This thesis investigates geostatistical methods to support forest planning. Measurements of the permanent inventory sampling in 2002 and 2016 from the forest of the Ludwig-Maximilians-Universität Munich served as input data. In early 2018 the permanent inventory points in a grid of 100 x 100 m were augmented by 15 measurement points in the intermediate fields. From this information values for the forest parameters dg (diameter of the mean basal area stem), stock in m3 of timber harvested and the percentage of Norway spruce (Picea abies [L.]), as well as of broadleaf trees per hectare were calculated for each inventory point. Combined with their GPS coordinates it was possible to identify different semivariogram models describing the spatial correlation within each variable. By using the statistical software R for block kriging, estimates for each 25 x 25 m block within the wood ground area of about 430 ha were interpolated. Results were imported into a GIS and displayed as thematic maps.
The geostatistical models determined spatial correlations for all considered variables, but the data basis for short distances (lags) of around 80 m and less was insufficient. Even the augmented measurement points in the intermediate fields did not produce a significant improvement. Integrating the silvicultural type, which was documented in 2002, as information about spatial pattern improved the model fit, which became visible in the derived maps. Together with the local forest ranger the map display was optimized for practical use. Nevertheless, the fairly wide confidence intervals of 95 % for the true values, relating to the estimates, must be considered when using the planning aids. Furthermore, it must be emphasized that even geostatistical models with poor validity allow the creation of seemingly plausible maps. Therefore, a verification of assumed spatial relations is strongly recommended.
The data basis could be improved by including additional inventory points in shorter lags, and by carrying out pre-stratification using aerial photographs or other remote sensing techniques. Combining forest parameters with soil or stand based information would also be possible. As an innovative concept, the forest planning aids within a GIS could be updated continuously by integrating data from the ongoing forest management processes.
«
This thesis investigates geostatistical methods to support forest planning. Measurements of the permanent inventory sampling in 2002 and 2016 from the forest of the Ludwig-Maximilians-Universität Munich served as input data. In early 2018 the permanent inventory points in a grid of 100 x 100 m were augmented by 15 measurement points in the intermediate fields. From this information values for the forest parameters dg (diameter of the mean basal area stem), stock in m3 of timber harvested and the...
»