AI empowers superintendents and general managers to remotely analyze the golf course turf condition and make informed decisions to optimize its maintenance. AI processes large amounts of data and returns precise insights, even when golf course decision-makers are not physically present. This data-driven approach leads to a better overall turf condition, optimized input applications, and a reduced environmental impact. In this article, we explain how that is possible.
The foundation to build accurate AI models based on deep neural networks is the quality of data used for training. Multi-spectral imagery, obtained from satellites or drones, and the values of reflectance coefficients obtained after calibration in particular wavelength ranges constitute the input data for SKIM’s classification and prediction models. The necessary benchmark data to build SKIM’s evaluation and prediction models was also meticulously gathered during three growing seasons in experimental fields, both in-situ and with tests performed by accredited laboratories. The entire process was orchestrated in cooperation with scientists from Wroclaw University of Science and Technology, the University of Agriculture in Cracow, and AGH University of Science and Technology.
Once quality data is gathered, AI models are trained until they provide accurate results. Accuracy is measured by comparing AI-generated results with benchmark data. In total, SKIM’s AI module is born from 33,000 satellite images, 7,000 drone images, 5,000 visual assessments performed by experts, 4,500 in-situ tests, and 24,000 laboratory results. Classification models (vegetation indicators, biomass, nutrients) and prediction models (water stress, diseases, deficiency alerts) achieve an average accuracy of 80%. At SKIM Turf Management, we continuously train and fine-tune our models. That’s why the process of AI improvement never stops.
SKIM users now benefit from these models to remotely manage their turf performance. The input data analyzed by the AI module is satellite or drone images. Compared with traditional laboratory analyses, AI does not focus on several points of the course where a sample is extracted for analysis. The sample is the entire golf course, and the granularity of AI results depends on the precision of the aerial imagery, ranging from 3x3m (low resolution) to 30cm x 30cm (high resolution). AI returns near real-time results with high frequency and precision on large surfaces.
The AI module is finally integrated with an application for visualizing results, planning agrotechnical treatments, and handling alerts related to threats on golf courses. Precise data is the foundation for precision turfgrass management. Thanks to AI, it's possible to observe trends at scale and anticipate deficiencies. The user is informed of these focus areas and receives recommendations. Therefore, AI has become an additional impactful aid in golf decision-makers’ toolkit.
How AI can empower golf courses with precise data and insights to optimize turf care?
