(Source: Nuttapong punna/stock.adobe.com)
Precision agriculture leverages technology to enhance crop productivity and inform decision-making on farms. Its core principles are utilizing data to understand variability and then targeting interventions to match specific needs found across fields.
By assessing and responding to nuanced differences within farms, rather than taking a blanket one-size-fits-all approach, precision agricultural practices aim to increase efficiency, productivity, profitability, and sustainability. This data-driven concept has seen increasing global adoption to help address growing food demands while conserving resources and the environment.
This blog dives deeper into the challenges farmers face, where technology is ready to solve them, and where it falls short. Mostly, farmers are keen to adopt precision technology and are counting on engineers to make it viable sooner rather than later.
Globally, the overarching challenge to farmers is producing enough food to feed ten billion people by 2050 while also reducing the impact on the farmer, the farm, and the environment.[1] Farming has few guarantees, and farmers face a barrage of variables that are out of their control—every season, day, and hour can look very different.
Speaking at the FCC's Precision Ag Connectivity Task Force,[2] Ryan M. Krogh, Manager of Production Systems and Program Management at John Deere, outlined the first critical challenge. Krogh explained that for farmers, many key variables lie outside their control. Therefore, as a fundamental criterion, agricultural technologies must enable managing uncertainties, helping farmers bend inevitable volatility in their favor.
A second major challenge is labor. According to the US Department of Agriculture (USDA) 2022 Census of Agriculture, farmers in the US are putting in long hours.[3] Based on 2016 data, principal farm operators' weekly on-farm hours varied significantly based on commodity type. Dairy farmers worked the most at 64 hours per week, while those managing cotton, peanut, and rice farms also put in over 40 hours. It’s important to note that many farmers also work off-farm to make ends meet.
Technology is a logical way to help farmers work smarter. Krogh said that it is important not to think of farm automation as a piece of equipment but instead as a group of engineers, farmers, and logistic experts packed into that machine to make something happen for a reason. In agricultural automation, technology must help the farmer make decisions based on what is happening at every given moment while increasing profitability, sustainability, and productivity.
One example of technology-based decision-making comes from Andrew Nelson, a fifth-generation farmer and software engineer, who leverages hyperlocal weather forecasting across his 7500-acre farm in eastern Washington state.[4] Nelson uses IoT devices such as sensors throughout his fields to measure soil and ambient air temperature. He feeds this field-level sensor data via a mix of satellites, LoRA®, cellular, and Wi-Fi® into Microsoft Research Project FarmVibes. Project FarmVibes then produces microclimate temperature predictions at the field level, helping Nelson generate tailored predictions of frost risk for each individual farm field. Because the models are trained on ground truth data from the farm's microclimate, they can forecast overnight freeze potential much more accurately than general weather reports.
Accurately predicting frost occurrences allows precise targeting of crop protection chemicals and treatments to maximize yields. If a late freeze threatens productive areas of a field, Nelson can use the system's guidance to know exactly when to apply preventative spray treatments.
The Project FarmVibes machine learning (ML) models provide similar hyperlocal guidance by pairing weather data with soil moisture sensors across fields. When trying to plan field operations, it can help assist with understanding how saturated the soil is across the field and determining if he should attempt field operations with heavy equipment. This allows proactive field management.
In essence, Nelson demonstrates how artificial intelligence (AI) and ML models can transform raw sensor data into actionable insights for the farmer by accounting for the unique conditions on that farm. As sensor networks grow in agriculture, increasing value from the data will require intelligent systems tailoring broad information to field-specific recommendations.
Nelson also uses drones with AI and visual, multispectral, or normalized difference vegetation index (NDVI) cameras (which capture plants' reflected red and near-infrared light) to better detect weeds and spray only areas requiring treatment. The imagery vividly shows locations where weeds are encroaching on the crop rows. Because it would be impractical to manually scan drone images across thousands of acres to identify weed hotspots, Nelson uses AI models trained on tagged drone imagery that can automatically analyze each image as it is captured. The models identify and map any weeds in the field down to precise geographic coordinates.
