Björn Reineking

Statistical inference for stochastic simulation models – theory and application

Statistical models are the traditional choice to test scientific theories when observations, processes or boundary conditions are subject to stochasticity. Many important systems in ecology and biology, however, are difficult to capture with statistical models. Stochastic simulation models offer an alternative, but they were hitherto associated with a major disadvantage: their likelihood functions can usually not be calculated explicitly, and thus it is difficult to couple them to well-established statistical theory such as maximum likelihood and Bayesian statistics. A number of new methods, among them Approximate Bayesian Computing and Pattern-Oriented Modelling, bypass this limitation. These methods share three main principles: aggregation of simulated and observed data via summary statistics, likelihood approximation based on the summary statistics, and efficient sampling. We discuss principles as well as advantages and caveats of these methods, and demonstrate their potential for integrating stochastic simulation models into a unified framework for statistical modelling.

Statistical inference for stochastic simulations models : theory and application

Statistical models are the traditional choice to test scientific theories when observations, processes or boundary conditions are subject to stochasticity. Many important systems in ecology and biology, however, are difficult to capture with statistical models. Stochastic simulation models offer an alternative, but they were hitherto associated with a major disadvantage: their likelihood functions can usually not be calculated explicitly, and thus it is difficult to couple them to well-established statistical theory such as maximum likelihood and Bayesian statistics. A number of new methods, among them Approximate Bayesian Computing and Pattern-Oriented Modelling, bypass this limitation. These methods share three main principles: aggregation of simulated and observed data via summary statistics, likelihood approximation based on the summary statistics, and efficient sampling. We discuss principles as well as advantages and caveats of these methods, and demonstrate their potential for integrating stochastic simulation models into a unified framework for statistical modelling.

Natural enemy interactions constrain pest control in complex agricultural landscapes

Biological control of pests by natural enemies is a major ecosystem service delivered to agriculture worldwide. Quantifying and predicting its effectiveness at large spatial scales is critical for increased sustainability of agricultural production. Landscape complexity is known to benefit natural enemies, but its effects on interactions between natural enemies and the consequences for crop damage and yield are unclear. Here, we show that pest control at the landscape scale is driven by differences in natural enemy interactions across landscapes, rather than by the effectiveness of individual natural enemy guilds. In a field exclusion experiment, pest control by flying insect enemies increased with landscape complexity. However, so did antagonistic interactions between flying insects and birds, which were neutral in simple landscapes and increasingly negative in complex landscapes. Negative natural enemy interactions thus constrained pest control in complex landscapes. These results show that, by altering natural enemy interactions, landscape complexity can provide ecosystem services as well as disservices. Careful handling of the tradeoffs among multiple ecosystem services, biodiversity, and societal concerns is thus crucial and depends on our ability to predict the functional consequences of landscape-scale changes in trophic interactions.

Integrating movement ecology with biodiversity research - exploring new avenues to address spatiotemporal biodiversity dynamics

Movement of organisms is one of the key mechanisms shaping biodiversity, e.g. the distribution of genes, individuals and species in space and time. Recent technological and conceptual advances have improved our ability to assess the causes and consequences of individual movement, and led to the emergence of the new field of ‘movement ecology’. Here, we outline how movement ecology can contribute to the broad field of biodiversity research, i.e. the study of processes and patterns of life among and across different scales, from genes to ecosystems, and we propose a conceptual framework linking these hitherto largely separated fields of research. Our framework builds on the concept of movement ecology for individuals, and demonstrates its importance for linking individual organismal movement with biodiversity. First, organismal movements can provide ‘mobile links’ between habitats or ecosystems, thereby connecting resources, genes, and processes among otherwise separate locations. Understanding these mobile links and their impact on biodiversity will be facilitated by movement ecology, because mobile links can be created by different modes of movement (i.e., foraging, dispersal, migration) that relate to different spatiotemporal scales and have differential effects on biodiversity. Second, organismal movements can also mediate coexistence in communities, through ‘equalizing’ and ‘stabilizing’ mechanisms. This novel integrated framework provides a conceptual starting point for a better understanding of biodiversity dynamics in light of individual movement and space-use behavior across spatiotemporal scales. By illustrating this framework with examples, we argue that the integration of movement ecology and biodiversity research will also enhance our ability to conserve diversity at the genetic, species, and ecosystem levels.

Modeling the impact of climate and vegetation on fire regimes in mountain landscapes

Assessing the long-term dynamics of mountain landscapes that are influenced by large-scale natural and anthropogenic disturbances and a changing climate is a complex subject. In this study, a landscape-level ecological model was modified to this end. We describe the structure and evaluation of the fire sub-model of the new landscape model LandClim, which was designed to simulate climate–fire–vegetation dynamics. We applied the model to an extended elevational gradient in the Colorado Front Range to test its ability to simulate vegetation composition and the strongly varying fire regime along the gradient. The simulated sequence of forest types along the gradient corresponded to the one observed, and the location of ecotones lay within the range of observed values. The model captured the range of observed fire rotations and reproduced realistic fire size distributions. Although the results are subject to considerable uncertainty, we conclude that LandClim can be used to explore the relative differences of fire regimes between strongly different climatic conditions.

