Neue Hilfsmittel zur Anwendung zweiphasiger Stichprobenverfahren in der Waldinventurpraxis

New tools for application of two-phase sampling in forest inventory practice Sample-based forest inventories provide reliable information on the status and development of the forest. The sample sizes are thereby chosen in order to meet high estimation precisions for forest attributes on the national, cantonal or forest district level. In order to provide sufficient precision for regions where the sample size is limited, the combination of the terrestrial information with auxiliary data such as derived from remote sensing sources can constitute a cost-saving alternative compared to increasing the terrestrial sample size. Design-based estimation methods developed at ETH Zurich which implement such two-phase inventory concepts have lately been provided by the open-source software package “forestinventory” in the statistical software R. This article explains the principle of a two-phase inventory of small-area estimation and illustrates the workflow by the example of a case study in Embrach (Canton of Zurich). Target variable to be estimated for 22 subregions (small areas) located in the study area (503 ha) was the average timber volume density. The main work steps were (A) the enlargement of the one-phase inventory to a two-phase inventory, (B) the derivation of explanatory variables from remote sensing data (mean vegetation height, standard deviation of the vegetation heights and broadleaf tree proportion), (C) the formulation of a regression model to link these explanatory variables to the terrestrial information, and (D) the calculation of small area estimates (average timber volume density) using forestinventory. Despite average model accuracy (R̂2 = 0.533), it was possible to reduce the variance (as a measure of the estimation accuracy) in the 22 small areas by 40% on average.