Analyse von Schälschadensinventuren in Harz und Solling

Abstract

The present work represents the continuation of a project that examined the causal relationships between environmental factors and the occurrence of bark stripping on spruce by red deer in the Harz and Solling mountains in Germany. In addition to a detailed representation and analysis of the data set of the two study sites, this study employs various statistical tests to estimate the bark stripping damage identified during three years by means of cluster sampling inventory. For this purpose, values of bark stripping are associated with the red deer hunting bag as well as with forest inventory data and terrain parameters in univariate models to calculate the odds ratios of the respective parameter classes. Subsequently, the potential influence factors are analyzed using binary logistic regression and generalized linear models in order to determine their effect as an explanatory variable in multivariate models. For the implementation of this study, appropriate working hypotheses are established at the beginning. The calculation of odds ratios yields a strong correlation between previous bark stripping and the emergence of new stripping at the respective sample point. Similarly, a high hunting bag is associated with increased bark stripping. While young forest stands seem to be more susceptible to bark stripping damage than older stands, the analysis of the influence of terrain characteristics does not give a clear picture. The results of the regression models vary in their level of model fit and their explanatory potential. In general, results for the Solling site are less predictive than models for the Harz site. Depending on the inventory year and area, variables reported as effect parameters change from one model to another. In addition, they partly contradict each other in their effect direction. Also, the presence of previous bark stripping damage is not included as an explanatory variable in all models. The generalized linear models for the Harz site show recurring effects for topography based parameters. Despite the varying effect size their characteristics over the entire series of observations is the same. Increasing slope inclination and slope aspect are associated with differences in bark stripping risk. While the risk of bark stripping decreases with increasing exposure of the sample point, locations with a higher potential radiation in January show higher stripping rates. The results of this study only partially correspond with the results of other authors. The contradictory effect values of this analysis raise fundamental doubts about the reliability of causal relationships generated by multivariate models. In addition to an appropriate sampling design, especially targeted parameter selection and collection are highlighted as crucial factors for subsequent statistical analysis. Unambiguous results requires data collection customized specifically to the scientific problem rather than the use of data collected for other purposes.