Application of Aerial Photographs in the Estimation of Standing Volume in the Swiss National Forest Inventory : Forest Inventories in Europe with Special Reference to Statistical Methods

1991

S. 176-191

7 Abb., 18 Lit. Ang.

Unselbständiges Werk

200086402

The study was conducted in order to test the suitability of two sampling designs, both of which employ aerial photographs as a source of information, for the estimation of standing volume in large-scale forest inventories. Two- phase sampling with regression estimators combines auxiliary variables computed from aerial photographs with the variables of interest measured in the field through regressions. In two-phase sampling for stratification, an auxiliary variable is determined from aerial photographs and used to allocate samples to strata (sub-units). The data on the auxiliary variables are projected over the separata strata. As in sampling procedures only data from a small part of the population are recorded and used as a basis for conclusions on the whole population, the estimates are necessarily inaccurate. Comparison of the standard error and the costs for a particular survey allows comparison of different sampling designs in terms of their efficiency. Since it costs far less to determine variables from aerial photographs (phase 1) than through field measurements (phase 2), two-phase surveys employ more aerial samples than terrestrial ones. In order to investigate the influence of the number of terrestrial samples on the inventory results, the results for various sample sizes were compared. For stratification, five variables were computed from aerial photographs. In this study considerably lower coefficients of determination were achieved through two-phase sampling with regression estimators than can be found in the literature on similar investigations. This is partly due to the scale of the aerial photographs employed (1:25 000) and the heterogeneity of the forest conditions in the study area. Although two-phase sampling for stratification in inventories of standing volume is less costly than two- phase sampling with regression estimators, comparison of the variances and costs shows that its efficiency is lower. Nevertheless, the computation of regression estimators requires much computer time and a carefully planned organization of the data for the interpretation of inventory results, especially when more than one variable of interest is to be determined. Forest inventories are not conducted solely to determine standing volume but are almost exclusively designed as multi-purpose surveys. As well as standing volume, timber quality, data on the silvicultural and ecological condition of the stands, and the protective functions of the forest are of importance. Rational relationships to the parameters which can be measured on aerial photographs cannot be established for all variables, so that regressions cannot be computed in every case. This is one of the main reasons that two-phase sampling for stratification, despite its lower statistical effectiveness in estimating standing volume, presents an interesting alternative to two-phase ......