Course material of the InterJIA workshop (Madrid, November 2026), by the GeoOcean Group (Universidad de Cantabria).
Slides: https://geoocean.github.io/BlueMath_Course_InterJIA/
The Santoña estuary (Cantabria, Spain) can be flooded by the sea, the Asón river and the rain, often at the same time. Simulating every possible day with a flood model is too expensive, so we build a hybrid model: HyDeCoFfEE, a simpler version of HyCoFfEE that uses SFINCS instead of SCHISM.
- pick 100 representative days with LHS + MDA,
- simulate them with SFINCS,
- learn the flood maps with PCA + RBF, a surrogate that answers in milliseconds,
- drive it with a climate emulator to get 3 × 100 years of floods and the 10-, 50- and 100-year flood maps.
In the browser (recommended): click Open in GitHub Codespaces above and
wait about a minute. Everything is installed. Open the notebooks in
notebooks/ and choose the Python kernel in /opt/conda.
On your machine, with Docker: open the repository in VS Code and choose Reopen in Container (Dev Containers extension). Or, from the repository folder:
docker run -it --rm -p 8888:8888 -v "$PWD":/workspace -w /workspace \
geoocean/interjia:latest jupyter lab --ip 0.0.0.0 --no-browserand open the http://127.0.0.1:8888/... link it prints.
On your machine, without Docker: with Miniforge:
mamba env create -f environment.yml
conda activate interjia
jupyter lab notebooks/SFINCS is not included this way: set run_sfincs = False in notebook 01 and
it uses our results.
Run them in order: each one uses what the previous one created.
| Notebook | What you do |
|---|---|
01_SFINCS_wrapper |
Pick 100 days (LHS + MDA), run SFINCS, compress the flood maps with PCA |
02_Surrogate |
Fit the RBF surrogate, check its error, predict new days |
03_Emulator |
Fit one Gaussian copula per weather type and simulate 3 × 100 years of rain, sea level and river flow |
04_Flood_climate |
Run the 110,000 emulated days through the surrogate: return periods and flood maps |
The weather types are given in notebooks/data/: the 25 daily weather types
(DWTs, dwt_model.pkl), and the 3 × 100 years of annual (AWTs) and daily
weather types simulated with the two ALR models (AWT_SIM_nsims3_2000_2100.nc,
xds_output.nc). The slides explain how they were built, but you do not run
that part: notebook 03 starts from the simulated DWTs.
Something failed, or you want to skip a step? Set USE_PRETRAINED = True at
the top of notebooks 02 and 04, and they use our results in
notebooks/pretrained/.
python -m http.server 8000 --directory slidesand open http://localhost:8000 (press S for the speaker notes).
To see your own models in the interactive slides, after running notebooks 01 to 04:
python slides/scripts/export_data.py --ownand refresh the page. Without --own, the slides go back to our models.
- BlueMath toolkit (GeoOcean): LHS, MDA, PCA, RBF and the climate emulator.
- SFINCS (Deltares) and HydroMT-SFINCS.
Licensed under the GNU GPL v3.0, like BlueMath and SFINCS (see LICENSE).