RésuméRapid anthropogenic environmental changes have a widespread detrimental effect on global patterns of biodiversity. Climate change and land use/land cover (LULC) change have long been recognized as two of the drivers of biodiversity loss and shifts in species’ distributions. Climate and LULC changes can alter species’ habitats through changes in temperature, rainfall and extreme weather patterns, and land conversions from areas rich in resources to areas with insufficient resources. Species are then forced to move into areas with tolerable conditions and adequate resources or face local extinction. To be able to interpret historical dynamics, recognize present day patterns, and project changes under potential futures, it is essential to understand in detail climate and LULC requirements of different species at a variety of different extents and locations.
Wild bees represent an ideal study organism to explore these themes. Wild bee species are needed to pollinate the majority of wild flowers and can greatly influence crop pollination, supporting food provisioning for humans. Wild bees have also experienced significant changes in many areas over the last 100 years, showing large shifts in their distribution patterns, declines in diversity and abundance, and many local extinctions. In order to protect wild bees and mitigate the influence of rapid global changes, it is necessary to quantify the influence of LULC and climate effects on wild bees. Consequently, the general objective of this thesis is to examine how LULC and climate conditions impact the diversity and distribution patterns of wild bee species at different spatial and temporal scales.
To achieve the general objective we focused on three aims: to (1) test the efficacy of using statistical modelling tools to understand wild bee distributions in the present and future, and suggest how to improve these methods; (2) provide novel understanding of how wild bee community assemblages are structured at large geographical scales and what drives this structure; and (3) quantify and compare how past, present, and future changes to wild bee and specifically, bumblebee distributions are expected to be influenced by LULC and climate changes.
In order to accomplish these aims a variety of statistical techniques were utilized throughout the thesis. In particular, a common theme of the thesis is the use of species distribution models (SDM) to model the relationship between wild bee occurrence records and the environment, and to use this relationship to project distribution patterns. Furthermore, species interactions, phylogenetic relationships and functional species traits were included in the analyses to provide more ecological detail in explaining the observed patterns of diversity and distribution. Firstly, we introduce the background and knowledge gaps in chapter 1, general introduction and then present the material and methods used in the thesis in chapter 2. The three aims are explored across four chapters (3-6) with narrower objectives each representing a separate scientific study. Finally, we explore the relevance and implications of the thesis in chapter 7, general discussion.
The objective of Chapter 3 was to quantify the performance of species distribution models when modelling wild bee distributions. Specifically, we examined how habitat suitability predictions for Dutch wild bees are contingent on the LULC context where a species is predicted to occur and the functional trait groupings of all species. Independent collections made after the construction of SDMs were used to test the models. In total 52 wild bees species, of the total 193 modelled species, were collected in independent collections from agricultural habitats, specifically, arable fields and orchards. The 52 wild bee species were grouped into 4 separate functional trait groups representing small intermediate specialist, small generalist, highly specialised, and large generalist species. Habitat suitability projections were significantly better for highly specialised species and species collected in orchard habitats. The results suggest that SDMs for wild bees can be more or less useful depending on the species modelled. Specifically, projections made for specialist species and within stable habitats are likely to be the most accurate.
The objective of chapter 4 was to build on the results and implications of chapter 3 and to quantify and visualize the influence of habitat filtering and co-occurrence when modelling the assembly patterns of wild bee species. Again, this study was focused on the Netherlands. Firstly, the spatial co-occurrence of all 297 wild bee species was analysed. Wild bee species generally showed a strong positive correlation in co-occurrence. Suggesting, that many wild bee species are found together significantly more than expected by chance alone. Following this, a joint SDM (JSDM) approach was used to classify the significance of habitat filtering, biotic interactions, functional traits and phylogenetic relatedness on the geographic patterns of wild bee assemblages. The results showed that habitat filtering explained the majority of the geographic distribution of wild bee assemblages. The relationship between wild bee species and the environmental conditions was only weakly explained by traits but showed a strong phylogenetic signal, suggesting closely related species have similar habitat filtering requirements. Including species co-occurrence matrices into the JSDM approach improved model performance signifying that there are unexplained factors that certain species pairs require not captured in the modelling process. Overall, the study provides a clear representation of the geographic distribution of wild bee assemblages, the factors influencing this distribution and provides clear implications for wild bee conservation. The results indicate potential conservation units in the form of spatially explicit community and habitat profiles as well as outlining potential indicator species, which are representative of diverse and distinct assemblages.
