Results
The observed abundances of pollinators at each sampling transect varied from 0 to 367 for honey bees (Apis ) and from 0 to 154 non-Apis pollinators (i.e., wild bees and hoverflies).
Mean ozone value per study site varied from 0.140 to 0.144 mol.m-2 in the Netherlands and from 0.142 to 0.147 mol.m-2 in UK (for reference, worldwide it varies from 0.079 to 0.222 mol.m-2). The mean tropospheric NO2 per study site ranged from 27.8 to 76.5 µmol.m-2 (from 0 to 2.14 mmol.m-2worldwide) and the gradient of fertilizer N input varied from 2.28 to 21.09 t.km-2 (2.28 to 12.32 in UK and 3.89 to 21.09 in the Netherlands). The risk index of pesticide exposure varied between 2 to 8 in the Netherlands and between 2 to 10 in UK, (i.e., between 2 and 10 high risk active molecules were applied in the 1km buffer around study sites). The proportion of natural and semi-natural habitats in the 1km2 surrounding buffer varied from 0.1 to 37% in UK (with a mean=8.9 and median=4.4) and from 0 to 47% in the Netherlands (with mean=7.0 and median=1.8).
We observed effects of pesticide risk exposure and ozone on crop pollinator abundance as well as interacting effects between these two environmental factors, but such effects differed between Apis and non-Apis . We found that abundance of honey bees (Apis mellifera ) in crops was negatively related to concentration of ozone but positively related to the risk of pesticide exposure (see Appendix S3). However, the positive relationship observed between the abundance of honey bee and the risk of pesticide exposure becames less strong with increasing ozone levels (Fig. 3A; see Appendix S3). In other words, the negative effect of ozone on Apis pollinators is more accentuated when this risk of exposure increases (see Appendix S6). As for non-Apis , the negative effect of pesticides on abundance was more accentuated at higher ozone exposition (Fig. 3B; see Appendix S4).
As hypothesized, we found a negative correlation between ozone and the contribution of pollinators to crop production (i.e., crop pollination assessed by the difference of production between open and close treatments; see Appendix S5 and S6), but also an interacting effect between ozone and the risk of pesticide exposure on crop production. While at low concentration of ozone the risk index (RI) of pesticide exposure was positively related to the contribution of pollinators to crop production, the relationship became negative when ozone levels were high (Fig. 3C; see Appendix S5).
Contrary to our expectations, we found no evidence of a relationship of any Nitrogen sources studied here (i.e., atmospheric nitrogen dioxide deposition and mean application rate of nitrogen fertilisers at 1km resolution) on the abundance of Apis and non-Apispollinators nor on the contribution of pollinators to crop production (see Appendix S3, S4 and S5). We also did not observe evidence of a correlation between the proportion of semi-natural habitats and the abundance of crop pollinators or on their contribution to crop production.