AI: Entering a new era of food production??
By Berenice Di Biase , 26.04.2024
Agricultural activity faces unprecedented pressures: unpredictable and extreme weather due to climate change, ecological degradation and biodiversity loss, increased food demand due to a growing population. To sustain agricultural activity while not contributing to the aggravation of climate and environmental threats, innovative solutions must be deployed. The advent of artificial intelligence (AI) has enabled us to analyse and learn from large sets of data coming in all shapes and forms, catalysing the development of precision agriculture for the optimization of agricultural management.
Satellite data for decision making
A resilient and sustainable global food system is undoubtedly reliant on the recovery and maintenance of soil health. Soils are extremely heterogeneous, even at small spatial scales, the physicochemical and biological properties of soil are highly varied. AI offers the opportunity to increase productivity by shifting from a one size fits all approach to a soil centric precision approach when it comes to the application of agricultural inputs, irrigation and cropping system.
Modern satellite missions offer a wealth of publicly available data which can be used for the monitoring of soil health. For instance, the European Sentinel 2 offers high resolution (10 meters x 10 meters) data from the high-resolution multispectral imager with 13 spectral bands. These individual spectral bands can be combined to derive a signal able to predict specific soil properties and soil health parameters, such as soil moisture, organic carbon content, and nutrient levels.
By integrating this data with AI algorithms, we can create detailed soil maps that provide real-time insights into soil conditions across vast agricultural landscapes. These maps enable farmers to apply the precise amount of fertilizers, water, and other inputs exactly where they are needed, reducing waste and enhancing crop yields. Additionally, AI-driven soil health monitoring helps in early detection of soil degradation, allowing for timely interventions that can prevent long-term damage and promote sustainable farming practices.
Yield and Ecological Management: Precision Pesticide Application
Annually it is estimated that 20-40% of global yields are lost to pests and pathogens, potentially threatening food security and exacerbating food inequality. Crop control agents, when appropriately administered, are a crucial defence against yield losses.
Global pesticide use has been steadily growing and has more than doubled since the 90s. In 2021 the FAO estimated t , which is the equivalent of over 600,000 elephants. However, the overuse of crop protection products like pesticides poses several ecological threats. Inappropriate pesticide use has been linked to biodiversity loss, pollinator decline, and contamination of soils and groundwater – thus potentially threatening the long-term sustainability and productivity of our food systems as well as human health.
Finding a solution aiming to decrease pesticide use while ensuring yield stability, environmental protection, and fostering a sustainable and regenerative food system is not trivial. The reduction of pesticide sprays during the cropping period could decrease yields, while decreasing the applied dose could decrease product efficacy and increase the risk of pest genetic resistance to the product. Technology offers a viable solution, where pesticide use could be reduced by adopting more targeted application schemes.
Advanced algorithms based on image recognition powered by AI allowed for the development of tractor mounted herbicide sprayers, able to recognize weeds within a fraction of a second, and spray the herbicide on it. Although weed recognition based on image analysis is the most developed product, these sprayers can be equipped with state-of-the-art AI systems and sensors that ensure the precise application of pesticides, herbicides, and fertilizers. Utilizing variable rate technology (VRT), these sprayers can adjust the amount of chemical applied based on real-time data about soil conditions, moisture levels, and pest presence. GPS guidance systems further enhance accuracy, enabling farmers to cover fields uniformly without overlaps or missed spots. Additionally, these sprayers collect and analyze data during operation, providing insights that help refine future applications, reduce waste, and minimize environmental impact.
Keeping up with the times: AI driven development of new crop protection products
General pesticides tend to have a broad range of target organisms, meaning that when applied they can affect organisms beyond the intended pest. Broad-spectrum pesticides, while effective at eliminating a wide range of pests, pose significant risks to non-target organisms, including essential pollinators like bees. The indiscriminate nature of these chemicals means they do not differentiate between harmful pests and beneficial organisms, leading to substantial ecological disruption.
