Mesoscale frontal structures in the Canary Upwelling System: New front and filament detection algorithms applied to spatial and temporal patterns

In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier's archiving and manuscript policies are encouraged to visit: a b s t r a c t a r t i c l e i n f o An improved automatic detection of mesoscale frontal activity is proposed, based on the edge detection algorithm initially presented by Cayula and Cornillon (1992). The performance and the sensitivity of their original method have been extensively tested on a very large MODIS SST data set at 1 km resolution, over the Canary Upwelling System, and compared to the results of a classic gradient based method. Its evaluation, including the companion method using a contour-following algorithm, shows that a significant part of the fronts was not detected by the earlier method despite an overall robustness. We propose here an improved implementation of the single image edge detection algorithm, from the best combination of multiple detections based on sliding windows (referred to as CMW). The results show a very significant increase of the overall performance: using a time series of 6 years (1,988 images), we observed an average increase of 140% in the edge detection and 30% improvement in the average length of the segments. Additionally, the sensitivity to the size of the detection window is lower, and the necessity of using the " following algorithm " is greatly reduced. We applied the CMW improvements to the Canary Current System, and showed frontal activity developed along the whole coast, with more intense fronts between Cape Ghir (30° 30′ N) and Cape Beddouza (32° 30′ N), and between Cape Juby (28° N) and Cape Bojador (26° 30′ N). The highest filament activity was associated with Cape Bojador followed by Cape Ghir and Cape Blanco (21° N). Considering the seasonality, two main areas were identified, north and south of Cape Juby, with marked seasonality in the fronts and filaments. No major interannual differences in frontal activity were observed in 2002–2007, except less season-ality during 2007. Segmentation and edge detection are fundamental issues in image processing to detection and map mesoscale features in the oceans. It is easier and generally more informative to detect the edges of surface structures than the structure itself. In oceanography, images from remote sensing techniques provide opportunities to detect a full range …