Leveraging remote sensing and machine learning for sustainable management of Hanoi’s Urban golf courses

Abstract In recent years, golf courses have significantly contributed to the economic growth of Vietnamese cities like Hanoi. However, their environmental and social impacts—particularly regarding land use, pesticide application, water resources, and environmental degradation—remain a concern. This study utilizes Sentinel-2 and Landsat satellite imagery combined with geographic information systems to monitor golf courses in Hanoi’s metropolitan area. Through time-series multispectral data, we evaluate two detection methods: normalized difference vegetation index (NDVI) analysis and feature recognition. NDVI analysis, with an average NDVI value of 0.55–0.75 for turf grass, initially showed potential but faced classification challenges due to overlapping vegetation signatures with parks and agricultural fields. Conversely, feature recognition demonstrated strong accuracy, correctly identifying over 85% of golf course areas by leveraging distinctive characteristics such as bunker shape and turf grass arrangement. The integration of Sentinel-2 imagery with spectral mixing analysis further improved boundary delineation, reducing misclassification rates from 18% (using Landsat) to 7%. This research establishes a remote sensing-based golf course database covering over 2,500 hectares in Hanoi, tracks and land conversion trends over the past decade, and provides quantitative insights into sustainable golf course management. Our findings underscore the effectiveness of combining high-resolution satellite data and advanced classification techniques to support environmentally responsible urban land use planning. This study highlights and the growing role of remote sensing technologies in urban planning and environmental conservation, paving the way for further applications in sustainable land management.