Figure legends
Fig.1 Schematic showing the different metrics of resilience applied to an example time series. Grey lines and filled points indicate simulated yield data for a single hectad. A) Illustrates relative function, being derived from the average difference (area between lines shaded in light grey) between the hectad (mid-grey lines and points) and national average yields (light grey lines and points) across the time series. B) Illustrates yield stability, derived from the difference (vertical black arrows) between yield in any one year and the average yield over the two years either side (horizontal, dashed black bars). C) Illustrates resistance, derived from the proportional drop in yield (vertical black arrow) in 2012 from the pre-2012 mean (horizontal, dashed black bar). Note that, for analyses, the inverse of the latter two metrics is taken, such that higher values indicate higher resilience in all cases.
Fig.2 Partial residual plots of the explanatory variables in the ‘best’-fitting models for each response variable. Plots show the effect of a given landscape composition or configuration variable after removing variance from random and other fixed effects in the model (Including environmental zone and potential yield) for each of A) relative function, B) yield stability, C) resistance. Abbreviations on axis labels are SNH = semi-natural habitat, SNG = semi-natural grassland
Fig. 3 Example hectads (10km squares) from a single environmental zone predicted to have the maximum and minimum resilience values by the ‘best’-fitting models for each response variable. For all plots, shading indicates land cover type, with pale grey being arable land, mid-grey being deciduous woodland and black being semi-natural grassland. White areas indicate other land cover types (e.g. urban areas, improved grassland).