Scientists can predict how climate change will affect crop yields using computer models calibrated according to past crop and harvest data. However, at the regional and global scales, a lack of consistent, coordinated data makes these calibrations difficult. In a new paper, Xiong et al. explore how calibration affects uncertainty in large-scale crop simulation.
The researchers used the Environmental Policy Integrated Climate model and global data sets on climate, soil, terrain, and crop management to simulate global corn crop yields between 1951 and 2099. They tested the influence of four different parameters—phenology, harvest index, optimum temperature, and sowing density—by running the model several times with different values for each parameter.
No matter the calibration strategy, all simulations predicted that climate change will reduce future crop yield. The team found that phenology—which is characterized by seasonal events like flowering and ripening—contributed the most uncertainty to crop yield predictions, whereas harvest index had little effect. Calibration caused about 26% of the uncertainty in crop yield predictions by 2100.
The authors caution that global crop simulation is not reliable at the local scale since small calibration tweaks can result in widely varied local predictions. However, noting which locations are affected the most by calibration changes can pinpoint them for more detailed investigation to drive decision making about the world’s future food supply. (Journal of Advances in Modeling Earth Systems (JAMES), doi:10.1002/2016MS000625, 2016)
—Sarah Stanley, Freelance Writer