Department of Mathematics and Computer Science, University of Osnabrück, 49074 Osnabrück, GERMANY.
Ecosystems provide important functioning and services, like biomass for bioenergy production or storage of atmospheric carbon. Two examples of such ecosystems are temperate grasslands and tropical forests. Both vegetation are rich of various species, whereby each of the respective ecosystem benefits from its species-richness concerning their functioning, i.e. productivity. In this thesis both vegetation are in the focus of the investigations. In the first chapter, a review of existing grassland and vegetation models provides an overview of important aspects, which have to be considered for modelling temperate grasslands in the context of biomass production. Based on the review, new conceptual modelling approaches for temperate grasslands are proposed. In the third chapter, derived from the suggested concept, the process-oriented and individual-based grassland model Grassmind is presented. In the fourth chapter, the model Grassmind is used in order to parameterize and simulate the annual dynamics of a typical Central European grass species. Grassmind is able to reproduce the structure and dynamics of a temperate grass species. With reference to the parameterized grass species, a simulation study using defined species groups is performed in order to investigate on the effect of the richness of species groups on aboveground productivity. We do not observe a significant positive effect of species group richness on productivity, which is explained by limitations of using the parameterized grass species as a reference. In the fifth chapter, comprehensive investigations are carried out on the example of stem size distributions in forests concerning their statistical analyses, i.e. by using maximum likelihood estimation. The effects of uncertainties, i.e. binning of measured stem sizes or random measurement errors, are examined in detail. Uncertainties bias the analyses of maximum likelihood estimations. It is shown, that the use of modified likelihood functions, which include either binning or measurement errors, reduce these biases to a large extent. For both studies, i.e. modelling of temperate grasslands and analysing stem size distributions of forests, the presented investigations are discussed and possible examinations are suggested for future research in the last chapter.