PARENT SESSION
Oral Session #60: Remote Sensing and GIS. Presiding: A. Gallant.
Thursday, August 9, 2001. 8:00 AM to 12:00 PM. Hall of Ideas E.
A practical technique for relating land use and land cover data: a case study in the Amazon basin.
CARDILLE*, JEFFREY1, FOLEY, JONATHAN1, 1
ABSTRACT- Many broad-scale ecological questions require knowledge of both land use and land cover throughout the study region. Though similar, these two concepts are quite different: whereas land cover across a large area can be determined through a classification of satellite imagery, it is less straightforward to infer land use from remotely sensed data. There is, however, land use information for many countries in agricultural censuses, which contain highly detailed ground-surveyed data across an entire state or nation at a particular point in time. While these censuses quantify the characteristics of agricultural practices in an area, they are typically reported at the county scale--substantially coarser than satellite imagery. We have successfully developed a technique that merges land cover classifications with land use data into a unified map having the spatial resolution of satellite imagery and key attributes from agricultural census data. This regression-tree-based method allows us to create a new high-resolution land use data set across the entire Amazon basin for the mid-1990s. Results indicate that the fused map is correlated well with both agricultural census information and independent limited-extent land use data. We believe this technique is generalizable, readily understood, and potentially useful to those seeking a practical technique for fusing polygon and raster information. Here we explain the method and its results in the context of the creation of the Amazon basin land use data set.
KEY WORDS: land use, pasture, regression tree, Amazon