AI in the fields: Revolutionizing agriculture with smart technology

Researchers from the University of Bonn are revolutionizing farming by leveraging AI and new technologies to increase efficiency and sustainability. In a recent paper, published in the European Journal of Agronomy, the researchers have outlined key research questions that must be addressed to advance the smart digitization of agriculture.

The modern agricultural challenge

Modern, high-performance agriculture has enabled the Earth to feed over eight billion people. However, this success comes with significant environmental costs. Current cultivation methods are endangering biodiversity, synthetic fertilizers contribute to greenhouse gas emissions, and agricultural chemicals are contaminating bodies of water and the wider environment.

Many of these issues can be mitigated through more precise agricultural methods. For example, applying herbicides only where weeds are problematic rather than across entire fields can reduce chemical use. Similarly, treating only diseased crops and applying fertilizers where truly needed can optimize resource use. However, such strategies are complex and challenging to implement on a large scale using traditional methods.

Embracing smart technologies

“One answer could be to use smart digital technologies,” explains Hugo Storm, a member of the PhenoRob Cluster of Excellence. The University of Bonn has partnered with Forschungszentrum Jülich, the Fraunhofer Institute for Algorithms and Scientific Computing in Sankt Augustin, the Leibniz Centre for Agricultural Landscape Research in Müncheberg, and the Institute of Sugar Beet Research in Göttingen on the large-scale project geared toward making farming more efficient and more environmentally friendly using new technologies and artificial intelligence (AI).

The team includes experts from various disciplines, including ecology, plant sciences, soil sciences, computer science, robotics, geodesy, and agricultural economics. In their paper, they outline the steps they consider a priority in the short term. “We’ve identified a few key research questions,” Storm says. One of these relates to monitoring farmland to spot any nutrient deficiency, weed growth, or pest infestations in real-time. Satellite images provide a rough overview, while drones or robots enable more detailed monitoring. The latter can cover a whole field systematically and even record the condition of individual plants in the process. “One difficulty lies in linking all these pieces of information together,” says Storm’s colleague Sabine Seidel, who coordinated the publication together with him: “For example, when will a low resolution be sufficient? When do things need to get more detailed? How do drones need to fly in order to achieve maximum efficiency in getting a look at all the crops, particularly those at risk?”

Data integration and analysis

The data obtained provides a picture of the current situation. However, farmers are primarily interested in comparing the various potential strategies and their impacts. Questions like the tolerable amount of weeds, necessary fertilizer amounts, and the consequences of reducing pesticide use require precise answers. “To answer questions like these, you have to create digital copies of your farmland, as it were,” Seidel explains. “There are several ways to do this. Something that researchers still need to find out is how to combine the various approaches to get more accurate models.” Suitable methods also need to be developed to formulate recommendations for action based on these models. Techniques borrowed from machine learning

Machine learning is a subset of artificial intelligence (AI) that deals with the development of algorithms and statistical models that enable computers to learn from data and make predictions or decisions without being explicitly programmed to do so. Machine learning is used to identify patterns in data, classify data into different categories, or make predictions about future events. It can be categorized into three main types of learning: supervised, unsupervised and reinforcement learning.

" data-gt-translate-attributes="[{"attribute":"data-cmtooltip", "format":"html"}]" tabindex="0" role="link" style="margin: 0px; padding: 0px; border-width: 0px 0px 1px; border-top-style: initial; border-right-style: initial; border-bottom-style: dotted; border-left-style: initial; border-top-color: initial; border-right-color: initial; border-bottom-color: rgb(0, 0, 0); border-left-color: initial; border-image: initial; font: inherit; vertical-align: baseline; text-decoration: none !important; color: rgb(0, 0, 0) !important;">machine learning and AI have a major role to play in both these areas.

Farmer involvement and adoption

For these digital advancements to be fully integrated into crop production, farmers must understand and accept the benefits. “Going forward, we’ll have to focus more on the question of what underlying conditions are needed to secure this acceptance,” says Professor Heiner Kuhlmann, a geodesist and one of the Cluster of Excellence’s two speakers alongside the head of its robotics group Professor Cyrill Stachniss. “You could offer financial incentives or set legal limits on using fertilizer, for instance.” The effectiveness of tools like these, either on their own or in combination, can likewise be gauged nowadays using computer models.

In their paper, the researchers from PhenoRob also use examples to demonstrate what current technologies are already capable of. For instance, a “digital twin” of areas under cultivation can be created and fed a steady stream of various kinds of data with the help of sensors, e.g. to detect root growth or the release of gaseous nitrogen compounds from the soil.

“In the medium term, this will enable levels of nitrogen fertilizer being applied to be adapted to crops’ needs in real time depending on how nutrient-rich a particular spot is,” Stachniss adds. In some places, the digital revolution in agriculture is already closer than one might think.