HR: 0800h
AN: B41A-0166    [Abstracts]
TI: An Improved Spatial Data Set of Tropical Deforestation Rates for the 1980s and 1990s
AU: * Gibbs, H K
EM: hkgibbs@wisc.edu
AF: University of WIsconsin-Madison, 1710 University Avenue, Madison, WI 53726 United States
AU: Ramankutty, N
EM: nramanku@wisc.edu
AF: University of WIsconsin-Madison, 1710 University Avenue, Madison, WI 53726 United States
AU: Foley, J A
EM: jfoley@wisc.edu
AF: University of WIsconsin-Madison, 1710 University Avenue, Madison, WI 53726 United States
AU: DeFries, R S
EM: rdefries@mail.umd.edu
AF: University of Maryland, College Park, 2181 Lefrak Hall, College Park, MD 20742 United States
AU: Houghton, R A
EM: rhoughton@whrc.org
AF: Woods Hole Research Center, PO Box 296, Woods Hole, MA 02543 United States
AU: Achard, F
EM: frederic.achard@jrc.it
AF: Joint Research Centre of the European Commission, CCR / TP 440, Ispra, VA I-21020 Italy
AB: Tropical land cover dynamics in the 1980s and 1990s are highly uncertain, with enormous implications for balancing the global carbon budget and understanding the impacts on ecosystem goods and services. Recent estimates of tropical deforestation during the 1980s and 1990s vary by +/-40 percent due in part to differences in domain, forest baselines, methods, and definitions. The 8km Advanced Very High Resolution Radiometer (AVHRR) satellite record provides the only spatially-explicit data with comprehensive global coverage for both the 1980s and 1990s. However, sensor calibration and degradation issues combined with the coarse spatial resolution of AVHRR data may mask more diffuse deforestation events and likely capture only net changes in forest cover, thereby underestimating both gross deforestation and forest regrowth. Higher resolution Landsat data can capture gross changes in forest cover, but the processed data products are currently limited to particular regions or sampling schemes and "wall-to-wall" coverage is not available for the total tropics during the 1980s and 1990s. We used 200+ classified Landsat scenes from the TREES project and the FAO's Forest Resources Assessment to develop a spatially-explicit regression tree model based largely on the AVHRR record. Inputs to the regression tree included demographic, biophysical, and land-use predictor variables such as population, fires, soils, elevation, and distance from roads, rivers, and urban centers. We used the regression model to create an improved spatially-explicit estimate of tropical deforestation rates and locations that incorporates the strengths of key regional to global-scale data sets.
DE: 0400 BIOGEOSCIENCES
DE: 0402 Agricultural systems
DE: 0414 Biogeochemical cycles, processes, and modeling (0412, 0793, 1615, 4805, 4912)
DE: 0428 Carbon cycling (4806)
DE: 0429 Climate dynamics (1620)
SC: Biogeosciences [B]
MN: Fall Meeting 2005