The CropSmart Digital Twin (CSDT) decision tool, a research project led by George Mason University research professor Liping Di, is one of seven multidisciplinary projects nationwide recently selected to advance to Phase 2 of the National Science Foundation’s Convergence Accelerator, which could provide the researchers with up to $5 million in funding.
The user-inspired CropSmart Digital Twin provides on-demand, decision-ready solutions to take the guesswork out of crop management decisions. The services will be easily accessible to users through both web portals and smartphone apps. The optimal solutions are derived from near-real-time remote observations of cropping systems with artificial intelligence or machine learning-based modeling and simulations.
The United States accounts for more than 25% of the total grain traded globally. Successful crop production involves smart management decisions. With the Accelerator, the NSF is investing in new technologies that address the complex challenges in tackling food and nutrition insecurity from population growth, increasing diet-related diseases, and disparities and climate resilience.
In Phase 1 of the Convergence Accelerator, Di and his team received $750,000 from NSF to design CropSmart. With Phase 2, the team will focus on an operational release of the tool. “So it can be tested and evaluated in a real operational environment and provide services to users,” said Di, who is the director of the Center for Spatial Information Science and Systems in George Mason’s College of Science.
Farmers often depend on their own experience when making decisions, but some will make large investments in commercial instrumentation to measure and collect data on their crops. Yet, after making that investment, they must also collect the data and possess the scientific knowledge to apply it effectively. Di’s team will do this using remote sensing and modeling, covering larger areas at no or very low cost to the farmers, agribusiness operators, and governmental agriculture decision makers.
Di said they were working with more than 20 end-user partners to test CropSmart and anticipates increasing that number to 2,000 over the course of 2024. These users represent different agricultural sectors ranging from individual farmers to farm credit bureaus and ag-related businesses, and include the USDA.
The team is focused on commodity crops, such as wheat, corn, soybeans, and rice. Co-principal investigators on the project include Cenlin He at University Corporation for Atmospheric Research, Jenny Du at Mississippi State University, Haishun Yang at University of Nebraska–Lincoln, and Juan P. Sesmero at Purdue University.
Sesmero said the first part of the release—CropSmart 1.0—focuses on certain decision points: when to apply irrigation and how much; crop condition and yield prediction for supporting farm loan and crop insurance decision making; and in-season crop mapping and statistics.
Over time, the teams plans to add additional decision points, such as when to plant and harvest and when to apply nitrogen fertilizer and how much. They will conduct comprehensive field testing to ensure a fully tested and operational solution by the time they release CropSmart 2.0.
“We use the results of our testing to refine the technology,” said Sesmero. “We collect data, process it, and deliver that information through the app.”
The CropSmart Digital Twin will provide three types of products to users: decision-ready information, a “what if” service, and direct management recommendations. The decision-ready information will be delivered to end users for free.
“The users will get a good amount of aggregated information at whatever the format they want for free,” explained Sesmero.
The CropSmart "what if” and direct recommendation services are provided with a subscription fee, intended to help fund additional product development and sustain the operation of CropSmart Digital Twin.
In a “what if” service, the CropSmart Digital Twin will give a prediction of outcomes. Users can experiment with different decision scenarios and management paths, see their effects, and compare potential outcomes, by posing such questions as “Should we plant today or wait for another week?”
Also, a third service is available where the tool provides a direct management recommendation. For example, the CropSmart suggests planting within the next four days. This recommendation, optimized for the user’s specific decision goal, is automatically generated by CropSmart.