Data from: A century of wild bee sampling: historical data and neural network analysis reveal ecological traits associated with species loss.

<b>Contemporary data (2017/2018): </b>An open area on the north side of the ESGR (GPS coordinates: 42.461808, -84.011128) was the primary site for this study as it corresponds to the location of “Evans’ Old Field”, one of the areas historically sampled for bees. The field was described by Evans as a 7.7 ha abandoned field with a mid-successional community of plants surrounded by oak-hickory woods. It is now 1.3 ha of semi-open habitat with significant encroachment of the surrounding oak-hickory woods and invasive autumn olive (<i>Elaeagnus umbellata</i> Thunb.). The site was visited every other week during the summers of 2017 and 2018 to sample bees. In 2017, the first sampling day was June 1 and the final sampling day was September 25. In 2018, the first sampling day was May 8 and the final day was October 3. We expanded sampling in 2018 to include a wider diversity of bees with narrower phenological periods.During each visit we sampled bees using three methods. First, we walked to the center of the open field and randomly selected a direction to start the first 25 meter transect. Three other 25 m transects were then established based on the first one, each at a 90-degree angle from the neighboring transect for a total of 100m sampled, with each transect segment moving away from a central location. Each transect was walked for 10 minutes each, a total of 40 minutes of sampling. We used aerial insect nets to collect bees found within 1.5m of the transect, and time was stopped for specimen processing. The host plant was recorded for all specimens captured from flowers. Flowering plants were identified to the lowest taxonomic level in the field using Newcomb’s guide and the PlantNet app, usually to species. Second, we spent 20 minutes collecting bees from plants of any species in the general vicinity of the open field. Third, to most closely match the methods used by Evans (see below), we spent 30 minutes sampling bees at each of the primary blooming plant species located in the field. Total time spent conducting this final sampling method varied based on the number of primary blooming plants at each visit, with a minimum of 30-minutes if there was only one primary plant. This sampling method was always done last, and included any plants that we collected more than one bee from that day. All bees were identified to species (or lowest possible taxonomic level) using relevant keys. All specimens collected in 2017 and 2018 are currently held in the Isaacs Lab at Michigan State University (as of 2024), and will eventually be deposited at the A.J. Cook Arthropod Collection at Michigan State University for long-term inclusion in that collection.<b>Historical data (1921-1999):</b> The University of Michigan Museum of Zoology Insect Collection (UMMZI), Ann Arbor, MI, holds over 4,000 bee specimens from the historical collections at the ESGR, and specimens were databased as part of this study. Historical data were checked for entry errors and outdated taxonomies. Specimens with questionable species determinations were re-examined and re-identified using relevant keys (see above) where possible. Bees that could not be confidently identified to the species level were excluded from the dataset, and entries that were missing the date of collection were also removed. Excluded entries accounted for less than 1% of the specimens. There were notable gaps in records at the ESGR, as there were no focused survey efforts since Evans’ last efforts in 1989, and only occasional specimen