Thèse de doctorat
Résumé : The diversity of organisms within an ecosystem is essential to its functioning. It ensures a stable provision of goods and services, along with the resilience of many communities facing disturbances. Pollinators are involved in the reproduction cycle of a vast majority of flowering plants and crops, and play an essential role in overall ecosystem functioning. Wild bees constitute the most important group of pollinators worldwide. However, this major role is currently under threat with recent reports describing a global decline of overall communities. There are many different factors involved in these declines, both biotic and abiotic, many of which are induced by human activities and act either isolated or in combination with other factors. Climate, landscape and local factors are the main drivers of bee decline. They tend to influence one another, with synergistic effects also influencing bees in different ways. Many experts are calling for meaningful conservation measures to avoid massive detrimental consequences of multi-stressor interactions on wild bee communities inevitably leading to weakened ecosystem functioning.In regard to the complex interactions between decline factors in the context of human-induced global change, and the central pillar of biodiversity that wild bees represent, this thesis focuses our attention on an in-depth exploration of these intricate links. More specifically, our main objective is to study the relative impacts of decline factors in shaping wild bee communities, at different geographic scales and in different human-modified landscapes. We aimed to (i) disentangle the effects of intertwined factors in different contexts, in order to (ii) question long-held beliefs, and finally (iii) find key drivers for future conservation of wild bee communities.To achieve these aims, we decided to investigate both at a continental (i.e., one crop) and a local scale (i.e., one city). We observed standardized sampling methods to collect data on wild bee species combining active samplings using entomological nets during transect walks and passive samplings using colored pan traps. We also used historical data to characterize the bee fauna of a specific urban green space. This data allowed us to characterize wild bee communities at different scales by computing the species richness, functional and phylogenetic diversity. We then extracted data on decline factors such as climate, landscape composition and local factors using existing public databases and analyzing pollen samples. The relative effects of decline factors on wild bee communities were explored by using regression techniques, more specifically by building generalized linear models and structural equation models.The first chapter is a broad introduction on the global ecological background necessary to appreciate our research, followed by an explanation of the scope and corresponding outline of the thesis. Secondly, we give a general overview of the methodology used to collect the different types of data, the metrics chosen to characterize this data, and the statistical tools employed to answer our main objectives. The three formulated aims of this thesis are then explored throughout four main chapters (3 to 6), each of which represents an independent scientific study with distinct research questions, and all progressively answering our main objective. The seventh chapter is a large discussion developing the general relevance of this thesis, along with future perspectives for wild bee and overall biodiversity conservation, also including leads for future research. Finally, a general conclusion is given in chapter 8.In chapter 3, we use generalized linear models (GLMs) to disentangle the effects of local factors (i.e., land cover, management practices and honey bee dominance) on overall wild bee diversity. More specifically, we used standardized field surveys to sample bee specimens in 36 commercial apple orchards (paired in organic and non-organic) in Western Europe during the 2019 flowering season. Considering simultaneously all these parameters, the presence of managed honey bees in the apple orchards appeared to be the main factor influencing all wild bee diversity metrics, far beyond climate and landscape variables. Our results also show that organic practices in commercial apple orchards, believed to be highly beneficial to wild pollinator communities, actually constitute mere substitutes of conventional practices, with no measured benefits on overall wild bee diversity.Chapter 4 builds on the previous one, by including supplementary aspects of apple production and pollination efficiency. Here, we explore the causal and embedded effects of multiple biotic and abiotic stressors of wild bees (i.e., climate, landscape, management practices and local factors) on overall wild bee diversity. We also studied the effects on apple crop yield and fruit set using piecewise structural equation models (pSEM). Building on the dataset used in Chapter 3, we extend our scope to reach a total of 46 commercial apple orchards (organic and non-organic) in Western Europe and Morocco, sampled during the 2019 flowering season. Additionally, we collected data on crop yield and fruit quality following the same 2019 flowering season to explore aspects of pollination efficiency. The results presented here show that apple yields are mainly driven by management practice, without evidence for a significantly superior contribution by managed honey bees. Total wild bee diversity, while negatively correlated with honey bee dominance, enhanced seed set.The articles presented in Appendix E and F represent parallel published studies, going deeper in the description of methods used to (i) measure bee diversity, and (ii) to survey bee communities along a large-scale climatic gradient. These two parallel studies underline the relevance of the standardized survey methods and the diversity metrics used in chapter 3 and 4. The bee sampling data collected was shared with fellow scientists to help build what is now the most comprehensive open global data set on measurements of crop flower visitors, crop pollinators and pollination to date, presented in Appendix G. Together, chapter 3 and 4 constitute the aspect of this thesis focusing essentially on a human- modified context at a large geographic scale.In chapter 5, we explore an entirely different human-modified context, focusing on the drivers of wild bee diversity at the local scale and in an urban environment. We use an innovative experimental setup to explore the role of urbanization in the distribution of different wild bee decline factors. More specifically, for 50 sites within the Brussels-Capital region, we used pollen samples collected by solitary females of two Osmia species as bioindicators of plant richness, pesticide residues and parasite load. We then computed GLM to disentangle the prevalence of these drivers along an urbanization gradient. Overall, our results from chapter 5 showed that urbanization in itself did not act as an environmental filter of decline factors, neither on host plants availability.Building on the previous findings, chapter 6 is a case study showing how a certain habitat type present within urban areas (i.e., urban wastelands) can constitute a refuge to wild bee species. We compile five years of citizen science and standardized field surveys to characterize the wild bee community of the Friche Josaphat, the largest urban wasteland in the Brussels-Capital Region. We then use diversity metrics to compare the species at our target site with the regional list of species, and use null models of community assembly to test the local species assemblages against random assemblages. Our results show a significant share of the functional and phylogenetic diversity of wild bees known from the Brussels-Capital Region is represented in the Friche Josaphat, making it one of the most species-rich localities known to date for wild bees in Belgium. We highlight the strong complementarity of citizen science and academic approaches in biodiversity surveys. We also reaffirm the potential of wastelands as essential components of urban biodiversity. This chapter further addresses the long-held belief that urbanization is bad for wild bee communities, by putting the accent on the complex mosaic of habitats represented in urban areas, with high associated potential for future conservation measures.Overall, our results show that interactions between wild bee species and their environmental living conditions are an extremely complex and dynamic phenomenon, largely depending on local conditions that differ according to the spatial scale considered. On one hand, we investigated the wild bee communities associated to one of the most widespread crops in the world, at a continental scale and using standardized sampling methods in a way that had never been done before on such a large climate gradient and all within the same flowering season. On the other hand, we designed a truly innovative experimental setup to explore stressors of wild bee communities found at a local scale within one of the greenest cities in Europe. We also show that using standardized sampling methods in real life conditions is crucial for comparisons between studies from different landscapes and to build consistent long-term monitoring of species.Finally, this thesis opens clear doors for future research. While our research focused on apple orchards, we think that modeling contrasting crops could be interesting to provide a more balanced insight on the main drivers of bee diversity at different spatial and phenological scales. We strongly suggest that coming studies use standardized methods to monitor overall wild bee communities for multiple flowering seasons and use that data to model temporal trends of the relative impact of decline factors for wild bees. We also believe that significant improvements of our models can be made by increasing the amount of data collected, in turn feeding and refining the causal models. Regarding the experimental setup developed in chapter 5, we think it should be seen as a cheap, efficient and logistically practical tool. It will allow to broadly monitor the environmental composition in elements strongly shaping pollinator communities at the local level and good for all landscape contexts. We suggest that future studies willing to reproduce similar types of monitoring use this tool and improve it.