Aim: Pollination plays a crucial role in the conservation of many plant species persisting in fragmented, human-dominated landscapes. Pollinators are known to be instrumental in maintaining genetic diversity and metapopulation dynamics for many plant species and are important for providing ecological services that are essential in agricultural landscapes where populations of native plants are highly isolated. Numerous studies have explored the value of remnant native vegetation for supporting pollination services to crop species; yet the effect of mass-flowering crops on the pollinator communities and the pollination services they provide to native plant communities persisting in fragmented landscapes are less well-understood. Here, we assess the influence of the presence and phenology of a mass-flowering crop to pollinator community structure, abundance and pollen load composition in remnant vegetation in complex agricultural landscapes. 

Location: South-west Western Australia, Australia.

Methods: We recorded the composition and abundance of insect flower visitors and their pollen loads in isolated remnants of York gum-Jam woodlands adjacent to canola (insect-attracting) or wheat (non-insect-attracting) fields over two years. 

Results: All bees were much more sensitive to adjacent crop type (neighbouring canola or wheat) than non-bee pollinators. Honeybees were the most abundant pollinators in canola fields during peak flowering. Honeybee abundance increased in canola-adjacent reserves post-canola bloom, potentially indicating a movement into reserves as crop flowering waned. Native bees were the most diverse in remnant vegetation.  Pollen loads of native bees were more mixed (increased pollen richness and evenness) when sampled next to canola fields compared to wheat fields. 

Main conclusion: The availability of potential insect pollinators to remnant wildflower communities in agricultural landscapes is context dependent. Whether sampled communities were adjacent to wheat or canola in a landscape significantly impacted the abundance of potential pollinators in certain landscape elements, but not others, and the composition of pollen loads carried by these insects. Results offer novel insights about the influence of landscape context on pollinator communities and the potential pollination services available for the conservation of native plant species in highly fragmented agricultural landscapes.

Study System

This study was conducted across 235 km of the central wheatbelt of SW Western Australia (distance from north-most to southeast-most remnant) in July-September of 2015 and 2016 (Supplementary Figure 1). The Eucalypt woodlands of the Western Australian Wheatbelt account for a substantial proportion of the native vegetation in the region (about 44.6 percent) though are made up of fragmented patches dispersed within an agricultural matrix of wheat, canola, barley, oats, and lupins (DAWE, 2015). A major sub-group of this woodland type in this region is the York gum-Jam woodlands (YGJW), currently listed as a Priority III Critically Endangered Ecological Community under the EPBC Act (DAWE, 2015). This woodland supports an overstorey dominated by two main tree taxa, Eucalyptus loxophleba subsp. loxophleba(York gum) and Acacia acuminata (Jam). The understorey in these woodlands is a mosaic of largely native shrubs, perennial grasses, and annual forbs (Dwyer, et al., 2015). Annual forbs in this system are dominated by representatives of the Asteraceae, Goodeniaceae, and Araliaceae families with some common species known to require insects for pollination (Lai et al., 2015; Staples et al., 2016). The canopy trees in this system do not flower at the same time as canola and wheat, but the wildflower understorey does, with flowering starting at approximately the same time as canola (in July), extending up to two months after canola and wheat flowering stops in late August. In this study region, we selected 24 YGJW remnants (12 each year: six canola (Brassica napus), six wheat (Triticum sp.)), each greater than 0.1km2 and in intact condition (DAWE, 2015), supporting diverse annual wildflower communities. Some remnants were included in both study years, but with swapped categories between years due to crop rotations, while others were only studied for one year (see Supplementary Table 1). Across the two years, we established 324 sampling plots; with 252 sampling plots inside remnants, which were then divided into two groups: 108 core plots (>200 m from the agricultural edge) and 144 remnant edge plots. Remnant plots were divided among those adjacent to wheat and those adjacent to canola fields (<50 m from agricultural edge; Supplementary Figure 1). The remaining 72 sampling plots were established in canola fields adjacent to each woodland remnant bordered by canola (Supplementary Figure 1; Supplementary Table 1). Initial surveys of wheat fields revealing no flower visiting insects, so these were not surveyed further. 

Plot-level variables

In the 24 remnants, vegetation surveys and woodland “remnant” characteristics were recorded in 15 x 15 m plots during the months of July-September in 2015 and 2016. Measurements included: % bare soil within plot, proportion woody debris (i.e., dead logs) within each plot (from 0 to 1 – estimated as the amount of the plot containing woody debris that could be used as nesting substrate for solitary bees), % flower cover in the plot, and a categorization of the understorey community by dominant flowering plant life-history strategy (i.e. predominately native forbs, exotic forbs, shrubs or grasses). These measurements were repeated three times in each plot, one survey conducted during each flowering season. The same data were collected for remnants adjacent to canola, as well as an estimate of the canola flowering phenological stage in the relevant adjacent field, using three categories: ‘peak’, >70% inflorescences flowering; ‘past-peak’, between 40-70%; and ‘late’, end of the flowering period with <40% open flowers. Other variables included size of remnant (km²), size of the agricultural field adjacent to the study remnants (km²) and distance of each plot within the remnant to the nearest agricultural edge (m). 

Sampling of potential pollinators

Potential pollinators were collected three times from each plot across the canola flowering season (once during early canola bloom, peak canola bloom and late canola bloom). We sampled only on sunny or partially cloudy, still days in 20-minute time periods during peak insect activity times (10 am – 3 pm). The 20-minute timer was paused during insect capture and handling time. Preliminary surveys found minimal (often no) insect activity outside of this peak activity window possibly due to low temperatures (mean minimum temp of 6.1°C for July) early in the morning and late in these winter afternoons. Potential pollinators were collected only if they landed on a flower and were observed making contact with the anthers or stigmas of a flower. We were not able to directly assess whether each insect was actively pollinating during each visit, and thus note that collected insects in this study are potential pollinators only.  For conciseness, however, we refer to them as “pollinators”. Insects were captured using sterile plastic containers or bags to ensure all body pollen was accounted for and no cross contamination of pollen from different individuals occurred. All insects were identified by experts to the lowest taxonomic level possible, using collected specimens. Given the limited taxonomic work that has been done on Western Australian bees and flies, most species were only identifiable to the family or genus level.  For species whose taxonomy was unclear, we categorized into morphospecies. Due to low samples sizes (n = 36 individuals), wasps were excluded from the study. Sampling efforts in canola fields were limited to the edge of the crop fields due to access restrictions. All honeybees collected in this study were considered to be feral (unmanaged), as managed honeybee hives were not used by the canola growers in this region during the study years. This is a very sparsely populated region of Australia (Perenjori Shire has a population density of 0.07 people/km2 (ABS, 2017)) and thus we are confident we were aware of all homesteads and know that no local farms or households were keeping honeybees during the study period. Though feral honeybees are known to reach very high colony densities in Australia (Cunningham, et al., 2022).

Pollen load collection 

We recorded pollen load (the composition and number of pollen grains) found on 893 of the 923 individual pollinators collected in the 2015 season. The remaining 30 insects had no pollen on their bodies. Pollen loads were assessed by swabbing each collection bag and insect body with Fuchsin jelly, which was then melted on a glass slide and prepared as per Kearns and Inouye (1993) (see Supplementary Material Methodology 1). We also created a pollen library to help with pollen compositional assessment, using the same Fuchsin jelly approach, where a voucher specimen of every flowering plant species in each remnant during each visit. Details regarding pollen identification and counting can be found in the Supplementary Material Methodology 2.