We used a full-factorial experimental design to test for the effects of pollen limitation on fruit production and foliage variables of whole trees experiencing four resource treatments: normal water and nutrients; reduced water/normal nutrients; no nutrients/normal water; and reduced water and no nutrients. In each of these resource input combinations, we applied three pollination treatments: supplemental hand pollination to maximise cross-pollination; open-pollination with flowers exposed to bees freely foraging in the field; and pollinator exclusion, accomplished by caging trees during flowering. The 12 treatment combinations were randomly assigned to individual trees and replicated five times in adjacent rows .Hand-pollination was carried out from 20 to 28 February using Padre pollen that had been harvested before bud opening and stored at 20 °C to maintain viability. Prior to application, pollen was thawed and used immediately to ensure viability. We hand-pollinated all open flowers using small brushes every 2–3 days until about 90% of all buds had opened. The last 5–10% of flowers that opened late in the blooming season were frequently characterised by deformed or missing female or male parts. For the pollinator exclusion treatment, we covered individual trees from shortly before blooming started in February to the end of bloom in early March with 1.5 m² 9 2-m tall cages constructed of aluminium tubing and cloth with a mesh size of 0.8–1.0 mm.
To test whether wind could carry pollen grains through the mesh, we conducted the following experiment. An almond branch with more than 50 flowers whose anthers were dehiscing was held between an electric fan and a new, square plastic plant pot unused cage free from pollen grain contamination. Inside the cage four microscope slides were placed at the same height as the flowers, to intercept any pollen grains that might have passed through the mesh. No pollen packets or single pollen grains of almond could be detected with light microscopy on the microscope slides, although using the same technique without a cage many pollen grains were caught. Cages were removed after blooming was completed, just before trees began to develop leaves. In winter trees were not irrigated and fertilised. The experimental water and nutrient treatments were conducted from January to August 2008. The following nutrients were applied every month by hand when irrigated: 521.6 g nitrate, 344.7 g potassium, 244.9 g sulphur, 158.8 g calcium, 158.8 g phosphorus, 54.4 g magnesium, 27.22 g boron, 27.22 g iron, 27.22 g manganese, and various micronutrients including zinc, cobalt, molybdenum . No nutrients were applied to trees in the no nutrient treatment. Water reduction of the typical irrigated volume for this region and age of the trees was accomplished by manipulating the irrigation system of tubing and emitters at each tree. For the water reduction regime, three out of the four emitters at each tree were closed, reducing water to 27 l every third day. The fungicide Rovral was applied at the rate of 0.0844 g m 2 before rain during blooming to avoid fruit fungal infections.
To quantify fruit set at different developmental stages, we counted the total number of withered flowers on each main branch of each experimental tree from 28 February to 10 March, and we then counted developing fruits four times every 3–4 weeks . On 2 July, we harvested and counted all fruits per whole tree for the last time and then kept 48 fruits per tree in the lab for further measurements. Fruits were randomly selected from the main branches . Freshly harvested fruits were dried on the ground for 7 days while protected from bird and mammal predation with metal cages. After fruit drying, the hulls were removed and shells cracked. We characterised kernel quality by counting the number of unfilled, single and double kernels and the number of kernels damaged by arthropod pests or fungal and bacterial diseases. We measured the length and weight of each of the 48 kernels per tree. On the same dates as developing fruits were counted, we counted the number of leaves, starting at the tip of the main branches for 20 cm and noted the length and colour of ten randomly selected leaves per main branch of each tree. Leaf loss was calculated as the proportion of leaves that dropped between full development of the leaves and fruit harvest .The effect of the treatments on the following response variables were analysed: fruit set and its decrease over time , estimated total number of harvested kernels, mean kernel weight based on the 48 kernels per experimental tree harvested for detailed measurements, and estimated total yield per tree at harvest . To quantify the vegetative response to treatments, we also analysed effects on the number of leaves, proportion of leaves lost from 4 weeks after blooming until harvest, and the proportion of yellowing leaves.
