Effects of data uncertainty on dynamic treatment units

The purpose of this thesis was to study the effects of uncertainty in the pixel level estimates concerning the potential benefits of dynamic, pixel-based treatment units compared to traditional stand-based treatment allocation. Because of automatic stand delineation having benefits resulting in the possibility to be more commonly used in forest industry. In thesis assumption is made that the original ALS-dataset exactly represents the “Ground truth”. The study area is a part of the Östad foundation (Östad “stiftelse”) located in Southern Sweden. The property is owned by Östad foundation and is mainly used for industrial and education purpose. The analysis area is comprised of 1 848 ha. The forest is managed in the traditional Scandinavian clearcutting regimes. The results shows that uncertainty effects potential incomes in all cases, especially on the stands where traditional stand delineation was used. Target harvest volume (share of harvested pixels) also has an impact on the potential losses. When the target volume goes from 60 000 m3 to 80 000 m3 the financial trend is changing and the potential losses per m3 is decreasing. Due to this phenomenon the biggest potential loss in SEK per m3 is reached at the target volume of 60 000 m3. In terms of dynamic treatment units (DTU) the difference per m3 between the “Ground truth” and created rasters are relatively low, in some cases less than -1 SEK/m3. The results concerning the average difference between DTU and “Original borders” variants with simulated dNPV in all cases favor the DTU’s. In all tested cases DTU had an economical advantage over the planning within existing boundaries. Even with target volume of 20 000 m3 results improve by -3.2 SEK/m3 in average between DTU and “Original borders”. The largest difference between simulated dNPV data of DTU and “Original borders” is - 5.9 SEK/m3, this result is achieved by 60 000 m3 as the target harvest volume. The conclusion is drawn that DTU planning is more efficient than forest management planning based on “Original stand” borders even when the effect of volume data errors is considered. The data errors have larger effect on the planning results within the framework of original borders that within the DTU framework. Highest difference was achieved in case of minimum segment size 0.5 ha and target harvesting volume 60 000 m3 where, the difference was -6.9 SEK/m3