Final Report Summary - BIRDCHANGE (Exploring bird distribution drivers across differentspatial scales and time to predict the potential impact of global change on bird distributions and diversity)
The project BIRDCHANGE proposes to study current bird distribution shifts in response to global change to predict more accurately the future expected changes as well the predicted variations of bird diversity in terms of taxonomic, functional and phylogenetic components. Its main research objectives are: (1) Exploring the drivers of bird distributions across spatial scales and time; (2) Predicting the potential impact of global change on North American bird distributions and diversity; (3) Predicting the worldwide impact of global change on birds.
To predict the worldwide impact of global change on birds, I used the latest species distribution modeling techniques (e.g. ensemble forecast) to predict the current and future (under climate change scenarios) distributions of all bird species worldwide (>9,000 species). Having all these projections, we were able to estimate current and future species richness of bird assemblages worldwide and thus to estimate the predicted change of species richness. As species richness is only part of biodiversity, we went further and analyzed our results with regard to functional and phylogenetic diversity. Indeed, bird assemblages fulfill a critical set of ecological functions for ecosystems that may be altered substantially as climate change induced distribution changes lead to community disaggregation and reassembly. We were able to show that assemblage functional and phylogenetic structure is projected to change highly unevenly across space. These differences arise from both changes in the number of species and changes in species’ relative local functional/phylogenetic redundancy or distinctness. They sometimes result in substantial losses of functional diversity that could have severe consequences for ecosystem health. Range expansions may counter functional losses in high-latitude regions, but offer little compensation in many tropical and sub-tropical biomes. Future management of local community function and ecosystem services thus relies on understanding the global dynamics of species distributions and cross-scale approaches that include the biogeographic context of species traits.
Falling under the scope of the first research objective, I assessed the relevant climate predictors of bird species for species distribution modeling in the USA. Indeed, species distribution models (SDMs) are increasingly used to address numerous questions in ecology, biogeography, conservation biology and evolution but surprisingly, the crucial step of selecting the most relevant variables has received little attention, despite its direct implications for model transferability and uncertainty. For 243 species, we used yearly data since 1971 (from the North American Breeding Bird Survey) to run SDMs (six different algorithms) with combinations of six relatively uncorrelated climate predictors (selected from 22 widely used climate variables). We then estimated the importance of each predictor - both spatially and over a 40 year time period – by comparing the accuracy of the model obtained with or without a given predictor. Our study showed that three temperature related variables (annual potential evapotranspiration, mean annual temperature, and growing degree days) produced significantly more accurate SDMs than any other predictors. Among precipitation predictors, annual precipitation provided the most accurate results. Albeit only rarely used in SDMs, the moisture index performed similarly strongly. Interestingly, predictors that summarize average annual climate produced more accurate distributions than seasonal predictors, despite distinct seasonal movements in most species considered. Encouragingly, spatial and temporal (over 40 years) evaluation of variables yielded very similar results. Furthermore, if the approach we developed allowed us to identify the statistically most relevant predictors for birds in the US and it can be applied to other taxa and/or in different parts of the world. Appropriately selecting the most relevant predictors of species distributions at large spatial scale is vital to identifying ecologically meaningful relationships that provide the most accurate predictions under climate change or biological invasions.
Falling under all three research objectives, I also tackled the model transferability issue. Indeed SDMs relate species presence / abundance and absence data to environmental variables based on statistically or theoretically derived response surfaces. Usually the response curves between climate predictor and probability of presence / abundance are estimated using data covering the widest range of the species distribution. ‘Space-for-time’ substitution is then commonly used to predict the potential impact of climate change on species distributions, but the validity of this approach has rarely been assessed. In this project, I thus aim to disentangle spatially and temporally derived climatic responses of species, using the BBS data within a Bayesian framework. Therefore, for 52 US bird species, I fitted both “spatial” and “temporal” models. All these models aimed at predicting variation in abundance of a given species with climate (mean annual temperature an annual precipitation). To further assess the transferability of these models either in time (for the “spatial” model) or in space (for the “temporal” models), I used the “spatial” model to predict temporal variations in abundance and compared predictions from either the “spatial” model or the “temporal” model with the observed temporal variation in abundance. Similarly, I used any “temporal” model to predict spatial variation in abundance and compared predictions from either the “temporal” model or the “spatial” model with the observed spatial variation in abundance. Overall the “spatial” models show a very good accuracy (mean R2=0.548 range 0.243-0.863) and the “temporal” models a lower accuracy (mean R2=0.142 range -0.138-0.318). “Space for time” and “time for space” substitutions however show overall poor predictive power. It thus appears that using spatial variation of abundance with climate won’t allow us to accurately predict how abundance will change in time in response to climate change, at least for bird species.
