Spatiotemporal mapping of soybean plantations in Rondônia, Western Brazilian Amazon

ABSTRACT Although soybean production has been increasing in the state of Rondônia in the last decade, soybean planted area has been estimated indirectly using secondary datasets, which has limited understanding of its spatiotemporal distribution patterns. This study aimed to map and analyze spatial patterns of soybean expansion in Rondônia. We developed a classification technique based on Spectral Mixture Analysis (SMA) derived from Landsat imagery and Decision Tree Classification to detect and map soybean plantations in 2000, 2005, 2010, and 2014. The soybean classification map showed 93% global accuracy, 23% omission and 0% of commission errors for soybean crop fields. The greatest increases of soybean cropped area in the state of Rondônia were observed between 2000-2005 and 2005-2010 time-periods (33,239 ha and 59,628 ha, respectively), mostly located in Southern Rondônia. The expansion of soybean areas to Northern Rondônia (25,627 ha) has mostly occurred in the 2010-2014 time period. We estimate that 95.4% of all newly created soybean plantations, detected by 2014, were established on lands deforested nine or more years earlier. We concluded that the incursion of soybean plantations on lands deforested for other land uses (e.g. ranching) is contributing to their displacement (pastures) from older colonization zones toward more remote frontier areas of the Amazon, exacerbating new deforestation there.