Turf optimization is inherently complex and multivariate, relying on multiple interdependent factors, including maintenance practices and environmental conditions. Consequently, optimizing turf maintenance through traditional methods such as domain expertise and physical experiments alone proves to be impractical. To address this, the ADORE project explored a digital, data-driven approach powered by Artificial Intelligence (AI) to identify sustainable and playability-optimized turf maintenance strategies across the vast array of possible configuration combinations.
In the project, AI time series modeling was employed to simulate future turf conditions, focusing on both playability and stress metrics, based on current turf state, maintenance practices, and environmental data. Given the high resource demands and cost of continuous data collection, the project aimed to answer two integrated research questions: (i) what types of data are essential?, and (ii) what volume of data is necessary to produce accurate predictions and assess the viability of AI-generated turf maintenance strategies?
A key focus was to determine whether AI and time-series modeling could be affective without significant investment in data measurement hardware or the need for large volumes of data. While AI has shown proof-of-concept success on courses equipped with soil sensors and local weather systems, the challenge was to assess its applicability for the average Scandinavian golf club, where data collection is limited to records of maintenance actions, occasional soil analysis, and opensource
weather APIs.
The ADORE project concluded that current state-of-the-art AI time series models struggle to deliver accurate predictions when relying on sparse, manually collected data and non-local weather information. Both playability and stress metrics were evaluated, with even basic target variables such as STIMP, which is strongly influenced by factors like mowing height and frequency, showing reduced accuracy.
