A Pan-European Multimodal Dataset Linking Field Photos with Satellite Image Time Series for Agricultural Mapping

AgroFusion AgroFusion is a continental-scale, multimodal, pixel-based dataset designed to support reproducible research in agricultural remote sensing and geospatial artificial intelligence. By providing high-resolution in-field photos that contextualize satellite-derived analyses, AgroFusion enables a wide range of applications, including multimodal and multiscale crop type classification and phenology retrieval. Data AgroFusion integrates multiple data modalities for approximately 200K validated LUCAS points, representing agricultural land-cover and land-use (LCLU) areas: 📷 In-situ field photos 📝 Land-cover and land-use annotations 🛰️ Sentinel-2, Sentinel-1, and Landsat-8 pixel-based time series 🌍 ERA5-Land agrometeorological variables 📍 Copernicus GLO-30 DEM topographic information 🌱 AI4SoilHealth soil property data 📚 AlphaEarth annual embeddings Code Code for downloading and processing in-situ field-level photos is available at: https://github.com/stemauro/agrofusion/tree/v0.1.1 A snapshot of the GitHub repository at version v0.1.1 is also included with this archive as: agrofusion-0.1.1.zip The ZIP file contains a fixed copy of the source code and supporting project files, enabling users to inspect and run the pipeline without downloading the repository from GitHub. To run the processing locally, clone the repository and install the environment git clone https://github.com/stemauro/agrofusion.git && cd agrofusion uv sync && source .venv/bin/activate then run the entire processing pipeline with uv run dvc repro For instructions on running specific stages of the pipeline, refer to the Download section in the code repository's README. License The code is released under the permissive MIT License. The dataset is distributed under the Creative Commons Attribution 4.0 International (CC‑BY 4.0) license.