The Sentinel2Height is an ESA-funded project that develops a new framework for mapping forest canopy heights using coherence images derived from Sentinel-1 Synthetic Aperture Radar (SAR) data. The project bridges existing gaps in space-based Lidar data, such as GEDI and IceSat-2., by using Sentinel-1's frequent and all-weather coverage to estimate forest heights. It aims to provide a wall-to-wall, cost-effective, reliable, and timely method for monitoring forest height, thereby facilitating better forest management and conservation strategies.
The project is structured around several key objectives:
- Create a Cross-Modal Forest Benchmark Dataset: The project aims to develop a benchmark database that includes Sentinel-1 SAR coherence images and forest height measurements from space-based Lidar systems such as GEDI and IceSat-2. The database will adhere to FAIR (Findable, Accessible, Interoperable, Reproducible) and Open Science principles.
- Develop a CNN Model Customized for Forest Height Mapping: The objective is to create a deep learning model based on a convolutional neural network (CNN) designed for SAR-derived coherence images. The model will be trained using the cross-modal benchmark database and used to enable forest height estimation using Sentinel-1 SAR data.
- Demonstrate and Validate the Cross-Modal Benchmark and CNN Model: The project will assess the performance of the cross-modal benchmark database and CNN model using four selected case study areas. The results will be shared publicly and published in open-access resources.
- Produce an Open Processing Tool: The project will develop an open-source tool, Sentinel2Height, which will allow users to process Sentinel-1 SAR images and estimate forest height. This tool will be made publicly available, accompanied by user-friendly documentation and open-source code, ensuring that it can be accessed and utilized by a broad range of users.
- Ensure the Long-Term Availability of the Benchmark Data and Tool: The project will ensure the sustainability and continued availability of the developed dataset and tool by storing them in open repositories (e.g., GitHub, Zenodo) and promoting their use for future research and advancements.
The project is funded from the European Space Agency (ESA) in response to the Tender Action Number 6-60008 and Activity Number 1000037670. It kicked off on January 15, 2025.
Consortium partners: The University of Twente, The Netherlands (prime contractor), Università degli Studi di Napoli Parthenope, Italy, Aalto University, Finland, and ETForestService group, Finland.