LBA - Land Cover Dataset for the Center for Sustainability and the Global Environment (SAGE)

These files are gridded at 5-minute latitude by longitude (~9km at the equator) resolution for the entire Amazon/Tocantins river drainage basins (21S - 6N; 45W- 80W). The files contain the estimated proportions of total agriculture ("totag"), cultivated area ("cul"), natural pasture ("pasnat"), and planted
pasture ("paspla") in each grid cell in either the mid-1980s or mid-1990s. Data is provided in the netcdf format (".nc" files) and the ARC/INFO ASCII format
(".txt" files). Proportions are expressed as parts per 10,000 so that, for example, a value of 8945 means that 89.45% of the given pixel is estimated by
this procedure to be used for agricultural activity.

The data were generated from a fusion of agricultural census data and satellite classifications, and are described in Cardille, J.A., J.A. Foley, and M.H.
Costa (2002). Characterizing patterns of agricultural land use in Amazonia by merging satellite classifications and census data. Global Biogeochemical Cycles 16(3), 10.1029/2000GB001386, 20. The fusion technique merges agricultural census data from countries of the basin (including Brazil) with land cover data from the University of Maryland Global Land Cover Facility 1-km classification. This technique was used to derive the mid-1990s total agriculture surface for the region, which was then apportioned according to agriculture census data proportions into cultivated area, natural pasture, and planted pasture.

The mid-1980s maps were created by scaling the mid-1990s snapshots backward in time using the relative increase or decrease in agriculture, as derived from
mid-1980s census data and United Nations Food and Agriculture Organization (FAO) data.

These data files are generated by a fusion of two data sources which may contain errors in position and value. In particular, census variables may be
susceptible to under- and over-estimation for political and economic reasons. While we do not treat these errors of estimation directly, it appears that the
fusion technique smooths extreme/unlikely values in the census data.

[Summary provided by SAGE.]