LAI and FPAR estimation and land cover identification with multiangle multispectral satellite data

Land cover, vegetation green leaf area index (LAI) and fraction of incident photosynthetically active radiation absorbed by vegetation (FPAR) are important variables in many environmental processes and are thus key parameters in general circulation and biogeochemical models. The estimation of these variables over large areas, or globally, is greatly facilitated by remote sensing techniques. Recent advances in remote sensing technology make feasible multispectral and multiangle remote sensing of the Earth's surface at moderate spatial resolutions (∼1km). This dissertation is focused on LAI and FPAR retrievals and land cover identification with multispectral and multiangle remote sensing data. A physically based LAI/FPAR algorithm designed for the Multiangle Imaging Spectroradiometer (MISR) has been tested with data from the Polarization and Directionality of the Earth's Reflectance (POLDER) instrument. Results from prototyping of the MISR LAI/FPAR algorithm with POLDER data over Africa demonstrate the feasibility of global LAI and FPAR retrievals from multiangle data. The quality of the LAI/FPAR product is mainly determined by band- and view direction-dependent uncertainties in atmospherically corrected surface reflectances. The use of multiangle data result in fewer LAI solutions that are consistent with observations, and the retrieval uncertainties in LAI and FPAR are consequently reduced. Further analysis with POLDER data from North America demonstrates that the display of vegetation angular signatures in spectral space is a powerful tool for visualizing the information content of multiangle data. Such signatures provide a cogent synthesis of information from both the spectral and angular domains. The angular signatures can be characterized by their location, inclination and length. Statistical analyses indicate that these metrics convey information valuable for identification of biome types. The physical basis of these angular signatures is provided by the three-dimensional radiative transfer equation, which expresses the law of energy conservation in the most general form, and provides the required consistency between land cover definitions and their signatures. Preliminary classification analysis using POLDER data suggests improvement in classification accuracy when multiangle data is used together with multispectral data. Further research with MISR is required to document the utility of multiangle data for land cover and biome classification.