The environmental consequences of climate-driven agricultural frontiers

formula 2a/2a+b+c where a = cells in common and b + c = cells in disagreement. (DOCX) S1 Fig. Flowchart showing the input data, modeling methods, frontier determination and impact analysis. Sources for data are indicated by reference number. (TIF) S2 Fig. Global map of climate-driven agricultural frontiers. Blue color ramp shows transi- tion from zero currently suitable commodity crops to one or more suitable commodities. Red color ramp shows transition from one suitable commodity to two or more suitable commodi- ties. Intensity of color denotes level of GCM agreement for the RCP8.5 2040–2060 climate sce- nario. Areas in grey are either currently suitable for two or more commodities or not suitable for any commodities in the projected climate. (TIF) S3 Fig. GCM agreement (green = low; blue = high) of modeled frontiers for individual commodity crops for RCP 8.5 2060–2080. (TIF) S4 Fig. Distribution across 17 GCMs of total area (million km2) in commodity frontiers for each modeled crop under RCP8.5 2060–2080 climate projections. Boxes are interquar- tile range; whiskers extend to 1.5 times interquartile range; outliers indicated by points. (TIF) S5 Fig. Comparison of modeled agricultural frontiers with areas of very severe soil con- straints (green). Soil constraints were defined as the union of GAEZ soil resource classifica- tions and crop specific limitations of pH and soil depth. Areas in green were used to reduce total frontier impact on soil carbon in Table 2. (TIF) S6 Fig. Comparison of agricultural frontiers of this paper (blue and red) vs. MAGPIE RCP

SSP-5 [91] cropland in 2070 (green). (TIF) S7 Fig. Comparison of modeled wheat suitability in available GAEZ gridded datasets (blue —Top panel) and this paper (red—Bottom panel). An exact match for GCM, scenario and time step was not available, but shows general spatial congruence of the modeling methods for late-century, high emissions scenarios. Existing agriculture is represented in grey for both panels. (TIF) S8 Fig. Comparison of existing cultivated land (green) with modeled current suitability (purple) and soil constraints (black overlay). Areas of visible purple are where the universal agreement among three models indicates that it is currently suitable for at least one modeled commodity crop but is not currently cultivated. The pole ward edges of the extent of modeled vs. realized agriculture are in good agreement, with modeled agriculture extending further than existing continuous agriculture. Quantification of frontiers and their potential impacts presented here are therefore conservative. (TIF) Author Contributions Conceptualization: Lee Hannah, Patrick R. Roehrdanz, Krishna Bahadur K. C., Camila I. Donatti. The environmental consequences of climate-driven agricultural frontiers PLOS ONE | https://doi.org/10.1371/journal.pone.0228305 February 12, 2020 14 / 19