Bees increase crop yield in an alleged pollinator-independent almond variety

Wild pollinators are declining and the number of managed honey bee colonies is growing slower than agricultural demands for pollination. Because of these contrasting trends in pollinator demand and availability, breeding programs for many pollinator-dependent crops have focused on reducing the need for pollinators. Although numerous crop varieties are now available in the market with the label of pollinator-independent, the real dependence of these varieties on pollinators is mostly unknown. We evaluated the hypothesis of pollinator independence in the Independence almond variety, the fastest growing variety in California that is the main almond production region in the world. In this presumed pollinator-independent variety, we measured the effect of honey bees on fruit set, yield, and kernel nutritional quality at tree level. Fruit set was 60% higher in bee-pollinated than bee-isolated trees, which translated into a 20% increase in kernel yield. Despite its effect on almond production, there was no evidence that bee visitation affected almond nutritional quality. Based on these results, we recommend the use of bees, whether they are wild or managed, to maximize yield even in self-fertile almond varieties.


Supplementary S2. Quantification of fatty acids
Since the main nutritional component in almond are fats 1 , we analyzed the fatty acid portion of the almond kernel. Particularly, we estimated the oleic and linoleic acid content, and the oleic to linoleic ratio. Samples from each tree were processed, and their fat was extracted following a modified Folch 2 technique as follows: 1. Extraction. Almonds randomly collected from sampled trees (see M&M "Kernel nutritional quality" in main text) were finely crushed. Then, we weighed 1 ± 0.1 g of the finely crushed almonds and added 4 ml of the extraction solution (chloroform: methanol 2:1). Samples were stirred in a vortex and let stand for 30 min; then they were sonicated 60 minutes and filtered. The filtrate was incubated in an over at 44 ± 2 ° C for 24 hours. The crystalized fat was re-suspended in 3 ml of pure chloroform (purity ˃ 98%). The resulting solution was then filtered using a syringe with a 0.45 µm filter and poured into a 4 ml vial. After the chloroform had evaporated, the vials were closed and stored at 4 ºC. 2. Esterification of fatty acids. One milliliter of acid methanol (1% HCl in methanol) was added, and then the vials were taken to the stove at 44 ± 0.1ºC for 24 hours. The next day, we added a 1ml of n-hexane (purity ˃ = 98%) and shacked the sample in a vortex. In a 2 ml vial, we added to the sample 950 µl of dichloromethane (purity 99.9%) and 50 µl of esterified hexane (upper phase of the 4 ml vial). The sample was then run in a gas chromatograph (HP 6890) with a FID detector, using a column Thermo 7HG-G033-10 FAME ZB and the following chromatographic conditions: (i) a constant flow of 2 ml/min, (ii) an injection volume of 1 μl, and (iii) a temperature of the detector of 275°C. The standard used was a mixture of 37 fatty acid esters (SIGMA®) Accustandard brand. 3. Quantification. Presence and retention times of 37 fatty acids were verified using the FAME (SIGMA®) standard run along each batch. To quantify the fatty acids in each sample, we located the peaks corresponding to each fatty acid according to their retention times and recorded the area of each peak. Values were reported as a percentage, so that the total peak area across all fatty acids for each sample added 100%.

