Crop–climate models need an urgent update

According to the researchers, many of the current crop-climate models used for estimating potential impacts of climate change on food supply are badly out of date. Also, many models do not agree on what future will bring.

"The majority of models that are applied to this problem were developed at least two decades ago. Many current models do not incorporate the latest knowledge about how crops respond to a changing climate. Also, they may not properly represent modern crop varieties and management practices," the researchers argue.

For example, the writers point out, field experiments have shown that when temperatures go above 30-36°C during flowering, crops such as wheat, rice and maize experience a sharp decline in grain set and yield. Most process-based models do not account for this, and thus tend to overestimate future yields in regions where hot days will become more frequent during the growing season.

Uncertainty often understated

According to Rötter et al., crop-climate modelers should focus on multi-model ensemble techniques in order to improve assessments and information on appropriate strategies of adapting crop production systems to climate change. These strategies are badly needed by agricultural policymakers: accurate information is of utmost importance in managing the risks and moderating the influence of climate change on future crop productivity.

Furthermore, the writers argue, crop-climate modelers often fail to identify and report the uncertainty of their models, which can promote mistrust in model results and make it difficult for policymakers to act on the information. Thus Rötter, Carter, Olesen and Porter argue that crop-climate modelers urgently need to embrace methods already prevalent in the climate modeling community. Instead of relying on a single model, these methods apply a number of models to see how they differ.

"In terms of research, we should start by reviewing existing models to see where the big deficiencies are in understanding crop growth and development processes."

Worldwide data and inter-comparison needed

Second, the writers argue, a renewed effort needs to be made to produce and compile high-quality field data for model testing at the scale of whole farms. An achievable goal would be to develop high-quality data sets for each major crop type at about ten test sites worldwide. Even if models are improved in the future based on more comprehensive data, intercomparison of different models will continue to play a vital role in testing model performance and sensitivity in different agro-environments.

The researchers argue that there is an urgent need for more quality, transparency and consistency in the modeling methods used to address climate change impacts on crop production. They suggest that their approach would provide a firm basis for delivering more robust and usable information, for everyone from farmers to policymakers.