1. Threats to bee pollinators such as land use change, high pesticide risk, and reduced floral diet diversity are usually assessed independently, even though they often co-occur to impact bees in agroecosystems.

2. We established populations of the non-native mason bee O. cornifrons at 17 NY apple orchards varying in proportion of surrounding agriculture and measured floral diet diversity and pesticide risk levels in the pollen provisions they produced. We used path analysis to test the direct and indirect effects of different habitats, diet diversity, and pesticide risk on emergent female offspring number and weight.

3. Our results showed that high proportions of agricultural habitat surrounding bee nests indirectly reduced the number of female offspring produced, by reducing floral diet diversity in pollen.

4. When proportion agriculture surrounding bee nests was high, bees collected increased proportions of Rosaceae in their pollen provisions, which marginally (0.05<p<0.1) increased fungicide risk levels in pollen, which, in turn, marginally reduced female offspring weight. In contrast, female offspring weight increased as proportion surrounding open habitat (wildflowers, grassland, pasture) increased, but this effect was not influenced by proportion Rosaceae or fungicide risk levels in pollen.

5. Synthesis and Applications: To promote healthy O. cornifrons populations in apple, we must maintain floral resource diversity and open habitats, while reducing fungicide risk levels and agricultural habitats. More broadly, our results show that land use change, in the form of increasing agricultural habitat, can negatively impact bee populations in agroecosystems indirectly through multiple, simultaneous threats. We must strive to understand these complex interactions between simultaneous threats to maintain healthy bee populations in agroecosystems, where we rely on them for pollination.

Please refer to the "Materials and Methods" section of the associated article, as well as Appendix S1 and S2 of the Supporting Information, where you will find all information related to methodogy and data processing.

Please see the “Materials and Methods” section of the associated article, as well as Appendix S1 and S2 of the Supporting Information for details on how these data were collected and processed. This “Read Me” information walks you through the data-set. The Site_Code is a two-letter code based on the apple orchard site where data were originally collected. We do not add lat/long information here, in order to respect the privacy of our participating growers. Actual orchard area in square meters (see column “Orchard Area m2”) was also measured by ground-truthing the area on a map in qGIS (see methods and Figure S3 in the Supplemental Materials). There were tree time points per site, corresponding to distinct calendar dates. Site letters here correspond to letters used in the Supplemental Materials of the MS. They were re-coded from the original 2-letter codes to further protect site identity.

The landscape metrics complied using the CropScape data layer are abbreviated here as follows: “Ag” for Agriculture and “WetShrub” for wetland and shrubland. “Other” refers to barren and open water habitats. Each habitat type (see method for what constitutes each type) was measured at different meter radii surrounding bee nests, denoted here by “-###m” after the habitat type. “Apples” refers to surrounding apple area at our sites, as measured by the crop-scape data layer.

The total emerged male and female offspring for each time point are shown here, as well as the proportion of female offspring per both male and female offspring. The average weight (in grams) of up to ten (or fewer if fewer were produced) male and ten female offspring per site are also recorded here. Finally, proportion larval mortality (as recording using x-radiography of bee nests – see methods) per time point is shown here. The average temperature between time points (see methods) in Celsius is shown here. As explained in the methods, the total nest tubes collected for reproduction and for pollen analysis per time point are recorded here, as well as the total number of pollen provisions complied for pollen (pesticide and diet diversity) analysis in equal amounts per nest tube (see methods).

The average fungicide and insecticide risk measured in percent hazard quotient (see methods) is shown here, along with the “herb risk”, or herbicide risk. For more information about pesticide risk, the chemicals tested and detected, and the parts per billion found, please see Table S2, S3, and S6 in the Supplemental Materials.

The proportion of pollen from each of the 11 floral families detected in pollen provisions is shown here, based on 300 grain counts per time point (see methods). Finally, floral richness, Shannon diversity index, and evenness at the family level were calculated and are presented here.

Missing values (denoted as “NA”) are also explained in the Materials and Methods and Appendix S1 and S2 of the associated article and its Supporting Information, respectively.