Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Israel → Shrublands Biomass Mapping by Integrating Polarimetric SAR Data and Multi-Spectral Data

Technion - Israel Institute of Technology (2017)

Shrublands Biomass Mapping by Integrating Polarimetric SAR Data and Multi-Spectral Data

Jisung Chang

Titre : Shrublands Biomass Mapping by Integrating Polarimetric SAR Data and Multi-Spectral Data

Auteur : Jisung Chang

Etablissement de soutenance : Technion - Israel Institute of Technology

Grade : Doctor of Philosophy (PhD) 2017

The south-eastern Mediterranean is a hot-spot of climate change and desertification. Further warming and drying of this region until the end of this century is predicted with potential of yielding significant deterioration of ecosystems around the Mediterranean Basin. Shrublands consist a primary element of these ecosystems and their resilience to climate change. High topographic, lithologic, and soil variability combined with natural and anthropogenic influences resulted in high spatio-temporal heterogeneity. Extracting information about shrublands response to climate change is difficult in such complex environments. The use of sensor systems that are sensitive to vegetation and soil properties in areas of varying habitat conditions is instrumental for this purpose. Numerous optical remote sensing studies were carried out aiming at mapping the Mediterranean and arid environments and the changes they undergo. Limitations on the information content of optical sensor data regarding biomass are well recognized by researchers in this field. Radar L-band systems with quad-polarization measurements (HH, VV, HV and VH) have the potential of extraction of volumetric information by remote sensing, however, experience in the use of polarimetric radar in highly heterogeneous landscapes is quite limited. Full polarimetric L-band PALSAR data facilitated one of the first empirical assessments of the relationships between radar polarization properties (RPPs) and ecological pattern properties (EPPs) in the desert fringe shrublands of the South-Eastern Mediterranean Basin. Correlations between radar parameters (polarized backscattering intensities, their ratios, entropy, and degrees of polarization) and remotely sensed ecological parameters (shrub cover, edge parameter, average shrub height, and biomass information) were analyzed for 52 least-disturbed sites along a climatic gradient between semi-arid and arid zones in the south-eastern Mediterranean. The degree of horizontal polarization (DOHP) was the RPP that was most highly correlated with all EPPs, while NDVIR biomass (NDVI multiplied by relative rainfall) was the EPP most highly correlated with all RPPs. Integrating NDVI with DOHP facilitated one of the first attempts to map shrublands across the whole Mediterranean Basin. Principle Component Analysis of NDVI and DOHP combinations indicated that DOHP has significant contribution for the discrimination between land cover types. Comparison of NDVI and DOP combination classes with existing Global Land Cover data bases created using remote sensing techniques, revealed that they agree regarding forest extents, while there is very limited information regarding shrublands in the global databases. NDVI and DOHP showed potential for mapping shrublands and for separating between desertified and vegetated surfaces. A new polarimetric radar vegetation index (PRVI) utilizing the degree of polarization and the cross-polarized backscattering coefficient, based on the theoretical volume scattering model and a semi-empirical model is proposed. The accuracy of the model was examined in semi-arid natural areas having high geodiversity, representing wide desert fringe zones around the world. It shows good agreement with the biomass data extracted using NDVI and rainfall data and it performed better than other radar parameters. Assessment of the PRVI based biomass estimation with allometric data from 67 sites across the desert fringe zone indicated moderate performance with Root Mean Square Deviation of 0.33 Kg/.


Page publiée le 26 novembre 2018