Industry needs matter – incorporating stakeholder interests in the selection of flower resources to support pollinators

Most pollinator policy initiatives are focused on habitat restoration and increasing the availability of floral resources, yet the choice of plant species is not always compatible with farming system cultivation and management needs. 

In this paper, we developed a framework for selecting plants to specifically meet stakeholder needs. We trialled 19 plant species and collected observational data on plant insect visitors, plant survival in the orchard environment and potential risks to crops and the environment. We used this framework to identify plants suitable to incorporate into blueberry cropping systems.

Practical implication: Our framework ensured plant choice based on informed decisions and allowed selection of two plant species that aligned well with industry needs. Different plants may be optimal for different conservation aims, hence plants selected need ideally to be evaluated for their use by the flower-visiting taxa, as well as align with industry growing practices and needs.

Methods
The data set consists of observations of flower visitors to candidate flowering plant species incorporated into a blueberry production system with the aim to: 

1) increase the activity of pollinators throughout the blueberry rows in polytunnels, and

2) support the health of managed honeybees and wild stingless bees.

19 plant species, attractive to pollinators, were establishment in the blueberry system in May 2019. The plants were first placed in blueberry blocks grown under polytunnels and thereafter transferred to open field blueberry blocks. For the 36 individual plants representing 11 plant species that survived in the blueberry system, and co-flowered with the blueberry plants, we conducted standardised observations of floral insect visitors. The aim of the standardised observations in field trials were to record the actual identity and number of visitors that visited the different plant species. The observations were conducted in polytunnels in March 2020, approximately 10 months after establishment and in open field blocks in March-May 2020. Individual plants were observed for five minutes, and all floral visitors were recorded, together with the number of open flowers. We conducted eight separate days of observations in the polytunnels and five days of observations in the open block. 

Statistics – We modelled the number of visitors to candidate plants using generalised linear mixed effect models with the glmmTMB package (Brooks et al., 2017) in R (R Core Team, 2017). Random effects reflected experimental design, with the effects of plant location nested within row, and the crossed effects of observation date. The natural logarithm of the mean number of flowers for each plant species was used as an offset. Models were specified with a negative binomial error distribution following model diagnostics implemented with the DHARMa package (Hartig & Hartig, 2017). Separate models were fitted for honeybees, stingless bees, and for all visitors combined. Due to low visitation to some plant species, only plant species receiving greater than one floral visit in aggregate were included in the models. Pairwise testing was performed using the emmeans package (Lenth & Lenth, 2018).