Final Report Summary - SESAM-ZOOL (A novel modeling framework for predicting animal communities under global change)
Humans are causing enormous global changes and responses to these modifications have been already observed on biodiversity. Models play a central role to foreseen the effects of human driven changes on ecosystems, but improvements are still needed to better account for assembly processes in the reconstruction of communities. In this research project I focused on fundamental questions regarding the prediction of species assemblages to develop new techniques for advanced analyses on complex systems. As a first step I revised the range of possible approaches to produce spatially explicit models of communities. These analytical processes may vary in their theoretical foundation and in the way they include one or more of the drivers known to influence the assembly of species in communities, which I categorized as (i) historical and evolutionary, (ii) environmental, (iii) biotic, and (iv) stochastic. By analysing the most recent research for community level modelling, it emerged that a promising solution to manage the complexity of the assembly process and to obtain more reliable community predictions could be the integration of multiple drivers in a unique modelling framework. I defined the “modelling frameworks” as methodological procedures made of a series of sequential analytical and/or modelling steps. This structure should be anchored in a solid theoretical background, still allowing the implementation of case specific settings according to the case study. The steps can use pre-existing methods independently developed to predict communities or incorporate any new technical advances and include static and dynamic models to make the most of their respective strengths. This research has been accepted for publication in the scientific journal Biological Review (D’Amen et al. In press, a).
In the most recent literature, some proposed solutions agree with the framework structure, and in the following development of the project I focused on one proposed approach, the SESAM framework (Spatially-Explicit Species Assemblage Modeling). The SESAM framework aims at reconstructing species assemblages by four successive steps: i) designing the species pool (the candidate species); ii) applying abiotic filter based on species’ ecological niche; iii) applying macroecological constraints to set the limit of coexisting species number; iv) applying biotic rules for selecting among the potential species from step ii, those actually able to coexist in the modelled community. Each of these modules can be implemented with different modelling options and assessed independently. I implemented and tested for the first time this framework on a very robust community data set of plants and animals collected for many years in the Swiss Alps from my hosting group. In particular, I run modelling analyses to obtain the potential composition of both plants and insects’ assemblages. I tested these two biodiversity components separately combining individual species distribution models, macroecological models (also based on species traits) and different approaches to set biotic rules.
In particular, as concerns the test of the SESAM framework on the plant dataset, I collaborated with my colleagues in developing two alternative implementations, either based on species traits or on the species’ probability of presence, now published in the Journal of Biogeography (D’Amen et al. 2015). First we investigated the use of species’ traits as macroecological rule: the predicted extreme values of three different functional traits were used to constraint a pool of environmental filtered species from binary SDMs predictions – “Trait Range” rule. Second, the community composition was determined by selecting the species in decreasing order of their rough probabilities from SDMs (assuming specie with higher probability being competitively superior, i.e. a simple ecological assembly rule) to match the species richness predictions (the macroecological constraint) - “Probability Ranking” rule.
In the second test, now in press in the journal Global Ecology & Biogeography, departing from the results obtained I introduced three novelties: 1) I applied the SESAM framework for the first time on animal communities, considering the two insects’ groups (butterflies and grasshoppers); 2) I explored the influence of different techniques to create presence/absence predictions for individual species in the implementation of the step ii) of the framework; 3) I developed and tested the integration of the new ecological assembly rules in the framework as biotic filtering and I compared their performance with the “Probability Ranking” rule. This last part is very important as I evidenced in my review work that understanding the complexity of inter-specific interactions remains a major challenge for community-level models. In particular, these rules are based on the spatial patterns of species presence absence in the study area (co-occurrence) considered as proxy of interaction intensity. I developed a new R function to test for non-random patterns of species co-occurrence applying an environmental constraint. According to the results obtained with this function for the considered groups, I developed two sets of assembly rules based either on estimating competition intensity for single species or interaction strength for species pairs.
Test of community modelling were also conducted on birds communities in collateral projects. These projects belonged to the PhD theses of two students who visited the laboratory of Prof. Guisan for a short stay (from the Università del Molise, Italia and the Centre Tecnològic Forestal de Catalunya, Spain) and whom I co-tutored during their time in Lausanne. From my results and the collaborative works several papers arose, which are accepted for publication or currently under review in top ranking journals (Diversity and Distribution, Global Change Biology, Ecography).
The new insights on the application of the SESAM framework lead us to draw interesting conclusions on the assembly mechanisms of the considered communities and to delineate the main routes to follow for the future research on community-level modelling. First, it is important to highlight the significance of producing good individual species distribution models to reconstruct communities. If these models are fitted in a biased manner or miss the relevant predictors representing the environmental filter, they offer little information about the ecological potential of the species, and when applying SESAM, the framework will be similarly biased. Second, according to our tests on multiple data sets, we can support the benefit of applying the SESAM framework to reconstruct community composition. The improvement in community prediction obtained in applying SESAM is higher when the individual models alone are not able to depict the community structure, and a sign of this lack is an over-estimation of the community richness when simply summing the individual species predictions. On the contrary, when this bias is not observed, the application of all steps of SESAM cannot further improve the prediction of community composition. Third, in relation to the biotic rules in SESAM, we can confirm the utility of the initially proposed “Probability Ranking” rule in improving community predictions, when the previous bias is detected. Finally, based on our results on the insects’ groups we cannot reach a definitive conclusion about the proposed biotic rules based on co-occurrence, but we identified the best performing one so far. The importance of explicitly taking into account biotic interactions is highly dependent on the ecosystem under study, thus further tests on species assemblages structured by competition could help clarify this issue. Other investigations of biotic interactions could complement or substitute this co-occurrence analysis, while also considering population densities when data are available.
