The contributions of flower strips to wild bee conservation in agricultural landscapes can be predicted using pollinator habitat suitability models

Sowing flower strips along field edges is a widely adopted method for conserving pollinating insects in agricultural landscapes. To maximize the effect of flower strips given limited resources, we need spatially explicit tools that can prioritize their placement, and for identifying plant species to include in seed mixtures.

We sampled bees and plant species as well as their interactions in a semi-controlled field experiment with roadside/field edge pairs with/without a sown flower strip at 31 sites in Norway and used a regional spatial model of solitary bee species richness to test if the effect of flower strips on bee species richness was predictable from the modelled solitary bee species richness.

We found that sites with flower strips were more bee species rich compared to sites without flower strips and that this effect was greatest in areas that the regional solitary bee species richness model had identified to be particularly important for bees. Spatial models revealed that even within small landscapes there were pronounced differences between field edges in the predicted effect of sowing flower strips.

Of the plant species that attracted the most bee species, the majority mainly attracted bumblebees and only few species also attracted solitary bees. Considering both the taxonomic diversity of bees and the species richness of bees attracted by plants we suggest that seed mixes containing Hieracium spp. such as Hieracium umbellatum and Pilosella officinarum; Taraxacum spp; Trifolium repens; Lotus corniculatus; Stellaria graminea; and Achillea millefolium would provide resources for diverse bee communities in our region.

Spatial prediction models of bee diversity can be used to identify locations where flower strips are likely to have the largest effect and can thereby provide managers with an important tool for prioritizing how funding for agri-environmental schemes such as flower strips should be allocated. Such flower strips should contain plant species that are attractive to both solitary and bumblebees, and do not need to be particularly plant species rich as long as the selected plants complement each other.

Methods
We used a paired design of study sites located in Southeastern Norway consisting of a field edge and an adjacent vegetated roadside, with or without a sown flower strip in the field. During early spring (April) 2022, we used data from Vestfold/Telemark County Governor’s office to identify sites where farmers had previously (in 2021 and sometimes also in 2020) sown a flower strip along the field edge. We included sites in our study if farmers were also planning to sow a flower strip along the same field edge in 2022. Species composition of the flower strips varied, but Phacelia tanacetifolia, Trifolium pratense and Trifolium repens were commonly used in the seed mixtures. We paired each flower strip site with a control site (i.e. study site without flower strip) with a road side of similar width, located between one and five kilometers from the flower strip site. We considered our samples as independent, i.e. that they sampled different bee communities, because distances of one km or greater are beyond the typical foraging range of most wild bees in our region. At one of the indended control sites, the farmer did sow a flower strip in 2022 and for another site we did not find a suitable control site. Our resulting study design consisted of 17 flower strip sites with a vegetated roadside and a flower strip, and 14 control sites with a vegetated roadside and without a flower strip.
At each site we sampled wild flower-visiting bees with an entomological net by walking slowly for 20 minutes along two 50m transects placed in the vegetated road- and field edge, respectively. To account for handling time, we added 30 seconds sampling time per collected specimen. For our samples to cover seasonally distinct parts of the local bee communities, we conducted three surveys during the summer of 2022: in late May (early summer), late June/early July (summer), and late July (late summer). Because of unstable weather we were only able to sample 25 of the 31 sites during the first survey. All flower-visiting bees collected were kept in 50ml falcon tubes filled with 96% ethanol, labelled according to date, collector identification, site, habitat (roadside vs. field edge), and plant species. Collected bees were identified by the lead author. Voucher specimens are stored in the entomological collections at the Norwegian Institute for Nature Research in Oslo. In July, we placed five 1m2 square vegetation plots regularly along the 50m transects with one plot per ten meters. In each 1m2 vegetation plot we recorded the occurrence of forb and shrub species in four 25 by 25 cm sub-plots. We recorded all species regardless of growth stage so that our single plant survey provided estimates of the relative frequency of plant flowering during and outside the survey period.

