The N technologies commercially available were grouped as (1) crop model-based, (2) remote sensing-based, (3) enhanced efficiency fertilizers, and (4) biologicals ( Figure 1 ). To reduce the barriers of testing a new technology partnering farmers were provided with financial compensation to negate the cost and risks associated with a new N management plan.

A novel OFPE design was used to test the N technologies ( Figure 2 ). The experimental design combined field length strip-trials and site-specific small N rate blocks with at least four repetitions. The strip-trials were used to compare farmers’ traditional management (“business-as-usual” N management) to the N technology they were interested in (“next-level” N management). For corn most of the participating growers applied a flat N rate either in the fall, pre plant or at planting with a proportion of the fertilizer going as sidedress. For winter wheat most of the growers applied the fertilizer at planting or early in the spring.

Nitrogen rate blocks were placed in contrasting zones within the field that were expected to have a contrasting yield response to N. For technologies that provided flat or variable rate N recommendations such as model-based and sensor-based N tools, the technology was evaluated in field-length strips and the N blocks were placed near by the strip trials in contrasting management zones ( Figures 2A , 3A, B ). Recommended N rates for the field-length strips and N blocks were assembled into a VRA that was implemented on-the-go using the farmer’s VRA controller ( Figure 3C ). Nitrogen rate blocks for the biologicals and enhanced efficiency fertilizers strip-trials were implemented in a split-plot design ( Figure 2B ). Technologies were changed manually (in the case of enhanced efficiency fertilizers) or applied with a “split planter” approach (in the case of biologicals). Nitrogen rate blocks were implemented as a prescription via the farmers VRA controller.

Before the implementation of the field trial, a soil characterization was performed to measure organic matter (OM) and soil texture stratified by depths at contrasting yielding areas of the field. During the growing season, soil moisture and temperature, soil nitrate, crop phenology, plant biomass, high-resolution imagery, and leaf area index was measured (LAI; Figure 3D ). Farmers provided as-applied and yield monitor data for the field-scale trials ( Figure 3E ).

An automated OFPE data processing, management, analysis, and reporting system was developed to enable robust and quick data processing ( Figures 3F–I ). This system aggregates data layers from various sources and implements data quality control methods to check for overlapping, misalignment, or outliers within yield and as-applied data. The system does not eliminate yield observation, instead, they get flagged when issues were found. The system was coded in R software (R Core Team, 2022) to enable rapid processing of the on-farm research trials.

Partnering farmers were engaged throughout the process by first letting them select the technology to test and by providing them with hands-on experience. To guide this process, we utilized in-depth discussions with farmers, their crop advisors, extension educators, graduate students, and specialists to first understand the farmer’s current N management and technology capabilities and then to guide the selection of technology. This customized, farmer-centric approach increased the potential for future adoption of the technology tested and allowed growers to progressively increase the complexity of their N management. For example, 25% of the growers had no in-season N application experience and they were more receptive to test enhanced efficiency fertilizers than in-season N management tools. However, the remaining participants with in-season N application capabilities were open to test remote-sensing or crop model-based tools with their available capabilities and logistics in place.

A total of five public-private partnerships with industry were established to provide growers with access to the N technologies ( Figure 1 ). Partnerships with N technology providers played a critical role in ensuring that technologies were available to the grower and implemented correctly. This facilitated technology transfer with expert input and allowed graduate students to be supported through industry collaborations. In addition, three on-line workshop training sessions were organized to train the PNP team and growers on how to use some of these technologies. Workshops were highly beneficial for the crop model and sensor-based N tools due to their input requirements and complexity compared to for example, enhance fertilizer and biological products.

From 2020 to 2022, the PNP project completed 80 trials in corn and winter wheat. Of these trials, technologies selected were 39% crop model-based tools, 34% remote sensing-based, 21% enhanced efficiency fertilizers, and 6% biologicals. Biologicals were offered as an option for the first time in 2022 and is expected to increase in 2023. 98% of the experiments were successfully established.

The automated data processing method developed for N technology testing and benchmarking with the site-specific optimal N rate encompassed yield and as-applied data processing, data aggregation and reporting of results. Each step is described in Sections 3.3.1 to 3.3.4.

