Back to DrivenData | Blog

Sentinel 1 RTC - lack of real-time data

I have noticed that the source s3://sentinel-s1-rtc-indigo/ of Sentinel 1 RTC stopped providing real-time data in April this year. Could you please confirm if I am correct? This could be a problem during the stage 2b of the competition.

Screenshot from AWS console:

5 Likes

Would be great to investigate the data availability for the Sentinel 1 data, it’s one of the few allowed remote sensing data sources providing unique measurements band. But without the recent data to match the inference phase timeline, it’s of very little to no use for this task, unfortunately.

I quickly checked for the alternative data locations, but could not find anything suitable, but maybe I have missed something?

Hi @rszostak, thanks for surfacing this, it’s unclear why this s3 bucket is no longer getting the latest data. I’d suggest looking into the Copernicus Open Access Hub. Let us know if you can find the continuation of this series there: https://scihub.copernicus.eu/dhus/#/home

If it’s not available there, we can consider supporting GRD instead: Sentinel-1 GRD | AIforEarthDataSets

Hi @emily, given that s3://sentinel-s1-rtc-indigo/ does not provide real-time data, does this mean that https://scihub.copernicus.eu/dhus/#/home becomes the approved data source for inference?

We’re still investigating options but will add an announcement when we’ve determined which dataset will be supported for inference.

Hello @emily , do you have any update regarding this topic ?

Thanks for your patience. Since Sentinel 1 RTC data is not currently available for 2022, Sentinel 1 GRD has been approved as a replacement data source that can be used for inference.

For information about this data source, please see the updated Approved features list. The approved data access location is: https://sentinel1euwest.blob.core.windows.net/s1-grd

2 Likes

Thank you for the update !

Deriving Radiometric Terrain Corrected (RTC) data from Ground Range Detected (GRD) data is an expensive (in terms of compute) operation. The processing steps for the first approved dataset are described here and it seems to be easily reproducible using the open-source SNAP toolbox for ESA. This could permit to use the RTC data from Indigo Ag up to April 2021 and process later images using GRD + SNAP.

Howevere steps 6 and 7 for terrain flattening and terrain correction require the SRTM 1Sec HGT DEM while only Copernicus DEM is allowed for the challenge so far. In addition, Copernicus DEM seems to be not supported by SNAP Toolbox (see forum post here).

Is it possible to use SRTM 1Sec HGT DEM within SNAP for this purpose ? It would avoid processing all data before April 2021 which is not feasible for people without access to large compute.

Thanks

I made a small experiment to test the value of RTC data :

  • Choose a RTC cell with a lot of in-situ site measurements (cell 10SGJ)
  • Retrieve backscatter measurements σVV, σVH and local incident angle θ for the 2018 - 2019
  • On a selected area, compute median(- log(σ * cos(40°) / cos(θ))))

Here is the selected area :

And here are the results :

Once smoothed and linearly rescaled, σVV gets a RMSE of only 4 which is quite promising ! On the bad news side, Sentinel 1B is malfunctioning (see in the news here) meaning that only Sentinel 1A will be available for the evaluation stage (half the data :frowning: ).

Hope this can help some of you !
Simon

1 Like

(post deleted by author)

Just noticed Google Earth Engine also process GRD data : https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S1_GRD?hl=fr#description

Why did you choose not propose this processed dataset instead of the raw GRD data ? The processing using SNAP looks quite compute and bandwidth intensive