There is an abundance of information available to farmers to help them manage their crops, but wading through data from multiple platforms and processing it into actionable information is often too time consuming or expensive to be practical. At the same time, the USDA’s Agriculture Innovation Agenda calls for a 40% increase in crop production while decreasing environmental impacts by 50% by 2050. In order to achieve this goal, a more data-driven approach to farming will be required. 

A multi-disciplinary team of researchers, including scientists at the U.S. National Science Foundation National Center for Atmospheric Research (NSF NCAR), aims to make agricultural data more accessible with a mobile app and web-based application called CropSmart. The goal of CropSmart is to revolutionize the way farmers, agribusiness operators, and government agricultural officials make decisions about irrigation strategies, fertilizer applications, and when to harvest. 

CropSmart will provide anyone in the agricultural industry with individualized, data-driven recommendations at no or minimal cost through the use of a digital twin. A digital twin is a virtual replica of something in the physical world and can run multiple simulations to help answer questions for “what if” scenarios. 

Accelerating solutions

CropSmart was one of seven projects selected earlier this year to move to phase two of the U.S. National Science Foundation Convergence Accelerator program’s Track J. The Convergence Accelerator funds teams to solve societal challenges through innovation and convergence research – a method that strives to solve complex research problems by merging expertise across diverse disciplines. 

Track J is a partnership with the U.S. Department of Agriculture (USDA) and focuses on tackling food and nutrition insecurity challenges. Being selected for phase two, provides the team with $5 million in funding to expedite the process of taking the research from the idea phase to implementation. 

The award from the NSF Convergence Accelerator provides the CropSmart team with three years time and funding to get the platform ready for market. Additionally, the Convergence Accelerator provides mentorship, training, and resources that will help the team transition their idea from a prototype to a solution. This includes training in human-centered design, team science, communication, storytelling, and pitching.

“The purpose of the convergence accelerator is to accelerate the transition from research to industrial production and we are leveraging our existing tools to do this,” said Cenlin He, co-PI on CropSmart and a researcher at NSF NCAR. “A lot of the teams in Track J are building physical products, whereas we are building a digital system. We hope to do this in a way that's really impactful and can benefit not just farmers, but consumers.” 

Liping Di of George Mason University (GMU) leads the The CropSmart team which is composed of researchers from Kansas State University, Mississippi State University, NSF NCAR, Purdue University, and University of Nebraska–Lincoln.  

Journey from pilot to proven technology

Currently, the CropSmart team is working with farmers in Nebraska and Michigan that grow commodity crops like wheat, corn, and soybeans to test CropSmart’s predictions against conventional practices. Utilizing neighboring fields, one field acts as the control field with the farmer following traditional practices and the second field follows advice from CropSmart. Across a year, the researchers will compare how the two fields' crop yields and economic gains differ. 

CropSmart’s digital twin uses a community model called Noah-MP-Crop, which is a land surface model that has crop and irrigation capabilities. The model combined with high-resolution remote sensing satellite data and machine learning techniques developed by GMU is able to create a digital copy of a farmer’s fields and deliver actionable information personalized to their farm. 

While each component of CropSmart has been evaluated individually – Noah-MP-Crop has been assessed in previous studies and the machine learning algorithm developed by GMU has been used already to predict things like in-season crop-type mapping – this is the first time the integrated system will be tested as a whole in real world contexts.The researchers hope that the reliability of each piece will result in successful recommendations, but anticipate that some fine-tuning will be needed to take CropSmart from a research tool to a ready-to-use product. 

A site visit highlighted one of the challenges of taking a product from the lab to the real world. Many on the team had never seen an irrigation wheel called a “center-pivot system”, a sprinkler system often used to water large fields, before the site visit. In the Noah-MP-Crop model, irrigation worked like an on/off switch with an entire field instantaneously being watered on demand. In reality, an irrigation wheel slowly moves across a field watering a section at a time and it can take two to three days to water an entire field. 

As a result of this site visit, He and his colleagues are exploring how they can better represent agricultural irrigation practices in their model, which is something no other crop modeling system has previously done.

“This project has been really exciting, but also a new experience for me because it is not just about the fundamental science, but applying it,” said He. “It’s important to see what you are representing in a model in reality because you don’t always know what is missing from a model until you see the thing you’re modeling in real-life. It’s a simple lesson, but it’s not always simple to represent that inside the model.” 

See all News