Browsing by Subject "jäännösvaihtelu"

Uneven aged forest management was allowed in the Finnish forestry law in 2014 and after that continuous cover management was added to the Finnish forest management recommendations. After the amendment, the popularity of uneven aged forest management has increased and changed the development of forest structures and silviculture. The changing climate has also effect on forest growth reactions. Because of this, previously created forest development models are not as useful as before in itself. Statistical models describing development of a stand give unbiased estimates if tree growth reactions differ from growth reactions of model fitting data. In general, it is also important to examine the goodness and functionality of the model outside of the model fitting data, to gain additional information about the reliability and suitability of the model. At the same time, reliability of decisions based on model estimates also improves. Aim of the study is to examine differences in the five-year growth estimates of the tree diameter model between two National Forest Inventory periods (NFI10-11 and NFI11-12). The aim is also to examine whether topography variables would explain residual variation of the model. In addition, the model has been tried to calibrate in a plot level with topography variables that have effects on the residual variation. The study dataset consisted of 299 permanent plots measured in the 10th, 11th, and 12th National Forest Inventories in Central Finland. Digital elevation model was acquired from open sources from National Land Survey of Finland. Statistical measures and graphical review were used in the model goodness of fit analysis. Calibration was based on linear variance component analysis where the logarithmic prediction error of the diameter growth model was estimated by topography variables. No significant differences were noticed in the model estimates between different time periods, i.e., the model worked well outside of the model fitting data in the NFI11-12 period. The model diameter growth estimates were higher than the measured diameter growths in all situations and in both periods. Bias was 13,5 % lower in the NFI11-12 period than in the NFI10-11 period, when the entire dataset was examined. No significant differences between periods were observed in the accuracy of the model estimates in different classes (species, habitat type, tree class etc.) The model estimates were the most accurate in pine dominated xeric heath forests. The bias increased when the model was used in more spruce dominated and nutritious stand. There was no unambiguous connection between topography variables and unexplained residual variation of the model. Model calibration based on topography variables reduced plot-specific bias when calibration dataset included at least one measured diameter of a tree from 20 plots and only trees belonging to the dataset were calibrated. Success of the calibration was due to the addition of terrain truth to the analysis. According to the results, the diameter growth model can be applied to all structured stands outside of the model fitting data. However, temporal, and geographical limitations of the dataset of the thesis must be considered when the model is used outside of the model fitting data. The differences between periods in the accuracies of the model estimates weren’t significant in terms of practical application of the model. No new information was obtained on the factors affecting the residual variation of the model. Some topography variables other than those used in the thesis may have an effect on the residual variation of the model. The topography variables worked well in the prediction error-based calibration of the model. Calibration can be used to get more accurate diameter growth estimates, which lead to a more reliable simulation scenario of the future development of the forest.