Pollen supplementation and hive olfactory conditioning as strategies to improve apple orchard productivity

This thesis evaluated two approaches to optimize pollination in three commercial ‘Cripp’s Pink’ apple orchards in southern Uruguay: (i) hive olfactory conditioning using floral volatile mimics and (ii) supplemental pollen application, tested individually and in combination against conventional hive management as a control. The study followed a randomized complete block design (RCBD) with georeferenced sampling grids of 100 trees per experimental unit. Measurements included bee activity at hive entrances and on the crop, fruit set, crop load, yield, seed number, and fruit size. Data were analyzed using generalized linear mixed models and post hoc contrasts, while fruit size categories were further assessed using a multinomial model. The combined treatment increased centre and fruit set by 35.3% and 32.1% relative to the control, respectively, and achieved the highest crop load (121 vs. 59 fruits plant⁻¹) and yield (18.65 vs. 9.2 kg plant⁻¹). Seed number increased in the pollen (+14.5%) and combined (+12%) treatments and was positively associated with equatorial diameter. Fruit size distribution depended on treatment (χ²(9) = 57.38; p < 0.001), and the multinomial model confirmed a higher probability of favourable commercial categories for the combined treatment (Sidak, p < 0.05). Taken together, increased hive-entrance activity, greater seed formation, changes in fruit size distribution, and particularly the increase in crop load provide convergent evidence supporting the effectiveness of the evaluated management strategies. Incremental cash-flow analysis showed that the combined treatment produced the highest economic benefit (ΔP = +US$7,651 ha⁻¹ relative to the control), while the pollen treatment was also profitable (ΔP = +US$3,074 ha⁻¹). The conditioning treatment produced a smaller positive margin that was sensitive to the cost of the service. Geostatistical analysis, including global Moran’s I, fitted semivariograms, and universal kriging with distance to the hives as a covariate, revealed spatial autocorrelation and heterogeneous crop-load patterns within experimental units.