Guardat en:
| Autors principals: | , , , , |
|---|---|
| Format: | Recurso digital |
| Idioma: | anglès |
| Publicat: |
Zenodo
2025
|
| Matèries: | |
| Accés en línia: | https://doi.org/10.5281/zenodo.14727432 |
| Etiquetes: |
Afegir etiqueta
Sense etiquetes, Sigues el primer a etiquetar aquest registre!
|
Taula de continguts:
- <h3>FORGENIUS-PP Repository: Deliverable D3.1</h3> <p>This Zenodo entry provides a DOI for the repository associated with the FORGENIUS-PP project, which focuses on the development and application of reproducible workflows for the spatial analysis and monitoring of <em>Pinus pinaster</em> ecosystems. This repository documents Deliverable D3.1 and supports the FORGENIUS initiative under the Horizon 2020 framework.</p> <h3>Key Contributions</h3> <ul> <li>Data: UAV-acquired images (RGB), georeferenced datasets, and environmental variables including Canopy Height Models (CHM), Digital Terrain Models (DTM), and soil moisture maps.</li> <li>Methods: Automated workflows for data processing, including photogrammetric reconstruction, tree crown segmentation, and geostatistical interpolation of soil moisture using Ordinary Kriging with External Drift (OKED).</li> <li>Deliverables: Comprehensive documentation and reproducible workflows that support the collection, processing, and analysis of field data.</li> </ul> <h3>Scope and Innovation</h3> <p>This repository addresses the challenges of monitoring forest ecosystems across multiple scales, integrating UAV-based photogrammetry, geostatistical modeling, and machine learning techniques. The project includes study sites in <strong>Lacanau (France)</strong> and <strong>Tocchi (Italy)</strong>, where detailed field data collection and UAV campaigns enabled centimeter-level georeferencing accuracy.</p> <p>The repository provides workflows and tools to generate:</p> <ul> <li><strong>3D models</strong> such as Digital Terrain Models (DTM) and Canopy Height Models (CHM).</li> <li><strong>Tree crown delineations</strong>, based on advanced segmentation methods like the Dalponte algorithm.</li> <li><strong>Environmental indices</strong>, including solar irradiance (direct, diffuse, and total), topographic wetness (SAGA-TWI), and slope/aspect derived from high-resolution digital surface models (DSM).</li> <li><strong>Cone classification</strong>, leveraging orthorectified image mosaics for machine learning-based detection of <em>Pinus pinaster</em> cones using two distinct workflows.</li> <li><strong>Soil moisture and temperature maps</strong>, generated through Ordinary Kriging with External Drift (OKED), incorporating terrain elevation data for precise interpolation of micro-climate conditions.</li> </ul> <p>Additionally, the repository integrates a <strong>Sentinel Time Series</strong>, providing access to approximately 350 cloud-free Sentinel-2 scenes spanning the last five years. Scripts for NDVI computation and further analysis are available, ensuring adaptability for various applications.</p> <p>By leveraging state-of-the-art UAV systems, advanced geospatial methods, and reproducible workflows, this project highlights scalable solutions for monitoring forest dynamics across diverse spatial and temporal scales.</p> <h3>Repository</h3> <p>The full dataset, workflows, and documentation are hosted in the GitLab repository: https://gitlab.uni-marburg.de/reudenba/forgenius-pp</p> <h3>Reproducibility</h3> <p>All workflows are implemented in R and QGIS to ensure reproducibility and accessibility for users. Scripts include detailed comments and are complemented by field protocols, making it possible to replicate data collection and analysis.</p> <h3>Acknowledgments</h3> <p>This project was developed by Dr. Chris Reudenbach at Philipps-Universität Marburg, as part of the Horizon 2020 FORGENIUS project (Grant Agreement No. 862221). The repository documents efforts to provide robust and reproducible monitoring methods for forest ecosystems, supported by innovative geospatial analysis and machine learning techniques.</p>