Environmental variability and system heterogeneity in terrestrial biogeochemical models

Ecosystems are highly heterogeneous systems subjected to important levels of environmental variability; however, it is common in terrestrial biogeochemical models to assume homogeneous properties of the elements of the system or constant environmental conditions. For some processes, heterogeneity in these models is treated very simplistically, but there is not much information about the advantages of including more complex representations in these models. By environmental variability I refer to the continuous changes of abiotic drivers in ecosystems, mainly climatic conditions. System heterogeneity is treated here as the diversity of elements that compose an ecosystem and respond differently to biotic and abiotic drivers. In this dissertation I performed a theoretical analysis to evaluate the consequences of ignoring heterogeneity and variability on the representation of carbon and nitrogen cycling in terrestrial biogeochemical models. For this purpose I used tools from probability theory and simulation models to test the hypothesis that ignoring heterogeneity and variability excludes a variety of system properties and behaviors that cannot be obtained with simpler models. Explicit treatments of climatic variability showed that changes in temperature variance alone can modify the amounts of respiration and carbon storage in ecosystems. Additionally, changes in temperature variance can modify predictions solely based on changes in temperature averages. This behavior is strongly dependent on the degree and nature of nonlinearity in ecosystems. Effects of system heterogeneity on carbon and nitrogen cycling are also strongly influenced by nonlinearities. Extrapolations of average system behavior are only valid when the system is linear and the elements of the system are distributed homogeneously or symmetrically around an average value. In all other cases, the nonlinearity of the system and the distribution of its elements produce complex behaviors that are impossible to predict with simple models.