Characterizing a Dynamic Land Cover Change Frontier Using MODIS Phenology Metrics: Cropland Expansion in Mato Grosso, Brazil

The state of Mato Grosso, Brazil experienced the most rapid agricultural expansion of any region in the Amazon Basin during the existing MODIS data record. Rapid conversion of Amazon forest, transitional tropical forest, and cerrado woodland-savanna for cattle ranching and grain production continues to fragment large tracts of these biomes. Tropical forest loss is estimated annually, yet the fate of cleared lands and losses of transitional forest or cerrado have not been well characterized in this region. Using phenological information from time series of MODIS 16-day composite data, it is possible to capture the temporal dynamism of land cover change and accurately separate primary and secondary land use transitions. We use time series of cloud-cleaned MODIS NDVI and EVI at 250 m resolution to characterize land cover based on metrics of wet-season, dry-season, and annual phenology from 2000-2004. Distinct phenological signatures for forest, pasture and natural grasslands, cerrado, and cropland enable accurate classification of land cover types when compared to field validation data (overall accuracy = 85%). We estimate that more than 1.6 million hectares were converted to cropland between 2000 and 2004. The majority of new cropland resulted from the direct conversion of cerrado (35%) or forest (29%); conversion of natural grassland areas or planted pasture accounted for 36% of new cropland areas. While secondary transitions from existing cattle pasture to cropland are an important source of new agricultural production, our findings contradict recent statements that cropland agriculture is not directly associated with new deforestation activities. Separation of more seasonal cerrado vegetation from transitional tropical forest based on vegetation phenology highlights land cover dynamics in regions with no previous deforestation monitoring. Phenological information from MODIS is extremely important to monitor land cover dynamics, separate forest types, and estimate impacts of tropical forest replacement by grassland (C4) or cropland (C3) on regional hydrology, energy balance, and ecosystem function.