flwr_visit_2019.csv

To determine whether the floral visitor community and the behavior of each floral visitor differed among cucurbit species, we filmed 142 flowers from 100 plants: in total <em>C. sororia </em>pistillate-22, staminate-46; <em>C. argyrosperma </em>pistillate-9, staminate-29; <em>C. moschata </em>pistillate-12, staminate-24. Each day, four to eight flowers were filmed using HERO5, GoPro cameras (San Mateo, CA, USA) from different plants. Most flowers were filmed for the entire anthesis which varied depending on weather (sometimes rain affected recording times) and cucurbit species (average of 2.7 - 3.56 decimal hour per staminate and pistillate flowers of each species). We selected plants separated from each other by at least five meters. Only floral visitors that entered the corolla tube were counted. For each floral visitor, we recorded morphospecies, time of arrival, duration of the visit, reason or motive for entering the flower, and contact with reproductive parts. The motive for entering the flower was separated into four categories: nectar collection, pollen collection, nectar + pollen collection and nothing. To calculate the visitation rate for each floral visitor group we divided the number of visits by the total duration of the video recording (decimal time = time x 1440) for each flower. For taxonomic identification of floral visitors, we collected specimens of each floral visitor observed in cucurbit flowers for 15-minute time intervals at three times; 07:00, 09:00 and 10:30 hours. We used taxonomical guides for insect identification (Michener et al<em>.</em>, 1994; Ayala and Griswold, 2012). We limited our analysis to the floral visitors that entered more than three flowers and/or collected flower rewards from the flower and not just a consequential passive visit. Floral visitors were separated in two categories; primary floral visitors, six morphospecies groups that visited more than 30 flowers, and secondary floral visitors, a combination of all floral visitor morphospecies that visited less than 30 flowers (average number of flowers visited = 10, SD = 7). <br> To analyze the visitation and behavior of floral visitors in staminate and pistillate flowers of cucurbit species, we conducted generalized linear models using the GLIMMIX procedure in SAS version 9.4 (SAS Institute Inc, 2014). All models included the interaction between cucurbit species, flower sex and floral visitor type as the predictive variables and site as random variable. To test if visitation was affected by the predictive variables, we used a binomial model (logit link) and a gamma model (log link) (Bolker, 2021). The binomial model tested the probability that the flower was visited (yes or no), while the gamma distributed model tested differences in visitation rate (amount of visits/duration of video in decimal time). To test whether the predictor variables affect the probability of contact of floral visitors with floral reproductive parts (yes/no) we used a binomial distributed model. To test whether anthesis time affects visitation rate of each floral visitor, we used time of visit, assigned to one-hour time range blocks from 0600h-1200h, as the nesting factor for the predictor variables assuming a negative binomial distribution to control for zero inflated overdispersion of data. Finally, to test whether the visitation rate and the visit duration were influenced by the motive of visit, we used the interaction between cucurbit species, flower sex, floral visitor type and motive as the predictive variables. For all analysis, we used type III Wald chi-square tests and specified the ILINK option in the LS-MEANS statement to obtain the least square means back-transformed to the original scale.