Use of Google Earth Engine in predicting future giant cane (Arundinaria gigantea (Walter) Muhl.) restoration sites

Abstract

The restoration of giant cane (Arundinaria gigantea) along riparian areas offers significant ecological benefits, including water quality improvement, riparian areas stabilization, and enhanced wildlife habitat. However, identifying optimal sites for restoration in modified landscapes remains a challenge. This study leverages the integration of Geographic Information Systems (GIS) and Remote Sensing (RS) technologies, combined with machine learning techniques, to identify suitable sites for giant cane restoration in Missouri (MO), USA. Sentinel-2 imagery, soil and environmental data, and custom spectral indices were utilized to assess site suitability. A Random Forest (RF) classifier was trained with ground truth data representing suitable and unsuitable giant cane sites, achieving an overall accuracy score of 95 %, with 0.04 MSE and 0.2 RMSE. The model identified favourable sites predominantly located near riparian zones, enabling targeted restoration efforts. Results reveal spatial patterns linked to environmental factors, such as soil texture, moisture and pH, that influence site suitability for giant cane growth. This research highlights the potential of GIS and RS in ecological restoration, offering a robust framework for future projects focused on habitat restoration and conservation in riparian ecosystems. By combining field data with remote sensing, this study may contribute to the restoration of vital habitats, supporting biodiversity conservation and water quality enhancement.