Adaptation of US maize to temperature variations

High temperatures are associated with reduced crop yields1,2, and predictions for future warming3 have raised concerns regarding future productivity and food security4–8. However, the extent to which adaptation can mitigate such heat-related losses remains unclear9–13. Here we empirically demonstrate how maize is locally adapted to hot temperatures across US counties. Using this spatial adaptation as a surrogate for future adaptation, we find that losses to average US maize yields from a 2 C warming would be reduced from 14% to only 6% and that loss in net production is wholly averted. This result does not account for possible changes in temperature variability or water resources, nor does it account for all possible forms of adaptation14–18, but it does show that adaptation is of first-order importance for predicting future changes in yield. Further research should be undertaken regarding the ability to adapt to a changing climate, including analysis of other crops and regions, the application of more sophisticated models of crop development, and field trials employing artificially increased temperature. Global maize yields are forecast to decline in response to increasing temperature, particularly as the upper range of growing season temperatures become hotter1,2,4–7,19. The sensitivity of crop yields to increased temperature is often estimated through analysis of variability in annual yield and growing season temperature1,2,7,19, but there is a potentially important distinction between year-to-year anomalies and changes in climate in that the latter can be more fully adapted to. For instance, US corn hybrids have a product half-life of about 4 years11, suggesting sufficiently rapid turnover to adapt to decadal changes in climate. To explore the adaptability of maize production to long-term differences in climate, we analyse the sensitivity of extant crops growing in a range of different climate conditions and use this spatial variation to develop a functional form for future adaptation. We explore yields within the US because relatively high-quality data and a highly adapted and managed agricultural demographic can be assumed. Yield data are available from more than 1,600 counties between 1981 and 2008 from theUnited StatesDepartment of Agriculture/National Agriculture Statistics Service20 in the Eastern US, and daily temperature is estimated for each county using a network of 534 weather stations21 for which daily minimum andmaximum surface air temperature is available. The influence of temperature on yield is parameterized using growing degree days (GDDs) and killing degree days (KDDs). GDDs are a commonly used measure for the cumulative warmth a crop has experienced over the growing season1,15,22,23, here defined as the sum of all daily average temperatures over the growing season in excess of 8 C. The threshold is in accord with previous studies1,15, but we use a new approach to define the growing season using average planting and harvest dates reported for each state on each year20, with the average weighted according to the amount of planted or harvested crop. Daily temperature is computed by taking the average of the maximum and minimum temperature at