Fruit set over time was modelled using generalised linear mixed models with a binomial distribution and a logit link. We accounted for non-independence of multiple measurements per tree and for extra-binomial variance by including tree and observation, respectively, as a random factor in analyses. Total number of harvested kernels, mean kernel weight and yield were analysed for differences among pollination and resource treatments using generalised linear models . The number of harvested kernels and yield were lntransformed to reduce variance heterogeneity. For analyses of number of kernels and yield, the number of flowers was included as a covariate in the models, since this is a pre-treatment variable that varies from tree to tree . The ln-transformed number of flowers was centred on its mean to make model interpretation easier. For analysis of mean kernel weight, the number of harvested kernels was included as a covariate. Treatment effects on number of leaves, the proportion of leaves lost and the proportion of yellow leaves were analysed using GLM. Average number of leaves per branch was analysed using a GLM for normal data, with the response variable untransformed. Leaf loss was analysed with a GLM for binomial data as a proportional variable. A quasi-binomial GLM was used to model the tree-level leaf colour outcome, identified as the most frequent leaf colour recorded on the tree, with a binary variable . We removed interactions that did not contribute at least marginally to the model . Non-significant main effects were retained. For individual variables, F and P-values in the text are from comparisons between the model with all main effects and significant interactions and the model with the tested variable dropped. All analyses were performed using R, version 2.8.1 for Windows . Mixed models were fit using lmer .Our experiment shows that pollination strongly limits almond fruit set and yield and therefore supports general expectations and previous results of high pollinator dependency in almond . The strong pollination effect on yield even in conditions of reduced water input and nutrient reduction was in contrast to descriptions of California almond production as dependent on high water and nutrient inputs . The negative effects of water reduction on yield, with only marginal negative effects on fruit set and mean kernel weight and no detectable effect on the number of kernels, in this study is supported by previous studies that showed negative effects of water stress on yield , but not on bud development, 25 liter square pot fruit abortion and kernel weight . Surprisingly, the initial benefit of pollination on yield components was not eliminated by reduced water and was not offset by the negative relationship between number and weight of kernels. Although leaf water potential was not measured in this study, as in other work , water stress was indicated as increased leaf loss occurring in the reduced water treatment. Such leaf loss is often observed in water-limited almond trees . The strong effect of reduced water on leaf loss, its marginal effect on mean kernel weight and the increased number of yellowed leaves in open- and hand-pollinated trees with reduced water indicate that when under water stress, almond trees may allocate resources selectively to maintain kernel quantity while reducing kernel quality and delivery of resources to leaves. The lack of any direct significant effects of the cut-off of nutrients on fruit set, yield or leaf loss suggests that the young trees may have already accumulated sufficient nutrients for fruit maturation from the previous summer’s nutrient applications. Nevertheless, the significantly higher proportion of yellowed leaves at harvest on trees receiving no nutrients and reduced water, and the significant interaction between the water and nutrient treatments on leaf colour indicate that the trees were stressed in this treatment combination, especially when pollination took place.
Trees from which pollinators were excluded were characterised by canopies consisting of dense, large and dark-green leaves, in contrast to hand-pollinated trees characterised by small, yellow-green leaves. These differences in foliage indicate that excess nutrients beyond those needed for nut production in the pollinator-excluded trees were used for canopy development. Thus, the positive effect of pollination on fruit production comes at the expense of vegetative performance features and may have long-term consequences for the tree. We found a significant interaction of pollination and irrigation on yield resulting from decreased yield in hand- and open-pollinated treatments receiving reduced water, but no effect of reduced water on yield in the pollinator exclusion treatment, indicating a threshold of pollination is needed before the negative relationship between pollination quantity and water reduction on yield manifests itself. Two other studies analysed the interactions between pollination and plant resources on fruit set in woody plants . Niesenbaum focused, in two consecutive years, on a dioecious, understorey forest shrub whose reproduction was highly limited by light, but not by pollination, with no interaction effect between pollination and light. In contrast, Groeneveld et al. manipulated pollination, light, nutrient and water input and tested for the single and interaction effects of these variables on fruit set and number of harvested cacao pods after 1 year. They found that shade increased the number of aborted fruits, and the interaction of hand-pollination with shade, as well as the interaction of hand pollination with nutrients, reduced the number of fruit abortions, but the interaction effects were not translated to losses or increases in fruit set or yield found in our study. To our knowledge, the present study is the first in which significant interactions between pollination and plant resources on fruit set and yield were found, highlighting the importance of studying pollination and plant resources in a full factorial design to understand their single and combined effects on plant performance in general and crop production in particular. Almond yield was extremely low when pollinators were excluded, although these trees produced large kernels, while yield of hand-pollinated trees were high with small kernels. The kernel size in the different pollination treatments is likely caused by resource allocation and availability rather than pollination quality. In the pollinator exclusion treatment, kernels are assumed to result from self-pollination with low quality and quantity pollen. These results are contrary to studies showing that fruit or seed size and weight are often positively related to pollination quality and quantity . It also indicates that intensive pollination management, such as simulated by our hand-pollination treatment, can result in low kernel quality . Future experiments conducted over consecutive years are needed, particularly because high fruit set in year one resulting from supplemental pollination in the previous year may impose limits on reproduction in subsequent years . We found that foliage was reduced by water stress and indirectly by pollination in our 1-year study, but this may influence fruit set in the following year because the number and size of leaves influences rates of photosynthesis and hence resources available to develop new flowers . Further, fruit load may be more strongly determined by the stress history of the trees rather than the current year’s irrigation treatments . Although the need to study pollination and resource limitation for several years in perennial plants is evident, the pollinator-dependent yield response determined with and without resource limitation of a single year can help growers to make ad-hoc decisions in years of pollinator and/or water shortages. Our results suggest that for almond, pollination of the crop should be a high priority, but that other resources must be concurrently monitored and managed because of their well known effects and potential interactions that can influence overall plant performance.