Pattern-oriented parameterization of general models for ecological application: Towards realistic evaluations of management approaches

Abstract

General ecological models address classes of systems instead of focusing on specific systems. However, a major challenge when using general models for ecological applications is parameterization. This process involves a trade-off between analyzing the entire parameter space, which might be misleading because unrealistic parameter combinations are likely to be included, versus analyzing the model for a specific parameter set, which limits its generality. Here, we present a parameterization strategy that excludes unrealistic parameter combinations, without focusing on specific systems. This strategy adapts pattern-oriented modeling (POM) for general models. We employ a set of qualitative patterns that describe and thereby define the class of systems to be represented with the general model. Each pattern is employed to filter parameter sets that would lead to uncharacteristic model behavior. As an example, we use a general model of semi-arid rangelands that links vegetation biomass dynamics, livestock grazing, and management. The purpose of the model is to compare constant and adaptive stocking strategies. Through the pattern-oriented parameterization method, we narrow the parameter space significantly, from one billion to approximately 11,000 parameter sets. The remaining parameter sets reveal interrelationships between model parameters and processes. This increased our understanding of the model and is therefore useful for addressing applied management questions. Using the parameterized model, we found that adaptive stocking is beneficial for livestock production in all cases. Storage biomass dynamics appear to be the most important process for evaluating stocking strategies. Consequently, adaptive stocking is particularly beneficial in rangelands that are vulnerable to storage degradation by overgrazing. Our pattern-oriented parameterization provides a new way to use general models to support decision making, while avoiding the two pitfalls of employing either unrealistic parameter combinations or having an excessively narrow focus. Additionally, this approach supports systems analysis by revealing interactions and trade-offs between parameters and their corresponding processes. In summary, our approach allows the use of general models supporting a realistic evaluation of management approaches. (C) 2013 Elsevier B.V. All rights reserved.