Sustainability science from space: quantifying forest disturbance and land-use dynamics in the Amazon.

T ropical deforestation is occurring at unprecedented rates, contributing as much as a fifth of annual global carbon emissions and imparting significant impacts on biodiversity, ecosystem function, livelihoods, and climate (1–3). Selective logging, removal of 1–25 canopy trees per hectare with associated levels of harvesting damage, is predicted to greatly exceed tropical deforestation in extent worldwide, but it remains largely unquantified (4). However, selective logging is highly heterogeneous not only in time and space but also in intensity of forest disturbance. Quantifying how forests are degraded by selective logging has been one of the major technical challenges in remote sensing applications to forest cover assessment. Widely available satellite imagery used to monitor global vegetation cover has a much coarser grain than that of forest disturbance created by logging. These sensor constraints are exacerbated by rapid vegetation growth that reduces indicator signals measured by optical sensors and, thus, previously prevented reliable and consistent detection of logging, especially over several years. In this issue of PNAS, Asner et al. (5) present innovative high-resolution remote sensing analyses that have generated the first automated, large-scale assessments capable of discriminating logged forest condition across vast areas of the Amazon. Asner et al. (5) develop a standardized canopy gap fraction metric for both unlogged and logged forests that quantifies spatially explicit logging intensity and canopy damage systematically and with high accuracy across a range of forest structural conditions. Even this baseline measurement of canopy gap fraction in unlogged forest across a diversity of Amazonian regions is a major contribution to our understanding of natural forest gap dynamics. Through a 5-year time series, Asner et al. (5) track the fates of these logged areas and find that selective logging significantly increases the probability of deforestation, rendering current management schemes unsustainable. This work advances largearea remote sensing capabilities; it both demonstrates the urgent need to improve forest management and provides an essential analytical tool that could be applied to enhancing governance across the tropics.