Assessing the response of the MODIS vegetation indices to landscape disturbance in the forested areas of the legal Brazilian Amazon

Abstract

In this study we assessed the impacts of forest fragmentation on the Amazon landscape using remote sensing techniques. Landscape disturbance, obtained for an area of approximately 3.5 × 106 km2 through simple spatial metrics (i.e. number of fragments, mean fragment area and border size) and principal component transformation were then compared to the MODIS (Moderate Resolution Imaging Spectroradiometer) NDVI (Normalized Difference Vegetation Index) and EVI (Enhanced Vegetation Index) seasonal responses. As expected, higher disturbance values prevail in the southern border of the Amazon, near the intensively converted deforestation arc, and close to the major roads. NDVI seasonal responses more closely follow human-induced patterns, i.e. forest remnants from areas more intensively converted were associated with higher NDVI seasonal values. The significant correlation between NDVI seasonal responses and landscape disturbances were corroborated through analysis of geographically weighted regression (GWR) parameters and predictions. On the other hand, EVI seasonal responses were more complex with significant variations found over intact, less fragmented forest patches, thus restricting its utility to assess landscape disturbance. Although further research is needed, our results suggest that the degree of fragmentation of the forest remnants can be remotely sensed with MODIS vegetation indices. Thus, it may become possible to upscale field-based data on overall canopy condition and fragmentation status for basin-wide extrapolations.