Environmental Monitoring with Large-Scale Land Cover Classification
summary
Produces continent‑wide, time‑consistent land cover maps for Australia using FAO LCCS level 3 classes applied across the Landsat archive, implemented on Open Data Cube / Digital Earth Australia with processing on ANU’s supercomputer and AWS. Combines fractional‑cover inputs, median absolute deviation summaries and existing products (water, mangroves, intertidal) with simple rule sets and supervised feature‑ranking (scikit‑learn decision trees) rather than black‑box CNNs to deliver transparent, policy‑ready change detection; next steps add level‑4 detail and radar.
AI-written summary · about
chapters (15)
- 0:00Introduction to storytelling
- 1:05Why land cover classification
- 4:38FAO classes and semantics
- 5:39Importance of open source
- 7:25Land cover classification system
- 9:10Dynamic land cover dataset
- 9:31Machine learning approach
- 12:24Data processing pipeline
- 13:06Biophysical classification method
- 15:58Level three classification challenges
- 17:21Classified examples and complexity
- 18:44Sustainable development goals
- 19:26Next steps and improvements
- 19:47Take-home message
- 20:28Q&A session
related talks
19mTime and tide. Mapping the changing 3D shape of Australia’s coastline using p
14mSatellite data monitoring for analytical models
34mGeoscience Big Data Computing with Python
48mTaking the planet’s pulse: AI to empower adaptation to global change
19mDigital Earth Pacific DEP cloud native approaches to detect mineral extraction
21mAquawatch Data Systems a cloud solution for water quality monitoring and forecasting
17mFree satellite imagery and DEM analyses enabling NRM in the developing world
36mClassification of Satellite Imagery