AI promises precision for agriculture, but farmers weigh intuition against technology

The agricultural sector is exploring the integration of artificial intelligence to refine critical tasks like predicting crop ripeness and optimizing harvest schedules. While AI offers the potential for increased efficiency and productivity, its adoption by farmers is tempered by considerations of cost, complexity, and the enduring value of human intuition.

Joel Carter of Okanagan Specialty Fruits, a Washington State-based company, highlighted the challenges faced during last year's apple harvest. Unforeseen extreme heat forced a halt to picking operations, underscoring the need for more dynamic planning tools. "You need to know more than just when your fruit is going to be ripe. How long do you have to pick it?" Carter stated, emphasizing the potential of AI models that can forecast ideal harvest dates by incorporating weather patterns.

Okanagan Specialty Fruits is already experimenting with technology from Vivid Machines, a Canadian firm. Their system uses AI-powered cameras mounted on tractors to analyze apple trees, identifying buds, flowers, and fruit. This technology provides crop estimates and potential harvest dates, with Carter noting its effectiveness in detecting small flower buds that are difficult to spot with the naked eye. However, he cautioned that the accuracy of these AI forecasts is heavily reliant on the quality of historical data specific to each farm, describing the process as "bespoke."

For crops like berries, where the harvest window can be as narrow as a few days, precise timing is crucial. Raymond Martin, co-founder of FruitCast, a UK company, explained that while experienced farmers generally know when their crops are ready, AI can provide this assessment on a much larger scale, analyzing entire farms with diverse growing conditions. FruitCast offers crop predictions for strawberries, raspberries, blackberries, blueberries, and tomatoes, with plans to expand to grapes. Their system analyzes imagery captured by drones, smartphones, or farm vehicles. Martin noted that this year's challenging weather, including heat and drought in the UK, has stressed fruit plants and slowed production, a factor AI can help monitor.

Beyond harvest timing, researchers are developing advanced methods for fruit analysis. Yasaman Ghasempour at Princeton University is exploring the use of millimetre waves, a type of high-frequency radio wave, to detect fruit ripeness. Unlike traditional brix meters that measure sugar content, millimetre waves can penetrate deeper and are responsive to humidity, water, and sugar levels. Her team has developed a millimetre wave-based ripeness detector that could be used by farmers or even consumers.

Despite these technological advancements, adoption remains a significant hurdle. Jing Zhang at North Carolina State University, who is developing a system to automatically count blueberries from smartphone images, noted that growers need to have confidence in the technology's reliability. The cost-effectiveness of these AI tools is also a key consideration. Kevin Wang at the University of Florida has created a crop-counting tool that utilizes imagery from affordable drones, suggesting that simpler systems may not require substantial investment.

Concerns also exist regarding the sharing of sensitive farm data with third-party AI models. Ben Palone of Western Growers, an association representing US farmers, acknowledged the potential of harvest forecasts as an "optimization tool" but suggested that farmers will likely continue to rely on human judgment for critical decisions, integrating technology as a supportive element rather than a complete replacement.