geolocation accuracy

[johntruckenbrodt]

The S1-NRB product defines a geolocation accuracy estimation as radial root mean square error (rRMSE) that is based on the accuracy of the source product and that of the DEM. This information needs to be written to the metadata.
The following aspects have to be considered:

• different values for the SLC accuracy have been reported and the source used for this product needs to be documented
• the accuracy must be adjusted when applying ETAD correction
• Copernicus DEM accuracy
  • values for different regions of the world can be taken from the Copernicus DEM Product Handbook
  • detailed values for individual DEM tiles can be read from an ancillary file *ACM.kml (accuracy mask); however, not all tiles contain this information.
  • the ACM file is not present in the freely available option with ID Copernicus 30m Global DEM
• the accuracy of other supported DEMs is yet to be investigated.

The implementation could be done in multiple stages:

• make use of the global LE90 value reported in the Copernicus DEM Handbook and the accuracy values of the SLCs reported in the S1-NRB product specification. The error at 1 sigma can be estimated from LE90 as LE68 assuming Gaussian distribution
• differentiate between global regions as reported in the DEM Handbook
• read tile-specific accuracy if existing and fall back to values above otherwise
• investigate other sources of SLC accuracy (for example annual Sentinel-1 performance reports) and refine the computation if possible
• include DEM resampling error into the rRMSE computation
• include SAR image processing error into the rRMSE computation

[johntruckenbrodt]

ALE comparison between different softwares:
s1tools/s1-etad#3 (comment)