Pollination services are affected by landscape context, farming management, and pollinator community structure, all of which impact flower visitation rates, pollen deposition and final production. We studied these processes in Argentina for Highbush Blueberry crops which depend on pollinators to produce marketable yields. 

We studied how land cover and honeybee stocking influence the abundance of wild and managed pollinators in blueberry crops, using structural equation modeling to disentangle the cascading effects through which pollinators contribute to blueberry fruit number, size, nutritional content and overall yield. 

All pollinator functional groups responded to landscape changes at a spatial scale under 1000 m, and the significance or direction of the effects were modulated by the field-level deployment of honeybee hives. 

Fruit diameter increased with pollen deposited, but decreased with honeybee abundance, which, had indirect effects on fruit acidity and sugar content. Honeybees had a positive effect on the number of fruit produced by the plants and also benefited the overall yield (kg plant^-1) through independent effects on both the quality and quantity components of fruit production.

Synthesis and applications:

Deployment of beehives in blueberry fields can buffer, but not compensate for the negative effects on honeybee abundance produced by surrounding large scale none-flowering crops. Such compensation would require high-quality beehives by monitoring their health and strength. The contribution of honeybees to crop production is not equal across production metrics. That is, higher abundance of honeybees increases the number of berries produced, but at the cost of smaller and more acidic fruits, potentially reducing market value. Growers must consider this trade-off between fruit quantity and quality when actively managing honeybee abundance.

Sampling design and pollinator observation: 

In total we sampled pollinators on nine farms, eight in 2020 and six in 2021 —five of these farms were sampled in both years. Five farms had honeybee hives to enhance crop pollination while four rely on wild pollinators or feral honeybees (Appendix 1 Figure S1a). On each farm, we established an edge area (25 m from the field margin) and an inner area (50 m from the field margin), each comprising two 70 m transects running parallel to the field margin and separated by eight meters. On each of the four transects we selected 25 plants in full flower on which we conducted three 30-second pollinator counts, one in the morning, one at midday, and one in the afternoon. All sampling took place between 10:00 and 17:00 on sunny days with low wind and a temperature above 15 °C. We recorded the abundance of honeybees, small wild bees —any bee smaller than a honeybee — hoverflies (Syrphidae), and butterflies. Hummingbirds and large wild bees (bumblebees Bombus spp. and carpenter bees Xylocopa spp., which are all larger than honeybees) showed behaviors that made them difficult to record as they fled when surveyors approached the plants. Therefore, after every 25 plant pollinator counts, we walked the same transect to record the abundance of hummingbirds and large wild bees for five minutes. This sampling protocol was repeated twice for each farm in each year (2020= 56 hours, 2021= 42 hours of pollinator counts).

Landscape classification:

All land use classifications were carried out using the interactive Google Earth Engine application ‘Your Maps Your Way’ (YMYW) (Morton & Schmucki, 2023). A detailed explanation of the classification process can be found in Appendix 1, Table S1. In summary, we classified the different land covers into three broad categories: (i) natural cover (natural primary forest, secondary forest, and shrubland), (ii) flowering crops (citrus and blueberry), and (iii) large-scale crops (sugarcane and soybean crops). We grouped flowering crops together because blueberries occupy a small area compared to citrus, although both are perennial and provide comparable resources (Emerald bloom in August and citrus in September). We used the terra (Hijmans 2022) and sf (Pebesma, 2018) R packages to define the area of the three land use classes within concentric circles with a radius of 200 to 4000 m around each site, progressively increasing the radius by 200 m (Appendix 1 Figure S1a). This was done to identify the spatial scale of pollinator responses to land use.

Cascading effects: from pollinators to production

Pollinator counts:

During the 2021 flowering season (July – August), we used a subset of the nine farms (N = 3) to examine the effects of pollinators on blueberry pollen deposition and productivity metrics. On these farms growers manage pollination services using honeybee hives. On each farm, we selected five to six plots (Emerald cultivar, plot size 1.31 ± 0.09 ha) distributed from the edge of the farm to the interior to capture the potential variation in underlying pollination services, soil condition, irrigation system and plant age (Appendix 1 Figure S1b). The average distance between plots was 384.4 ± 197.5 m. In each plot we randomly selected five plants and conducted a 5-minute pollinator count per plant under the same weather conditions as mentioned above (26.6 ± 2.8 plants per farm). Previously, we estimated the floral display size per plant by combining the flowering percentage of the plant, the total fruits produced (see production metrics section) and the fruit set of the Emerald cultivar at each farm (see the detailed procedure in Appendix 1 Table S3). We recordedthe abundance of honeybees and small wild pollinators. We also used the plant as a central point to record the abundance of hummingbirds and large wild bees in an area of 12 m² (area occupied by 14 blueberry plants). The density of honeybee hives within 200 m (12.5 ha) of each sampled plot was recorded (Appendix 1 Figure S1b). Each plant was sampled twice during the ~30 day flowering period (13.3 hours of pollinator observation).

Pollen deposition:

After flower anthesis, we collected three styles per plant from senescent flowers (Ntotal = 240). These were placed on a microscope slide after a transversal cut of the style at the stigma height and stained with Alexander’s solution (Alexander, 1969). We then counted the number of pollen grains as a measure of the stigmatic pollen load.

Production metrics:

 We considered the following metrics to assess the quality and quantity of blueberry production: 1) We visited each sampled plant and counted the total number of fruits on two randomly selected primary branches. The product of the average number of fruits per primary branch and the number of primary branches was used as an estimate of the total number of fruits per plant; 2) We randomly selected ten mature fruits per sampled plant and measured their equatorial diameter; 3) Plant yield (kg plant^-1) was the product of average fruit weight (Nfruits per plant = 10) and the total number of fruit produced; 4-7) Nutritional content based on a 30g fruit sample in the form of degrees Brix (measure of sugar content), total acidity, and concentration of anthocyanins (Appendix 1 in Table S4 for the laboratory protocol).