Statistical error analysis for biomass density and leaf area index estimation

Dimension analysis uses fitted allometric relationships to estimate whole-tree characteristics from dimensional measurements. Errors in dimension-analysis estimates of forest stand biomass and leaf area are often unreported and rarely analyzed. Errors may arise from measurement error, choice of model, fitting of parameters in models, and spatial variation within a sampled stand. Examination of these components separately will show how efficiency can be most effectively increased and balanced. We used dimension analysis to estimate biomass density and leaf area index for stands of trembling aspen (Populustremuloides Michx.) and black spruce (Piceamariana (Mill.) B.S.P.) in the Superior National Forest, Minnesota, United States. Estimates had average coefficients of variation ranging from 11% (aspen biomass density) to 23% (spruce leaf area index). Partitioning of these errors showed that error sources varied with species, variable, and site conditions, but most error was due either to parameter estimation or to spatial variation; model inaccuracies contributed only trivially to total error. The most cost-effective means of increasing precision of estimation would be to sacrifice more trees (even with less precise measurements of individual trees) to fit the models, or sample larger areas in stands; more sophisticated models would have little effect on error.