at c()
Turf Talk
AI on the Fairway
The next frontier of weed management
by JUAN R. ROMERO & SHAWN D. ASKEW, Ph. D.
Graduate student Juan Romero lines up for a broadcast comparison treatment among an array of plots recently treated by the ARA vision sprayer( top left). Note the patterns of blue dye indicating the ARA is only spraying targeting weeds.
In the 1890s, hand weeding was standard practice on golf courses. The advent of herbicides in the 20th century transformed this back-breaking work into a more efficient process, allowing superintendents to deliver consistently pristine playing surfaces.
Despite those advances, Virginia’ s golf course superintendents now face a new challenge. Rising demand for organic food has driven agriculture to reduce pesticide use, and stricter herbicide regulations have slowed the development of new products. At the same time, troublesome weeds such as goosegrass and annual bluegrass continue to develop resistance to existing chemistries. The overall sum of these issues leaves superintendents with fewer effective tools to manage weeds and greater concern about the long-term quality of fairways and greens.
Exploring an AI-driven solution Fortunately, technology may offer a path forward. Over the past year, researchers at Virginia Tech have tested an innovative sprayer that uses artificial intelligence( AI) and machine vision to detect and treat weeds on a plant-by-plant basis. Known as the ARA and developed by Swiss company Ecorobotix, the system was originally designed for vegetable crops. We wanted to evaluate whether the same technology could be adapted for golf course turf, where the height of cut, grass varieties, and weed species differ significantly from agricultural systems.
Field trials and key findings In summer 2024, we conducted field trials at the Pete Dye River Course of Virginia Tech to assess the ARA’ s accuracy and its potential to reduce pesticide use. Even though the system’ s training data were primarily based on vegetables, results were promising. On creeping bentgrass fairways, the ARA sprayed less than five percent of the total area; on tall fescue roughs, about 25 percent— only where weeds were detected.
This translated to a more than 90 % reduction in herbicide use on fairways and a 70 % reduction on roughs, offering potential savings of up to $ 500 per hectare. Targeted spraying also minimizes environmental impact and reduces unintended turf injury.
When identifying specific weeds, the ARA successfully treated 88 % of thin paspalum( Paspalum setaceum) in roughs and 100 % of broadleaf weeds. Goosegrass proved more difficult, with only about 20 % of plants treated. This likely stems from the limited“ turf-specific” training data available— the system occasionally confused goosegrass with desirable turf species. As the company refines its models to include turf imagery, accuracy is expected to improve.
What’ s next for turfgrass research These findings align with previous research showing that machine-vision sprayers can significantly reduce chemical and labor inputs, but performance depends heavily on the quality and diversity of training data. The more images and field conditions the system can learn from, the better it performs.
Looking ahead, we plan to collaborate further with Ecorobotix to optimize the ARA’ s software for golf course environments, with a particular focus on improving detection of goosegrass and other problem species.
Ultimately, this technology could help superintendents reduce costs, minimize chemical use, and maintain healthier turf, while meeting the growing demand for sustainable course management. From hand weeding to AI, weed control continues to evolve. With ongoing collaboration between researchers, industry partners, and superintendents, machine-vision sprayers may soon become an integral part of golf course management in Virginia and beyond.
VIRGINIA TECH
14 V IRGINIA G OLFER | N OVEMBER / D ECEMBER 2025 vsga. org