Characterizing patterns of agricultural land use in Amazonia by merging satellite imagery and census data.

Jeffrey Alan Cardille - 2002
[UMI Proquest Full Citation]

In recent decades, millions of hectares of Amazonian primary forest, cerrado, and secondary forest have been cleared to support a dramatically increasing number of cattle and humans. With plans proposed for major new highways and utilities in the basin, development is highly likely to continue in coming years. Conversion to human use threatens to change the climate, ecosystems, and natural resources of Amazonia, and these effects are due not only to changes in land cover but to the land use management practices that follow. Unfortunately, we lack basin-wide information about land use across Amazonia. A key reason for this dearth of information is that earth-observing satellites designed to interpret land cover are prone to miss the land use changes within; in an area encompassing millions of square kilometers, it is impossible to visit more than a small portion of the study region to quantify land use activities. Agricultural censuses suggest a strategy to fill this gap: in Amazonia, they provide the only ground-surveyed land use information — yet because they are not easily reconciled with satellite-based land cover information, census data are underutilized. The research forming this dissertation presents a new, basin-wide depiction of land use in Amazonia by developing and applying new tools for understanding the past, current, and future impact of agricultural development. Specifically, this dissertation: (1) presents a new detailed understanding of the distribution and density of agricultural land use practices in Amazonia in the mid-1990s by fusing agricultural census data with satellite-derived land cover classifications; (2) assesses historical changes in agriculture of the previous decades; and (3) describes and applies new general techniques for the rapid update of land use data sets and maps using satellite imagery and census data. The fusion of census and satellite data described here advances our understanding by uniting the strengths of two distinct types of information about human activity in Amazonia. The contributions presented here provide demographic, economic, and ecological studies with new knowledge of the location, density, and history of crucial agricultural land use processes in this highly threatened and rapidly changing region.