By knowing exactly which field portions require intervention, Nelson can leverage GPS-based auto-nozzle control technology on his spraying equipment to target applications exclusively to weed-infested rows. This enables an incredible gain in efficiency compared to broad blanket spraying. The technology allows more precise field treatment while using a fraction of herbicides, which translates directly to cost savings and better protection of lands and communities.
It’s a better solution, but Nelson points out it’s not without challenges. Nelson said that a significant problem is connectivity: "We’re working on this with Microsoft, but you have to get your device to connect in a low-cost way over very long distances; that is a very hard problem. The middle ground of bandwidth is very hard to cost-effectively deploy on a farm."
Outside of cost and the need to protect components against the harsh environments of agricultural settings, the most significant impediment to incorporating this kind of technology is connectivity. An engineer can only go so far in solving that issue.
High-speed internet is available in only some rural communities, and internet options are limited in general. Even if a farmer does get excellent high-speed service delivered to the house or barn, it won’t solve the fundamental problem: a command center must be mobile because the farmer does not sit still.
For example, an autonomous tractor will have on-machine edge computing with all the intelligence to execute the job. Still, autonomy needs continuous communication and connectivity because if the tractor doesn’t know what it’s seeing, it will need intervention. That intervention might be a person checking a mobile app and directing the tractor to move around the object, drive over it, or pause until someone can move it. So, that’s just a simple wireless signal, right?
But getting consistent wireless signals can be very difficult in many types of farmland. The connectivity infrastructure relies on cell and radio towers being in sight range. As farm equipment moves through hills, valleys, trees, or even vast spaces, it can lose line of site with a tower. This means the signal gets disrupted or lost altogether.
The same issue applies to sensors placed in the fields to collect data. If a hill is blocking the view of a tower, that sensor may not be able to transmit its data back to the farming systems, or that connection could be intermittent as signals come in and out of range. This is a huge problem if a farmer wants to leverage all equipment and field sensor data for operational insights. The inconsistent connectivity renders much of that data inaccessible when it's needed to make critical decisions.
Cost is an issue even if the farm is spread across miles of flat land. A remote crop monitoring system that requires the placement of hundreds or thousands of cameras and soil sensors at US$5–10 per month per cellular connection quickly becomes quite costly.
This dynamic severely limits the viability of sensor-heavy solutions for agriculture. Only some farmers can justify spending hundreds of thousands yearly in cellular data plans to enable sensor-based software or analytics.
However, several new wireless technologies offer hope to cut these connectivity costs radically:
Over the coming years, these and related innovations could eliminate connectivity as a barrier to technology adoption in agriculture. With low operating costs, almost any sensor or hardware solution then becomes viable based on productivity gains rather than data costs. But today, solving the fundamental issue of reliable, low-cost connectivity over long ranges in the undulating countryside remains an active challenge.
Technology in agriculture stands poised to transform productivity and sustainability on the farm through greater connection, automation, and data utilization. Cloud computing and AI are fueling tremendous progress in leveraging technology tailored to agriculture. However, critical barriers still restrict widespread adoption by average farmers. Solving persistent challenges to reliable connectivity, interoperability, usability, and costs remains imperative.
The operations side of agriculture also warrants more focus from technologists. Testing innovations under actual working conditions on farms allows designing for pragmatic needs. As innovators continue addressing pressing pain points, technology promises to usher agriculture toward a more precise, efficient, and environmentally responsible future—maintaining productivity while stabilizing the foundation that food and communities depend on. Realizing technology's full potential in agriculture requires embedding it in the context of modern farming.
Sources
[1] https://www.wri.org/insights/how-sustainably-feed-10-billion-people-2050-21-charts. [2] https://www.fcc.gov/news-events/events/2023/11/precision-ag-connectivity-task-force-meeting-november-2023 [3] https://www.nass.usda.gov/Publications/AgCensus/2022/Full_Report/Volume_1,_Chapter_1_US/usv1.pdf [4] &Andrew Nelson, interview with the author