Geolocation Accuracy

Item 4.3 of the CARD4L NRB specification requires, as minimum, an estimate of the absolute location error (ALE) “as bias and standard deviation, provided in slant range/azimuth, or Northing/Easting” [3]. As desired target the accuracy is less or equal 0.1 pixels radial root mean square error (rRMSE), which can be defined as:

\[RMSE_{planar} = \sqrt{RMSE_{SLC,Az}^2 + (\frac{RMSE_{SLC,Rg}}{sin(\theta_{i,min})})^2 + RMSE_{DEM,planar}^2 + RMSE_{proc}^2}\]

The error induced by the DEM can be described as:

\[RMSE_{DEM,planar} = \frac{\sigma_{DEM}}{tan(\theta_{i,min})}\]

where

\(\theta_{i,min}\) = The minimum possible angle of incidence

\(RMSE_{SLC,Az/Rg}\) = Error induced by SLC source data in azimuth/range

\(RMSE_{DEM,planar}\) = Error induced by DEM inaccuracy

\(RMSE_{proc}\) = Error induced by other processing steps

\(\sigma_{DEM}\) = DEM accuracy at \(1\sigma\) (LE68)

Limitations

Currently, the following simplifications need to be considered for the calculation of rRMSE values found in the metadata of each S1-NRB product:

  • Processing induced errors (\(RMSE_{proc}\)) and the error term related to DEM interpolation are not further considered and assumed to be 0.

  • The DEM accuracy (\(\sigma_{DEM}\)) is estimated on the global mean accuracy LE90 reported for the COP-DEM [1] under the assumption of gaussian distribution:

    • Global: LE90 = 2.57; LE68 \(\approx\) 1.56

  • rRMSE is only calculated if a COP-DEM was used for processing, otherwise the value is set to None

Development Status

The development status is tracked and discussed in the following Github issue: https://github.com/SAR-ARD/S1_NRB/issues/33