For example, organophosphate based pesticides are one of the most used pesticide classes. The primary mode of action of organophosphate insecticides is the inhibition of a biochemical pathway common to many organisms. The pesticide inhibits the enzyme acetylcholinesterase, leading to the accumulation of acetylcholine, a key neurotransmitter crucial for the functioning of the nervous system, eventually killing the insect. Because the targeted biochemical pathway is conserved across different forms of life, organophosphates, if ingested, can also harm wildlife.
However, crop protection development is undergoing a shift both methodologically and ideologically. There is growing demand and interest for pesticides whose active principle targets biochemical pathways only common to the target pests, thus reducing any negative impacts on wider organisms. Similarly to drug discovery in the pharmaceutical industry, AI has revolutionized the development and discovery of crop protection chemicals. By leveraging AI and machine learning, scientists can sift through vast chemical libraries, identifying small molecules that specifically target pest species without affecting non-target organisms.
AI-driven methods streamline the pesticide discovery process by using computational screening of extensive virtual libraries. Machine learning models, particularly those trained on DNA-encoded small molecule libraries, analyse binding affinities and structural motifs to predict promising candidates. These models utilize next-generation sequencing data, which enables the rapid and efficient identification of molecules with desired properties. Consequently, AI can uncover subtle patterns and interactions that human analysis might miss, thus enhancing the precision of target molecule selection. This approach not only accelerates the development of environmentally safe pesticides but also significantly reduces costs and time, making it a pivotal innovation in sustainable agriculture.
Conclusion
In conclusion, the integration of artificial intelligence into agriculture presents a transformative opportunity to address the challenges faced by the sector today. From optimizing soil health and enhancing crop yields to reducing pesticide use and discovering safer crop protection products, AI-driven technologies are at the forefront of sustainable and precision farming and can
As we look to the future, the continued advancement and adoption of AI in agriculture will be crucial in meeting the growing food demands of an expanding global population while mitigating the adverse effects of climate change and ecological degradation. We at Soilytix are riding this wave of technological change. We are using AI to understand how to improve agricultural productivity at field level while safeguarding our soils via targeted interventions, while also developing our proprietary remote sensing algorithms to inform soil sampling strategies. By embracing these technological innovations, we can pave the way for a more resilient, sustainable, and equitable global food system.
Resources:
- Douglas, M. R. et al. Putting pesticides on the map for pollinator research and conservation. Sci Data 9, 571 (2022).
- H. Kashani, M., Ghorbani, M. A., Shahabi, M., Naganna, S. R. & Diop, L. Multiple AI model integration strategy—Application to saturated hydraulic conductivity prediction from easily available soil properties. Soil and Tillage Research 196, 104449 (2020).
- Lushchak, V. I., Matviishyn, T. M., Husak, V. V., Storey, J. M. & Storey, K. B. Pesticide toxicity: a mechanistic approach. EXCLI J 17, 1101–1136 (2018).
- McCloskey, K. et al. Machine Learning on DNA-Encoded Libraries: A New Paradigm for Hit Finding. Med. Chem. 63, 8857–8866 (2020).
- Savary, S. et al. The global burden of pathogens and pests on major food crops. Nat Ecol Evol 3, 430–439 (2019).
- Sun, Z., Liu, F., Wu, H. & Zhang, G.-L. Developing a national black soil map of China through machine learning classification. CATENA 240, 107993 (2024).
- Zanin, A. R. A. et al. Reduction of pesticide application via real-time precision spraying. Sci Rep 12, 5638 (2022).
- https://www.esa.int/Applications/Observing_the_Earth/Copernicus/Sentinel-2/Introducing_Sentinel-2
- https://www.fao.org/pest-and-pesticide-management/about/understanding-the-context/en/
- https://www.unep.org/news-and-stories/story/bees-bans-and-broad-spectrum-pesticides