Valencia flood mapping notebook - coregistration

[clausmichele]

https://eopf-toolkit.github.io/eopf-101/64_flood_mapping_valencia.html#coregistration

Since the GRD images we are working with already have the same dimensions (cropped during the initial steps of the tutorial), we can perform our own coregistration. Such process involves: - Keeping the coordinates of one intensity dataset as the reference - Resetting the coordinates of the other ones - Reassigning the coordinates to match the reference dataset and that they can all overlay perfectly.

I would like to point out that the content of the notebook is misleading, since there is no co registration occurring. The data has the same shape, but the corresponding pixel lat/lon pairs are not the same for each date. The data is forced on the first acquisition coordinates.

It should be highlighted that this might work in this case because the selected point is in a large flooded area, but a re-gridding (geocoding) on a regular grid would be the right way to align the data and work correctly with a time series.

Moreover, please test the notebook launching it with the button, as it is right now it doesn't work.

[ciaransweet]

Hi @clausmichele - Thanks for raising the issue, I'll raise this with the team, as to:

Moreover, please test the notebook launching it with the button, as it is right now it doesn't work.

Can you elaborate a bit on this?

[clausmichele]

xarray_sentinel is not available in the default image of https://jupyterhub.user.eopf.eodc.eu/hub/ , you should add a cell to install it!

Same for 23_S1_basic_operations.ipynb

[ciaransweet]

xarray_sentinel is not available in the default image of https://jupyterhub.user.eopf.eodc.eu/hub/ , you should add a cell to install it!

Same for 23_S1_basic_operations.ipynb

Okay good to know. If you come across other issues, thorough explanations are much more useful, as we can then identify the causes quicker.

You are also welcome to contribute fixes if you see a need! But I shall raise this with the team.

You can also use: 4zm3809f.c1.de1.container-registry.ovh.net/eopf-toolkit-python/eopf-toolkit-python:latest as an image to use on the JupyterHub rather than their default image, that's out of our control unfortunately.

We'll make a point to advertise our images we use to develop the notebooks though.

[beatrizbsperes]

Thanks for pointing that @clausmichele, which is in fact true!

We wanted to perform an actual co registration but unfortunately there are yet no zarr tools to do sentinel 1 co registration. This was a DYI way we found to continue with the times series analysis - choose one image as the master and assign its coordinates to all the others.

Maybe we should have been more clear about this DYI co registration implementation.

By any chance do you know any zarr working library that performs such operations?

[clausmichele]

You can also use: 4zm3809f.c1.de1.container-registry.ovh.net/eopf-toolkit-python/eopf-toolkit-python:latest as an image to use on the JupyterHub rather than their default image, that's out of our control unfortunately.

We'll make a point to advertise our images we use to develop the notebooks though.

@ciaransweet you could easily modify the link and make JupyterHub use your image, like I did for this notebook: https://eopf-sample-service.github.io/eopf-sample-notebooks/gdal-explore-zarr/
which using a custom image called clausmichele/eopf-zarr-driver:latest

We wanted to perform an actual co registration but unfortunately there are yet no zarr tools to do sentinel 1 co registration. This was a DYI way we found to continue with the times series analysis - choose one image as the master and assign its coordinates to all the others.

Maybe we should have been more clear about this DYI co registration implementation.

By any chance do you know any zarr working library that performs such operations?

@beatrizbsperes I totally understand, since I'm also trying to use the Zarr GRD data for a time series analysis, but I wouldn't use shortcuts like this. Otherwise some users might get confused and use the data in the wrong way for other applications, leading to even more confusion on the topic. I wouldn't even call it DIY co registration, since there's no co registration at all.

Anyway, unfortunately no, I'm not aware of pre-existing solutions.

[ciaransweet]

@clausmichele can you link to some documentation around that functionality of adding the image into the link? I didn't find anything when I originally looked into doing that 😄

[clausmichele]

@clausmichele can you link to some documentation around that functionality of adding the image into the link? I didn't find anything when I originally looked into doing that 😄

I also made it via several trial and error. Check the discussion here: EOPF-Sample-Service/eopf-sample-notebooks#171

[clausmichele]

FYI I published a notebook performing calibration + geocoding of a time series here: https://eopf-sample-service.github.io/eopf-sample-notebooks/radar-vegetation-index

Please note that the calibration required some special attention due to an issue with the EOPF product:
https://eopf-sample-service.github.io/eopf-sample-notebooks/radar-vegetation-index/#calibration-rvi

Reference issue and discussion about the calibration problems can be found here: https://forum.step.esa.int/t/constant-offset-for-grd-calibration/45194/9?u=clausmichele