Plant-pollinator interactions were sampled at each plot in each meadow using 15- minute 'watches' on each of five days in each summer (mid-June to mid-August) from 2011 to 2018. Each meadow contained ten 3 x 3 m plots, spaced along two parallel transects, with five plots along each transect. The plots were spaced 15 meters apart along each transect, and transects were spaced 20 meters apart. The two transects were aligned along the major axis of each meadow, near the center of each meadow. Transects in some of the small meadows were curved slightly to avoid proximity to forest edges. Thus, 90 m2 were surveyed in each meadow regardless of meadow size.
Each meadow was visited five times approximately weekly over the course of each summer season, starting as soon as snowmelt permitted access to the meadows (typically late June), and continuing through until meadows had dried out and most flowers ceased blooming (typically early to mid-August). A ‘watch’ was conducted at each plot during each visit. Watches were conducted between 0900 and 1700 hours, on days that were sunny or partly cloudy, had little to no wind, and experienced no precipitation. At the beginning of each watch, every flowering plant in the plot was identified to species, the number of stalks (that contained at least one flower) was counted, and the numbers of flowers on each stalk were counted for each species (up to ten stalks per species). Thus, flower abundance of each species was counted or (in the cases of species with more than ten stalks) estimated based on the average number of flowers per stalk multiplied by the number of stalks present. Compound flowers and inflorescences were counted as a single flower. In 2015, a student found some potential differences in counting methodology among observers (Sanders 2015a,b,c); see supplemental documents for details. Therefore, analyses that include flower abundance should use the number of flower stalks rather than the total number of flowers per plot. Individual plants that could not be identified were photographed from outside the plot, harvested, pressed in the meadow, and transported to the lab for identification later in the day.
Each 'watch' consisted of observations for a period of 15 minutes. At each minute, the observer recorded the weather (sunny or partly cloudy and still or breezy), whether the plot was in the shade or not, the air temperature, and all visits by a flower-visitor species to each flower species. An 'interaction' consisted of an individual flower-visitor making contact with the flower; in 2016, the datasheet was revised to require observers to distinguish whether the visitor was waiting (W), on reproductive structures (P), or unsuccessful at contacting reproductive structures (U). This change resulted in a decrease in the frequency of observations of Epicauta puncticollis, a beetle that rests on flower petals but rarely pollinates. If a particular flower-visitor visited different individuals of the same flowering species, or different flower species in a given minute, each visit was recorded as a separate interaction for that sampled minute. If multiple flower-visitor individuals of the same or different flower-visitor species in the plot visited a flower in the plot during the same minute, each interaction was recorded separately. If a particular flower-visitor individual visited multiple flowers of the same flowering species during the same minute, those visits were counted and recorded as a single interaction, and the number of visits was recorded in a separate column. The timer was stopped between successive minutes to enable completion of note-taking.
Flower-visitors that could not be visibly identified in the field were captured using a net after the first visit and euthanized in the field using a small jar of ammonium carbonate and a small amount of moisture (to generate carbon dioxide). From the field they were transported to the lab, given unique ID numbers, pinned, and sent to an expert entomologist for identification (Andy Moldenke). During the capture and euthanization process, the 15-minute timer was stopped, and no observations were made. When the observer was ready to continue the watch, the timer was restarted and the 15-minute watch continued from the point where it stopped.
In the summer of 2021, bee and wildflower data was collected every other week over a six-week period by Melinda Vickers and her field assistant, Logan Kary. The data was collected according to the field methods defined by Helderop (2015) and Sanders (2015) using their datasheets and definitions. The only difference was that in 2021 only visitations by bees were recorded rather than visitations by all pollinators.
In 2021, we did surveys in 10 meadows (all Carpenter and Frissell meadows as well as Lookout Main and Lookout Bog). In each meadow, we began our survey by counting all the flowers in antithesis in each plot (10, 3x3m plots per meadow). To get a measure of relative abundance, we counted flowers by their stalks (often inflorescences) and then counted the number of flowers by stalk (see Sanders, 2015 for specific definitions of these terms that are used in the dataset).
After completing the flower survey, we conducted a bee survey according to the prior protocol used for pollinator surveys. In each plot, we conducted a 15-minute survey where we recorded the species of bees and the species of flowers that they visited. We recorded the time they entered the plot and the number of flowers they visited in a minute. Because we were unable to capture and identify each individual bee, harder to identify bees (mainly solitary ones), were only identified to genus level, if possible.
In each meadow, this process was repeated three times throughout June and July, with the exception of the Lookout Meadows where bee watches were only done two times in summer 2021.
At the end of each day, the data from the day’s watches were entered into Excel and checked for potential errors by the field crew leader. At the conclusion of the study, the entire dataset (2011-2018) was checked using specific R programs for several types of errors with help from Eugene Seo, Suzanne Remillard, Andrew Moldenke, Rebecca Hutchinson, and Julia Jones. These included: discrepancies between the interaction dataset and the flower survey dataset (for example, a flower that appeared in the interaction dataset but not the flower survey), abnormally high numbers of stalks or flowers (for example, an observer attempting to count the individual flowers of an inflorescence), and data entry errors (incorrectly entered temperature, incorrectly coded weather, etc.). These errors were corrected when possible.
For the first few years of the study, observers recorded temperature data in the field. Due to inconsistencies in these measurements, a data user should be advised to refer to official Andrews Forest climate stations (MS001) at VANMET, CENMET, or UPLMET to obtain accurate temperature and other climate data.
The study was conducted in 18 meadows located in five meadow complexes located between Lookout Mountain, Bunchgrass Meadow, and Cone Peak in the western Cascades of Oregon, within or adjacent to the HJ Andrews Experimental Forest (all meadows located near 44.2° N and 122.2° W). Study meadow size varied from 0.26 ha to 4.44 ha as of 2005. Elevation ranged from 1,308 to 1,536 m, and slope ranged from 0 to 55 percent. The climate of the region is maritime, with wet winters, and warm and dry summers. Mean annual precipitation is 2,221 mm and the mean annual temperature is 6.7°C. Mean annual snow water equivalent (at Vanilla Leaf MET station, 1,273 m) is 370 mm.
Vegetation communities in these meadows tend to be xeric or mesic (Miller and Halpern, 1998). Soils are shallow and bedrock is exposed in some locations, especially in xeric meadows. The meadows may have been maintained by Native American burning over at least the past 6,000 years (Highland, 2012). Decimation of Native Americans by the late 1700s and fire suppression since arrival of Europeans in the 1800s is associated with significant contraction of meadow habitat throughout the Cascade Range of Oregon (Miller and Halpern, 1998). In the Andrews Forest, montane meadows have contracted by 45% from 1948 to 2005 (Takaoka and Swanson, 2008, Rice, 2009, Highland, 2012).
