Linking Remote Sensing with a State-of-the-Art Terrestrial Biosphere Model to Better Predict Ecosystem Dynamics

Until recently, estimates of forest structure and composition have been based on ground-based forest inventories that are limited in their spatial extent, and do not provide a comprehensive estimate of the current state of the above-ground ecosystem. Previous studies have shown how active remote sensing measurements can provide information on forest structural attributes, and how such remotely-sensed estimates of forest structural attributes can be used to help constrain the predictions of terrestrial biosphere models. However, to date, these remote sensing estimates of forest structure have focused on developing estimates for single forest metrics, such as lidar estimates of canopy height or radar estimates of above-ground biomass. In this study, we investigate how a combination of remote sensing data over Harvard Forest, MA, can be used to best estimate forest above-ground ecosystem state and subsequently constrain and test the dynamics of the ED2 state-of-the-art terrestrial biosphere model. Following on from previous work, this fusion of forest structural attributes is more than just canopy height and biomass, and aims at using remote sensing signals more completely. Recent results show that waveform lidar data from LVIS is successful (R2 = 0.8, RMSE = 0.7) in determining LAI over 40 plots at Harvard Forest from canopy gap probabilities. This was better than just considering NDVI-LAI relationships from optical remote sensing techniques such as EO-1 Hyperion (R2 = 0.45). We then use the remote-sensing derived estimates of forest structure to constrain the state of above-ground ecosystems within ED2, a state-of-the-art terrestrial biosphere model, and quantify the impacts of this reduction in uncertainty regarding the current ecosystem state on predictions of current and future biophysical and biogeochemical functioning at Harvard Forest.