Satellite data can be used for flood monitoring, however, it is difficult to interpret optical imagery as floodwaters do not look blue, they are usually brown due to sediment loads. This can hinder the fast response needed for dangerous flood events.
To combat this, we can use machine learning to process flood images much quicker in order to generate vitally needed flood maps. This was completed by Oxford researchers and ESA’s Phi-lab, who developed the WorldFloods AI model.
WorldFloods is a training data set that combines, in ‘machine-learning ready form,’ several existing databases of satellite imagery of historical flood events. The dataset contains pairs of Sentinel-2 images and flood extent maps covering 119 global flood events. This is an expert dataset that was labelled by flood experts – which means it has a higher confidence level and a higher accuracy. Typically, this kind of expert labelling takes about six months of work.
You need experts to label the dataset because it can be hard to identify what is a flood and what isn’t a flood. For example, a flood can look different depending on what colour composite you’re looking at.
Worldfloods is specifically designed for deployment in specialized hardware in space on low-cost satellites in low Earth orbit. So, in the future, flood segmentation mapping will be able to be run on-board satellites. The new CubeSat satellites such as PhiSat include chips that can run the algorithm – which would significantly speed up response times. This is ultimately what the WorldFloods dataset will be used for.
Jupyter Notebook Task: ML4Floods: Training
In this last Jupyter Notebook you will have a go at flood mapping with machine learning using WorldFloods data which combines, in “machine-learning ready form”, several existing databases of satellite imagery of historical flood events. The dataset contains pairs of Sentinel-2 images and flood extent maps covering 119 global flood events.
ML algorithms have the potential to offer significantly faster and more accurate flood mapping than traditional methods.
You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.
For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.
If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below!
Jupyter Notebook Task - ML4Floods: Inference
This notebook guides you step-by-step through the inference process of a flood extent segmentation model using Sentinel-2 data.
You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.
For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.