Automated Mapping of Artificial Drainage in Peatlands Using Deep Learning and Very High-Resolution Aerial Imagery

Date:

This presentation demonstrates the application of deep learning semantic segmentation models for automated mapping of artificial drainage networks in Irish peatlands using very high-resolution aerial imagery.

Research Highlights

  • First national-scale automated drainage mapping in peatlands
  • Semantic segmentation using sub-meter aerial imagery (0.5m resolution)
  • Processing of 2TB+ of aerial imagery data
  • Deep learning implementation in R using Keras and TensorFlow
  • National conservation policy applications

Technical Innovation

The study developed scalable GPU-based workflows for processing massive aerial imagery datasets, enabling efficient feature detection at unprecedented spatial scales for peatland monitoring.

DOI

https://doi.org/10.5194/egusphere-egu24-21619