records from 1990-1999. There were no surveys and no records for the ESGR after 1999 and prior to this study in 2017/2018. All specimens from the ESGR were included in this dataset, not only those specifically collected at the Evans’ Old Field.In addition to the 4,000 plus records from the ESGR since 1921, we also include Evans’ dataset from his 1972 and 1973 collection effort. Evans’ original dataset from 1972/1973 was available through UM records. The dataset is unique compared to the records from the museum, because Evans did not always collect observed bees if he was confident in their identification (especially <i>Bombus</i> spp. and oligolectic species, e.g., <i>Andrena rudbeckiae</i> Robertson, 1891 and <i>Dufourea monardae</i> (Viereck, 1924)), and these records come only from the site now called Evans’ Old Field, whereas the exact sampling locations within the ESGR of many other specimens in the collection are not known. Therefore, his original dataset provides a more complete representation of the community he encountered at the Evans’ Old Field location.Evans describes his sampling as: “records of the dates and duration of flowering were made at frequent intervals (2-3 days every week) throughout the flowering season…Observation of visitation by bees was usually made between 9:00 am and 4:00 pm and on any given day was limited to a maximum of 30-40 minutes per flower species…no orderly system of monitoring was developed. More attention was given to abundant resources when they were being heavily visited than was paid to them near the beginning or end of their flower periods or to less frequently encountered species".<b>Please open the README file first, which has descriptions of each included data file.</b><b>当代数据（2017/2018年）：</b>本研究的核心采样点位为ESGR北侧一片开阔区域（GPS坐标：42.461808, -84.011128），该区域即“埃文斯老样地（Evans’ Old Field）”——历史上开展蜂类采样的区域之一。埃文斯曾将该样地描述为一片面积7.7公顷的撂荒地，植被处于中期演替阶段，四周被栎-山胡桃林环绕。如今该区域仅存1.3公顷的半开阔生境，周边栎-山胡桃林已大量侵入，同时伴生入侵种秋橄榄（<i>Elaeagnus umbellata</i> Thunb.）。 2017与2018年夏季，研究团队每两周对该点位开展一次蜂类采样。2017年的首次采样日期为6月1日，末次采样为9月25日；2018年首次采样为5月8日，末次采样为10月3日。2018年，我们扩大了采样范围，以覆盖物候期更窄的更多蜂类类群。 每次采样均采用三种方法：第一种为样线法：先抵达开阔样地中心，随机选取一个方向作为第一条25米样线的起始方向；随后以该样线为基准，依次设置另外三条25米样线，每条与相邻样线呈90度夹角，最终形成总长100米的采样样带，所有样线段均从中心向外延伸。每条样线耗时10分钟完成采样，总时长40分钟。采样时使用手持捕虫网收集样线两侧1.5米范围内的蜂类，处理标本时暂停计时。所有从花上采集的蜂类标本均记录其寄主植物。野外鉴定开花植物时，我们借助《纽科姆野花指南》（Newcomb’s guide）与PlantNet应用程序，通常可鉴定至最低分类阶元，多数情况下可到物种级。 第二种方法：在开阔样地周边区域，针对任意物种的植株采集20分钟蜂类样本。 第三种方法：为最大程度复刻埃文斯的采样方法（详见下文），针对样地内每一种主要开花物种，分别开展30分钟的蜂类采样。该方法的总耗时随每次采样时主要开花物种的数量变动，若仅存在1种主要开花植物，则至少耗时30分钟。该采样环节始终放在最后开展，且涵盖当日采集到1只以上蜂类的所有植株。 所有蜂类标本均借助相关分类检索表鉴定至物种级（或可达到的最低分类阶元）。截至2024年，2017与2018年采集的所有标本均存放于密歇根州立大学艾萨克斯实验室（Isaacs Lab），最终将移交至密歇根州立大学A.J.库克节肢动物标本馆（A.J. Cook Arthropod Collection）进行长期馆藏。 <b>历史数据（1921–1999年）：</b>密歇根大学动物学博物馆昆虫标本馆（UMMZI，位于安娜堡，密歇根州）馆藏了超过4000份采自ESGR的历史蜂类标本，本研究将这些标本的数据进行了数字化整理。研究人员对历史数据进行了核查，修正了录入错误并更新了过时的分类学信息。对于物种鉴定存疑的标本，尽可能借助前述分类检索表重新检视并鉴定。无法可靠鉴定至物种级的蜂类标本被排除出数据集，缺失采集日期的记录也一并移除。被排除的标本占总标本量的比例不足1%。 ESGR的标本记录存在显著缺口：自埃文斯1989年最后一次采样后，未开展过针对性的调查工作，仅在1990–1999年间有零星标本记录；1999年至2017/2018年本研究启动前，ESGR未开展任何调查，也无相关记录。本数据集纳入了所有采自ESGR的标本，而非仅限定于埃文斯老样地内采集的标本。 除1921年以来ESGR的4000余条记录外，本数据集还纳入了埃文斯1972与1973年采集工作得到的数据。埃文斯1972/1973年的原始数据集可通过密歇根大学档案获取。该数据集与博物馆馆藏记录存在显著差异：埃文斯若能可靠鉴定观测到的蜂类（尤其是熊蜂属（<i>Bombus</i>）类群以及寡食性蜂类（oligolectic species），如<i>Andrena rudbeckiae</i> Robertson, 1891与<i>Dufourea monardae</i> (Viereck, 1924)），则未必会采集这些标本；且该数据集仅涵盖如今被称为埃文斯老样地的区域，而馆藏中其他多数标本的精确采样点位已不可考。因此，埃文斯的原始数据集更完整地呈现了其在埃文斯老样地观测到的蜂类群落组成。 埃文斯对其采样方法的描述如下：“在整个开花季，以高频次（每周2–3次）记录开花植物的花期与持续时间……蜂类访花观测通常在上午9:00至下午4:00之间开展，单日针对单个开花物种的观测时长最多不超过30–40分钟……未建立标准化的监测体系。相较于开花初期、末期或较为罕见的物种，当大量蜂类访花的丰富蜜源植物时，我们会投入更多关注。” <b>请先阅读README文件，其中包含所有纳入数据集的文件说明。</b>