Check final Ocean Regimes notebooks

[AndreaGarciaJuan]

If you have a moment, could you please check the notebooks in your computer? And read the explanations to be sure they are clear? Thanks :-)

[AndreaGarciaJuan]

You can download the notebooks in the master branch: ModelDevelopment_OceanRegimes.ipynb and PredictandPlot_OceanRegimes.ipynb. You will need BIC_calculation_OR.py, Plotter_OR.py and preprocessing_OR.py files. You will also need a datasets/ and models/ folder and the logos for saving the figures

[quai20]

I tested both notebooks locally, some notes : Dev notebook :

• In the text before plotting the mask : mask_path=None should be mask_path='auto'
• I had an error when plotting the mask : The input coordinate is not sorted in increasing order along axis 0. This can lead to unexpected results. Consider calling thesortbymethod on the input DataArray. To plot data with categorical axes, consider using theheatmapfunction from theseabornstatistical plotting library.
• I had an error on the scatter plot : The palette list has the wrong number of colors.
• At the end of the notebook, we should put a text introducing the predict notebook

Predict notebook

• If I don't provide mask with the example dataset, I get a "trash" class (K=2) that hide other time series

[AndreaGarciaJuan]

Development notebook

• [x] In the text before plotting the mask : mask_path=None should be mask_path='auto'
• [x] I had an error when plotting the mask I added a sortby after unstacking the mask and I check dimensions order
• [x] The palette list has the wrong number of colors. When not all classes are represented in the subdataset we use to do this plot, there are so many colors in the palette. Now colors in palette are chosen depending on the unique values of labels in the subdataset
• [x] At the end of the notebook, we should put a text introducing the predict notebook Predict notebook
• [x] If I don't provide mask with the example dataset, I get a "trash" class (K=2) that hide other time series The problem was that I was making an interpolation. Now there is an option interp in OR_delate_NaNs function: if interp=True an interpolation is applied, if interp=False, all time series with any NaN value are deleted. Using interp = False we do not find trash classes any more

[AndreaGarciaJuan]

Now we provide a histogram and a box plot in log scale of the dataset in both notebooks:

[AndreaGarciaJuan]

• [x] Now we use .isocalendar for doing the weekly mean so the user needs to have the last version of xarray.
• [x] The maks created in the 'auto' option is a 2D mask now
• [x] There is an arrow indication the min in the BIC plot