InSAR decorrelation to assess and prevent volcanic risk

SAR has several strong key features: fine spatial resolution/precision and high temporal pass frequency. Moreover, the InSAR technique allows the accurate detection of ground deformations. This high potential technology can be invaluable to study volcanoes: it provides important information on pre-eruption surface deformation, improving the understanding of volcanic processes. As a downside, SAR measurements are influenced by artifacts such as atmospheric effects or bad topographic data. Correlation gives a measure of these interferences, quantifying the similarity of the phase of two SAR images. Different approaches exist to reduce these errors but the main concern remains the possibility to correlate images with different acquisition times: snow-covered or heavily-vegetated areas produce seasonal changes on the surface. Working with images with very fast acquisition time partly limits decorrelation. Though, images with a short temporal baseline aren't always available and some artifacts affecting correlation are time-independent. This work studies correlation of pairs of SAR images focusing on the influence of surface and climate conditions, especially snow coverage and temperature. Furthermore, the effects of the acquisition band on correlation are taken into account, comparing L-band and C-band images. All the chosen images cover most of the Yellowstone caldera (USA) over a span of 4 years, sampling all the seasons. Interferograms and correlation maps are generated. To isolate temporal decorrelation, pairs of images with the shortest baseline are chosen. Correlation maps are analyzed in relation to snow depth and temperature. Results obtained with ENVISAT and ERS satellites (C-band) are compared with the ones from ALOS (L-band). Results show a good performance during winter and a bad attitude towards wet snow (spring and fall). During summer both L-band and C-band maintain a good coherence with L-band performing better over vegetation.