Water Detection Based on the Australian WOfS Algorithm
Documentation
Surface water is detected from satellite images using an automated water mapping algorithm created by Geoscience Australia. The water detected for each location is summed through time and then compared to the number of clear observations of that location. The result is a percentage value of the number of times water was observed at the location. The colouring of the summary layer indicates how often water was observed. Possible floods appear in the low values (red to yellow colours) while consistent water bodies such as lakes and dams have high values (blue to purple colours).
A multi-spectral decision tree application to detect water. This application was developed in Australia (Mueller, N., et al, 2016) and has yielded improved water detection over standard water quality flags in the Landsat product. When run over a time series, it produces a product that shows the frequency of water at every location.
Tutorials
-
Head to the ODC Sandbox (click here) and login with GitHub Open Authorization.
-
Open the /examples folder.
-
Open and run the /examples/SDG_6.6.1_Change_Extent_Water_Ecosystems.ipynb notebook block by block.
-
In block [2], use the map interface to draw a small area of interest, and select a time frame.
-
Click "Query Cube" and review the number of tiles that will be accessed. Less than 100 will run quickly.
-
Run block [3], 'water_stackplot_aws_landsat8', and select a region to investigate within your area of interest, using the WOfS product.
-
Run block [4], 'water_max_min_aws_landsat8', and select a region to investigate within your area of interest, using the WOfS product.
Code/Python Notebooks
GitHub Resources
-