4. Comparing sample designs
An effective way to compare sample designs is via simulation if a forest area model is available. The model may be obtained from a previous inventory or from satellite image-based estimation of variables of interest. An example of the standard errors obtained from sample designs for estimating mean growing stock volume is shown in Figure 3. The test site is in North Finland and has land area of 6.47 million hectares, forest land area of 4.19 million hectares, and mean volume on forest land of 52.7 m3/ha.
A pixel level, border-to-border forest map has been produced using field data from the preceding inventory, satellite images, and digital map data (Tomppo & Halme 2004, Tomppo 2004). Satellite images of different resolution provide one information source, in addition to existing maps. In addition, a pilot inventory may be used to collect information for planning the final sample design. Representative sub-areas can be selected from the population where pilot inventories may be conducted. However, these pilot inventories must be acknowledged and accepted as less than optimal. In addition, new sample designs can be created using information from previous inventories as has been the case in countries where forest inventories have been conducted since the 1920s and 1930s (e.g. Ilvessalo 1927).
Figure 3. Standard errors based on sampling simulations with different distances between field plots and with numbers of plots per cluster ranging from 9 to 17. The distance between clusters is 10 km.