Test Agroptimizer across the NC Region: A significant challenge in identifying and adopting the optimum cropping system at the field scale is that multiple factors (more than we can evaluate simultaneously in replicated trials) interact and affect yield. Therefore, the management practices that result in the greatest yield can and will vary among different fields, even within the same region.
An important issue with current research approaches is that the highly variable and farm-specific management costs (e.g., different farmers often pay different prices for similar seeds) should be considered. Therefore, the effect of a management practice (e.g., seeding rate) on yield is disconnected from its associated cost (e.g., $/seed bag). Although this does not negatively affect the recommended rate for optimum yield, it makes profit optimization at the farm level challenging. Consequently, input overapplication, or simply input application when there is no need, can suppress farmer profit. Therefore, results from traditional replicated field trials cannot be used to thoroughly recommend actionable knowledge at the field level since they cannot account for all sources of variability and uncertainty.
Given all the well-known deficiencies of current agricultural research methods, a new machine learning cloud-based decision support tool (Agroptimizer, www.agroptimizer.com) was developed to identify optimum corn and soybean cropping systems for maximum yield and profitability from among thousands of possible cropping systems a farmer can choose from in a single field. Agroptimizer estimates yield by accounting for field location, soil type, weather conditions, and several management practices and uses a combination of methods to evaluate all possible management combinations. Then, the most profitable cropping system is identified by utilizing the estimated yield and production costs specific to each field. Eventually, the cropping systems with the highest probability of success are recommended to the farmer. The spatial coverage of the Agroptimizer is extensive and includes the entire NC soybean production region.
In a recent study, Agroptimizer recommendations were compared against cropping systems generated by University of Wisconsin researchers (experts) across Wisconsin between 2021 and 2023 (Mourtzinis and Conley, 2024). Agroptimizer recommendations increased soybean yield and yielded similar profit compared to experts’ recommendations. Overall, the results showed that Agroptimizer recommendations were similar to those of local experts and identified cropping systems that resulted in high yield and profit.
To date, no other decision-support tools, public or private, are available to US farmers seeking optimal cropping systems (a combination of multiple management practices) at the field level. Evaluating such tools under field conditions involving unexpected and unmanageable yield adversities across a range of growing conditions is important for determining their effectiveness in helping US farmers increase yield and profit across major agricultural regions.
Validate a satellite-assisted field scouting alert system and a digital crop defoliation tool to integrate into the OCM. Even with an extensive field history, it is still difficult to predict where abiotic and biotic stressors will be present within a field due to uncertainty about the anticipated pest-related yield pressure during the growing season and weather conditions. Additionally, there is no rapid method to evaluate the presence and distribution of these yield-limiting factors across entire fields; thus, intensive in-season scouting is needed. Unfortunately, the agricultural workforce trained in scouting is depleted. As a result, many farmers apply inputs prophylactically across the entire field, which, in the absence of pest pressure, results in a negative return on investment, potentially increases pest resistance, and decreases long-term sustainability (i.e., pesticide tools don’t work as well when truly needed). A tool that guides users on where to scout across a large field would be highly valuable to soybean farmers.
As part of the Data-Driven project, we developed a new tool that uses Sentinel-2 satellite images and automatically extracts the Normalized Vegetation Difference Index (NDVI) for every 60 x 60 ft section in a field. We hypothesize that this information can guide precision scouting efforts throughout the growing season. Field areas with low NDVI values may be associated with yield-limiting factors (e.g., pest pressure, weeds, etc.) and should be scouted.
Open Crop Manager (OCM):We have also developed Open Crop Manager (OCM), a cloud-based management decision support tool, with the support of our Data-Driven project. We leveraged this investment with internal and federal funding support to help add additional layers to the platform and begin planning for new tools. The OCM provides a user-friendly platform for in-field crop scouting and data collection, including information on crop production practices at the farm and field scale. The OCM enables data collection from scouting reports for 32 pests, 37 diseases, 48 weeds, and 28 abiotic issues. This intensive scouting method was tested in 2022 and 2023 across ten states, generating 3,695 observations in its first year of application (Cucak et al., 2022) and 4,195 in its second year (unpublished data). Field-scale scouting also resulted in the collection of 7,743 images.
The OCM protects data privacy by requiring user authorization, utilizing data perturbation methods, and generalizing results. Access to a field’s data can be controlled by allowing growers to determine individual collaborators’ level of access to that data. Users of OCM are also made aware of our data privacy policies and their data privacy rights with access-appropriate privacy notices/waivers.
The OCM platform was developed to continuously add new applications. In this funding cycle, we will add a digital crop defoliation tool that uses existing, open-source leaf defoliation estimators (e.g., Leafbyte) to rapidly assess the level of herbivory and return an estimate of whether the sampled leaves are at or approaching an action threshold (i.e., “red light = treat, yellow light = re-sample within a week, green light = well below threshold”). These thresholds are soybean growth stage-specific; some have been in place for decades. They were recently re-evaluated in a multi-state NCSRP-funded objective led by co-PI Nick Seiter, which revealed that the pest complex contributing to soybean herbivory rarely exceeds these thresholds in the North Central region. The integration of this defoliation estimator into the OCM platform would provide a quick, easy, and repeatable way to confirm whether intervention for defoliating insects is necessary – adding a new dimension to its ability to assist in scouting and decision-making.