The EO4NOWCAST project aims to develop an AI model to predict flood risk based on Earth Observation. Nowadays, flood prediction is mainly based through complex physical models that have some limitations. For examples, most of them does not include estimation of soil moisture in the algorithms, while soil moisture is a crucial parameter to determine risk of flood. Indeed, the EO4NOWCAST project exploit Copernicus product, in particular Normalized DIfference Water Index (NDMI), to predict flood risk. However, the great limitation of EO resources is that the information are available only when Sentinel satellites overcome the area of interest, indeed every 5-10 days.
To overcome this limitation, EO4NOWCAST develops an AI model to update the most recent NDMI map from Copernicus through rainfall map obtained by in situ monitoring network. In this way, even if the last map of Copernicus has been produced some days ago, the model estimates the current level of soil moisture with high spatial resolution (around 100mx100m).
In this asset, you can find the Jupyter notebook of the AI model developed within the EO4NOWCAST project. The model is structured to receive two inputs:
- a NDMI map from Copernicus service at time t0
- a cumulative rainfall map from t0 to t1
And, as an output, the model produces a NDMI map updated at t1.
You can train and test with your own dataset (see the Readme inside the asset for more information) or you can access to the code to adapt or modify the model according to the problem you want to solve.