J. D. Sauerländer's Verlag, Bad Orb

In this study we create a new forest inventory design considering repeated measurement of sample plots from former stratified sampling inventories in an anonymised private forest enterprise. In order to meet certain gains of accuracy for distinct target variables under minimum costs we apply a double sampling scheme for stratification. Initially, the sampling error of total timber volume over bark of hardwood trees and also of spruce and fir with dbh larger than 30 cm should not exceed 5% of the mean estimation. For the mean volume estimation of the smaller trees a 10%-precision was conceded. These precision demands were to be met in each of the 6 districts of the whole enterprise. The variance estimation is obtained by means of sample plot data from the prior inventories in 1998 and 2000. Due to the coarse stratification for the prior inventories on the basis of the forest management plan, variance estimations fail for several strata and districts. Therefore, we search for an optimal allocation on the enterprise level. In order to achieve an optimal allocation for phase-II-plots in the forthcoming inventory, there is a need to give up existing plots. However, we can minimise the sampling error for the mean estimations of growing stock changes by considering volume change as additional target variable for the optimisation process. Variances of volume change are estimated from a neighbouring repeated inventory. Finally, the phase-II sample plot allocation is obtained by a compromise between the optimal allocations for five target variables, which comprise the four volume estimates for two species groups and two diameter classes and, additionally, the mean change of growing stock. It is shown that it is impossible to meet the postulated precision within the forest districts with acceptable costs. Hence, the total phase-II sample size is fixed at 2200 plots. The allocation compromise of the double sampling for stratification proves to be very beneficial compared to simple random sampling. Estimations on the basis of double sampling for volume of larger spruce and fir trees is cost-saving in the amount of 19% instead of simple random sampling. For the volume estimation of larger hardwood trees with same precision costs are reduced by 59% using the double sampling scheme instead of an unstratified simple random sampling.

Download