Succession of semi-natural grasslands : spatially-explicit, mechanistic simulation considering various forms of land use

Degradation of natural and semi-natural landscapes has become a matter of global concern including for habitats in Western and Central Europe. For maintenance and restoration of grasslands belonging to the most species-rich biotopes in Central Europe, elaborate manage-ment is crucial. The Eifel National Park contains vast areas of semi-natural grasslands on the plateau Dreiborner Hochfläche in the area of the former military training site Vogelsang. To support decision-making for the management of this area, we built the GraS-Model (Grassland Succession Model), which mechanistically simulates the dynamics of grassland vegetation depending on the form of land use. When dealing with a complex landscape, entities acting at different scales have to be considered. While trees can be distinguished as single individuals, grasses and herbs are rather perceived as the sum of plants building up a certain cover. To cope with these unequal plant types, the GraS-Model is set up using a multimodeling approach. Representative species and plant functional groups of the herbaceous layer are modeled as compartments; their abundance is expressed as cover and only vegetative spread is considered. Difference equations are used to simulate their growth, which is adjusted using Briemle’s utilization numbers for cutting, grazing and trampling. Competitive power arises from the growth rate of each species in relation to that of the others. Herbaceous species growth was calibrated based on an extensive community data set of the study site. Trees, in contrast, are modeled using an individual-based approach. Each individual tree germinates, grows, disperses seeds, and ages in a spatially-explicit environment. As factors determining wood encroachment, inhibition by the grasslayer, interference of wild boar (Sus scrofa) and of ungulate browsers is considered. The plants are embedded in a landscape that is simulated as a detailed grid (100 m² cells), which allows inserting spatially-explicit input data from a GIS (geographic information system) so that the model can be easily connected to data of an existing landscape. Due to this raster-based approach, neighborhood relationships can be accounted for and different management regimes can be applied to distinct areas of the landscape. Pattern-oriented model evaluation revealed that the model produces results that are in line with general literature and experiences of local experts, and emulates observed successional pathways on the Dreiborner Hochfläche on the spatial as well as the temporal scale. After successful evaluation, the GraS-Model was used to predict the landscape developments of the Dreiborner Hochfläche under different management regimes for up to 100 years. It provides a highly detailed, spatially explicit prognosis integrating the initial vegetation composition and resulting neighborhood interactions which would not have been possible without the model. Forest encroachment on the Dreiborner Hochfläche was most strongly delayed by non-interference with the given high abundance of red deer (apart from mowing), whereas grazing by bison promoted a diverse landscape mosaic. Due to its high complexity, the GraS-Model initially exceeded the available computer main memory of current workstations when applied to the modeled Dreiborner Hochfläche (ca. 1500 ha). The landscape had to be fragmented and modeled separately. Therefore, the model was parallelized so that it can be run on a cluster, improving practicability and enhancing the ecological accuracy of the simulation results. The benefits of using the GraS-Model as decision support system are multifaceted: It gives highly detailed spatially explicit prognoses providing a strong basis for decision support and facilitating effectiveness and efficiency control; also it combines knowledge of different disciplines and enhances communication between scientists and stakeholders. Due to its modular design, the model can be continuously updated by integrating latest scientific knowledge, which can then be easily communicated. In the future, light, nutrients or moisture as environmental factors could be added in. Furthermore, other raster-based dynamic models (e.g. models of forest succession, nitrogen cycling, climate change, or animal mmovement) could be coupled.