A vegetation-based indicator to assess the pollinationvalue of field margin flora

Conservation of pollinator abundance and diversity is an important issue because it contributes to maintaining a diverse community of plant species in agroecosystems. The presence of semi-natural areas favorable to pollination is a key factor for achieving this objective of sustainability. Sowing mixtures of dicotyledonous plants that are rich in pollen and nectar as flower strips along field margins is an efficient solution to attract pollinators and to support their foraging activity on arable land. The enhancement of agroecosystems requires operational methods that make it possible to assess the impact of existing and sown semi-natural areas on pollination. We developed here a new predictive indicator that can be used at the field margin and floral levels, which predicts the pollination value of floral diversity and abundance of field margins on arable land. We based the predictive indicator on decision trees using "if-then" linguistic rules because of the lack of sufficient quantitative knowledge about the relationships between floral traits and pollination. This approach makes it possible to use quantitative and qualitative information. We associated fuzzy subsets to the decision trees and the classes of variables in order to avoid the knife-edge effect of class limits. At the species level, the indicator depends on three criteria: (i) visual attractiveness in terms of flower size, color and UV reflection; (ii) flower accessibility according to the botanical family, the symmetry and the shape of the flower; and (iii) the reward linked to pollen and nectar quantity and quality. An aggregation procedure allows us to obtain a value at the field margin level for each month as a function of the flowering period and pollination activity. Examples of calculations for honeybees, wild bees, bumblebees and hoverflies are shown. The evaluation of the predictive quality yielded significant correlations between pollinator abundance and the indicator value. The level of correlation is satisfying for this type of indicator, which might be further improved with additional data on plant traits. Coupling this indicator with a model that assesses the impact of management on plant diversity and abundance will be a further step to help agronomists who work on the improvement of arable farming management in order to lower its negative impact on pollination.