AgMIP Wheat Team Succeeds in Critical Model Improvement

AgMIP_wheat_smallThe AgMIP Wheat team at a team workshop in Suzhou, China, 9-11 May 2017.

By Senthold Asseng

The international Agricultural Model Intercomparison and Improvement Project (AgMIP) reached another milestone recently with a publication in Nature Plants. Led by Dr. Enli Wang, CSIRO, Australia, and Dr. Pierre Martre, INRA France, along with the AgMIP-Wheat team, the publication makes significant advancements in reducing wheat model uncertainty by improving temperature response functions.

The publication, entitled ‘The uncertainty of crop yield projections is reduced by improved temperature response functions’ (Nature Plants, 2017, Vol. 3), demonstrates how model intercomparison can encourage model improvement. In total, 60 wheat crop modelers, field crop experimentalists, a climate scientist and a biostatistician contribute to the AgMIP Wheat team’s research.

Graphic_for_blogTemperature response functions for biomass growth from wheat models. After Wang, Martre et al. (2017, Nature Plants, Vol. 3).

Since 2011, the AgMIP Wheat team has worked diligently to compare and improve wheat models in an effort to better simulate and understand the impacts of climate change on wheat production. Almost immediately, the AgMIP Wheat team recognized when comparing 27 widely used wheat models, model simulations differed mostly in climate impact assessments due to different approaches in modeling temperature change. These differences greatly contributed to model-simulated yield uncertainty and significantly reduced the reliability of model outputs

This new study addresses the issues regarding temperature change reduces its uncertainty by isolating each temperature function for phenology and biomass growth (see Figure 1) from 29 different wheat models and then incorporating these functions into two wheat models. The simulations with these two models and the 29 model temperature functions showed that the temperature response functions account for >50% of the uncertainty in simulating measured grain yields.

By proposing a new temperature response function (similar to the black line in the Response Type 4 panel – right lower panel in above figure) based on the latest crop physiology research and incorporating this function into several wheat models, the simulation error for grain yields across international field sites could be reduced by up to 50%. The improvement in temperature response functions will significantly improve wheat crop modeling in rising temperature scenarios. These temperature response function improvements can be transferable to other crop models. This will benefit many decision makers – farmers and extension workers will have better knowledge to improve intervention strategies, while policy makers will be better informed as they prepare for the future.

The publication by Wang, Martre and colleagues shows that model improvement is critical to improve the accuracy of impact model simulations. This is an important step forward in assessing climate change impacts and preparing for adaptation and future food security.