The automated data method enabled the PNP team to reduce the data processing time, cost, manpower, and enabled more efficient result sharing with farmers right after harvest ( Figures 4 – 9 ). 90% of the trials were analyzed using the automatic process and the remaining trials were analyzed manually due to data quality issues. The PNP team was able to analyze site-specific yield response to N curves for winter wheat and corn in more than 70% of the experiments.

Results were published in the annual on-farm research results extension publications (Thompson et al., 2021; Thompson et al., 2022). Results were shared with over 200 individuals annually through the on-farm research meetings and 12 presentations. Additionally, individual meetings were held at the end of the growing season to share results with the partnering farmers.

Farmer comments and stories revealed they were more comfortable using technology because of participating in this project. One producer noted, “I’ve had crop canopy sensors for years but didn’t feel confident using them. Now that I’ve seen the results, I will use them farm wide.”

Growers also benefitted from seeing the results of the NUE analysis for their management practices. One producer commented, “I’m shocked that our NUE is 1.1 [lb N/bu grain]. I want to push the efficiency below 1 [lb N/bu grain]. I was planning on purchasing some more fertilizer for the upcoming year, but now that I see these results, I think what I have is enough.”

Although the research still on-going and while limited, farmer statements provided evidence that the proposed N technology testing approach was having immediate impact in growers’ NUE and in their willingness to adopt technology. For example, a central Nebraska producer participated in a study looking at sensor-based fertigation technology. Using the responsive sensor-based fertigation approach, the grower was able to reduce N fertilizer rate by nearly 100 kg N/ha compared to his normal management with no impact on yield. Reflecting on his results, the farmer stated, “the project has proved to me that with the sensor-based approach, we can get by with a lot less nitrogen.” In the following year, he adopted this technology across his farm. Quantitative data supporting the impact from the described N technology testing framework will be published future publications.

The unique OFPE methodology presented here allowed researchers, agronomist, and growers to test a combination of DA technologies to direct N rate applications and/or to enhance the efficiency of the N fertilizer ( Figures 1 , 2 ). Compared to small plot N trials (Bullock and Bullock, 1994; Makowski et al., 2002; Kitchen et al., 2017), large-scale N trials (Bullock et al., 2019), and traditional strip trials (Sela et al., 2016; Kyveryga et al., 2018), the PNP framework provided three major novel aspects: on-farm testing of commercially available N technologies, side by side comparison of the N technology with growers’ current N management, and unbiased benchmarking of both N management approaches using site-specific EONR ( Figures 8 , 9 ).

Major advances in N management have been made by employing OFPE. For example, the Data-Intensive Farm Management Project (DIFM) (Bullock et al., 2019), enhanced the utilization of small-plot trials at a commercial field level, allowing site-specific determination of the EONR and evaluation of novel modeling strategies (De Lara et al., 2023; Hegedus et al., 2023). While this method allowed growers’ current N management to be benchmarked against the observed EONR, there is no tool available for growers to utilize trial results to direct future site-specific N recommendation. Thus, the applicability for these trials to result in immediate change in growers’ management and adoption of more efficient N management practices is limited. Although the framework presented here is still under evaluation for one more year, grower attitudes following participation revealed an immediate increase in the grower’s knowledge of the technology, willingness to continue using the tested practices beyond the life of the trial, and rapid diagnosis of their current NUE. This supports the hypothesis that solving N related environmental issues in agriculture cropping systems will need to first rely on determining NUE of current N management practices (Tenorio et al., 2020; Cassman and Dobermann, 2022).

Strip trials, on another hand, are often used for testing products (e.g., N inhibitors, N sources), limited numbers of N rates (Kyveryga and Blackmer, 2012) and N tools (Sela et al., 2016) compared to growers’ standard management. However, in most cases these trials do not have sufficient treatment layout and data to determine EONR, thus unbiased benchmarking of the proposed N management is not possible. It is important to note, that the PNP framework took the evaluation of special fertilizers and biologicals to the next level ( Figure 2B ). First, site-specific performance of the product can be evaluated by aggregating metadata and auxiliary information about soil and crop characteristics (Section 2.2.3). Second, testing N technologies under an array of N rates below and above grower’s traditional N rate, allows to determine impact of these technologies on EONR and under what N level the technology offered the biggest benefits. This is particularly true for the case of biological products that are aimed at reducing total N rate but often applied at excessive N rates, undermining potential benefits. Similarly, N inhibitors aimed at protecting the synthetic N fertilizer source from losses, were beneficial when the N rate is already optimized. The PNP approach enhanced the testing of these technologies on farmer fields, at commercial scales, and across diverse landscapes.

