Picture (above): Young wheat. Credit: Pixabay (Pexels).

Assessments of the impacts of future climate changes on crop yields are subject to significant uncertainties. This is partly due to imperfections in computer models used to simulate the climate and crop growth. Different climate models can simulate different climate changes when given the same scenario for future atmospheric greenhouse gas concentrations and different crop models can simulate different changes in crop yields when given the same changes in climate.

Previous research has concluded that the uncertainty due to crop models is more important than the uncertainty due to climate models. However, this work does not cover all the world’s agricultural regions and tends to rely on a particular “change factor” method of generating realistic future climate data by combining climate model output, which gives an imperfect representation of the climate, and observations.

Therefore, the researchers examined the contribution to uncertainty in simulated future changes in crop yields using the change factor method and an additional, more sophisticated, statistical downscaling method of generating realistic future climate data from climate model output. They examined wheat simulations for two important climatically different zones of rain-fed wheat production – southeast Australia, with a Mediterranean-type climate, and northern China, with a continental monsoon climate.

The researchers simulated changes in wheat yields during the 21st century at two sites in southeast Australia and two sites in northern China. For each site, they ran simulations of eight different crop models with climate data derived from 32 different climate models using both the change factor and statistical downscaling methods. For each site, the simulations produced a range of different future changes in wheat yields.

The researchers found that the relative contribution to this uncertainty due to climate models and crop models varied between the southeast Australian and northern Chinese wheat belts. The primary reason for this was the influence of future changes in growing season rainfall on the wheat yield simulations.

Crop model uncertainty was more important in China, where the influence of rainfall is relatively weak and climate models agreed on a future increase in growing season rainfall. However, climate model uncertainty was more important in Australia, where the influence of rainfall is stronger, because some climate models simulated future rainfall increases and some simulated future rainfall decreases. These results held for both the change factor and statistical downscaling methods.

Because of the variation of results across climate models and crop models, our study supports the practice of using multiple climate models and crop models in assessments of the impacts of future climate changes on crop yields and highlights the need for thorough evaluation and further development of the models.

  • Paper: Wang, B., Feng, P., Liu, D.L. et al. Sources of uncertainty for wheat yield projections under future climate are site-specific. Nat Food 1, 720–728 (2020). https://doi.org/10.1038/s43016-020-00181-w