Final Report Summary - MATES (Understanding and predicting multispecies assemblages and interactions in space and time) Rapid biodiversity loss is one of the most vital challenges of the 21st century. Ecosystems have experienced dramatic changes over the last decades, with main drivers in the past being land use change and pollution whereas the impact of climate change is predicted to increase drastically in the next decades. In recognition of steady biodiversity loss, in 2011 the European Commission adopted the new Europe 2020 strategy to halt the loss of biodiversity and ecosystem services in EU by 2020, and biodiversity research has been an integral part of the European Union’s 7th Framework Programme for Research.For making quantitative predictions of expected future biodiversity distribution, we increasingly rely on computer-based models to predict potential biodiversity impacts under specific scenarios of climate change. Many factors complicate the assessment of climate change impact on biodiversity that are not well understood including for example uncertainties about complex biotic interactions within and across trophic levels. Most applications and publications have been concerned with predicting climate change-induced range shifts and range dynamics rely on correlative species distribution models (SDMs). These fit statistical relationships between species distribution data (occurrence or abundance) and prevailing environmental variables to characterise the environmental niche and to describe the species’ range limits in geographic. One criticism when using SDMs in climate change projections is that interspecific interactions are modelled only implicitly not explicitly, meaning that SDMs are not able to disentangle abiotic and biotic niche components, and we now see a development towards multi-species modelling frameworks (so-called joint species distributions models, JSDMs) that account for these interactions. Against this background, the main research objective of MATES was to improve the prediction of species assemblages in changing environments. Using a set of diverse modelling tools including extensive statistical analyses of spatial distribution data, implementation and estimation of JSDMs by means of Bayesian computation, and design and implementation of a multispecies simulation model for JSDM testing, we were able to shed some light on important challenges in prediction of multispecies assemblages in space and time, and to provide some guidelines to deal with these challenges. (1) Using a individual-based community model, we produced a set of consistent benchmarking data in a multi-scale design with an array of important community assembly and demographic processes. These benchmarking data can be used as a new standard for testing novel approaches, including single-species and multi-species, static and dynamic approaches. Further, we exemplified this by benchmarking novel single-species range dynamic models ranging from correlative species distribution models to complex population dynamic approaches. Our results clearly show that predictive performance of single species models is strongly influence by interactive effects of the species’ dispersal ability and by complex community interactions, and we give guidelines to address these challenges. (2) We tested for typical bias in predictions of species richness and community assemblage from stacked, single-species SDMs and from species richness regressions. Using data on Swiss breeding bird communities, we show that prediction accuracy is strongly affected by species’ habitat and resource requirements, and thus by scale decisions. This is an important notion for JSDM development as these are also affected by sub-scale environmental heterogeneity, which may distort the true underlying relationships between species. (3) We developed a novel analyses framework for identifying pairs of facilitating and facilitated species from simple co-occurrence and co-abundance data, which allows assessing the relative strength and importance of facilitation along environmental gradients. Exemplifying this method with data on 1242 plant community plots along a steep temperature-moisture gradient in Switzerland, we were able to show unexpectedly high complexity of facilitative interactions, with facilitation intensity being strongest under cold stress whereas facilitation frequency was higher under drought stress, and with distinct variation in functional distance between facilitating and facilitated species along the gradient. (4) We designed a new functional-trait based approach for complexity reductions. In this framework, we simultaneously estimate biotic interactions between species within functional groups and between functional groups. For this purpose, recently published JSDMs were adapted to include two hierarchical levels, species and functional groups. Currently, these functional group-based JSDMs are being estimated and evaluated for the Swiss Breeding Bird data. With these methodological advancements and comprehensive analyses and benchmark tests, MATES provides the basis for more targeted future research on multispecies interactions and community dynamics. Specifically, further methodological and theoretical developments are needed towards understanding and integrating the scale dependence of interspecific interaction mechanisms and how these scale up to affect community patterns and range dynamics, and towards disentangling the interactive effects of demography, dispersal and interspecific interactions on species’ niches. Understanding these complexities will be crucial for predicting potential effects of climate change and the emergence of novel communities. Also, more efforts are needed towards efficient and standardized model benchmarking. Taken together, these points will greatly help improving and setting standards for integrated biodiversity assessments. A project website is available at http://www.wsl.ch/fe/landschaftsdynamik/projekte/MATES/index_EN
