The environmental consequences of climate-driven agricultural frontiers

Critical maximum and minimum temperatures critical to crop success [26] were modeled with gridded global daily observations for maximum and minimum air temperature were obtained from the NOAA Earth System Research Laboratory Twentieth Century Reanalysis Version 2: 4-times daily and daily average monolevel dataset [27] at 2˚ x 2˚ resolution. Global daily observation grids (2˚ x 2˚) were summarized by month to create a count of days where temperature exceeded each integer degree in the range -50C to +50C. Tempera- tures were rounded to the nearest degree C and summarized in one-degree intervals. Counts of temperature exceedances were generated over a 20-year period (1980–2000) and resampled to 30 arc seconds (see S7 Table). Binary suitable vs. not-suitable grids were then created by thresholding the critical temperature counts at 20% of all days within the relevant month over the 20-year period using a cropping calendar dataset [28]. Agricultural frontier ensemble Suitability for each crop was determined for current climate and in each of 34 future scenarios (17 GCM, 2 radiative forcing) as the agreement of the three suitability methods (EcoCrop, Maxent, Critical temperatures). Frontiers are defined as areas that transition from zero crops suitable in current climate to one or more crops suitable in the future climate scenario. Addi- tionally, we mapped frontiers that transition from one crop suitable in current climate to two or more crops suitable in the future climate scenario. Validation statistics as measured by

Dice-Sorenson spatial congruence for all crop models are shown in S8 Table. Water quality impacts The analysis was implemented in a hydrological model, WaterWorld version 2, globally at 10km spatial resolution [29]. To calculate the population within the new agricultural areas, who may experience water quality declines, we sum the Landscan global population dataset 2011 [30] over the area of new agriculture. The WaterWorld hydrological footprint (HF) [29– 31] is a measure of the influence of an area on downstream water and is calculated by cumu- lating a global water balance calculated by the WaterWorld model [31], along a HydroSHEDS [32] based flow network. The hydrological footprint at a point along this network is defined as the contribution of an upstream area (such as a new agricultural area) to flow of water at this point. We calculate the Agricultural Water Quality (AWQ) index as the land area affected by hydrological footprint >0% and >50% as proxies for all contamination and for significant contamination respectively. Further, we calculate the population affected by this hydrological footprint as the sum of Landscan 2011 [30] people in pixels with a hydrological footprint of the new agricultural areas >0% and >50%. We use a global database of 38,000 dams [33] to iden- tify which lie on rivers with a hydrological footprint>0 as a surrogate for agricultural, hydro- power and urban water supplies potentially affected by this new agricultural land. Footprints are calculated for the current distribution of cropland fractional cover [34], the current distri-

bution of land suitable for cropland and each ensemble member of the frontiers. We used a “difference method” where the number of people affected is a function of the agricultural water footprint of current farming systems, minus the agricultural water footprint of future farming practices. We use this difference to calculate the fraction of water, at any point in the hydrological network, that fell as rain on particular land uses upstream and thus is likely to be contaminated by new cropland. Soil organic carbon impacts To account for soil organic carbon stocks in agricultural frontiers, we obtained a gridded global dataset of estimated soil organic carbon (tonnes ha-1) in the top 100 cm [35]. Individual The environmental consequences of climate-driven agricultural frontiers PLOS ONE | https://doi.org/10.1371/journal.pone.0228305 February 12, 2020 4 / 19