An improved fusion method for detecting upwelling off the coast of northwest Africa from chlorophyll-a and sea surface temperature satellite images

Abstract

ABSTRACT In this letter, a novel method is proposed for extracting upwelling along the Moroccan Atlantic coast using the fusion of satellite observations of biological and physical parameters during the time period 2000–2016. The developed approach first adopted a pre-processing step based on the interpolation to predict the missing values of pixels affected by clouds. Then, three fusion methods defined in the literature were applied to the corrected data to provide three different classifications to each satellite image. The majority voting method was investigated for combining the three classified images and therefore having one fused result more informative than the inputs and detecting well the productive, cold inshore waters. The performance of our procedure is compared to a popular fusion method used to identify upwelling and is evaluated by an index of validation over 769 weekly sea surface temperature ‘SST’ and 769 eight-day chlorophyll-a ‘chl-a’ images. The findings show the robustness of our proposed approach versus the latter method in terms of segmentation accuracy and computational efficiency. Developing on the proposed approach, the spatial and temporal variations of the main physical and biological upwelling patterns are studied.