How to deal with the presence of weather affected data is an unavoidable topic in the processing of optical imagery. Clouds and cloud shadows significantly alter the spectral signatures obtained from satellite data, which often leads to problems for any kind of scientific analysis. In this research there has been elaborated on two different kind of problems: The detection of clouds and cloud shadows and the mitigation of the effect caused by cloud shadows. Most of existing operational cloud detection algorithms are so-called rule-based. Their performance is highly variable and they have their limitations. A new promising research was done by Mohajerani and Parvaneh (2019), where a convolutional neural network (CNN) named ’Cloud-Net’ was developed. In this study we have elaborated on this CNN, by converting the analysis to Sentinel-2 data and making significant modifications on the model setup. The results have been compared to the ESA Scene Classification Map (SEN2COR algorithm). It was found that for the detection of clouds the overall CNN accuracy outperforms the ESA Scene Map (95.6% vs. 92.0% respectively). For the detection of cloud shadows the modified Cloud-Net model also gave better results (90.4% vs. 84.4%). Previous work on cloud shadow correction algorithms show rather complex and inconvenient methods, where the only goal was to remove the effect of the shadow. If one is interested to also correct for illumination effects, to make it more aligned to a predetermined ground truth, new possibilities arise which allows for simpler and more direct methods. Two proposed methods have been investigated in this study. The first method, called ’decomposition of components’, investigated the use of a single formula. The affected cloud shadow pixel is corrected based on the RGB difference with a ground truth image, and a single correction factor that was determined based under the assumption that cloud shadows cause a homogeneous alteration effect in a small area. The second method, called the ’CNN based method’, presents a totally new idea by changing the Cloud-Net model to a regression model, in order to correctly alter cloud shadow affected pixels. The performance of both methods was quantified by the structural similarity index measure (SSIM). It was found that the decomposition of components method has the most potential, showing significant improvements on the correction of cloud shadow affected areas. ... Read more

The GRACE Follow-On mission is using classical K-Bandranging (KBR) and a new laser-ranging interferometry (LRI) method. The lattergives ranging data two orders of magnitude more accurate compared to theclassical K-Band ranging data (Dahl C et al. 2016). This gives a newopportunity for analyzing KBR noise by defining KBR noise as the differencebetween the KBR and LRI ranging data. In order to get a realization of KBR noise, the KBR andLRI epochs should be aligned and outliers have to be removed. An interpolation on the LRI data to make the LRI epochsaligned with the KBR epochs did not give sufficient results. Therefore only theoverlapping epochs were used as input for the next step of outlier detection, atotal of 2,250,000 epochs. The outlier detection method starts by taking theabsolute difference between the KBR and LRI range-rates. The largestdifferences were investigated where the original range-rate was compared to anestimation of the true value. This ‘true’ value is found by the use of a thirddegree polynomial function through the six range-rates that lie next to thesuspected outlier. The outlier detection method removes LRI range-rates thatdiffer more than 2.0*10^-8 m/s from the estimated true value and KBRrange-rates that differ more than 7.7*10^-7 m/s. 3764 LRI epochs and362 KBR epochs were labelled as outliers. After the outlier detection a histogram was made for theKBR range-rate noise. It was found that in order to get a normal distributionof KBR range-rate noise, the interval had to be in the range of <-3*10^-7 m/s, 3*10^-7m/s>. 59,000 still fell out of this noise range. 54,000 of these pointsformed 10 major clusters together where a subtraction of the KBR and LRIrange-rates did not give a realization of noise, but a trend similar to theoriginal signal and are therefore likely related to a clocking error. These54,000 points were therefore also removed from the dataset. To the remaining KBR range-rate noise values a thresholdof 3  was applied, removing an additional 4972 noisevalues that were larger than 4.03*10^-7 m/s. Finally , an estimation of the PSD of the KBR range-ratenoise  was made and compared to an oldPSD image of December 2008 of  the GRACEmission. The PSD plot of this project shows three peaks between 10^-4Hz and 10^-3 Hz. These peaks are not present in the PSD plot found inliterature of the GRACE mission. Except from the second and the third peak, thePSD values of the KBR range-rate noise of this project were lower, up to twoorders of magnitude in the lower frequency range (around 10^-4 Hz)and one order of magnitude in the higher frequency range (around 10^-1Hz). ... Read more