Fig. 3: Global impact of climate change on staple crops.
From: Impacts of climate change on global agriculture accounting for adaptation
a, Time series of projected climate change impacts on global average yields (area-weighted) in a high-emissions scenario (RCP 8.5), accounting for producer adaptations and adaptation costs. Box and whisker plots show total distribution of end-of-century projections accounting for joint statistical and climate model uncertainty (in each subplot: boxes, 25th to 75th percentiles; whiskers, 5th to 95th percentiles). To the right are projections adjusted for the estimated effect of CO2 fertilization from ref. 9 (shaded) and projections for moderate emissions (RCP 4.5) without CO2 fertilization (unshaded). See full results for moderate emissions in Supplementary Fig. 8, CO2 fertilization in Supplementary Information, section J, and uncertainty in Supplementary Figs. 4 (RCP 8.5) and 5 (RCP 4.5). b, Projected end-of-century yield impacts, by deciles of present-day average temperature over 24,378 regions. Impacts across crops weighted within region by cropped area and caloric content; distributions across regions weighted by total cropped area. c, As in b but by decile of present-day income. d, Empirical end-of-century global damage function describing calories (kcal) lost as a quadratic function of global mean surface temperature anomaly (ΔGMST). Each point represents a single climate-model-by-Monte Carlo run for RCP 4.5 (blue) or RCP 8.5 (red) in 2093–2097, including gains from CO2 fertilization. Grey bands indicate 10th–90th and 25th–75th quantile bands, conditional on ΔGMST. Bottom panel shows the distribution of warming under each RCP across 33 climate models and model surrogates (Supplementary Information, section G). Box plots to the right show distribution of damages collapsed to RCP. Right axis describes calorie losses normalized by 2015 global calorie production for the six crops studied here. Projected log(yield) impacts for panels a–c winsorized at the top and bottom 1% of the impacts distribution over all region–GCM–years, by RCP–crop, then converted to percentages.
