VBoS training: ML for climate change and natural disasters - ML training for Climate Change and Natural Disasters

Welcome to this Vanuatu Bureau of Statistics (VBoS) training course! This Jupyter book contains notebooks on how to apply machine learning and remote sensing methods for Climate Change Monitoring and Natural Disaster Preparedness.

Overview of tutorials¶

Part 1¶

Introductory lessons to satellite remote sensing and basic machine learning.

Lesson 1A

Intro to Selecting and Processing Raster Data

Lesson 1B

Land Use / Land Cover Segmentation Using Sentinel-2 and Random Forest

Lesson 1C

Segmentation workflow using PyTorch for flood and water mask detection

Lesson 1D

Road segmentation

Part 2¶

Various environmental and climate indicators covered during an in-person workshop.

Lesson 2A

Coastal changes

Lesson 2B

Assessing Flood Risk from Past Weather

Lesson 2C

Climate indicators

Lesson 2C (continued)

Marine heat waves

Lesson 2D

Coral reef mapping

Lesson 2E

Soil health

Extra scripts¶

Additional processing scripts to scale ML models country-wide for input into VBoS’s Management Information System (MIS). More at https://github.com/Vanuatu-National-Statistics-Office/climate-change-natural-disasters/tree/main/scripts

Lesson 1E

Full country LULC Segmentation Using GeoMedian Composite Sentinel-2 and GPU-accelerated Random Forest

Lesson 1E (inference)

Land Accounts inference on Sentinel-2 GeoMAD with Random Forest

Lesson 2D (alternative)

Coral Reef Health Monitoring around Vanuatu’s Coastline