Pesticide input data
To estimate average level of pesticide applied per crop at each field site (1km buffer), we calculated a pesticide risk assessment (RI ), including herbicides, insecticides, molluscicides and fungicides, using the methodology described by Yasrebi-de-Kom et al. (2019) as:
\begin{equation} RI=\ \sum{HQ=\left(\frac{Application\ rate\ (\text{g.ha}^{-1})}{Toxicity\ (\text{LD}_{50}\text{\ in\ μg\ per\ bee})}\right)>50}\nonumber \\ \end{equation}
with HQ the hazard quotient (HQ) of each active molecule and the median lethal dose per bee (LD50 ). The median lethal dose is one way to measure the short-term poisoning potential (acute toxicity) of a substance. The LD50 is the amount of a substance, given all at once, which causes the death of 50% of a group of test animals. The hazard quotient ratio gives an approximation of how close the likely exposure of bees is to a toxicologically significant level. The pesticide risk index (RI ) was defined as the number of high risk active ingredients (HQ>50 ; see EPPO, 2010) that were applied. IfHQ<50 , the active ingredient was categorized as low risk to bees.
The LD50 of 390 active ingredients used in the UK and the Netherlands were extracted from the “Pesticides Properties DataBase ” (PPDB) from the University of Hertfordshire, UK (https://sitem.herts.ac.uk/aeru/ppdb/en/index.htm; Lewis et al., 2016; Lewis and Tzilivakis, 2019) (see list of active ingredients in Appendix S1 in Supporting Information). As proposed by EPPO (2010), the risk assessment was carried out selecting the lowest of the oral and contact LD50 values available across the different bee species (honey bees, bumble bees and other wild bees), to take the most conservative approach for the entire bee community (see Table S2). However, LD50 values were mainly available for honey bees, sometimes for bumble bees, and much less frequently for other pollinators (Lewis et al., 2016; Lewis and Tzilivakis, 2019; Yasrebi-de Kom et al., 2019).
For the UK, the average annual application rate at 1km resolution (in kg/km2/year) was obtained for 130 pesticide active ingredients from the “CEH Land Cover® plus: Pesticides 2012-2016” (Jarvis et al. 2019) of the Centre for Ecology and Hydrology (CEH, Wallingford, UK; https://www.ceh.ac.uk), across a four-year period (from 2012, 2013, 2014 and 2016). For the Netherlands, we created the average annual allowed application rate at 1km2 (in kg/km2/year) across a two-year period (2015 and 2016) for 179 pesticide active ingredients, combining allowed application rates produced by Yasrebi-de-Kom et al. (2019) and the BRP shapefiles for crops in 2015 and 2016 obtained from the RVO (https://english.rvo.nl).