AI Harvest Forecasting Helps Farmers Navigate Timing and Profit Challenges

Artificial intelligence is emerging as a critical tool for agricultural producers, offering precise forecasts on when crops will reach peak ripeness. As climate variability complicates scheduling, companies like Okanagan Specialty Fruits and UK-based FruitCast are deploying AI models to help growers optimize harvest windows, manage labor, and maximize profits.

Joel Carter of Okanagan Specialty Fruits highlighted the practical challenges of harvesting in Washington State, where extreme heat recently forced a halt to apple picking at 10:00 AM for worker safety. “You need to know more than just when your fruit is going to be ripe. How long do you have to pick it?” Carter said. His company, which manages over 1,250 acres of orchards producing sliced apples, is experimenting with cameras from Canadian firm Vivid Machines. Mounted on tractors, these systems use AI to identify buds, flowers, and fruit, providing both crop estimates and harvest dates.

Carter noted that the technology is particularly adept at detecting tiny flower buds invisible to the naked eye. However, he emphasized that these models require farm-specific historical data rather than generic internet information to be effective. While apples offer a relatively long harvest window, perishable berries such as strawberries and blueberries present a much tighter timeframe, where delays can lead to rapid disease spread and significant financial loss.

Raymond Martin, co-founder of FruitCast, explained that while experienced farmers often have an intuitive sense of ripeness, they cannot simultaneously monitor every plot across large or mixed indoor-outdoor operations. His company analyzes imagery captured by drones, smartphones, and vehicle-mounted cameras to predict yields for strawberries, raspberries, blackberries, blueberries, and tomatoes. FruitCast reports high accuracy rates, with forecasts landing within 10% of actual volume one week out and within 17% three weeks out.

The urgency for such tools has increased due to recent severe droughts and heatwaves in the UK, which have induced thermal dormancy in fruit plants and slowed production. Angus Soft Fruits, a client of FruitCast, acknowledged that while progress is significant, the industry has not yet achieved a fully integrated forecasting ecosystem.

Beyond visual analysis, researchers are developing more granular detection methods. Yasaman Ghasempour at Princeton University has created a millimetre-wave ripeness detector that penetrates deeper into fruit than traditional infrared brix meters. By measuring responses to humidity, water, and sugar, the device could allow consumers and farmers to assess ripeness without cutting the fruit. Early tests at a New Jersey market revealed initial skepticism from vendors, but the technology holds promise for future integration with yield models.

Despite the technological advances, adoption remains a complex hurdle. Jing Zhang from North Carolina State University pointed out that growers must have confidence in the research before investing. Meanwhile, Kevin Wang of the University of Florida developed a low-cost crop-counting tool using $100 drones, though he noted concerns among farmers about sharing sensitive commercial data, such as fertilization and irrigation strategies, with third parties.

Ben Palone of Western Growers described the technology as an “optimization tool” but observed that human judgment remains indispensable for critical decisions. “Growers like to have people in the mix to make some of those very critical decisions – especially when it comes to harvests,” he said, suggesting that AI will likely serve as a complement to, rather than a replacement for, traditional farming expertise.