Monitoring riparian vegetation from space: developing a cloud-based application in Google Earth Engine

Implementing objective and reproducible tools for vegetation monitoring is fundamental to effective conservation efforts. Ecosystems located at the interface between terrestrial and aquatic environments, such as riparian ecotones, provide essential ecosystem services but are highly vulnerable to anthropogenic pressures. Moreover, inadequate monitoring approaches that can support decision-making processes often hamper their conservation. In this context, satellite remote sensing has the potential to address data gaps by providing information with extensive spatial and temporal coverage. However, the large volume of satellite datasets typically requires significant computational resources and specialized expertise, limiting their usability by non-specialists. To address these challenges, this work aims to develop a user-friendly application for monitoring riparian vegetation over time using freely available remote sensing data. The application has been developed on Google Earth Engine, a cloud-based computing platform that has been increasingly applied for large-scale geospatial analysis in recent years. Development and testing are carried out in a pilot area corresponding to the Bolsena volcanic lake (central Italy). The tool consists of three main modules: (1) selection and pre-processing of the satellite image collection, (2) extraction of water bodies, and (3) identification and quantification of riparian vegetation. Sentinel-2 multispectral imagery is selected as the remote sensing dataset due to its high spatial resolution (10 m) and revisit frequency (5 days). In the first step, the image collection is filtered for a user-defined time range and cloud cover. Subsequently, water bodies are identified based on automatic thresholding of specific spectral indices, such as NDWI. Then, riparian vegetation is identified within a buffer surrounding water bodies using various approaches, including the thresholding of spectral vegetation indices, and the application of automatic image classification techniques based on machine learning. Additionally, the feasibility of discriminating between different types of riparian vegetation will be explored. Ultimately, the tool will be validated using ground truth data collected from the study area, with further testing for generalization to additional sites. The tool will allow the generation of maps and quantitative assessments of riparian vegetation extent and changes over time. Its simple interface will ensure accessibility to a wide range of users, and its cloud-based nature will allow the analysis of large amounts of multi-temporal data without requiring advanced computational resources. This approach will therefore provide critical data to support monitoring, planning and decision-making, and ultimately the conservation of riparian ecosystems, while fostering local community engagement and promoting the social benefits of conserving these vital habitats.