Supplementary S3. Fruit components
To estimate kernel production at tree level, we multiplied the total fruit weight produced per tree by the weight proportion that the kernel represented in the samples (i.e., total fruit weight x kernel weight proportion). To this end, we randomly sampled 70 fruits from each experimental tree from which we measured pericarp, endocarp and kernel weight separately. We examined components of fruit weight of a total of 2,100 fruits (70 fruits/tree x 10 trees/treatment x 3 treatment).
We evaluated the effects of the pollination treatment (Isolation, Open, and Control) on pericarp, endocarp and kernel weight with general linear mixed-effects models. Data analysis was carried out using the lme function from the nlme package 3 4 of the R software 5 . Because of the response variables were continuous (i.e.,weight), we assumed a Gaussian error distribution in the three models. In all cases, the pollination treatment was included in the model as a fixed effect and plot and tree within plots as random effects, allowing the intercept to vary among plots and trees.
Overall fruit weight differences were not identical/equivalent for each constituting part. First, pericarps from isolated trees were ~18 % heavier than those from bee-pollinated trees. The pericarp of fruits from the isolated trees weighted 2.62 ± 0.03g, while pericarps from open and mesh-control trees weighted 2.17 ± 0.03g and 2.14 ± 0.03g, respectively ( Figure  S3). Second, endocarps from isolated trees were ~12% heavier than those from trees pollinated by bees. The fruit endocarp from isolated trees weighted 0.61 ± 0.008g, while endocarps from open and mesh-control weighted 0.54 ± 0.007g and 0.53 ± 0.005g, respectively ( Figure S3). Third, kernels produced by isolated trees were 8% heavier than those by trees open to bee pollination. Kernels from isolated trees weighted 1.50 ± 0.02g, while kernels from open and shaded trees weighted 1.40 ± 0.01g and 1.38 ± 0.01g, respectively ( Figure S3). Thus, in relative terms, differences in kernel weight between bee and non-bee visited trees were much smaller than the differences found between the other fruit components.
As we showed here, the fruit weight proportion represented by the kernel is different in isolated trees than in those pollinated by bees. In isolated trees, the weight of the kernel represented, on average (± SE), 31 ± 0.2% of the total fruit weight, while in open and control trees the weight of the kernels represented in both cases 34 ± 0.2%. For this reason, to estimate kernel production at the tree level we multiplied total fruit weight per tree times 0.31 for isolated trees, and times 0.34 for open and control trees. Figure S3. Effects of pollination treatments (i.e., isolation, open and control) on the weight of the different parts of the almond fruit (i.e., pericarp, endocarp, and kernel). Thick bars represent mean values, while the thin black bars two standard errors.

Supplementary S4. Yield profits.
To calculate the profit associated with almond production, we first categorized all sampled kernels by class size (also called "ounce count"). This classification reflects the number of kernels per ounce, where fewer kernels per once mean higher quality (larger) kernels. Second, we estimated the proportion of kernels in each class size or marketable size category (Table S4) for each tree and multiplied these proportions by total kernel production per tree. This allowed us to estimate weight of kernels in each size category produced by a tree. Third, the profit per tree for each treatment was translated into US dollars by multiplying the estimated weight of kernels in each kernel size category by the 2018 Harris Woolf's pricing list of the Independence variety for each kernel size category, and then adding these values up (see Table S4). Finally, we calculated profit at the plot level (per ha) by multiplying the average profit per tree times the average number of trees planted per ha (i.e., ~336 trees/ha). The rental cost of honey bee colony was included in the final calculation of the profit per ha.
Kernel production quality in terms of size did not differ strongly between non-bee and bee-pollinated trees. While non-bee pollinated trees produced 94% of the kernels from the first three sizes category (i.e., high quality), bee pollinated trees produced 83%. After accounting for the differential price related to kernel size category, the revenue per tree was ~20% lower in isolated trees than in those trees pollinated by bees (Table S4). Non-beepollinated trees generated a revenue $25.04 USD, while bee-pollinated trees ~ $30 USD. Assuming 336 trees . ha -1 at the field level, young (~ 4 years) non-bee plantations would produce ~$8.400 USD . ha -1 , while bee-pollinated plantations would produce ~$10.000 USD . ha -1 .
Adding the cost of five colonies per ha (on average, 180 USD per colony), the costbenefit balance is still positive, even in young, low-productive trees like we studied here. Also, although here we could not estimate the optimal range of bee visits needed to maximize yield, two colonies per acre, as used in self-incompatible varieties 6 , probably provide more bees visits than needed to maximize yields in this self-compatible almond variety. Table S4. Production quality classification and profit from the tree and plantation levels for each pollination treatment. "Avg." denotes averages; "Class size" indicates the number of kernels per ounce (i.e. 28.38 gr); "Price" indicates the USD price for one kilogram (kg) of almond kernels using the Independence variety pricing list from Harris Woolf 2017. Production was calculated in kilograms (kg), and profits are in USD. Totals per ha were calculated assuming a tree density of 336 trees . ha -1 . In this study, experimental trees were in their fourth year of production.