To track the progression of the flowering period, the dates were converted into Julian dates and grouped so that the observations were either during the “Early”, “Middle”, or “Late” season. The early dates (July 9, 13, 16) would represent the scenario if climate change shifted C. pitcheri flowering season earlier than current or historical flowering dates. Similarly, later dates (July 27, 29, August 4, 2) represent what would happen if C. pitcher responded to increased temperatures with delayed flowering times. The middle season would represent the scenario that C. pitcheri did not shift its flowering period in response to climate change. For my observations of individual plants, a plant was identified as an early flowering plant, if the terminal head had past and there were several secondary heads. A middle flowering plant either had to have had a past terminal and one or two secondary heads open. Plants identified as late flowering were still in bud at the start of the observation or the terminal was still open. Multi-stemmed plants were not included in this study because determination of the terminal head is difficult to distinguish.The data were collected at Sturgeon Bay (45.72oN, -85.00oW) in the Wilderness State Park in Emmet County, northwest lower Michigan (Fig. 1a, b), 179 m above sea level. Fieldwork was conducted July 9- August 4, 2016 during the main flowering season of C. pitcheri (late June to early August, Loveless 1984). I located subpopulations of C. pitcheri in 10 different locations (Fig. 1c) along the fore dunes of Sturgeon Bay. At each site I selected three plants that represented either Early, Middle, or Late flowering plants. For every location, I assigned a specific number to the plant to help identify which stage it represented and which group I surveyed it from (i.e., 10L means a late flowering plant at site 10).

Visitor Observations

For 5 weeks, I surveyed a total of 30 individual Cirsium pitcheri (n=30) twice a week and monitored the visiting insects. The insect was identified as a pollinator if it was seen actively foraging for pollen or nectar from the open inflorescences. If an insect visited and could not be identified during an observation, an aspirator or a kill jar was used to take it back to the lab for further investigation. I used three stopwatches at one time to track how long an individual stayed on one head to record the times of multiple insects visiting. The close proximity of plants were intentionally planned to make these observations as accurate as possible. I stood far enough so that I could see all three plants at once but close enough to notice where an insect landed on a plant. I noted whether the pollinator left to go to another C.pitcheri (Outcross), and or, stayed and visited the other open heads visited (Self). The observations of each plant took 10 min to complete. I rotated the times that I surveyed the groups on a given day to reduce bias. In addition, the number of open heads were recorded to test whether the amount of floral display increased the attractiveness of C.pitcheri to pollinators. It was recorded whether or not an individual pollinator visited within the heads of one C. pitcheri (S = Self) and or among the population (O = Outcross).

Statistical Analysis

A Chi-square test was used to test differences in variables in the flowering seasons. Specifically, I looked at the number of pollinator visits, the amount of outcross and selfing behavior performed, and length of time in seconds spent on a head. There were unequal amount of dates within the flowering seasons. Thus, the expected values were calculated by the taking the sum of the observed and dividing it by the proportion of dates for that given flowering season. A one-way ANOVA in IBM SPSS Desktop Statistics 21 was used to test differences in mean pollinator visit time. Since the times were not normally distributed, the times was transformed.