Integrating Survey and Remote Sensing Data to Analyze Land Use at a Fine Scale: Insights from Agricultural Households in the Brazilian Amazon

Abstract

The ability to model factors influencing land use can be significantly improved by incorporating variables derived from geographic information systems with more detailed survey data. While remote sensing data have the advantage of providing land cover measurements for large areas, survey data collected from households provide a more detailed account of land use. We estimate land use decisions in the Brazilian Amazon with land cover data derived from satellite images merged with observations from a household panel. We focus on the rapidly expanding cattle industry or ``pecuarizaçao'' (cattleization) as well as a potentially competing land use strategy: the planting and harvesting of annual and perennial crops. We identify spatial error in our initial estimations and as a result use corrected models of land use with greater explanatory power and efficiency. Estimation results indicate that both pasture creation and agriculture are determined by similar household and spatial characteristics; however, the impacts of these determinants are in opposing directions for both land uses estimated. We conclude from our survey and satellite comparisons that while remote sensing data may overestimate the extent of pasture, the overestimation is minor. Finally, based on the current extent of pasture, we also conclude that policy that focuses on reducing deforestation within these long-established settlements is most likely inefficient. Rather, policy that addresses the land use practices employed by recent settlers may be more effective in reducing further pasture creation.