National remote sensing-derived aboveground biomass yield curves for Canada

Abstract Accurate and current estimates of aboveground biomass (AGB) are essential outputs of forest inventories and are critical for carbon accounting. To obtain current estimates and provide projections into the future, growth simulators or yield curves are applied. Remotely sensed time series of AGB estimates across large areas provide a novel opportunity to generate representative AGB yield curves for a broader range of tree species, age classes, and regions, including noncommercial and unmanaged forests using a single, consistent data source and approach. In this study, remote sensing yield curves (RSYC) are demonstrated as a means to extend AGB modeling across Canada’s forested ecosystems. To ensure national coverage while maintaining regional representativeness, yield curves were developed at a 150 × 150 km tile extent for 27 individual species, along with three additional multi-species models (generic, coniferous, and broadleaf). Evaluation against an independent set of plot data showed that among the multispecies models, the coniferous model achieved the lowest relative bias (2.06%) and root mean square error (RMSE) (41.03 t/ha), while species-specific models exhibited variable performance. Comparison to existing national species-specific yield curves indicated that the RSYC models generally achieved lower bias and RMSE than existing national models, with particularly notable improvements for key coniferous species such as black spruce and jack pine (e.g. RMSE% reductions from 55.89% to 34.69% for black spruce, and from 112.79% to 35.41% for jack pine), demonstrating enhanced accuracy and consistency for national-scale applications. Remotely sensed time series data enabled the development of a pool of species-specific and generic yield curves for Canada that are generated from a nationally consistent (and freely available) data source and approach—regardless of the management, ownership, or protection status of the forest.