The AI telling farmers when to harvest

When the apple harvest began in Washington State last year, the fruit was ready for picking but the scorching weather posed a risk to workers. Artificial intelligence (AI) models that predict optimal harvest times, considering weather conditions, could be invaluable. Joel Carter, from Okanagan Specialty Fruits, explains that knowing the precise moment to harvest is crucial, more than just the ripeness of the fruit.

His company owns over 1,250 acres of apple orchards in Washington, producing sliced apples for hotels and schools. These apples are genetically modified to resist browning after cutting. However, harvest planning is a challenge due to the volatile fruit prices, especially for high-value berries like strawberries or blueberries. Okanagan Specialty Fruits is testing cameras from Vivid Machines that capture images of apple trees as they pass by a tractor, using AI to identify buds, flowers or fruit in the footage.

This tool provides crop estimates and harvest dates, though accuracy relies on the quality of historical data fed into the system. Raymond Martin, co-founder and COO of UK-based FruitCast, offers harvest forecasts for strawberries, raspberries, blackberries, blueberries and tomatoes. FruitCast's forecasts are accurate within 10% one week out and 17% three weeks out.

The company guarantees less than 20% error. However, experienced farmers often know when their fruit is ripe, but FruitCast provides this information on a larger scale. Angus Soft Fruits has collaborated with FruitCast. While AI technology for fruit ripeness forecasting is not yet perfected, researchers are exploring millimetre wave-based ripeness detectors, which could help inform forecasts.

Despite the potential benefits, many growers question whether investing in such technology is worthwhile, as they need confidence in the research's efficacy.