Accurate uncertainty assessments of plot-level live tree biomass stocks are an important precursor to estimating uncertainty in annual national greenhouse gas inventories (NGHGIs) developed from forest inventory data. However, current approaches employed within the United States' NGHGI do not specifically incorporate methods to address error in tree-scale biomass models and as a result may misestimate overall uncertainty surrounding plot-scale assessments. We present a data-driven, hierarchical modeling approach to predict both total aboveground and foliage biomass for inventory plots within the US Forest Service Forest Inventory and Analysis (FIA) program, informed by a large multispecies felled-tree dataset. Our results reveal substantial plot-scale relative uncertainties for total aboveground biomass (11-155% of predicted means) with even larger uncertainties for foliage biomass (27-472%). In addition, we found different distributions of total aboveground and foliage biomass when compared with other generalized biomass models for North America. These results suggest a greater contribution of allometric models to the overall uncertainty of biomass stock estimates than what has been previously reported by the literature. While the relative performance of the hierarchical model is influenced by biases within the fitting data, particularly for woodland and conifer species, our results suggest that poor representation of individual tree model error may lead to unrealistically high confidence in plot-scale estimates of biomass stocks derived from forest inventory data. However, improvements to model design and the quality of felled-tree data for fitting and validation may offer substantial improvements in the accuracy and precision of NGHGIs.
- Bayesian hierarchical models
- Data assimilation
- Forest biomass
- National greenhouse gas inventory