KLIWAS 
Seite 14 
KLIWAS 
Climatology 
of North Sea 
Fronts 
3 Alg o rithm De sc rip tio n 
3.1 Approaches to feature detection 
The advent of remote sensing from satellites has enabled global monitoring of 
oceanic fronts from space. The first parameter used for this purpose was the sea 
surface temperature (SST). SST fronts are caused by different physical processes like 
convergence of different water masses, river run-off, or up- and down-welling etc. 
These SST gradient zones can be detected in SST images by objective methods. 
Two approaches became widely accepted: the gradient method, e.g. the Canny edge 
detector, mainly due to its simplicity and the histogram method, due to its robustness 
and comprehensive worldwide validation (Canny 1986, Cayula and Cornillon, 1992). 
Other methods have been developed as well, notably the cluster-shadow method, 
wavelet methods and classification of water masses (Belkin and O’Reilly, 2009). 
Gradient-Algorithm 
The search-based methods detect edges by first computing a measure of edge 
strength, usually a first-order derivative expression such as the gradient magnitude, 
and then searching for local directional maxima of the gradient magnitude using a 
computed estimate of the local orientation of the edge, usually the gradient direction 
(Canny 1986, Jahne 2005). Different published edge detection methods mainly differ 
in the types of applied smoothing filters, in the types of the filter used for computing 
of the gradient and by the way to determine the edge strength. 
Histogram-Algorithm 
Histograms as graphical representation of the probability density function of the 
underlying variable can be used for edge detection as well. Histogram algorithms are 
search-based methods that detect edges by testing if more than one population (in this 
case water mass) is present in the area under consideration. Edge detection methods 
mainly differ in their requirements on noise removal and the way to decide if the 
hypothesis of more than one population can be accepted (Cayula and Cornillon 1992, 
1995 and 1996). 
3.2 The ore tic a 1 Algorithm Description 
Miller (2009), Shimada et al. (2005), Belkin and O’Reilly (2009), Vazquez et al. 
(1999) and others have demonstrated the general feasibility of detecting SST and OC 
fronts in satellite data. In their studies, the automatic detection of fronts in large data 
volumes is done either by a gradient algorithm, which exploits spatial gradients 
within a satellite image, or a histogram algorithm, which works on the frequency 
distribution of the values within image subsets. The strength of the gradient 
algorithms is that they enable the detection of any front regardless of its strength, as