Fragmentation in the Legal Amazon, Brazil: Can landscape metrics indicate agricultural policy differences?

Abstract

Large-scale land conversion for agriculture in Brazilian Amazonia is occurring at persistently high rates. Basin-wide net land use and land cover changes imply substantially different situations between distinct regions and states due to different agricultural policies. This research used eight landscape metrics to quantify and investigate the spatial patterns of cattle pasture and cropland throughout the states of Pará, Mato Grosso, Rondônia, and Amazonas. These metrics were patch density (DEN), mean patch size (MPS), largest patch index (LPI), mean edge density (MED), mean twist number (TWI), corrected perimeter-to-area ratio (CPA), fractal dimension (FDI), and fragmentation index (FRG). A total of 1852 patches were analyzed, originating from 86 samples in 71 different plots, covering a total of 177,500 km 2 throughout all four states. Principal component analysis showed a partial overlap in the spatial pattern of agricultural patches between all states. The largest percentage of variance was explained by patch area metrics, which can be related to the different approaches in agricultural policies, but no clear division between the states was identified in this dimension. The metrics quantifying patch shape were de facto independent of deforestation area, and related to the second principal component axis. Although some overlap in this dimension was present as well, these metrics proved a possible measure for discerning the patterns of agriculture attached to a certain state. Different land use policies are hypothesized to lead to more heterogeneity in landscape patterns in an early stage, yet the increasing influence of both cropland and pasture agriculture eventually leads to more uniform landscapes in which spatial differences gradually disappear.