Environmental determinants of lightning- v. human-induced forest fire ignitions differ in a temperate mountain region of Switzerland

Understanding the environmental and human determinants of forest fire ignitions is crucial for landscape management. In this study, we consider lightning- and human-induced fires separately and evaluate the relative importance of weather, forest composition and human activities on the occurrence of forest fire ignitions in the most fire-prone region of Switzerland, the Canton Ticino. Independent variables included 14 drought and fire weather indices, forest composition and human influences. Logistic regression models were used to relate these independent variables to records of forest fires over a 37-year period (1969-2005). We found large differences in the importance of environmental and human controls on forest fire ignitions between lightning- and human-induced events: lightning-induced fires occurred in a small range of weather conditions well captured by the Duff Moisture Code from the Canadian Forest Fire Weather Index System and the LandClim Drought Index, and with negligible influence of distance to human infrastructure, whereas human-induced fires occurred in a wider range of weather conditions well captured by the Angstroem and the Fosberg Fire Weather Index, mainly in deciduous forests, and strongly depending on proximity to human infrastructure. We conclude that the suitability of fire indices can vary dramatically between ignition sources, suggesting that some of these indices are useful within certain regions and fire types only. The ignition source is an important factor that needs to be taken into account by fire managers and when developing models of forest fire occurrence.

Modelling Forest α-Diversity and Floristic Composition — On the Added Value of LiDAR plus Hyperspectral Remote Sensing

The decline of biodiversity is one of the major current global issues. Still, there is a widespread lack of information about the spatial distribution of individual species and biodiversity as a whole. Remote sensing techniques are increasingly used for biodiversity monitoring and especially the combination of LiDAR and hyperspectral data is expected to deliver valuable information. In this study spatial patterns of vascular plant community composition and α-diversity of a temperate montane forest in Germany were analysed for different forest strata. The predictive power of LiDAR (LiD) and hyperspectral (MNF) datasets alone and combined (MNF+LiD) was compared using random forest regression in a ten-fold cross-validation scheme that included feature selection and model tuning. The final models were used for spatial predictions. Species richness could be predicted with varying accuracy (R2 = 0.26 to 0.55) depending on the forest layer. In contrast, community composition of the different layers, obtained by multivariate ordination, could in part be modelled with high accuracies for the first ordination axis (R2 = 0.39 to 0.78), but poor accuracies for the second axis (R2 ≤ 0.3). LiDAR variables were the best predictors for total species richness across all forest layers (R2 LiD = 0.3, R2 MNF = 0.08, R2 MNF+LiD = 0.2), while for community composition across all forest layers both hyperspectral and LiDAR predictors achieved similar performances (R2 LiD = 0.75, R2 MNF = 0.76, R2 MNF+LiD = 0.78). The improvement in R2 was small (≤0.07)—if any—when using both LiDAR and hyperspectral data as compared to using only the best single predictor set. This study shows the high potential of LiDAR and hyperspectral data for plant biodiversity modelling, but also calls for a critical evaluation of the added value of combining both with respect to acquisition costs.

Predicting tree death for "Fagus sylvatica" and "Abies alba" using permanent plot data

Question: How well can mortality probabilities of deciduous trees (Fagus sylvatica) and conifers (Abies alba) be predicted using permanent plot data that describe growth patterns, tree species, tree size and site conditions? Location: Fagus forests in the montane belt of the Jura folds (Switzerland). Method: Permanent plot data were used to develop and validate logistic regression models predicting survival probabilities of individual trees. Backward model selection led to a reduced model containing the growth-related variable ʻrelative basal area increment ̓ (growth-dependent mortality) and variables not directly reflecting growth such as species, size and site (growth-independent mortality). Results: The growth-mortality relationship was the same for both species (growth-dependent mortality). However, species, site and tree size also influenced mortality probabilities (growthindependent mortality). The predicted survival probabilities of the final model were well calibrated, and the model showed an excellent discriminatory power (area under the receiver operating characteristic curve = 0.896). Conclusion: Mortality probabilities of Fagus sylvatica and Abies alba can be predicted with high discriminatory power using a well calibrated logistic regression model. Extending this case study to a larger number of tree species and sites could provide speciesand site-specific tree mortality models that allow for more realistic projections of forest succession.

Can they keep up with climate change? – Integrating specific dispersal abilities of protected Odonata in species distribution modelling

Abstract.  1. The effects of climate change on the distribution of species are typically inferred using bioclimatic envelope models, assuming either no or unrestricted dispersal abilities. Information on species‐specific dispersal abilities, especially of animals, is rarely incorporated.

Do small-grain processes matter for landscape scale questions? Sensitivity of a forest landscape model to the formulation of tree growth rate

Process-based forest landscape models are valuable tools for testing basic ecological theory and for projecting how forest landscapes may respond to climate change and other environmental shifts. However, the ability of these models to accurately predict environmentally-induced shifts in species distributions as well as changes in forest composition and structure is often contingent on the phenomenological representation of individual-level processes accurately scaling-up to landscape-level community dynamics. We use a spatially explicit landscape forest model (LandClim) to examine how three alternative formulations of individual tree growth (logistic, Gompertz, and von Bertalanffy) influence model results. Interactions between growth models and landscape characteristics (landscape heterogeneity and disturbance intensity) were tested to determine in what type of landscape simulation results were most sensitive to growth model structure. We found that simulation results were robust to growth function formulation when the results were assessed at a large spatial extent (landscape) and when coarse response variables, such as total forest biomass, were examined. However, results diverged when more detailed response variables, such as species composition within elevation bands, were considered. These differences were particularly prevalent in regions that included environmental transition zones where forest composition is strongly driven by growth-dependent competition. We found that neither landscape heterogeneity nor the intensity of landscape disturbances accentuated simulation sensitivity to growth model formulation. Our results indicate that at the landscape extent, simulation results are robust, but the reliability of model results at a finer resolution depends critically on accurate tree growth functions.