Inhaltszusammenfassung

Ecological patterns vary in space and time. Therefore, when using dynamic models in ecology, the spatial aspect should not be neglected prematurely since it could possibly change the model outcomes to a considerable extent. In view of this problem, we describe here a method how to construct a non-spatial version from a spatially explicit simulation model. The principle idea is to suppress the spatial correlations of cells in a grid in time by continuously re-assigning a random neighbourhood f...Ecological patterns vary in space and time. Therefore, when using dynamic models in ecology, the spatial aspect should not be neglected prematurely since it could possibly change the model outcomes to a considerable extent. In view of this problem, we describe here a method how to construct a non-spatial version from a spatially explicit simulation model. The principle idea is to suppress the spatial correlations of cells in a grid in time by continuously re-assigning a random neighbourhood for each cell on the grid. Since this procedure actually eliminates the spatial dimensions, it allows to quantify the unadulterated impact of spatial processes on the model results. To illustrate an important application of this approach in the context of forest management we use a grid-based model that simulates succession of Norway spruce (Picea abies (L.) Karst.) at mountainous sites after blowdown events. The output of this model is compared with the results of the deduced non-spatial version of this model regarding the predicted amount of re-growing trees. The non-spatial version dramatically overestimates the number of spruce trees on different microsites. Thus, the uncritical use of the non-spatial model might give reason to wrong management decisions that are based on too optimistic predictions. In practice, this may lead to dangerous situations, especially in mountain forests serving as protection against avalanches and landslides. This example demonstrates the successful applicability of our approach. Our method can be interpreted as a contribution to an extended sensitivity analysis: it analyses the sensitivity of the results due to structural changes of the model. This sensitivity allows one to estimate the redundancy or the necessity of spatially explicit processes in a model with regard to the parsimony principle of modelling. Since our approach is not dependent on special features of the simulation model used here, it is assumed to be applicable for other spatial models, too, and can thus be considered of general interest for a diligent model analysis.» weiterlesen» einklappen

Autoren

Klassifikation

DDC Sachgruppe:
Biowissenschaften, Biologie