Flower power: Modeling floral resources of wild cherry (Prunus avium L.) for bee pollinators based on 3D data

Pollinator declines pose a threat to ecosystems and food production. Agriculture contributes to, but also suffers from, the erosion of pollination services. Our study explores the potential of trees in agricultural landscapes to support pollinators by providing floral resources. Our overarching objective is the quantification of floral resources produced by wild cherry (Prunus avium L.) that can be used by flower-visiting and pollinating insects such as bees. Using an innovative approach, we combine pollen measurements with manual counts of flowers on branches and 3D data derived from terrestrial laser scanning. This approach allows us to scale up flower numbers from branches to entire trees. The derived models for estimating the probability of flower occurrence (R² _c = 0.52, R² _m = 0.50) and the number of flowers per branch (R² _c = 0.88, R² _m = 0.84), as well as the number of flowers per tree (R² = 0.83), show good model fits with only a small set of predictors. The model fits indicate that, at the branch level, predicting flowering probability is more challenging than predicting flower abundance. We found differences in the number of flowers per branch in different crown sections, suggesting that floral resources are heterogeneously distributed. Furthermore, we demonstrate that the number of flowers per tree increases exponentially with tree dimension (stem diameter, crown volume). Therefore, large trees provide disproportionately more floral resources than small trees and are particularly worthy of conservation efforts. For example, our models estimate that a single tree with a stem diameter of 25 cm carries 195,535 flowers (95% CI: 159,991-237,318), thus providing about 57 cm³ (95% CI: 32-88 cm³) of pollen and producing 170 g (95% CI: 48-345 g) nectar sugar per 24 h. This amount of pollen is sufficient to rear, for example, 5202 larvae (95% CI: 2886-8022) of Lasioglossum laticeps, a common and generalist sweat bee of cherry trees. In contrast, a smaller tree with a stem diameter of 10 cm provides only 8% of these resources. In conclusion, we demonstrate how our results contribute to the broader single-large-or-several-small debate in nature conservation by highlighting the value of large trees. Additionally, we show how information gathered at the branch level may be nondestructively upscaled to entire trees.