Retrieving Biome Types from Multi-angle Spectral Data

Mitchell Schull and Liang Xu and Pedro Latorre-Carmona and Arindam Samanta and Ranga Myneni and Yuri Knyazikhin

Many studies have been conducted to demonstrate the ability of multi-angle spectral data to discriminate plant dominant species. Most have employed the use of empirically based techniques, which are site specific, requires some initial training based on characteristics of known leaf and/or canopy spectra and therefore may not be extendable to operational use or adapted to changing/unknown land cover. An ancillary objective of the MISR LAI/FPAR algorithm is classification of global vegetation into biome types. The algorithm is based on the 3D radiative transfer equation. Its performance suggests that is has valid LAI retrievals and correct biome identification in about 20% of the pixels. However with a probability of about 70%, uncertainties in LAI retrievals due to biome misclassification do not exceed uncertainties in the observations. In this poster we present an approach to improve reliability of the distribution of biomes and dominant species from multi angle spectral data. The radiative transfer theory of canopy spectral invariants underlies the approach, which facilitates parameterization of the canopy bidirectional reflectance factor in terms of the leaf spectrum and two spectrally invariant and structurally varying variables recollision and directional escape probabilities. Theoretical and empirical analyses of ground and airborne data acquired by AVIRIS, AirMISR over two sites in New England and CHRIS/PROBA over BARAX site in Spain suggest that the canopy spectral invariants convey information about canopy structure at both the macro and micro scales. These properties allow for the natural separation of biome classes based on the location of points on the total escape probability vs the proportional escape ratio log-log plane.

(BibTeX entry) (PDF)