Pollinator ecological traits mediate the loss of pollination services with agricultural intensification

Background/Question/Methods

Research on the functional consequences of biodiversity loss is dominated by small-scale experimental studies that assume random extinction patterns. In nature, however, extinction is generally a nonrandom process with risk determined by ecological traits such as rarity or body size. Here, we present data on crop pollination services provided by native, wild bees to explore the role of species traits in determining how rapidly pollination declines with increasing agricultural intensification. Analyses were conducted in parallel for four crop systems (watermelon, cranberry, blueberry, and tomato) located within the same geographical region (mid-Atlantic USA) but pollinated by distinct bee communities. We first determined the traits associated with bee species that provide high levels of pollination (effect traits). Next, we assessed which traits predict species’ sensitivity to agricultural intensification (response traits). Lastly, we used simulations of species loss order with respect to response traits, to predict the effect of land use intensification on pollination function under different extinction scenarios.

Results/Conclusions

For all four crops, the most important effect trait is body size, with bigger bees providing more function per capita (GLMM all p-values < 0.01). In contrast, the response traits associated with sensitivity to land-use change varied across crop systems. For example, in the watermelon system, body size is negatively correlated with agricultural intensification (p < 0.01), while in cranberry it is positively correlated (p = 0.02). We found additional ecological traits that were negatively affected by agricultural intensity in particular systems; for example, diet specialization in blueberry (p = 0.01) and below-ground nesting in watermelon (p = 0.03). Hence response and effect traits are positively, negatively or not correlated depending on the crop system studied. Our simulations confirm that removing large-bodied bees first in watermelon erode function faster than expected under random losses. However, when using the same approach to remove specialist species first in blueberry, function is lost slower than expected under random removals. Thus we show that while understanding the correlation between response and effect traits is key to predicting effects on pollination services, there is as yet little generality as to which response and effect traits drive the biodiversity-ecosystem functioning relationship.