Statistical analysis of timber stock evaluation using the data of state forest inventory and forest planning

The State Forest Inventory (SFI) has been performed in Russia since 2007. The methodology of SFI includes stratification of the stands and limitation of the area of sample plots (SP). The projecting of the SFI is based mainly on the data of the forest planning (FP), which was applied earlier. The differences in the evaluation of timber stock based on the results of SFI and FP and the SFI methodology are a subject of discussion. The background statistics of SFI and FP data of Uvatsky forestry district in Tyumen oblast are considered. A feature characteristic of evaluating timber stocks by SFI and FP is a marked asymmetry of their distribution, which makes parametric estimation of the stock accuracy ineffective in both cases. It is shown that the limitation of the area of SFI SP did not influence the final stock evaluation. An attempt was made to estimate the error of timber stock evaluation using non-parametric statistics of resampling, generating a set of pseudo-samples. The use of resampling of SFI SP and FP forest inventory surveys at SFI points made it possible to determine the dependence of the accuracy on the number of SP and the value of the intra- forest compartment variation of timber stock, which did not depend on the number of SP and turned out to be much higher than that provided in the SFI projecting methodology. Generation of pseudo-samples of SP on the forest map considering the obtained values of intra- compartment variation allowed to determine the expected dynamics of the accuracy of evaluation of timber stock with increasing SP. Underestimation of the intra- compartment variation of timber stocks partly explains the large difference in the evaluation of timber stocks using SFI and FP data. The low influence of the applied stratification of the stands on the accuracy of SFI results is shown.