Flying by night: Comparing nocturnal pollinator networks over time in the Colorado Rocky Mountains

Because pollen-transport networks tend to vary widely over short periods of time but remain consistent over longer periods of time, it is important to account for study length when characterizing pollen-transport network structure. The study of nocturnal pollen-transport networks independently from diurnal pollinator networks has also been emphasized, as nocturnal pollen-transport networks are an often overlooked ecosystem component. 

Here, we systematically compare the properties of moth pollen-transport pollinator networks in Colorado, USA, over the course of a season, and test for a link between specialization and pollinator morphology. We sampled moths for pollen over nine weeks in June through August 2021. We then constructed pollen-transport networks, and used mixed models to test for differences in network indices between different parts of the 2021 season; we also tested for a relationship between moth proboscis length, moth presence across seasons, and the number of flower species pollinated.

As expected, we found high rates of species turnover and plant-pollinator interactions over the summer, and we found that longer proboscis lengths – as well as length of presence over the summer – were predictive of higher generalization within individual species. However, variation in the overall structure of pollen-transport networks was relatively low over the course of the season, with linkage density and robustness being the only metrics that changed.

We suggest that even with high rates of species turnover, nocturnal moth pollen-transport networks may maintain their overall structure over a season, with variations in some metrics possibly resulting from variations in the availability of floral resources. Our study highlights the necessity of continuing to research moth pollen-transport networks over different temporal scales, as aggregating pollination data over time can obfuscate patterns in temporal variation in network structure.

Methods
Study sites

In Gunnison County, Colorado, USA, summer 2021, we sampled a total of four sites around the Rocky Mountain Biological Laboratory (RMBL) (38°57′5”N, 106°59′6”W, 2912 m.a.s.l.) (Fig. 1). This is a high-altitude montane ecosystem, with dry meadows bordering aspen forests. Evening temperatures obtained from a weather station located within 600 meters of sites 1-3 ranged from 5°C to 15°C, with occasional rainy periods. Sampling never took place on a day with more than 0.5mm mean rainfall over the sampling period. There was no significant difference in temperatures recorded over our early, middle, and late-season sampling periods; the mean temperature during sampling hours in the early season was 9°C (SD = 2.8), 8.6°C (SD = 2) in the middle season, and 10°C (SD = 2.9) in the late season. Our sampling encompassed the peak wildflower bloom and the post-snowmelt growing season; the number of overall species flowering in the area increased from approximately 60 species at the start of our sampling to 494 species at the peak of our last period of sampling (unpublished data, collected in 2021 as described in Inouye 2008). Site 1 was located on the border of a dry meadow and a forested stretch of aspen trees (2910 m.a.s.l.). Site 2 was located in a dry meadow, adjacent to mixed spruce-aspen forest (2920 m.a.s.l.). Site 3 was located in a dry meadow, adjacent to willow trees growing near a creek (2900 m.a.s.l.). Site 4 was located in a dry meadow, adjacent to beaver ponds and mixed spruce-aspen forest (3060 m.a.s.l.). Sites 1, 2, and 3 were all located within 720 meters of one another; site 4 was 1620 meters away from site 1. 

Moth collection and identification:

We used a blacklight on a white sheet to attract moths (Macgown, 2022). Sampling occurred between 21.30 and 23.30h (MDT) between 9-Jun-2021 and 9-Aug-2021 and was conducted manually, as some studies have found that manual trapping is broadly more effective than automatic trapping (Axmacher & Fiedler 2004, Tikoca et al 2016). As we observed a negligible number of moths arriving at the light between 23.00h – 24.00h in 2020 (unpublished data), we focused on sampling the early peak of moth activity in 2021. For ease of identification, we generally did not sample microlepidoptera. Site 1 was sampled nine times; sites 2 and 3 were sampled seven times each, and site 4 was sampled once, for a total of 24 sampling nights over the course of the study. We collected no more than fifteen moths per night due to processing constraints. When more than fifteen total moths arrived at the blacklight, moths were randomly sampled at regular intervals over the course of the night, to avoid biasing sampling towards the species that arrived the earliest. Site 4 was excluded from sampling after one night due to time constraints and safety concerns related to presence of mountain lions; data from site 4 was only used in cumulative species lists and calculating Jaccard indices.

During collection, moths were stored in individual containers to avoid cross-contamination, and chilled on ice prior to being brought to the lab for pollen swabbing and further processing. We measured proboscis length using electronic calipers, and recorded the sex and wing wear of all moths (wing wear rated from 1, ‘no scale damage’, to 3 ‘extensive scale loss and wear’). When we encountered a new moth species or morphotype, we collected and pinned two voucher specimens; we released identifiable specimens on-site. Individual moths were identified to species where possible, and to family when it was not possible. Morphospecies were treated as unique species in subsequent analyses.

Pollen collection and identification:

To identify moth pollen samples, we referred to an existing pollen library at RMBL, as well as one period of selective sampling from blooming flowers in mid-June at site 3. To obtain pollen samples from all captured moths, we used fuchsin jelly cubes to swab the thorax, head, unrolled proboscis, and maxillary palps, which are most likely to carry pollen. We then placed the fuchsin jelly on a glass slide and covered it with a cover slip, and sealed the edges (Beattie 1971). We used a USB microscope (magnification x400 - x1000) to count and identify the pollen. We only recorded pollen counts when we observed at least five pollen grains of the same morphotype on a given moth. We chose five pollen grains as the cutoff, in line with other studies (Banza et al 2011, Banza et al 2015, Devoto et al 2011, Forup & Memmott 2005) which use this as a threshold to reduce the probability that observed pollen grains are the result of cross-contamination from light traps and handling.