The new analytic tools that I tested are available in the “ecospat” R package and may have wide-ranging applications to many other study systems, especially in ecosystem science e.g. in the conservation field to explore the possible consequences of biodiversity changes due to human activities; for the industrial sector to explore routes for mitigating the impact of their activities. In fact, the development of a new generation of ecological projections can support the resolutions of biodiversity managers to efficiently plan conservation strategies, which both preserve the functioning of current ecosystems and take future changes into account.
In the most recent literature, some proposed solutions agree with the framework structure, and in the following development of the project I focused on one proposed approach, the SESAM framework (Spatially-Explicit Species Assemblage Modeling). The SESAM framework aims at reconstructing species assemblages by four successive steps: i) designing the species pool (the candidate species); ii) applying abiotic filter based on species’ ecological niche; iii) applying macroecological constraints to set the limit of coexisting species number; iv) applying biotic rules for selecting among the potential species from step ii, those actually able to coexist in the modelled community. Each of these modules can be implemented with different modelling options and assessed independently. I implemented and tested for the first time this framework on a very robust community data set of plants and animals collected for many years in the Swiss Alps from my hosting group. In particular, I run modelling analyses to obtain the potential composition of both plants and insects’ assemblages. I tested these two biodiversity components separately combining individual species distribution models, macroecological models (also based on species traits) and different approaches to set biotic rules.
In particular, as concerns the test of the SESAM framework on the plant dataset, I collaborated with my colleagues in developing two alternative implementations, either based on species traits or on the species’ probability of presence, now published in the Journal of Biogeography (D’Amen et al. 2015). First we investigated the use of species’ traits as macroecological rule: the predicted extreme values of three different functional traits were used to constraint a pool of environmental filtered species from binary SDMs predictions – “Trait Range” rule. Second, the community composition was determined by selecting the species in decreasing order of their rough probabilities from SDMs (assuming specie with higher probability being competitively superior, i.e. a simple ecological assembly rule) to match the species richness predictions (the macroecological constraint) - “Probability Ranking” rule.
In the second test, now in press in the journal Global Ecology & Biogeography, departing from the results obtained I introduced three novelties: 1) I applied the SESAM framework for the first time on animal communities, considering the two insects’ groups (butterflies and grasshoppers); 2) I explored the influence of different techniques to create presence/absence predictions for individual species in the implementation of the step ii) of the framework; 3) I developed and tested the integration of the new ecological assembly rules in the framework as biotic filtering and I compared their performance with the “Probability Ranking” rule. This last part is very important as I evidenced in my review work that understanding the complexity of inter-specific interactions remains a major challenge for community-level models. In particular, these rules are based on the spatial patterns of species presence absence in the study area (co-occurrence) considered as proxy of interaction intensity. I developed a new R function to test for non-random patterns of species co-occurrence applying an environmental constraint. According to the results obtained with this function for the considered groups, I developed two sets of assembly rules based either on estimating competition intensity for single species or interaction strength for species pairs.
Test of community modelling were also conducted on birds communities in collateral projects. These projects belonged to the PhD theses of two students who visited the laboratory of Prof. Guisan for a short stay (from the Università del Molise, Italia and the Centre Tecnològic Forestal de Catalunya, Spain) and whom I co-tutored during their time in Lausanne. From my results and the collaborative works several papers arose, which are accepted for publication or currently under review in top ranking journals (Diversity and Distribution, Global Change Biology, Ecography).
The new insights on the application of the SESAM framework lead us to draw interesting conclusions on the assembly mechanisms of the considered communities and to delineate the main routes to follow for the future research on community-level modelling. First, it is important to highlight the significance of producing good individual species distribution models to reconstruct communities. If these models are fitted in a biased manner or miss the relevant predictors representing the environmental filter, they offer little information about the ecological potential of the species, and when applying SESAM, the framework will be similarly biased. Second, according to our tests on multiple data sets, we can support the benefit of applying the SESAM framework to reconstruct community composition. The improvement in community prediction obtained in applying SESAM is higher when the individual models alone are not able to depict the community structure, and a sign of this lack is an over-estimation of the community richness when simply summing the individual species predictions. On the contrary, when this bias is not observed, the application of all steps of SESAM cannot further improve the prediction of community composition. Third, in relation to the biotic rules in SESAM, we can confirm the utility of the initially proposed “Probability Ranking” rule in improving community predictions, when the previous bias is detected. Finally, based on our results on the insects’ groups we cannot reach a definitive conclusion about the proposed biotic rules based on co-occurrence, but we identified the best performing one so far. The importance of explicitly taking into account biotic interactions is highly dependent on the ecosystem under study, thus further tests on species assemblages structured by competition could help clarify this issue. Other investigations of biotic interactions could complement or substitute this co-occurrence analysis, while also considering population densities when data are available.
The new analytic tools that I tested are available in the “ecospat” R package and may have wide-ranging applications to many other study systems, especially in ecosystem science e.g. in the conservation field to explore the possible consequences of biodiversity changes due to human activities; for the industrial sector to explore routes for mitigating the impact of their activities. In fact, the development of a new generation of ecological projections can support the resolutions of biodiversity managers to efficiently plan conservation strategies, which both preserve the functioning of current ecosystems and take future changes into account.