KLIWAS 
Seite 16 
KLIWAS 
Climatology 
of North Sea 
Fronts 
3.2.1 Pre-processing 
Frequent cloud cover over the North Sea is the main factor which limits the 
availability of suitable data. It is important that clouds are accurately identified in 
order to properly retrieve SST and OC parameters. The cloud mask is an integral part 
of standard level 2 products, but its quality varies between different sensors and their 
cloud detection algorithms. The AATSR 2 cloud mask for example can be used 
without any modifications, whereas the cloud mask of the MERIS and MODIS is not 
precise enough for the purpose of front detection (residual clouds are detected as 
fronts). In this case, the cloud detection has to be improved for both sensors by the 
application of additional algorithms. The AVHRR data provided by the BSH already 
include a manually improved cloud mask. Because the edges of clouds are often not 
well defined, a cloud buffer or border has been additionally introduced. The following 
figure (Fig. 3) shows the resulting cloud, cloud shadow and land masks in a MERIS 
image. 
Fig. 3: a) RGB; b) corresponding cloud and land mask for MERIS RR 2012.02.19, Orbit: 52163 
Sensor electronics, spectral resolution, digital quantification and other effects cause 
noise which can hamper image processing algorithms and especially edge detection 
algorithms (Fisher et al. 2003). Hence a high degree of noise reduction is necessary 
before applying the gradient and histogram front detection algorithms. Due to its 
edge-preserving nature a median noise filter was used in the histogram algorithm 
while a Gaussian noise filter is better suited to the gradient algorithm because it 
removes small scale details. 
2 For different sensor types see Table 1 in chapter 4!