The site-specific EONR determination provides a unique and unbiased capability to evaluate both new technology and growers’ current N practice on a whole-field and site-specific basis ( Figure 2 ). Soil types, management zones, and other crop and soil measurements can be used to derive where the technology performed the best and where the recommended N rates were closer to the observed EONR. For example, in the field shown in Figure 9 , we observed EONR values higher than both grower and model-based N management in two of the three replications demonstrating that both approaches were potentially under-recommending N in these areas. This illustrates the potential for the insights derived from benchmarking to be utilized for further technology improvements.

This overview of the PNP introduced the main steps for the trial design, data processing, and analysis as part of a novel framework for conducting on-farm experiments to test N technologies and benchmark them with grower’s management practices and the observed EONR. However, in some cases, lack of access to GPS to enable site-specific applicators and planters limited the growers that could participate in this project. Despite this, the future of these testing strategies is promising as national trends suggest that use of GPS and VRT increased steadily since 1990 (McFadden, 2023).

There is a general agreement that requisite data handling, processing, and interpretation are major bottlenecks limiting advancements in digital agriculture (Kutter et al., 2011; Wolfert et al., 2017; White et al., 2021). Usually, data sets derived from OFPE trials are characterized by unstructured, heterogenous data coming from various sources which requires interaction with skilled data scientists and domain experts. The protocols applied in PNP for data processing, analysis, and reporting (Section 2.2), constitute important steps toward reducing these barriers. In addition, the open-source software has the potential to expand the approach to more users, increasing knowledge of N technologies performance and promoting adoption. The PNP is planning to release the source code for public use by the end of the project in 2024.

The PNP developed and validated a successful framework to accelerate adoption of proven and industry available solutions (Scharf et al., 2005; Weber and McCann, 2015; Sela et al., 2016; Morris et al., 2018). The recorded trends in this project indicated that more than 50% of the growers continue to use the N technology tested during this project. Although a more comprehensive assessment is needed to confirm this observation, we conclude that several barriers are eliminated once the technology is explained in detail to growers and their advisers and when agronomic, economic, and environmental benefits from technology are clearly communicated ( Figure 8 ). Access to training, close relationships with industry partners, compensation for participation, and reduced cost of technology are among other factors that support the proposed framework.

Growers are pulled in different directions due to a vast offering of DA tools and often do not have the support to evaluate the effectiveness of a particular N technology in their own operation (Bolfe et al., 2020; Puntel et al., 2022; McFadden, 2023). This is evidenced by the larger proportion of growers testing and/or implementing products compared to other technologies (models and sensors) (Castle et al., 2015; Weber and McCann, 2015). In addition, products such as N inhibitors or biologicals are simpler to utilize than other technologies. Crop models and remote sensing-based technologies require larger data inputs, more knowledge and investment. However, as already presented, the PNP removed some of these limitations. This confirms that incentive programs that provide financial and/or technical support like those existing for other management practices (e.g., cover crops, no-till, etc.) could greatly impact the adoption of N technologies. Incentive programs to increase adoption of N technologies should be considered as: (1) increasing NUE and reducing environmental N impacts while maintaining production remains a major challenge of modern agriculture and (2) most of the commercially available technologies tested by the PNP provided an increase in efficiency compared to typical grower management. Thus, the effectiveness of the PNP model serves as evidence to support incentive programs for DA adoption in agriculture.

In addition to the immediate benefit to the growers of testing N technologies in their own fields, this state-wide method is building a substantial dataset for future technology testing with a standardized protocol. This confirmed that public-private partnerships could lead to major contributions to develop or validate N fertilizer management practices (Kitchen et al., 2017; Ransom et al., 2021). Similar public-industry partnerships should be encouraged. Although this research was conducted in the U.S., the methodology and research outcomes will be applicable to agronomic studies across larger regions throughout the world. Further, the framework presented here can be expanded to other practices such as sensor- and model-based irrigation management or chemical use.