The final results will show how biodiversity in a larger sense (including species richness, functional diversity and phylogenetic diversity) is projected to change under climate change for birds, which will be of great value for policy makers and society in general (e.g. to help raising awareness of climate change issues and potential consequences among the general public). The project results will also be important from a methodological point of view as they highlight the importance of finding the best climatic drivers of bird distribution before predicting the potential impacts of climate change and raise the potential issue of model transferability, a key assumption for most of the “climate change” predictions.
To predict the worldwide impact of global change on birds, I used the latest species distribution modeling techniques (e.g. ensemble forecast) to predict the current and future (under climate change scenarios) distributions of all bird species worldwide (>9,000 species). Having all these projections, we were able to estimate current and future species richness of bird assemblages worldwide and thus to estimate the predicted change of species richness. As species richness is only part of biodiversity, we went further and analyzed our results with regard to functional and phylogenetic diversity. Indeed, bird assemblages fulfill a critical set of ecological functions for ecosystems that may be altered substantially as climate change induced distribution changes lead to community disaggregation and reassembly. We were able to show that assemblage functional and phylogenetic structure is projected to change highly unevenly across space. These differences arise from both changes in the number of species and changes in species’ relative local functional/phylogenetic redundancy or distinctness. They sometimes result in substantial losses of functional diversity that could have severe consequences for ecosystem health. Range expansions may counter functional losses in high-latitude regions, but offer little compensation in many tropical and sub-tropical biomes. Future management of local community function and ecosystem services thus relies on understanding the global dynamics of species distributions and cross-scale approaches that include the biogeographic context of species traits.
Falling under the scope of the first research objective, I assessed the relevant climate predictors of bird species for species distribution modeling in the USA. Indeed, species distribution models (SDMs) are increasingly used to address numerous questions in ecology, biogeography, conservation biology and evolution but surprisingly, the crucial step of selecting the most relevant variables has received little attention, despite its direct implications for model transferability and uncertainty. For 243 species, we used yearly data since 1971 (from the North American Breeding Bird Survey) to run SDMs (six different algorithms) with combinations of six relatively uncorrelated climate predictors (selected from 22 widely used climate variables). We then estimated the importance of each predictor - both spatially and over a 40 year time period – by comparing the accuracy of the model obtained with or without a given predictor. Our study showed that three temperature related variables (annual potential evapotranspiration, mean annual temperature, and growing degree days) produced significantly more accurate SDMs than any other predictors. Among precipitation predictors, annual precipitation provided the most accurate results. Albeit only rarely used in SDMs, the moisture index performed similarly strongly. Interestingly, predictors that summarize average annual climate produced more accurate distributions than seasonal predictors, despite distinct seasonal movements in most species considered. Encouragingly, spatial and temporal (over 40 years) evaluation of variables yielded very similar results. Furthermore, if the approach we developed allowed us to identify the statistically most relevant predictors for birds in the US and it can be applied to other taxa and/or in different parts of the world. Appropriately selecting the most relevant predictors of species distributions at large spatial scale is vital to identifying ecologically meaningful relationships that provide the most accurate predictions under climate change or biological invasions.
Falling under all three research objectives, I also tackled the model transferability issue. Indeed SDMs relate species presence / abundance and absence data to environmental variables based on statistically or theoretically derived response surfaces. Usually the response curves between climate predictor and probability of presence / abundance are estimated using data covering the widest range of the species distribution. ‘Space-for-time’ substitution is then commonly used to predict the potential impact of climate change on species distributions, but the validity of this approach has rarely been assessed. In this project, I thus aim to disentangle spatially and temporally derived climatic responses of species, using the BBS data within a Bayesian framework. Therefore, for 52 US bird species, I fitted both “spatial” and “temporal” models. All these models aimed at predicting variation in abundance of a given species with climate (mean annual temperature an annual precipitation). To further assess the transferability of these models either in time (for the “spatial” model) or in space (for the “temporal” models), I used the “spatial” model to predict temporal variations in abundance and compared predictions from either the “spatial” model or the “temporal” model with the observed temporal variation in abundance. Similarly, I used any “temporal” model to predict spatial variation in abundance and compared predictions from either the “temporal” model or the “spatial” model with the observed spatial variation in abundance. Overall the “spatial” models show a very good accuracy (mean R2=0.548 range 0.243-0.863) and the “temporal” models a lower accuracy (mean R2=0.142 range -0.138-0.318). “Space for time” and “time for space” substitutions however show overall poor predictive power. It thus appears that using spatial variation of abundance with climate won’t allow us to accurately predict how abundance will change in time in response to climate change, at least for bird species.
The final results will show how biodiversity in a larger sense (including species richness, functional diversity and phylogenetic diversity) is projected to change under climate change for birds, which will be of great value for policy makers and society in general (e.g. to help raising awareness of climate change issues and potential consequences among the general public). The project results will also be important from a methodological point of view as they highlight the importance of finding the best climatic drivers of bird distribution before predicting the potential impacts of climate change and raise the potential issue of model transferability, a key assumption for most of the “climate change” predictions.