Final Report Summary - SPATIODIVERSITY (Towards a Unified Spatial Theory of Biodiversity)
One of the persistent challenges in contemporary ecology is to explain the high diversity in ecological communities such as tropical forests. The objective of the SPATIODIVERSITY project was to understand the relative importance of processes and factors that govern the composition and dynamics of species-rich communities. The reach this ambitious goal the project took a radically different approach than previous attempts in adopting a spatially explicit perspective. Most processes which are thought to contribute to species coexistence have strong spatial component. We therefore developed novel and unconventional methods to analyze and extract the ecological information hidden in spatial patterns. We applied these methods to large and high quality data sets of forest dynamic plots, each comprising several hundred of species and >100000 trees that are mapped and repeatedly monitored (the CTFS-ForestGEO network).
The SPATIODIVERSITY project rests on three pillars: (1) development of methods of spatial point pattern analysis to quantify the observed spatial patterns at different hierarchical levels, (2) development of dynamic and spatially explicit vegetation simulation models ranging from simple neutral models to more complex process based forest models, and (3) development of methods of statistical inference and parameterization for stochastic simulation models that can deal with high number of species. The models with different complexity (2) produce spatial patterns to be analyzed with (1), and we systematically compared them with the patterns observed in real forest communities (3) to identify a minimal set of processes needed to explain the observed spatial and non-spatial structures in the vegetation communities. The following mayor achievements should be highlighted.
We developed new methods of spatial analysis for detailed quantification of spatial structures in species, phylogenetic and functional diversity and used them for multi-forest comparisons. An important product is the publication of the “Handbook of spatial point pattern analysis in ecology” that summarizes this methodological development of the project. One mayor achievement of our cross-forest comparisons using spatial analysis is posing of the “dilution hypothesis” that states that stochastic dilution effects due to increasing species richness overpower signals of deterministic species interactions. This hypothesis can explain the empirical observation that in species rich communities little signatures of species interactions can be detected whereas species poorer plant communities tend to show stronger signals of species interactions. For fitting our stochastic vegetation models describing the dynamics of hundreds of species to the data we developed an approach of stochastic inference (based on Approximate Bayesian Computation; ABC) which has been so far only used for regression models and simpler equation models. We could solve this task and applied it to several of the 25 or 50ha forest dynamics plots. While it is known that neutral models (where all species have the same properties) can explain non-spatial biodiversity patterns such as the species abundance distribution we expected that our spatially-explicit extension of the “Hubbell model” would be too poor in biological processes and therefore not able to approximate spatial diversity patterns such as the species area relationship at the same time. However, to our surprise we found that the spatially-explicit neutral model could already well approximate several biodiversity patterns except the species-area relationship together with distance decay of similarity. Addition of species differences in dispersal kernels, negative density dependence, and habitat associations produced a good match of biodiversity patterns, but with high variability among simulation replicates.
We could thus confirm the central hypothesis of the SPATIODIVERSITY project that the data on spatial patterns contain independent information to identify biodiversity processes. A key result of the project emerging from our analyses is that much simpler models than anticipated can already explain multiple aspects of the complex spatial and non-spatial structure of diverse vegetation communities. This result has important consequences for the theoretical foundation of ecology. Our spatially-explicit approach moves previous theory towards a dynamic spatial theory of biodiversity and demonstrates the value of spatial data to identify ecological processes.
The SPATIODIVERSITY project rests on three pillars: (1) development of methods of spatial point pattern analysis to quantify the observed spatial patterns at different hierarchical levels, (2) development of dynamic and spatially explicit vegetation simulation models ranging from simple neutral models to more complex process based forest models, and (3) development of methods of statistical inference and parameterization for stochastic simulation models that can deal with high number of species. The models with different complexity (2) produce spatial patterns to be analyzed with (1), and we systematically compared them with the patterns observed in real forest communities (3) to identify a minimal set of processes needed to explain the observed spatial and non-spatial structures in the vegetation communities. The following mayor achievements should be highlighted.
We developed new methods of spatial analysis for detailed quantification of spatial structures in species, phylogenetic and functional diversity and used them for multi-forest comparisons. An important product is the publication of the “Handbook of spatial point pattern analysis in ecology” that summarizes this methodological development of the project. One mayor achievement of our cross-forest comparisons using spatial analysis is posing of the “dilution hypothesis” that states that stochastic dilution effects due to increasing species richness overpower signals of deterministic species interactions. This hypothesis can explain the empirical observation that in species rich communities little signatures of species interactions can be detected whereas species poorer plant communities tend to show stronger signals of species interactions. For fitting our stochastic vegetation models describing the dynamics of hundreds of species to the data we developed an approach of stochastic inference (based on Approximate Bayesian Computation; ABC) which has been so far only used for regression models and simpler equation models. We could solve this task and applied it to several of the 25 or 50ha forest dynamics plots. While it is known that neutral models (where all species have the same properties) can explain non-spatial biodiversity patterns such as the species abundance distribution we expected that our spatially-explicit extension of the “Hubbell model” would be too poor in biological processes and therefore not able to approximate spatial diversity patterns such as the species area relationship at the same time. However, to our surprise we found that the spatially-explicit neutral model could already well approximate several biodiversity patterns except the species-area relationship together with distance decay of similarity. Addition of species differences in dispersal kernels, negative density dependence, and habitat associations produced a good match of biodiversity patterns, but with high variability among simulation replicates.
We could thus confirm the central hypothesis of the SPATIODIVERSITY project that the data on spatial patterns contain independent information to identify biodiversity processes. A key result of the project emerging from our analyses is that much simpler models than anticipated can already explain multiple aspects of the complex spatial and non-spatial structure of diverse vegetation communities. This result has important consequences for the theoretical foundation of ecology. Our spatially-explicit approach moves previous theory towards a dynamic spatial theory of biodiversity and demonstrates the value of spatial data to identify ecological processes.