Choosing the right moment to harvest can significantly impact agricultural outcomes. During the apple harvest in Washington State last year, high temperatures forced Okanagan Specialty Fruits to halt operations early. Joel Carter, from the company, stated that artificial intelligence (AI) models that forecast ideal harvest dates considering weather conditions would be beneficial. He emphasized the need for farmers to understand not just when fruit is ripe, but how long they have to pick it.
Okanagan Specialty Fruits operates over 1,250 acres of apple orchards, primarily growing fruit for sliced portions sold to hotels and schools. The company is investing in technology to enhance productivity, including genetically engineered apples that resist browning after cutting. Harvest planning is complex, as new tools that analyze fruit on trees and predict ripening times are becoming available. Incorrect harvest timing can lead to unnecessary labor costs and lost profits, particularly for high-value fruits like strawberries and blueberries.
The company is testing cameras from Vivid Machines, which collect imagery of apple trees as tractors pass by. AI analyzes this footage to identify buds, flowers, or fruit. Carter noted that the accuracy of these forecasts depends on the quality of historical data input into the system, which must be tailored to individual farms.
The harvest window for Granny Smith apples is three weeks, while berries require more immediate attention. Raymond Martin, co-founder of FruitCast, a UK-based company, explained that their system provides crop predictions for various fruits, including strawberries and raspberries, and will expand to grapes next year. He acknowledged that while experienced farmers have a general sense of ripeness, they may not have the same insight across large or diverse growing areas.
FruitCast’s model incorporates weather and irrigation data, claiming to achieve forecasts within 10% accuracy one week out and 17% accuracy three weeks out. Angus Soft Fruits has collaborated with FruitCast, with operations director Neill Finlayson noting that AI technology for harvest forecasting is still developing.
Driscoll's, a California-based fruit seller, confirmed that some of its UK growers have utilized FruitCast's technology. Researchers are also investigating advanced methods for analyzing fruit ripeness. Yasaman Ghasempour from Princeton University is developing a millimeter wave-based ripeness detector that could be beneficial for farmers or consumers.
Despite interest in new technologies, some farmers remain cautious about the investment. Jing Zhang from North Carolina State University highlighted the complexities of adoption and the need for growers to trust the research. Kevin Wang from the University of Florida has created a cost-effective crop-counting tool using imagery from drones.
Ben Palone, senior director at Western Growers, noted that while harvest forecasts can serve as optimization tools, farmers will likely continue to rely on human judgment for critical decisions, especially during harvests.