(Production Estimates and Crop Assessment Division) of the Foreign Agricultural Service of the U.S. Department of Agriculture (USDA). The USDA’s Cropland Data Layer, developed using Landsat 7 and Advanced Wide Field Sensor (AWiFS) data, is an excellent example of the use of remote sensing to monitor crop patterns and the implications for environment and society (http://www.nass.usda.gov/research/Cropland/SARS1a.htm).
Another recent successful application of satellite data in agricultural applications is the Famine Early Warning System Network (FEWS NET). This program was set up in 1985 by the U.S. Agency for International Development, initially in the Sahel and Horn of Africa and now extends to a few other arid developing nations, to incorporate satellite data in famine early warning (Hutchison 1998). This program uses AVHRR data to obtain vegetation conditions and rainfall estimates from the European Meteosat satellite. In FEWS NET, satellite information forms an important component of a multipronged approach to forecasting famines that includes both biophysical information and socioeconomic information to develop indicators for food supply, food access, and levels of development (Hutchison 1998). These and other achievements exemplify the benefits that can be gained from combining satellite observations with other available information (see Box 10.2, Figure 10.2).
Over the past few decades there has been increasing concern about tropical deforestation and the associated biodiversity loss and environmental consequences. Satellite data have played a crucial role in measuring both the rates and the patterns of forest loss. The first large-scale deforestation map using satellite imagery was made by Tardin and colleagues (1980) for the Brazilian Amazon. The NASA Pathfinder Humid Tropical Deforestation project has since made repeat assessments for the Amazon (Tardin and Cunha 1989, Skole and Tucker 1993) and for much of the tropics (Chomentowski et al. 1994; Figure 10.3).
Deforestation rates have been estimated for the entire tropics in several recent studies. Using a sampling of Landsat scenes, the Food and Agriculture Organization (FAO) mapped tropical deforestation for the 1980s and 1990s (FAO 2001), while the TREES II project of the Joint Research Center of the European Commission mapped deforestation rates for the humid tropics for the 1990s (Achard et al. 2002, 2004). While it is generally acknowledged that high-resolution remote sensing data are needed to identify deforestation, DeFries and colleagues (DeFries et al. 2002, Hansen and DeFries 2004) showed recently that it is also possible to estimate tropical deforestation over large areas using coarse-resolution weather satellite data (8-km resolution AVHRR Pathfinder data) calibrated against high-resolution estimates. Regardless of the specific methods used, all of these satellite-based estimates of deforestation rates were lower than those previously reported by ground-based inventories or national surveys (DeFries and Achard 2002, Hansen and DeFries 2004). The consequence of these new studies has been a lower estimate of carbon emissions from deforestation, with important implications for our understanding of the present-day carbon budget (DeFries and Achard 2002, Houghton 2003, Foley and Ramankutty 2004, Ramankutty et al. 2007).
While satellite data have been widely used to map deforestation around the world, good estimates of selective logging have not been available until recently. Asner and colleagues (2005) developed a method to estimate selective logging over the Amazon Basin using Landsat data (Figure 10.4). The study found that the area of forest damage from selective logging matched or exceeded rates of clear-cut deforestation. This implied a 25 percent increase in the estimate of gross annual anthropogenic emissions of carbon from Amazon forests over that estimated previously from deforestation alone. This has been a remarkable advance in our ability to map fine-scale patterns of land-use practices.
Even though monitoring and identifying regions of rapid land-cover change is a priority for the scientific community (for example, Box 10.3, Figure 10.5), baseline characterization of global land cover and land use is also important, especially for global analysis and modeling of ecosystems and their impacts. As described earlier, it is expensive and laborious to use Landsat data for large-area land-cover mapping. Therefore, moderate-resolution weather satellite sensors (~1-km resolution) have been used to characterize land-cover patterns globally (see Table 10.1). The University of Maryland pioneered the development of global land-cover classification data sets using AVHRR data. Since then there have been at least three other efforts to characterize global land cover (Table 10.1). These efforts have grouped the Earth’s landscape into numerous land-cover classes (Figure 10.6). In contrast to the discrete classifiers, the MODIS Vegetation Continuous Fields product provides a continuous description of the landscape (percentage tree cover, herbaceous and bare ground, as well as leaf type and phenology). These global data sets have provided a comprehensive global view of Earth’s land surface. They have become valuable inputs for global climate and ecosystem models used to study the influence of land-cover changes on the Earth system (DeFries et al. 1999, Feddema et al. 2005).
Fires are an important component of ecosystems; many natural communities depend on fires for their regeneration. Natural fires have been around since the presence of oxygen in the atmosphere, and humans have managed fire for more than a half-million years. However, only recently has the