KLIWAS 
Seite 17 
In order to increase the number of useable image pixels, small gaps of 1-2 pixels due 
to small clouds or otherwise missing data were filled with the average value of the 
neighbouring cloud-free water pixels. Furthermore, the OC parameter distribution is 
approximately lognormal and therefore the OC parameter data were converted to 
logarithmic values before processing (Campbell 1995, Gregg and Casey 2004). 
3.2.2 FYont Detection 
The first part of the front detection is based on the Canny edge detection algorithm 
which belongs to the group of gradient algorithms (Canny 1986, Castelao 2006). The 
application of the edge detection algorithm is based on a structured 4-step procedure 
involving noise reduction, calculation of the gradient, non-maximum suppression, 
and tracing edges through the image by hysteresis thresholding. The second part of 
the front detection works on the principle of the Single-Image Edge Detection 
algorithm (SIED) according to Cayula and Cornillon (1992, 1995 and 1996) which is 
a widely used histogram algorithm. The application of this edge detection method 
includes histogram analysis, the application of a cohesion algorithm using a fixed 
investigation window size and a contour-following algorithm. In GRADHIST, we 
used all steps except the contour filling. In the final step of the GRADHIST 
processing chain the resulting fronts of both algorithms are merged into a final front 
map (Kirches et al. 2013). 
The SST field from a scene of NOAA 17 AVHRR sensor and its SST fronts 
identified by GRADHIST are shown in Fig. 4. The advantages of GRADHIST are the 
equally good detection of strong and weak fronts, the determination of the gradient 
magnitude as well as the gradient direction, and the ability to process large data 
volumes fully automatically. 
280.0 281.3 282.7 284.0 285.5 0.02 .05.06 .1 .2 .4 
Fig. 4: SST field and SST frontal zones in the northern part of the North Sea: fronts identified by 
GRADHIST (AVHRR of NOAA 17) 
KLIWAS 
Climatology 
of North Sea 
Fronts