Practical Part 2

Fusion of Sentinel 1 and Sentinel 2 Data to produce denser NDVI time series with deep learning

Objective:

Develop skills to:

  • Read satellite images and Train a U-Net Model 1_UNET_TRAIN.ipynb
  • Predicted NDVI timeseries data extraction and comparison with ground truth 2_CentrePixelTmeseriesMaxField.ipynb
  • Construction of denser timeseries combining Sentinel 1 predicted and Sentinel 2 ground truth NDVI data 3_Timeseries_Unet_DemoPlot_Compare.ipynb.

The relevant data and codes can be accessed through the link https://github.com/zalf-rpm/KIKompAg/tree/main/Lesson1

The repository also provides the materials to pre-process and download Sentinel 1 and Sentinel 2 data from google earth engine. However, considering the limitation to host such large dataset, only a smaller dataset is uploaded. Interested parties might run the preprocessing and downloading codes to download the data from scratch. Each of the notebooks mentioned above is further explained in the following subsections.