We apply a resource landscape approach to map the spatial distribution of floral resources across landscapes using neighbourhood modelling and empirical data on floral availability at specific land-use types. In each of 25 Mediterranean landscapes, spanning a gradient of natural-agricultural land use intensity, we established a pair of arrays of potted phytometer plants (White Wall-rocket, Diplotaxis erucoides) that were either aphid-infested or aphid-free. In each array, we recorded the activity of insect flower visitors and subsequent seed-set. At the same time, we also recorded the relative flower abundance in dominant land uses at 1 km radii around each array.

Fieldwork was conducted for one month between early March to Early April in 2017 and 2018, coinciding with the peak activity season of pollinators. In each array of phytometer plants, we recorded the activity of insect flower visitors and subsequent seed-set. At the same time, we also recorded the relative flower abundance in dominant land uses at 1 km radii around each array.

Seed-set data were collected from phytometer plants in 2017 and 2018. We monitored the seed set by opening mature pods and recording the number of fully developed seeds. The seed set was averaged for each combination of experimental patch and year.

The frequency of pollinator visits to the phytometer plant flowers was recorded in 2017. To measure pollinator visit frequency, phytometer flowers in each patch were observed for two consecutive days, beginning 24 hours after plants were placed in the field. At each site, the total number of flowers in the patch was counted in the early morning hours, before commencing data collection. On each day, 7 to 15 observation rounds of 15 minutes each were conducted per patch between 8 AM to 4 PM, alternating between the aphid-infested and aphid-free patches. Flower visitors were classified as honeybees, wild bees, or non-bee pollinators (mainly flies). Pollinator visitation frequency was then calculated as the total number of visits per experimental patch divided by the number of flowers and observation rounds in the patch.

Flower cover in each of the six land-use types was estimated visually in 96-159 randomly-placed 1m-diameter circles per land-use type, distributed across sites (15-40 circles per site); a total of 339 and 380 individual circles were sampled in 2017 and 2018, respectively. We calculated mean flower availability for each land-use type and combined these estimates with land-use maps within a 1 km radius of each study site.