Principal Investigator: John Sulik
Research Institution: University of Guelph
Timeline: April 2021 – March 2022
- Evaluate data aggregation approaches for delta-yield nitrogen (N) fertility on-farm trial data.
- Determine which sensor/field data is essential for scaling up on-farm trials.
- Draft manuscripts about on-farm experiment findings; submit to peer reviewed journals.
- Provide in-house expertise for ongoing long-term cropping systems trials.
- This ‘wrap-up’ funding will facilitate speedy and relevant project deliverables for two on-farm projects aimed at improving N fertilizer management in corn. Over 2 years considerable amounts of crop and soil data were collected on 11 farm fields. While the current data provide value for participating farmers, for wider conclusions to be drawn more advanced data analyses are needed.
- Advanced analysis of these data will take considerable time and skill but will lay an important foundation for the direction of future on-farm research aimed at improving environmental/economic performance of fertilizer N in corn.
- Ken Janovicek’s expertise will ensure that lessons learned, and future research directions reflect the diversity of Ontario corn production environments (e.g., soil texture, GDDs) and the types of tools/technologies that are practical and realistic for on-farm implementation.
- For field-scale and on-farm N fertilizer research to expand in Ontario, a protocol and data management system for farmer-researcher collaboration is needed. Ken’s expertise will help with this.
- Matching funding from the Department of Plant Agriculture will ensure continuity of long-term rotation and tillage projects at the Elora Research Station until a replacement for Bill Deen is found. Ken’s continued work will allow ongoing cropping systems research projects to conclude within a much shorter timeframe and use those project findings to inform new project ideas and funding opportunities.
Considerable data collection occurred in a 2-year on-farm experiment involving 11 farmer-collaborators. Drawing conclusions from this voluminous dataset, which can benefit GFO farmer-members via actionable advice, required a statistician with expertise in field crop agronomy/Ontario corn production. With the retirement of Bill Deen, Ken Janovicek, a long-time corn researcher, filled this gap. The long-term aim of this research is to develop and refine N fertilizer decision support tools for corn tailored to Ontario’s soil and climate conditions.
Progress was made on writing code that will be generally useful for analyzing on-farm strip trial data for “delta-yield” experiments. The software summarizes the deltas using the standard approach of spatially adjacent replicates as well as stratified comparisons by soil type or other management zone scheme. A Bayesian Graphic Interface as a N management decision support system (DSS), which estimates sidedress N rate based on in-season weather and other field data (manuscript in preparation), was also developed.
We tested various machine learning algorithms such as Random Forest analysis and Bayesian Networks, to estimate corn yield and economically optimum N rates. A large dataset of on-farm and research station trials (n=341) was used for this purpose. Our best performing algorithm, a Bayesian network recommendation tool, predicts economically optimum N rates that are within $25 ha-1 of maximum potential profit in 64% of cases. This manuscript was submitted in 2022 for publication in Agronomy Journal and is currently under review.
Further, we also developed an algorithm that provides economically optimum N rate estimates from delta-yield and it produced returns that are within $25 ha-1 of maximum potential profit for 70% of cases (Janovicek et al., 2022).
We completed an analysis of a 22-year (1996-2017) dataset of soil NO3– concentrations in the surface 30 cm collected in early May and mid-June from the long-term Elora Rotation-Tillage trial, which revealed:
- On average, Ontario-recommended fertilizer N credits for first-year corn following alfalfa and underseeded red clover produced using tillage are valid.
- Although soil NO3 concentrations tended to be lower in no-till, they were still great enough to justify the use of at least 75% of Ontario’s recommended alfalfa and underseeded red clover N credits.
- Soil NO3 concentrations can change quickly throughout the spring, especially following alfalfa and underseeded red clover. So it is critical to take PSNT samples during recommended time periods to increase the chance of obtaining an accurate fertilizer N recommendation.
- Under tillage, the fertilizer N credit for second year corn in the corn-corn-soy-wheat (plus under seeded red clover) and corn-corn-alfalfa-alfalfa rotation, may be about 50 kg N/ha; especially in fields with a long history of regular inclusion of these forage crops in rotations.
- Under no-till, the fertilizer N credit for second year corn in fields that have regularly included alfalfa or underseeded red clover may be about 35 kg N/ha.
Our first recommendation is that for farmers to really get a handle on estimating the optimal N rate they need (each year), in-season weather should be used to inform the decision. Weather affects both N loss potential, but also crucially crop N uptake requirements. For farmers who are applying the full N rate at the start of the season, they will have to accept a limitation on how well they can predict fertilizer N requirements.
Secondly, we recommend the use of NNI (Nitrogen Nutrition Index) to guide in-season N rate decision-making. NNI is a crop tissue-based measurement taken anywhere from V6 to V16, that can help guide N rate decisions. We measured NNI on some fields in the on-farm experiment and found that it worked well at identifying areas with low N supply. Many commercial labs have the ability to measure NNI and provide an N recommendation to farmers, and the results are very easily interpretable. Commercial availability of the NNI is another potential benefit.
External Funding Partners:
Department of Plant Agriculture, University of Guelph.
Project Related Publications:
Sulik, J., Banger, K., Janovicek, K., Nasielski, J., Deen, B. 2022. Comparing Random Forest to Bayesian Networks as nitrogen management Decision Support Systems. Agronomy Journal (under review).
Janovicek, K., Banger, K., Sulik, J., Nasielski, J., & Deen, B. 2021. Delta yield–based optimal nitrogen rate estimates for corn are often economically sound. Agronomy Journal : 113(2), 1961–1973. https://doi.org/10.1002/agj2.20521.