An Approach for Developing Earth Science Data Records of Global Forest Cover Change

There is long-standing recognition of the need for global forest change detection at Landsat-class resolutions. Previously this was not feasible, because of the absence of well registered multi-temporal data sets, variations in sensors, the need for intensive human input during post-processing, variations in spectral responses of forests, the efforts needed to create validation data sets and the computational and storage demands in carrying out the analysis. We will demonstrate in this presentation that these problems have now been largely overcome by the availability of the Global Landsat Survey (GLS) data sets, our ability to create atmospherically corrected reflectance products, much improved classifiers, collection of automated dense training sets, the availability of ultra-fine resolution datasets and much lower computational costs. We will present an approach for producing the following Earth Science Data Records (ESDR) at fine and moderate spatial resolutions: - Global fine resolution (< 100 m) surface reflectance ESDR for four epochs centered around 1990, 2000, and 2005; - Fine resolution (< 100 m) forest cover change (FCC) ESDR between the three epochs; - Fragmentation products derived from the fine resolution FCC products; - Global 250-m vegetation continuous field (VCF) based FCC ESDR from 2000 to 2005; - FCC ESDR products aggregated from the fine resolution and the 250 m FCC products to 250 m, 500 m, 1 km, and 0.05 degree grids for use by carbon, biogeochemical and hydrological modelers.