For decades, sorghum has quietly disappeared from US fields. Once grown across much of the country, the drought-tolerant grain is now concentrated primarily in the southern Great Plains, especially Kansas, Oklahoma, and Texas. Since the 1970s, its planted area has steadily declined, as corn expanded into many of the regions where sorghum once dominated.

A new study led by Dr. Jasdeep Singh, a recent Ph.D. graduate from the UF/IFAS Horticultural Sciences Department, suggests that this decline is not just the result of limited breeding progress, but that modern US sorghum hybrids are already approaching the biophysical limit of how much grain can be produced for a given amount of water. In other words, sorghum may not be stagnating, but hitting its ceiling.

Using UF’s HiPerGator supercomputer, Singh ran hundreds of millions of crop simulations that linked genetics, environment, and management (GxExM) across the sorghum belt. When those simulated outcomes were compared with decades of field trial data, a clear pattern emerged: sorghum’s slower rate of improvement is not due to weak breeding, but to biological limits shaped by water availability.

The results of this study, Understanding rates of genetic gain in sorghum [Sorghum bicolor (L.) Moench] in the United States, published in The Plant Genome, makes the case that simply selecting from the same pool of genetics will no longer be enough. To keep improving sorghum, breeders will need access to new sources of genetic diversity and better ways to connect plant traits to field performance.

By combining crop growth models, genomics, and large-scale simulation, the research offers a way to test what might work before years of breeding and field trials begin. In a future where water is increasingly limited, this approach gives breeders a practical way to design crops that are not just higher yielding, but better matched to the environments they’re grown in.

That kind of work, Singh said, is only possible in a research environment that encourages cross-disciplinary problem solving.

“UF’s mission as a land-grant university is to tackle the most urgent challenges facing our environment and food supply,” Singh said. “Crop models aren’t new, but supercomputing is finally giving them the power they need.”

Understanding the Breeding Gap

For years, sorghum’s slower yield gains were often blamed on lower investment compared with crops like corn. Singh and his co-authors at the UF/IFAS Crop Transformation Center, led by Dr. Charlie Messina, tested a different idea. The concept that modern sorghum hybrids may already be performing near the best that current genetics and physiology allow under drought conditions.

To explore this, they used a method called breeding gap analysis (BGA). Instead of asking how much yields have increased over time, BGA asks a more practical question: how close are today’s crops to the best yields that are physically possible for the amount of water they use?

Because that ceiling cannot be measured directly in the field, Singh applied a crop growth model called CERES-Sorghum, which simulates how a plant grows day by day based on weather, soil, and key plant traits. Meteorologists have used similar approaches to simulate weather. By running the model across many realistic combinations of variables (weather, soils, planting times, and plant traits), Singh could estimate the best yield for a given amount of water, assuming the crop was not limited by pests, disease, or nutrient shortages.

These simulations created what the researchers describe as a “cloud of the biophysically possible”. It is a visual way to show the full range of yields sorghum could reach for a given amount of water. From this, the team could define an upper boundary that represents the maximum potential for productivity.

Designing Crops for Environmental Uncertainty

Once the biological boundaries were defined, the focus shifted from why gains have slowed to how breeding decisions can change moving forward.

Traditional breeding relies on field trials placed in a limited number of locations. Those sites shape which environments are sampled and which traits are selected, but breeders rarely know whether their testing network truly reflects the range of conditions their crops will face.

This framework provides a way to compare field trial data, allowing researchers to see whether a location, year, or management practice is revealing a meaningful genetic difference. In this way, the model becomes not just a tool for analysis, but a guide for decision-making.

This is where artificial intelligence moves beyond simply processing data.

“Building on a legacy of research and learnings from the maize industry, sorghum breeders can use AI to define plant attributes that improve drought tolerance, design experiments to evaluate germplasm, identify genes, and make more informed selections,” said Dr. Charlie Messina.

Rather than replacing human judgment, the system acts as a decision-support tool as it links genetics, management, and environment. It can also be used to evaluate cropping systems through the lens of long-term risk.

“This kind of infrastructure can essentially act like a risk analyst,” Singh said, “helping us understand where corn performs better, where sorghum might be the stronger option, and where there’s an opportunity to develop products that can transition between the two.” Focusing on stability over time is essential as climate uncertainty increases and water becomes a more limiting resource.

Looking Ahead

While sorghum is the focus of this study, the framework is designed to extend to other crops and regions. Instead of improving crops one generation at a time, this work points to a new strategy: first define what is physiologically possible, then design breeding pathways that move toward that future. One that may be essential as agriculture faces climate uncertainty, water scarcity, and rising global demand.

The approach is already being expanded through broader collaborative efforts that connect public research with industry and global development programs. Supported through the Climate Resilient Crops Innovation Lab (CRCIL) and partners including Corteva Agriscience and the US Department of State, the same tools are now being adapted for use in breeding programs globally.

“CRCIL helps move climate-resilient crop research from the laboratory into real-world impact by supporting tools that allow breeders to test ideas virtually and make better decisions for water-limited environments,” said Dr. Jared Crain, associate director of the CRCIL, a US Department of State–funded research program focused on developing climate-resilient cereal crops.

Postdoctoral researchers Drs. Sam Ipniyomi, Gustavo Visentini, and Juan Panelo are working to help translate these methods for both US industry breeding programs and smallholder-focused systems in Africa to show how the same science can serve very different agricultural contexts.

“When AI is used ethically, responsibly, and with clarity of purpose, its benefits can extend far beyond any single crop or region,” stated Messina.

2