Forest structure and composition from remote sensing to constrain predictions of forest ecosystem carbon dynamics

Background/Question/Methods

Forest structure and composition are key attributes of the terrestrial biosphere, affecting its biogeochemical and biophysical functioning, including current and future carbon fluxes and the exchange of water and energy between the land-surface and the atmosphere. Information on the structure and composition of forest ecosystems has traditionally come from ground-based inventories that provide detailed information on the size and species identity of stems within small plots, but do not provide a comprehensive, spatially-consistent measure of the current state of the above ground ecosystem. In this study, we investigate how active and hyperspectral datasets can provide appropriate information that can be used to constrain terrestrial-biosphere model carbon fluxes. 

Results/Conclusions

Full waveform lidar (Light Detection and Ranging) data were used in conjunction with a radiative transfer modeling approach to derive forest structure. Hyperspectral imagery was used with a spectral mixture analysis approach to derive forest composition. Three markedly different sites at Harvard Forest were correctly estimated as being either low or high tree density sites. The proportion of plant functional types in these sites was also correctly identified. The analysis subsequently showed that carbon fluxes from the remote sensing-derived forest structure and composition initialized using the ED2 biosphere model showed very close agreement between both ground-based initialized fluxes, as well as to the flux tower measurements (NEP within 0.02 kgC m^-2 yr^-1 of each other). A second analysis showed that traditional methods of simulating grid cells with a single plant functional type until the forest reaches a potential vegetation state with its climate forcing data, produces carbon fluxes which do not accurately reflect the disturbed locations at Harvard Forest (NEP 0.05-0.15 kgC m^-2 yr^-1 less than flux tower). Results from this study suggest that regional and potentially global-scale terrestrial carbon cycle and ecosystem model simulations could be constrained with satellite active and passive remote sensing data (e.g. ICESat-2/ HyspIRI), to produce more accurate predictions of how the earth system will change in the future.