It is expected that analysis on primary productivity and management in the ocean environment will be greatly enhanced by this new finding.
Key words: Satellite, in-situ, sea-WiFS, blending, corrector factor, pseudozeroes, noisy data, kernel smoothing.