Pollen limitation studies are scarce in entomophilous crops, as it can be very tedious to supplement plants with hand pollination at the relevant plant scale. To overcome this, recent studies have used pollinator gradients over several fields to assess whether crops were pollen-limited. But the plant maternal resources and thus the yield potential may vary between fields of the same crop. If these resources are not properly controlled, it may affect the conclusions of the study

We investigated the relationship between yield, measured as seed number, seed mass, and seed oil content per head, and pollinator density in sunflower (Helianthus annuus L.) by using a gradient of pollinator densities set both across and within fields. We analysed the yield and the % of yield potential, the latter being assessed with pollen supplementations applied at the whole plant scale. We analysed the data both with and without random effects including year, cultivar, field, and sampling plot. We also compared the yields of open and pollen-supplemented sunflowers to sunflowers isolated under tulle bags to assess autonomous self-fertilization and self-production rates in each field.

Without random effects included in models, yields increased with pollinator density but only up to 0.3 bees per head, implying pollen limitation only below this threshold, while with random effects included, the yields were marginally pollen-limited across the whole range of pollinator densities observed (P = 0.072 and 0.037). Yet, on the other hand, there was no relationship between bee density and % of yield potential (assessed with pollen supplementation treatments), with or without random effects included, implying no pollen limitation across the complete range of bee densities observed in our study (0.1-1 bees per sunflower head). These last results are explained by variation in the yield potential between and within fields, and by the positive correlation found between yield potential and pollinator density. Insect pollination contributed on average to 53.7% and 42.8% of the seed number and mass per head, respectively, and also increased the seed oil content by 5.5% on average.

Our study showed that plant maternal resources and random effects are crucial to consider when quantifying pollen limitation using pollinator gradients and yields. Pollinator gradients on a per-flower basis are a useful tool to identify target pollinator densities to maximise yields, but should be combined with pollen supplementation treatments at the relevant scale in order to correctly assess pollen limitation.