Upwelling Detection in AVHRR Sea Surface Temperature (SST) Images using Neural-Network Framework

In this paper we present a novel method for the detection and segmentation of upwelling regions in AVHRR SST data. We demonstrate the effectiveness of the algorithm on data from the Monterey Bay region, with prominent upwelling regions. A byproduct of the upwelling detection algorithm is the detection of frontal boundaries. The process is started with the training of a feed-forward back-propagation neural network for the purpose of finding regions of "uniform" temperatures, resulting in labeled clusters. Then statistical information is gathered from the various clusters. A quantitative criterion is developed that is used to test the existence of prominent upwelling region followed by detection and segmentation. The algorithm is applied on data from July 2003 to September 2003 and their results presented.