Research

Better understanding farmer-centric on-farm experimentation and the role of data in this process.

Survey results identified key barriers, such as lack of time, limited equipment, and difficulties tracking and analyzing results, as well as strong interest in collaborating with researchers. These findings helped clarify where research and extension can add value, guiding future farmer-centric programs. A key issue identified through the survey was that farmers face significant time and resource constraints, which limits their ability to document, track, and analyze experiments despite widespread engagement in experimentation. This finding suggests some benefit in identifying support roles for research and extension that reduce the burden of experimental complexity and complement existing farmer practices, rather than trying to replace them. By documenting how farmers experiment and identifying barriers that limit learning and collaboration, the project supports the development of research and extension approaches that help farmers progress towards their sustainability goals (e.g., improve input use efficiency, build soil health) while maintaining productivity.

Farmers benefited from this project by having their on-farm experimentation practices formally recognized, validating their role as active contributors to agricultural innovation, rather than passive recipients of research.  Extension educators and researchers benefited from a clearer, evidence-based understanding of how farmers experiment, what limits their ability to do so, and where support would be most valuable. This project contributes to public trust in agricultural research by promoting collaborative models that respect farmers’ knowledge, data ownership, and decisionmaking. These approaches support more transparent, inclusive, and socially responsible agricultural innovation systems that ultimately benefit food security, environmental quality, and rural communities. 

Monitoring crop phenology by remote sensing using the ND-Space

Crop phenology (i.e., crop growth stage) informs multiple decisions in agricultural management at both global and farm scales. For instance, it is important to know the growth stage to decide on the best weed control strategy. Following crop phenology in time can also inform farmers, agronomists, and governments on the normal or abnormal growth of the crop, helping to detect anomalies or anticipate yield. 

Monitoring crop phenology consists of visiting the field and inspecting the crop. Remote sensing can be used to estimate crop phenology but is prone to error. With remote sensing, we cannot distinguish a small healthy canopy from a tall senescing canopy, for instance, because both will display similar chlorophyll concentrations. To address this limitation, we propose to look at the data in 2 dimensions rather than the conventional one-dimension (e.g., single vegetation index) approach. When we use one dimension for detecting chlorophyll concentration (e.g.,  NDVI) and one for water concentration (e.g., NDSW), we can now distinguish phenological stages throughout the crop growing season. 

This project was done using hyperspectral data, allowing us to use any spectral band, and we are now working to see if we can get similar results using free multispectral data such as Landsat imagery.

Assessing corn N status using dynamic data

Monitoring corn N status in real time to inform fertilizer management is the Holy Grail of precision agriculture. It is difficult because N is dynamic in the plant (i.e., N can be reallocated from the lower leaves to the top of the canopy) and in the soil (i.e., it can be washed away with rain and replenished with fertilizer or microbial activity). Yet, it is important to get N fertilization right because if in shortage, yield will take a hit and if in excess, it will be released in the biosphere where it is very reactive and can cause multiple environmental issues (e.g., eutrophication or nitrous oxide emissions). 

To tackle this complex and important problem, we propose to use dynamic information provided by Planet Lab who launched a fleet of shoebox sized satellites that can revisit your field daily. We are looking at crop growth (i.e., height), ground-based vegetation indices, as well as satellite-based vegetation indices at a high frequency (e.g., 3 times per week). With this information, we hope to determine if this dynamic information will bring new insights that were not available when looking at static data (e.g., remote sensing on just one date).

GeoDART, a practical context data retrieval tool for on-farm experimentation

The Farmers DataLab works with farmers to support their experimentation process. This means that we visit several sites each year from which we collect data to learn new information on a site and a network basis. Indeed, when working with a network of farmers experimenting on the same topic (e.g., optimizing N fertilizer in corn after cover crops), we can bring the data from each of these experiments in the same database and start learning things that may not have been possible to detect with just one site on one farm (e.g., the effect of weather pattern). 

When bringing these data into the same database, it is important to document the context in which each experiment took place. Retrieving data layers for dozens of sites can be daunting and complex, requiring expertise in data science. Yet, this step is important for better interpretation if observations can be extrapolated to other sites and seasons. 

The Geographic Data Retrieval Tool (GeoDaRT) was developed to simplify this process and allow non-experts to retrieve site-specific data when conducting on-farm experimentation systematically.

Interseeding cover crops in corn with a farm robot.

Cover crop is an important management practice to improve soil health. For crops that are harvested in the middle of the summer like wheat or that are short like soybean, it is convenient to give cover crops and early start before Fall freezing. However, for tall crops like corn, planting your cover crop can become a headache. Farmers either go early in the season while they can circulate in the field with their tractor or after harvest when they are growing sileage corn. For grain corn, the grain is just harvested too late for any chance of planting a cover crop for it to grow before freezing. What if we could use a robot to plant the cover crop right at the time when the corn stops maximizing light capture to focus allocating resources to grain filling?

In this project we are experimenting with the Rowbot, a farm robot that can circulate between the corn rows and seed the cover crop at any time during the crop growing season. We are comparing in-season interseeding to other farmers’ practices (i.e., early or late planting) for three cover crops species: red clover, cereal rye and triticale.