Accounting for tree species composition in landscape-scale forest carbon storage estimates
Accurate measurement of carbon storage in forests is crucial to determine the impact of changes in forest cover on the global carbon cycles. Carbon storage estimates interpolated from forest inventory data are widely-used, but are resource-intensive and highly inaccurate in heterogeneous landscapes where interpolation is not appropriate. Remotely-sensed data products, combined with novel software (Dinamica EGO v2.4), provide an opportunity to improve carbon assessments in such landscapes. This study examines the impact of differing remote sensing inputs on carbon assessments of Vermont forests. Specifically, we compared 1) remote sensing total biomass maps (National Biomass and Carbon Dataset, WHRC) to carbon storage results calculated from 2) LANDFIRE (LANDFIRE Existing Vegetation Type Layer, U.S. DOI) maps (with carbon estimates specific to FIA-based forest subtype classifications) and stand age (Forest Age Maps for Canada and the U.S.A., NACP) maps, and carbon storage results calcuated from 3) general land cover (tree cover delineated as coniferous, deciduous, and mixed only) and stand age.
Examined over five ten-year timesteps, the biomass map indicated that Vermont total forested landscape stored 124.0 million Mg C in aboveground live dry biomass (AGLB); in contrast, the Dinamica models resulted in significantly higher estimates (between 129.0 million and 176.3 million Mg C depending on the year and the level of specificity of the forest classifications). Specific forest subtype classifications consistently generated lower carbon storage estimates than NLCD classifications. While independent validation against ground truth data is necessary, this indicates that more specific forest subtype classifications may be necessary for a more accurate accounting of forest carbon storage in heterogeneous landscapes.