Crowd-Sourced Calibration: The GEDI Strategy for Empirical Biomass Estimation Using Spaceborne Lidar

Abstract:

The central task in estimating forest biomass from spaceborne sensors is the development of calibration equations that relate observed forest structure to biomass at a variety of spatial scales. Empirical methods generally rely on statistical estimation or machine learning techniques where field-based estimates of biomass at the plot level are associated with post-launch observations of variables such as canopy height and cover. For global-scale mapping the process is complex and leads to a number of questions: How many calibrations are required to capture non-stationarity in the relationships? Where does one calibration begin and another end? Should calibrations be conditioned by biome? Vegetation type? Land-use? Post-launch calibrations lead to further complications, such as the requirement to have sufficient field plot data underneath potentially sparse satellite observations, spatial and temporal mismatches in scale between field plots and pixels, and geolocation uncertainty, both in the plots and the satellite data. The Global Ecosystem Dynamics Investigation (GEDI) is under development by NASA to estimate forest biomass. GEDI will deploy a multi-beam lidar on the International Space Station and provide billions of observations of forest structure per year. Because GEDI uses relatively small footprints, about 25 m diameter, post-launch calibration is exceptionally problematic for the reasons listed earlier. Instead, GEDI will use a kind of "crowd-sourced" calibration strategy where existing lidar observations and the corresponding plot biomass will be assembled from data contributed by the science community. Through a process of continuous updating, calibrations will be refined as more data is ingested. This talk will focus on the GEDI pre-launch calibration strategy and present initial progress on its development, and how it forms the basis for meeting mission biomass requirements.