Data and R code from: Downscaling species to individual-level networks reveals the importance of population-level processes in mediating generalized community-wide interaction patterns

Patterns and consequences of community-wide interactions have been evaluated at the species-level, thus ignoring differences in individual behavior. In plant–pollinator networks, downscaling interactions to the individual level may further help reconcile the high level of generalization related to community stability with the specificity needed for pollination functioning. Here, we built species-level and individual-level networks using pollen loads on individual pollinators collected in a diverse serpentine seep community in Northern California. Plant–pollinator interactions were almost two times more specialized at the individual- compared to the species-level, suggesting a higher level of niche partitioning among individual insects which may increase pollination functioning within the community. However, we observed differences in individual specialization even among generalist pollinator species, which may differentially impact conspecific pollen movement and pollination efficacy despite similar niches at species level. Furthermore, intraspecific trait variation (i.e., body size) did not impact pollinator niche, suggesting that other population-level factors related to resource use may drive plant–pollinator network structure. We did find that female bees are more specialized than males, suggesting that sex-based differences may contribute to variation in individual specialization with potential consequences for community-wide pollination success. Overall, using individual-level networks this study links individual foraging patterns with population-level processes that may scale up to mediate the structure of species-level plant–pollinator networks. In doing so, this study further aids in our understanding of perceived conflicts between specialization and generalization in plant–pollinator communities.

Methods
To build individual- and species-level networks we surveyed flower-visiting insects foraging on the serpentine seeps between 09h00 and 15h00 using entomological nets between May 9th and June 1st, 2021. Insects were collected by 2-3 people simultaneously walking at a steady pace while observing all plant species and collecting all insects observed visiting flowers and making contact with plant reproductive structures. Specimens were stored in tubes under cold temperatures in the field immediately after collection to prevent them from moving and losing pollen. We then sampled pollen loads from each insect in the lab by swabbing multiple body parts (head, dorsal and ventral thorax, and fore- and mid-legs) with a fuchsin jelly cube that was later mounted on a microscope slide. Pollen carried on corbiculae or scopae in bees was excluded because it represents a resource that is not typically available for pollination and neither present in male bees or other pollinator functional groups sampled. All pollen grains found in insect pollen loads were identified and counted under a microscope. Pollen identification was conducted with the aid of a pollen library previously established from anthers collected for each plant species at the study site. Insect specimens were identified at the species or morphospecies level (species hereafter). To avoid overestimating specialization we constructed plant–pollinator networks only considering insect species that were represented by at least five individuals (n ≥ 5). To characterize plant–pollinator interactions we constructed pollen-transport networks, which represent not only interactions occurring during the observed foraging bout, but may also capture flower visitation events that took place over previous foraging bouts. Plant–pollinator networks constructed based on pollen loads on pollinator bodies have been shown to more broadly capture realized flower visitor interactions in the field. All insect specimens are preserved in the pollinator collection at East Tennessee State University (ETSU).

学界此前多在物种层面评估群落水平种间互作的模式与效应，却忽略了个体行为间的差异。在植物-传粉者网络（plant–pollinator networks）中，将互作尺度细化至个体层面，或可进一步调和与群落稳定性相关的泛化程度与传粉功能所需的特化程度之间的矛盾。本研究依托美国北加州一处多样的蛇纹岩渗流群落（serpentine seep community）中采集的个体传粉者携带的花粉载荷，构建了物种层面与个体层面的两类传粉网络。相较于物种层面，个体层面的植物-传粉者互作特化程度提升近一倍，这表明昆虫个体间的生态位分化程度更高，或可增强群落内的传粉功能。然而本研究发现，即便在泛化传粉者类群内部，个体特化程度仍存在差异；尽管这类类群在物种层面的生态位相似，但其个体差异或会对同种花粉传播与传粉效率产生差异化影响。此外，种内性状变异（intraspecific trait variation）并未对传粉者生态位产生影响，这表明与资源利用相关的其他种群层面因素，或是塑造植物-传粉者网络结构的驱动因子。本研究确实发现雌性蜜蜂的特化程度高于雄性，这表明基于性别的差异或许会导致个体特化程度的变异，并可能对群落整体的传粉成功产生潜在影响。总体而言，本研究通过个体层面的网络，将个体觅食模式与种群层面的过程联系起来，而这些过程可通过尺度上推介导物种层面植物-传粉者网络的结构。通过这一研究思路，我们得以进一步理解植物-传粉者群落中特化与泛化之间看似存在的矛盾。

研究方法
为构建个体层面与物种层面的传粉网络，本研究于2021年5月9日至6月1日期间，在每日9时至15时使用昆虫网对蛇纹岩渗流生境上觅食的访花昆虫进行采样。采样由2-3名研究人员同步以稳定步伐行进开展，过程中观察所有植物类群，并采集所有被发现访花且接触植物繁殖结构的昆虫。采集后的标本即刻置于冷藏管中存放于野外，以防止昆虫活动导致花粉脱落。随后在实验室中，我们使用碱性品红胶（fuchsin jelly）块擦拭每只昆虫的多个身体部位（头部、前胸背板与腹板、前足及中足）以采集其携带的花粉载荷，之后将胶块固定于载玻片上。我们排除了蜜蜂花粉篮（corbiculae）或花粉刷（scopae）上携带的花粉，因为这类花粉属于蜜蜂储备的食用资源，通常无法参与传粉过程，且雄性蜜蜂及其他采样的传粉者功能类群并不具备此类结构。在显微镜下对昆虫花粉载荷中发现的所有花粉粒进行鉴定与计数。花粉鉴定工作依托此前在研究区域内为各植物类群采集花药所构建的花粉参考库开展。昆虫标本以物种或形态种水平进行鉴定（后文统称物种）。为避免特化程度被高估，本研究仅纳入采样个体数不少于5只的昆虫类群（n≥5）以构建植物-传粉者网络。为表征植物-传粉者互作，本研究构建了花粉转运网络，这类网络不仅能反映观测到的单次觅食过程中的互作，还可记录昆虫此前觅食过程中发生的访花事件。已有研究表明，基于传粉者体表花粉载荷构建的植物-传粉者网络，可更全面地反映野外环境中访花者的实际互作情况。所有昆虫标本均保存于东田纳西州立大学（East Tennessee State University, ETSU）的传粉者标本馆藏中。