The evaluation uses the longest available time series for beech and oak defoliation in Germany. The data from Hesse, starting from 1984, show a typical pattern: for the first 12 years, a continuous increase in defoliation was observed ranging from an average value of 14% in 1984 to a peak value of 30%. This was followed by a subsequent decrease in the loss of foliage accompanied by a high variability, until the last monitoring in 2003, where an average value of 25% defoliation was observed. For both tree species, the years of trend reversal were identical. The same pattern was observed in the German federal states: Rhineland-Palatinate, North Rhine-Westphalia, and Bavaria. The year of trend reversal was identical in Hesse and Rhineland-Palatinate. In North Rhine-Westphalia, it occurred 1 year earlier and in Bavaria 3 years earlier. Whereas defoliation trends were clearly demonstrated, tree mortality did not appear on a large scale. The sample trees were grouped into four discrete clusters according to their annual defoliation values from 1984 to 2003. In 1996, the clusters represent 15, 25, 35, and 50% defoliation values. Regarding beech in Hesse, there was no overlap in the defoliation curves observed among the different clusters. These four clusters having different degrees of defoliation over the whole time span of 20 years were used for a further detailed statistical analysis. For discrete variables like crown spacing and—in the case of beech—fruit bearing, mosaic plots were applied in order to visualize relations of low dimensional contingency tables, with defoliation trends being used as the response variable. The data show for beech a very clear relation between defoliation and age, relative crown spacing, stand composition, and fruit bearing. Regarding oak, besides age and relative crown spacing, the years with significant appearance of biotic stress factors—leaf eating insects—show a clear relation to trends of defoliation. The statistical model used in this study—logistic regression—allows applying a multinomial response variable and a number of continuous or categorical explanatory variables. With this approach, an iterative optimized selection of effect variables was used to test the relevance of different variables on the defoliation pattern of the same four clusters mentioned above. For this, the variables were grouped in an iterative process with five steps, starting with a few basic variables of tree and site information, and ending with a total of more than 20 variables in the fifth step. The process selects first the variables which are of significance on the defoliation, and calculates the possible errors in the grouping of the different trees to the four clusters. In this analysis of beech, the basic tree and stand variables: age, relative crown spacing, stand composition and fruit bearing proved to be the most relevant group of parameters, with the other variables explaining the variation of defoliation only to a minor extent. More complex model levels do not change any basic selected variables; however, Cation Exchange Capacity (CEC), C/N-ratio, Al- and Ca-proportion of CEC are additionally selected and give a hint of the relevance of soil conditions. Regarding beech, the errors of the statistical model are lower compared to oak.