Komorebi is a Japanese word that describes how light filters through leaves—creating a shifting, dappled “sun flecks” that illustrate plants’ ever-changing light environment. Crops harness light energy to fix carbon dioxide into food via photosynthesis. In a special issue of Plant Journal, a team from the University of Illinois reports a new mathematical supercomputer model that is used to understand how much yield is lost as soybean crops grapple with minute-by-minute light fluctuations on cloudy and sunny days. 

“Soybean is the fourth most important crop in terms of overall production, but it is the top source of vegetable protein globally,” said Yu Wang, a postdoctoral researcher at Illinois, who led this work for Realizing Increased Photosynthetic Efficiency (RIPE). “We found that soybean plants may lose as much as 13 percent of their productivity because they cannot adjust quickly enough to the changes in light intensity that are standard in any crop field. It may not sound like much, but in terms of the global yield—this is massive.” {module INSIDE STORY}

RIPE is an international research project that aims to improve photosynthesis to equip farmers worldwide with higher-yielding crops needed to ensure everyone has enough food to lead a healthy, productive life. RIPE is sponsored by the Bill & Melinda Gates Foundation, the U.S. Foundation for Food and Agriculture Research (FFAR), and the U.K. Government’s Department for International Development (DFID).

Past models have only examined hour-by-hour changes in light intensity. For this study, the team created a dynamic computational ray-tracing model that was able to predict light levels to the millimeter across every leaf for every minute of the day in a flowering soybean crop. The model also takes into account two critical factors: photoprotection and Rubisco activase.

Photoprotection protects plants from sun damage. Triggered by high light levels, this process dissipates excess light energy safely as heat. But, when light levels drop, it can take minutes to hours for photoprotection to relax, or stop—costing the plant potential yield. The team evaluated 41 varieties of soybean to find out the fastest, slowest, and average rate from induction to the relaxation of photoprotection. Less than 30 minutes is considered “short-term,” and anything longer is “long-term” photoprotection. 

Using this new model, the team simulated a sunny and cloudy day in Champaign, Illinois. On a sunny day, long-term photoprotection was the most significant limitation of photosynthesis. On a cloudy day, photosynthesis was the most limited by short-term photoprotection and Rubisco activase, which is a helper enzyme—triggered by light—that turns on Rubisco to fix carbon into sugar. 

The RIPE project has already begun to address photoprotection limitations in soybean and other crops, including cassava, cowpea, and rice. In 2016, the team published a study in Science where they increased the levels of three proteins involved in photoprotection to boost the productivity of a model crop by 14-20 percent. In addition, the RIPE team from the Lancaster Environment Centre at Lancaster University is seeking better forms of Rubisco activase in soybean and cowpea.

The RIPE project and its sponsors are committed to ensuring Global Access and making these technologies available to the farmers who need them the most.

“Models like these are critical to uncovering barriers—and solutions—to attain this crop’s full potential,” said RIPE Director Stephen Long, Ikenberry Endowed University Chair of Plant Biology and Crop Sciences at Illinois’ Carl R. Woese Institute for Genomic Biology. “We’ve already begun to address these bottlenecks and seen significant gains, but this study shows us that there is still room for improvement.” 

With some reports predicting the precision agriculture market will reach $12.9 billion by 2027, there is an increasing need to develop sophisticated data-analysis solutions that can guide management decisions in real-time. A new study from an interdisciplinary research group at the University of Illinois offers a promising approach to efficiently and accurately process precision ag data.

"We're trying to change how people run agronomic research. Instead of establishing a small field plot, running statistics, and publishing the means, what we're trying to do involves the farmer far more directly. We are running experiments with farmers' machinery in their own fields. We can detect site-specific responses to different inputs. And we can see whether there's a response in different parts of the field," says Nicolas Martin, assistant professor in the Department of Crop Sciences at Illinois and co-author of the study.

He adds, "We developed a methodology using deep learning to generate yield predictions. It incorporates information from different topographic variables, soil electroconductivity, as well as nitrogen and seed rate treatments we applied throughout nine Midwestern cornfields." {module INSIDE STORY}

Martin and his team worked with 2017 and 2018 data from the Data-Intensive Farm Management project, in which seeds and nitrogen fertilizer were applied at varying rates across 226 fields in the Midwest, Brazil, Argentina, and South Africa. On-ground measurements were paired with high-resolution satellite images from PlanetLab to predict yield.

Fields were digitally broken down into 5-meter (approximately 16-foot) squares. Data on soil, elevation, nitrogen application rate, and seed rate were fed into the computer for each square, with the goal of learning how the factors interact to predict yield in that square.

The researchers approached their analysis with a type of machine learning or artificial intelligence known as a convolutional neural network (CNN). Some types of machine learning start with patterns and ask the computer to fit new bits of data into those existing patterns. Convolutional neural networks are blind to existing patterns. Instead, they take bits of data and learn the patterns that organize them, similar to the way humans organize new information through neural networks in the brain. The CNN process, which predicted yield with high accuracy, was also compared to other machine learning algorithms and traditional statistical techniques.

"We don't really know what is causing differences in yield responses to inputs across a field. Sometimes people have an idea that a certain spot should respond really strongly to nitrogen and it doesn't or vice versa. The CNN can pick up on hidden patterns that may be causing a response," Martin says. "And when we compared several methods, we found out that CNN was working very well to explain yield variation."

Using artificial intelligence to untangle data from precision agriculture is still relatively new, but Martin says his experiment merely grazes the tip of the iceberg in terms of CNN's potential applications. "Eventually, we could use it to come up with optimum recommendations for a given combination of inputs and site constraints."