Assessing the radiometric impact of the Sentinel 2 orthorectification process

2021 
The Sentinel 2 (S2) orthorectification process converts the Level-1B (L1B) radiance data generated at sensor geometry into orthorectified top-of-atmosphere (TOA) reflectance data, which is the Level-1C (L1C) product distributed to users. However, the spatial resampling operations involved by the orthorectification also transform the radiometric quality of the data. In this work, we evaluate the impact of the S2 orthorectification process on the radiometric quality of the data, with focus on the radiometric uncertainty budget. In particular, the study reports the variation of the S2 noise model attached to the S2 L1C metadata and the effects on the L1C uncertainty products produced by an offline processor named Radiometric Uncertainty Tool (RUT). This assessment is divided into three steps, namely noise propagation, interpolation uncertainty, and covariance impact. For the first of them, the results show that noise is reduced by a factor of 0.65 from L1B to L1C data, both by simulations and an empirical approach that estimates noise variance from real L1B and L1C acquisitions at different sites. Regarding the evaluation of the interpolation uncertainty, aerial orthoimages are convolved by the S2 Point Spread Function (PSF) and upsampled into the S2 spatial resolution. The study shows that, from this S2-like image, a distribution of interpolation errors (i.e. uncertainty) can be associated to the standard deviation of the neighbouring pixels. Finally, the covariance change due to the spatial resampling has been simulated with a propagation of the L1B radiance errors in order to understand the increase of correlation between neighbouring pixels
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    0
    References
    0
    Citations
    NaN
    KQI
    []