Open AccessDetermination of subpixel fractions of nonforested area in the Amazon using multiresolution satellite sensor data
Author(s) -
Hagen, S. C.,
Braswell, B. H.,
Frolking, S.,
Salas, W. A.,
Xiao X.
Publication year - 2002
Publication title -
journal of geophysical research: atmospheres
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.67
H-Index - 298
eISSN - 2156-2202
pISSN - 0148-0227
DOI - 10.1029/2000jd000255
Subject(s) - subpixel rendering , thematic mapper , remote sensing , land cover , amazon rainforest , data set , image resolution , satellite , thematic map , pixel , environmental science , geology , satellite imagery , cartography , geography , land use , computer science , mathematics , statistics , ecology , civil engineering , artificial intelligence , aerospace engineering , engineering , computer vision , biology
We present the results of an analysis that combines coarse and fine spatial resolution remote sensing data to reconstruct time series of nonforested area in a region of the Brazilian Amazon. We used a 10‐year sequence (1989–1998) of Landsat Thematic Mapper data for one scene (∼30,000 km 2 ) in Rondônia, Brazil, to parameterize a regression model that uses subsets of AVHRR GAC reflectance data. This model detected information on interannual changes in nonforested area with a combination of detail and coverage greater than would normally be available using either data set alone. Within this domain we retrieved nonforest cover fraction in a cross‐validation test with coefficients of determination ( R 2 ) of 0.32 at 8–km resolution and 0.64 at 48‐km resolution. At the 48‐km resolution the model captures interannual variability and trends of fractional cover within individual pixels that changed considerably over the 10 years of this analysis. The results for this region suggest that retrospective analyses will benefit from further development of techniques combining AVHRR, TM, and other prior information.