The objective of chapter 5 was to look at aspects of habitat filtering at broader temporal and spatial scales and precisely to quantify the influence of dynamic land use/land cover projections on the projected distributional change of bumblebees under climate change. Using three model types, (1) only climate change covariates, (2) climate change and static LULC covariates and (3) climate change and dynamic LULC covariates the distribution of 48 bumblebees were modelled at the European and BENELUX scale. There were clear differences in the projections of range changes produced by the different model types. The implication of these results for modelling wild bee species under changing climate are that when available LULC change projections should be utilized in prospective biodiversity scenarios. Furthermore, the results indicate the need for improved and detailed LULC change projections that take into account smaller scale natural habitat types and land management.
Chapter 6 presents a historical look at the impacts of environmental changes in the Pyrenees with the objective to measure a specific case of how the composition and distribution of a wild pollinator group has changed over time due to the influence of LULC and climate changes. Using two collections datasets, one from 1889 and a follow-up conducted in 2005-06, the composition and distribution of the bumblebee, day-flying Lepidoptera and their visited plants were compared. Overall, all groups show an upward shift in mean elevation, but this shift is not evenly spread across all species. For the bumblebees, specialist species are found higher up the mountain and also experience greater shifts in their elevation. Furthermore, community composition does not change drastically. There is also an indication that pollinators and their visited plants are shifting in unison. The results lend support to predicted climate change effects on biodiversity, and indicate certain specialized species that could be in danger of significant declines if conservation efforts are not implemented.
In conclusion, this thesis highlights the significance of historical wild bee occurrence data and the utility of SDMs for investigating key environmental requirements of wild bee species and assessing long-term trends in distribution. We show that wild bees distribution patterns are highly dependent on LULC conditions in the present and future. The work also emphasizes the strong interaction between climate and LULC and how necessary it is to incorporate both in future biodiversity scenarios. It also shows for the first time influence of co-occurrence patterns on the formation of national wild bee assemblages. Which in turn increased our knowledge the processes behind patterns of distribution and multiple measures of diversity, including community, functional and phylogenetic. Finally, this thesis provides significant advice to conserve wild bee species individually and collectively.
The results clearly indicate areas of interest for future studies, which should focus on the complexities and the interactions of the relationships shown here. The drivers of wild bee decline strongly interact and therefore should be examined simultaneously. In particular, greater focus is needed on the ecological drivers of wild bee distribution patterns, including dispersal capabilities, biotic interactions with flowering plants, other bees and pathogens, as well as how physiological tolerance will influence the impacts of global change. Additionally, future LULC maps and projections which incorporate high-resolution depictions of natural areas and differences in land management will improve our ability to analyse and understand the environmental requirements of wild bees. As wild bee species are expected to continue to decline globally this thesis increases the knowledge and tools available to ensure that high diversity wild bee communities continue to persist.
|la date de réponse||26 oct. 2018|
|Sponsors||Service Public Federal de Programmation Scientifique - BELSPO|
|Superviseur||Nicolas DENDONCKER (Promoteur), Jacobus C. Biesmeijer (Copromoteur), Catherine Linard (Président), Jesús Aguirre-Gutiérrez (Jury), Luísa G. Carvalheiro (Jury), Nicolas Titeux (Jury) & Nicolas J. Vereecken (Jury)|