Phd in INFORMATION AND COMMUNICATION TECHNOLOGY (ICT)

Thesis title: Geolocating Low-Earth-Orbit satellite data from next-generation millimeter-wave radiometers using natural targets

The main goal of this work is to perform the geolocation error assessment of the channel imagery at 183.31 GHz of the Special Sensor Microwave Imager/Sounder (SSMIS). The frequency around 183.31 GHz still represents the highest channel frequency of current spaceborne microwave and millimeter-wave radiometers. The latter will be extended to frequencies up to 664 GHz, as in the case of EUMETSAT Ice Cloud Imager (ICI). This use of submillimeter observations unfortunately prevents a straightforward geolocation error assessment using landmark-based techniques. This work uses SSMIS data at 183.31 GHz as a submillimeter proxy to identify the most suitable targets for geolocation error validation in very dry atmospheric conditions, as suggested by radiative transfer modeling. Using a yearly SSMIS dataset, 3 candidates landmark targets are selected: i) high-altitude lakes and high-latitude bays using a coastline reference database; ii) Antarctic ice shelves and Arctic shorelines using coastlines derived from Sentinel-1 Synthetic Aperture Radar (SAR) imagery; iii) high altitude mountains using digital elevation model as reference. Data processing is carried out by using spatial cross-correlation methods in the spatial frequency domain and performing a numerical sensitivity analysis to contour displacement. Cloud masking, based on a fuzzy-logic approach, is applied to automatically selected clear-air days. Results show that the average geolocation error is about 6.2 km for mountainous lakes and sea bays and 5.4 km for ice shelves, respectively, with a standard deviation of about 2.7 and 2.0 km. Results are in line with SSMIS previous estimates, whereas annual clear-air days are about 10% for mountainous lakes and sea bays and 18% for ice shelves. The second goal of this work is to investigate ICI channels, focusing on 243 GHz at horizontal polarization (ICI-4). The results of the simulations using radiative transfer model and artificial neural network (ANN) confirm that ICI-4 will be the best candidate to validate the geolocation of the future ICI radiometer. At 243 GHz the atmosphere is less opaque and the surface could be more visible with respect to other frequencies. This work proposes an artificial neural network to reconstruct the 243 GHz starting from real data at 150 GHz and 183 GHz. ANN provides an average value of about 5.8 km with a standard deviation of about 2.7 km. These numbers are in line with those obtained for 183 GHz, but at 243 GHz the number of images that contains visible surface targets are much more with respect to 183 GHz.