Assessing land use/cover dynamics and exploring drivers in the Amazon's arc of deforestation through a hierarchical, multi-scale and multi-temporal classification approach

摘要: Abstract Land use and land cover (LULC) are intrinsically tied to ecological social dynamics. Still, classifying LULC in ecotones, where landscapes commonly heterogeneous have a wide range of physiognomies, remains challenge. Here we present three-level hierarchical classification approach, using both Landsat MODIS images, pixels objects as units information. We applied this multi-temporal -spatial approach classify the Upper Xingu River Basin (∼170,000 km2), located arc deforestation Brazilian Amazon. The first level includes five classes differentiates managed from native vegetation with high overall accuracy (93%). second has 11 (overall accuracy = 86%) separates main uses domains. third 16 accuracy = 83%) addresses productivity natural systems. find that new method presented here is more efficient than existing regional global products. Applying assess transitions basin 1985 2015, agricultural production increased, yet manifested itself differently northern (Amazon biome) southern (Cerrado portions basin. Analyzing change different levels, identify intensification occurred mainly Amazon while Cerrado undergone an expansion area. can be adapted other regions, improving efficiency landscapes.