A Cloud-Based Geospatial Intelligence Framework for Alpine Environmental Monitoring: Integrating Optical Time-Series and Alpha Foundation Satellite Embeddings in Valle d'Aosta (NW Italy)

Accurately monitoring phenological shifts and geomorphological transformations in highly dynamic alpine environments like Valle d'Aosta (Italy) requires scalable Earth Observation tools. Although massive multitemporal datasets and state-of-the-art deep learning models exist, their operational use by local stakeholders is often hindered by steep learning curves and high technical barriers.

To address this, we developed an interactive, zero-code Google Earth Engine (GEE) web application. The tool enables flexible spatial querying, allowing analyses across the entire regional extent or within any specific one of the 74 local municipalities. It seamlessly integrates classical optical time series (Sentinel-2 and NASA HLS Landsat/Sentinel) with 64-dimensional latent feature spaces from Alpha Foundation Satellite Embeddings. Crucially, the application empowers non-experts in GIS or GeoAI to compute eight traditional spectral and physical indices and directly compare them with deep learning-based change detection. Users can effortlessly perform multitemporal Euclidean distance mapping utilizing all or a targeted subset of the 64 embedding dimensions and execute interactive dot-product similarity assessments.

The platform executes computationally heavy pipelines in seconds with dynamic visual feedback. In Valle d'Aosta, comparing classical index differentials (T minus T-1) with the multitemporal Euclidean latent shift effectively isolated severe structural modifications, such as localized urbanization and forest disturbances, which can be subtle in traditional single-index analyses. The intuitive interface allowed users to seamlessly toggle between standard Landsat/Sentinel-2 indices and targeted foundation model bands to refine spatial pattern matching.

By fusing classical multispectral remote sensing with cutting-edge AI embeddings, this framework democratizes advanced GeoAI. It provides a robust, highly accessible decision-support tool, allowing users without coding expertise to leverage the power of foundation models for precise landscape assessment and environmental protection in vulnerable alpine regions.