Analysis of imagery of Belize, Central America, recently acquired by the National Aeronautics and Space Administration (NASA) Earth Observing-1 (EO-1) satellite, shows that this experimental imagery can be used effectively to classify a diverse set of land cover/land use types. The Bureau of Economic Geology and Center for Space Research are investigating the new imagery as part of a NASA-sponsored program to evaluate the capabilities of technologically advanced sensors onboard the EO-1 satellite to image the Earth's surface. Classification of land cover/land use using the multispectral Advanced Land Imager (ALI) on the satellite indicates that classifications were similar but superior to those of Landsat 7 Enhanced Thematic Mapper (ETM+) for several difficult classes in test data, such as thicket, regrowth, orchards, and cleared land. In addition, ALI data appear to be superior to Landsat TM data in delineating some coastal land-cover classes, such as mangrove and marshes. New statistical classification methods developed during the study yielded improved discrimination between difficult classes in both ALI and Landsat TM data. ALI data were also effectively used to determine impacts of Category Four Hurricane Iris, which made landfall in southern Belize on October 8, 2001. Comparison of post-hurricane ALI data with pre-hurricane Landsat 7 data indicated that broadleaf forest had been extensively damaged in the Monkey River area approximately 130 km south of Belize City. On ALI imagery acquired after the hurricane, more than 98 percent of areas previously classified as broadleaf forest using Landsat TM data were classified primarily as savannah and other grasslands, indicating extensive broadleaf destruction and defoliation. A similar analysis in inland mountainous areas that were affected by the hurricane also showed large areas of downed and defoliated broadleaf trees. The ALI data clearly delineated changes in spectral signatures and textures as a result of Hurricane Iris.

Spectral data are being classified using both existing statistical methods and new contextual and multisensor algorithms currently being developed at The University of Texas at Austin for multispectral and hyperspectral data. Classification results are entered into a geographic information system (GIS) for analysis of land-cover and land-use distribution and change. Classified areas are checked for accuracy and consistency using existing maps and previously collected land cover/land use data at Global Positioning System (GPS)-located field survey sites and overflights, supported by additional field verification sites using GPS coordinates. Results include an evaluation and validation of the capabilities of EO-1 and Landsat 7 ETM+ data for classifying a diverse set of land cover/land use types and analyzing trends such as rates of deforestation and regrowth.