在农田边缘播撒花带是农业景观中保护传粉昆虫的通用手段。在资源有限的条件下最大化花带的保育效果，需要具备空间明确性的工具，以优先确定花带布设点位，并筛选出适宜混入种子混合物的植物物种。

我们在挪威31个样点开展半控制田间试验，设置有/未播撒花带的路边/农田边缘配对样地，对蜂类、植物物种及其互作关系进行采样；同时利用独居蜂物种丰富度区域空间模型，验证花带对蜂类丰富度的影响是否可通过该独居蜂丰富度模型进行预测。

研究结果显示，布设花带的样点其蜂类物种丰富度显著高于未布设花带的样点，且该提升效果在区域独居蜂丰富度模型判定为对蜂类尤为重要的区域最为显著。空间模型还表明，即便在小型景观中，不同农田边缘播撒花带的预期效果也存在显著差异。

在吸引最多蜂类物种的植物中，多数仅主要吸引熊蜂，仅少数物种同时可吸引独居蜂。综合考虑蜂类的分类多样性与植物所吸引的蜂类丰富度，我们建议种子混合物可包含以下类群：山柳菊属（Hieracium spp.）物种，如伞花山柳菊（Hieracium umbellatum）、猫耳菊（Pilosella officinarum）；蒲公英属（Taraxacum spp.）；白三叶（Trifolium repens）；百脉根（Lotus corniculatus）；禾叶繁缕（Stellaria graminea）；以及欧蓍草（Achillea millefolium），这些植物可为研究区域内多样的蜂类群落提供采食资源。

蜂类多样性空间预测模型可用于识别花带有望发挥最大保育效果的点位，从而为管理者提供重要工具，以优先分配花带等农业环境计划的资助资金。此类花带应包含同时对独居蜂与熊蜂具有吸引力的植物物种，且无需包含过多植物种类，只要所选物种间能够相互补充即可。

## 研究方法
本研究采用配对样地设计，研究点位位于挪威东南部，包含农田边缘与相邻的植被化路边，且农田内播撒或未播撒花带。2022年早春（4月），我们通过西福尔-泰勒马克郡总督办公室的资料，筛选出此前（2021年，部分样点为2020年）由农户在农田边缘播撒过花带的点位。若农户计划在2022年于同一片农田边缘播撒花带，则将该点位纳入本研究。花带的植物组成存在差异，但天蓝紫草（Phacelia tanacetifolia）、红三叶（Trifolium pratense）与白三叶（Trifolium repens）是种子混合物中常见的物种。我们为每个花带样点匹配一个对照样点（即未播撒花带的研究样点），对照样点的路边宽度相近，且与花带样点的距离介于1至5千米之间。我们认为采样相互独立，即各点位的蜂类群落互不重叠，因为1千米及以上的距离超出了本研究区域内多数野生蜂的典型觅食范围。其中1个预设对照样点的农户在2022年播撒了花带，另有1个点位未找到合适的对照样点。最终本研究的样地设计包含17个带有植被化路边与花带的花带样点，以及14个带有植被化路边但未播撒花带的对照样点。

在每个样点，我们采用昆虫网采样：分别沿设置于植被化路边与农田边缘的两条50米样带缓慢行走20分钟，采集野生访花蜂类。为补偿标本处理耗时，每采集1份标本额外增加30秒采样时间。为覆盖本地蜂类群落的季节性差异，我们于2022年夏季开展3次调查：5月末（初夏）、6月末/7月初（仲夏）以及7月末（晚夏）。受天气不稳定影响，首次调查仅完成了31个样点中的25个。所有采集到的访花蜂类均保存于装有96%乙醇的50ml离心管（falcon tube）中，并按采集日期、采集者信息、样点、生境类型（路边vs农田边缘）以及寄主植物物种进行标记。采集的蜂类均由第一作者进行鉴定，凭证标本保存于奥斯陆挪威自然研究所的昆虫馆藏中。7月时，我们沿50米样带每隔10米设置1个1平方米的方形植被样方，共设置5个。在每个1平方米的植被样方内，我们通过4个25×25厘米的亚样方记录杂类草与灌木物种的出现情况。我们记录所有物种的生长阶段，因此单次植物调查即可估算植物在调查期内与非调查期内的开花相对频率。