Comparison of AVIRIS and AISA for chemistry mapping
ABSTRACT Hyperspectral sensing of forest chemistry can provide indicators of forest health. Foliar pigments are directly involved with the photosynthetic process and, therefore, are intimately tied to vegetation vigor. AISA and AVIRIS hyperspectral datasets were acquired over the Greater Victoria Watershed District test site in 2006 and 2002, respectively. AISA was calibrated to AVIRIS to facilitate sensor comparison. The data were used to generate a forest species classification, endmember fractions and chemistry for test plots. The hyperspectral products were used to separate ground cover (Salal) from the forest overstory and chemistry was estimated for both layers. Classification accuracies exceeded 89% in mapping major forest species. AVIRIS predicted chemistry agreed with measured chemistry (R2: 0.98). Incorporating an understory stratification step was anticipated to increase the accuracy of chemistry estimates; however, R2 values were unchanged. While plot data suggested AISA chemistry prediction performed well, significant bidirectional reflectance effects were evident; this effect was absent in the AVIRIS data.
- SourceAvailable from: Daniel Schläpfer[Show abstract] [Hide abstract]
ABSTRACT: A method for the radiometric correction of wide field-of-view airborne imagery has been developed that accounts for the angular dependence of the path radiance and atmospheric transmittance functions to remove atmospheric and topographic effects. The first part of processing is the parametric geocoding of the scene to obtain a geocoded, orthorectified image and the view geometry (scan and azimuth angles) for each pixel as described in part 1 of this jointly submitted paper. The second part of the processing performs the combined atmospheric/topographic correction. It uses a database of look-up tables of the atmospheric correction functions (path radiance, atmospheric transmittance, direct and diffuse solar flux) calculated with a radiative transfer code. Additionally, the terrain shape obtained from a digital elevation model is taken into account. The issues of the database size and accuracy requirements are critically discussed. The method supports all common types of imaging airborne optical instruments: panchromatic, multispectral, and hyperspectral, including fore/aft tilt sensors covering the wavelength range 0.35 - 2.55 µm and 8-14 µm. The processor is designed and optimized for imaging spectrometer data. Examples of processing of hyperspectral imagery in flat and rugged terrain are presented. A comparison of ground reflectance measurements with surface reflectance spectra derived from airborne imagery demonstrates that an accuracy of 1 - 3 % reflectance units can be achieved.International Journal of Remote Sensing 01/2002; 23(2002-13):2631-2649. · 1.14 Impact Factor
Conference Paper: Calibration of forest chemistry for hyperspectral analysis[Show abstract] [Hide abstract]
ABSTRACT: A primary advantage of hyperspectral sensors is the ability to provide measurements of canopy chemistry. Canopy chemistry can be used to estimate new and old foliage, detect damage, identify trees under stress, and map chemical distributions in the forests. We have begun a new EO-1 project, Evaluation and Validation of EO-1 for Sustainable Development of forests (EVEOSD). NASA's EO-1 satellite was successfully launched on November 21, 2000. In preparation for airborne and spaceborne data collection and calibration, we collected in September 2000 foliar canopy and ground cover chemistry samples from 54 plots distributed across the Greater Victoria Watershed (GVWD) test site. Treetop samples were collected from helicopters. Differential GPS was used to provide sample positioning to within 1 m. The foliar samples were divided into new and old foliage. Organic and inorganic chemistry analyses were done. Spectral calibration samples were collected over ground targets, over stacks of foliar samples, and over ground vegetation. Landsat-7 and Radarsat data were collected at the same time. The chemistry samples were placed into a database and integrated with GIS files of topography and forest cover. We obtained 1 m aerial orthophotography that allowed us to investigate the spectral components making up the Landsat-7 and EO-1 pixelsGeoscience and Remote Sensing Symposium, 2001. IGARSS '01. IEEE 2001 International; 02/2001
- [Show abstract] [Hide abstract]
ABSTRACT: Leaf chemical and spectral properties of 162 canopy species were measured at 11 tropical forest sites along a 6024 mm precipitation/yr and 8.7 degrees C climate gradient in Queensland, Australia. We found that variations in foliar nitrogen, phosphorus, chlorophyll a and b, and carotenoid concentrations, as well as specific leaf area (SLA), were expressed more strongly among species within a site than along the entire climate gradient. Integrated chemical signatures consisting of all leaf properties did not aggregate well at the genus or family levels. Leaf chemical diversity was maximal in the lowland tropical forest sites with the highest temperatures and moderate precipitation levels. Cooler and wetter montane tropical forests contained species with measurably lower variation in their chemical signatures. Foliar optical properties measured from 400 to 2500 nm were also highly diverse at the species level, and were well correlated with an ensemble of leaf chemical properties and SLA (r2 = 0.54-0.83). A probabilistic diversity model amplified the leaf chemical differences among species, revealing that lowland tropical forests maintain a chemical diversity per unit richness far greater than that of higher elevation forests in Australia. Modeled patterns in spectral diversity and species richness paralleled those of chemical diversity, demonstrating a linkage between the taxonomic and remotely sensed properties of tropical forest canopies. We conclude that species are the taxonomic unit causing chemical variance in Australian tropical forest canopies, and thus ecological and remote sensing studies should consider the role that species play in defining the functional properties of these forests.Ecological Applications 02/2009; 19(1):236-53. · 3.82 Impact Factor