Advancement of Earth Observation Algorithms for Forest Monitoring

DASOS' Vision is a multi-disciplinary R&D upcoming lab at the University of Exeter, UK. We develop scalable & transferable solutions for better monitoring forest ecosystems.

Environmental Intelligence

We develop robust, interpretable, and scalable methods to improve forest monitoring.

Earth Observation

We make large Earth Observation datasets more accessible for research and decision-making.

Forest Ecology

We study forest structure, composition, and health.

Neurology Services
763249.92 EUR Funding Received
3 Active Collaborations
Modern medical facility

Research Interests

  • Development of scalable algorithms for processing large-scale Earth Observation (EO) and LiDAR datasets
  • 3D forest structure modelling and analysis, including voxel-based LiDAR representations and cross-sensor structural harmonisation
  • Feature engineering and representation learning from EO and LiDAR data to support robust, interpretable machine learning models
  • Advanced machine learning applications for forest monitoring, spanning classification, regression, anomaly detection, and change analysis
  • Phenological analysis using multi-sensor EO time-series, with particular emphasis on Synthetic Aperture Radar (SAR)
Funded Projects
Active Collaborators
Alumni

Featured Papers

Highlighting the latest and most impactful studies, carefully chosen for their significance and insight. Full list of peer-reviewed papers is available here.

PlotToSat Paper

PlotToSat: A tool for generating time-series signatures from Sentinel-1 and Sentinel-2 at field-based plots for machine learning applications

  • PlotToSat automates the extraction of Sentinel-1 and Sentinel-2 time-series signatures at multiple regions.
  • PlotToSat is designed to be user-friendly, with clear documentation and examples.
  • Its first application is tree genera classifications. More here.
Time-series Google Earth Engine
Learn More
Dead Trees paper

Detection of Dead Standing Eucalyptus camaldulensis without Tree Delineation for Managing Biodiversity in Native Australian Forest

  • It is possible to detect dead standing trees without tree delineation using object detection approaches
  • The most important feature for detecting dead standing trees is the standard deviation of heights.
  • Multi-scale windows improves detection in native Eucalypt forest as shown here
LiDAR Object Detection Biodiversity
Learn More
Dead Trees paper

Time-series SAR data analysis for finding the drivers causing phenological changes in Paphos Forest, Cyprus

  • In 2018, the warmer spring was associated with a delayed summer peak, coinciding with the leafing period of Pinus brutia.
  • After the extended drought of winter 2015–16, the average backscatter coefficient dropped.
  • Cold autumn temperatures kill Thaumetopoea pityocampa, which feeds annually on pine needles. A cold November was associated with a change in the SAR annual phenology.
time-series SAR phenology
Learn More

Projects

Explore our ongoing and completed projects.

People

Meet the researchers, and alumni.

Tools

Explore the open-source tools developed by our research group.

Join us

Become a member or collaborate.