J. D. Sauerländer's Verlag, Bad Orb

We evaluated the single tree-based growth simulator SILVA 2.2 with approximately 19,700 trees from 49 long-term experimental growth and yield plots in south-western Germany, consisting of European beech (Fagus sylvatica [L.]) and Norway spruce (Picea abies [L.] Karst.) as mainly pure stands. We provided single-tree information including tree positions, dimensions and removal dates as input values for SILVA. Main evaluation variables were mean and dominant diameter (d, d100), mean height and standing volume. As evaluation criteria we computed the absolute and relative bias, the absolute and relative decennial bias, relative precision and accuracy. Based on relative bias analysis, SILVA underestimated diameter development by about 5% in relative terms per decade and 10% in total, since the observed period varied between 12 and 35 years. d100 was slightly more underestimated than d. Further statistical and sensitivity analyses showed that heavier thinnings had the tendency to cause stronger underestimation. After adjusting potential height growth to the regional conditions by Chapman-Richards functions, SILVA proved to be a useful tool for general growth predictions in south-western Germany. However, before SILVA might be used as a tool for long-term virtual silvicultural treatment optimization in this region, growth allocation to individual trees needs adaptation. These results underline that methods for fine-tuning to regional and recent yield levels are needed when a model is used outside the range of its calibration data. Currently, such methods are being included in the growth model SILVA.

Download