Land cover and forest mapping in boreal zone using polarimetric and interferometric SAR data: Dissertation

Remote sensing offers a wide range of instruments suitable to meet the growing need for consistent, timely and cost-effective monitoring of land cover and forested areas. One of the most important instruments is synthetic aperture radar (SAR) technology, where transfer of advanced SAR imaging techniques from mostly experimental small test-area studies to satellites enables improvements in remote assessment of land cover on a global scale. Globally, forests are very suitable for remote sensing applications due to their large dimensions and relatively poor accessibility in distant areas. In this thesis, several methods were developed utilizing Earth observation data collected using such advanced SAR techniques, as well as their application potential was assessed. The focus was on use of SAR polarimetry and SAR interferometry to improve performance and robustness in assessment of land cover and forest properties in the boreal zone. Particular advances were achieved in land cover classification and estimating several key forest variables, such as forest stem volume and forest tree height. Important results reported in this thesis include: improved polarimetric SAR model-based decomposition approach suitable for use in boreal forest at L-band; development and demonstration of normalization method for fully polarimetric SAR mosaics, resulting in improved classification performance and suitable for wide-area mapping purposes; establishing new inversion procedure for robust forest stem volume retrieval from SAR data; developing semi-empirical method and demonstrating potential for soil type separation (mineral soil, peatland) under forested areas with L-band polarimetric SAR; developing and demonstrating methodology for simultaneous retrieval of forest tree height and radiowave attenuation in forest layer from interferometric SAR data, resulting in improved accuracy and more stable estimation of forest tree height