Toward error analysis of large-scale forest carbon budgets

Quantification of forest carbon sources and sinks is an important part of national inventories of net greenhouse gas emissions. Several such forest carbon budgets have been constructed, but little effort has been made to analyse the sources of error and how these errors propagate to determine the overall uncertainty of projected carbon fluxes. We performed an error analysis for estimates of tree volume and volume change determined by repeated measurements of permanent sample plots for the South‐eastern United States as a step toward assessing errors in the carbon budget constructed by the USDA Forest Service. Error components recognized were: sampling error for sample plot selection; measurement error for tree height and diameter; and regression error for tree volume. Most of the propagated error in volume and volume change estimation was due to sampling error. Error estimates depended on the size of the area examined (single state to region) and the degree to which tree growth and recruit‐ment balanced mortality and harvesting. Approximate regional 95% confidence intervals were 3 455 073 000 ± 39 606 000 (1.1%) m³ for current growing‐stock volume, and 10 616 000 ± 4210 000 (39.7%) m³ years^?1 for growing‐stock volume change. These methods should be useful in further analysis of the sources of error and overall uncertainty in national efforts to quantify carbon fluxes associated with forests and land cover dynamics.