Dong Pham

Overview

I am a remote sensing scientist specializing in continental-scale land-cover mapping and deep-learning methodology development. Currently, I work at the EO Lab, University of Greifswald. I obtained my Ph.D. in 2025 (magna cum laude), where my research focused on novel techniques to map land cover consistently across large spatial and temporal domains.

Expertise at a Glance

• 6+ years working with large-scale land cover applications with Earth observation data using Data Cube.
• Developing and applying Deep-learning algorithms with EO data.
• Land cover applications: Urban, Forests, Croplands, Wetland vegetations.
• Deploying workflows on High Performance Clusters (HPC), as well as cloud plattform such as CODE-DE.
• I am the creator of:

BSRLC+: the first large-scale, annual land cover dataset in Europe that contains detailed information on croplands and peatlands. For this application, I developed a universal method called Temporal Encoding for handling irregular time-series remote sensing data for long-term mapping.

BSRLC-U: the first tri-annual 10-m maps of built-up types over two decades using Landsat and Sentinel-2. To create this dataset, I developed a deep learning super-resolution approach to upscale historical Landsat data to 10 m.

Technical Skills

• Programming & Scripting
  • Python ★★★★★ – pandas, rasterio, PyTorch, matplotlib
  • Javascript ★★★★☆ – leafletJS, ReactJS, frontend
  • Bash / Shell ★★★★☆ – HPC job automation, Docker, Git
• Deep-Learning Frameworks
  • PyTorch ★★★★★ – CNNs for EO, 1D-CNN, 2D-CNN, LSTM, Transformers
  • TensorFlow / Keras ★★★★☆ – rapid prototyping
• Remote-Sensing & GIS
  • GDAL / Rasterio ★★★★★ – data ingest & preprocessing
  • CDSE OData API ★★★★☆ – I developed a search/download tool CDSE-S2 Downloader
• Data Engineering & Cloud
  • FORCE datacube ★★★★★ – large-scale processing
  • Docker / Singularity ★★★★★ – reproducible environments
  • Slurm ★★★★☆ – cluster scheduling

Scientific works

List of Scientific publications