LC-19 Abstract

Use of Airborne Remote Sensing for Uncertainty Assessments in Regional Extrapolations of Ground LBA Ecology Measurements with MODIS Data

Laerte Guimaraes Ferreira, Federal University of Goias (UFG) (SA-PI)
Alfredo R. Huete, Terrestrial Biophysics and Remote Sensing Lab, University of Arizona (US-PI)
Edson Eyji Sano, EMBRAPA (SA-PI)

      The Cerrado and Amazon Forest Cerrado transition are the most intensely stressed areas in Brazil, with rapid and aggressive land use conversions.  This study aims to scale-up and extrapolate ground-measured ecological parameters, such as fraction of absorbed photosynthetically-active radiation (fAPAR) and leaf area index (LAI), over this fragmented and heterogeneous region in Brazil.  Synoptic and repetitive fAPAR and LAI estimations are crucial as inputs in regional-scale productivity and carbon budget models, prompting many remote sensing investigations for the development of techniques for satellite-based LAI/fAPAR retrievals.

      We propose low level AVIRIS flights for scaling-up field measured, ecologic parameters (fAPAR, LAI, %cover) across an environmental gradient spanning the diverse Brazilian cerrado biome (dry grassland woodland cerrado cerrado-forest transition) to the seasonal Amazon forest biome, including their land use and land converted classes.   As part of the scaling-up process, we will use AVIRIS hyperspectral reflectances to identify and discriminate the natural and altered land cover classes along the gradient of vegetation types using mixture models based on 'pure spectral signatures' of 'endmember' grassland, woodland, forest classes. This refined land cover characterization will be used in the scaling up of fAPAR, LAI, and % cover to AVIRIS pixel size (5 m) and the coarser Landsat ETM+ (30 m) and MODIS (250 m to 1 km) sensors. Our scaling and extrapolation approach will result in regional mapped products, with an accuracy and uncertainty fields, that account for distinct, land cover dependent, functional relationships between ecologic parameter (linear and non-linear) and vegetation indices. The scaled-up regional products will account for inter-sensor translation and continuity and will provide an independent 'validation' of the LAI/fPAR, land cover, and surface reflectance/VI MODIS products (both Terra and Aqua) for the cerrado,  transition, and forest study areas.  The results of this study will enable the detection of 'hotspots' of land conversion activities with the MODIS sensor;  enable for their fine-scale analysis with the Landsat ETM+ sensor or ASTER sensor; and allow a time-series analysis with the AVHRR historical data. This will greatly aid seasonal and inter-annual carbon balance studies in the Cerrado and Amazon transition regions.