...
- Difficult to aim for a single forward operator that works for everything
- Global vs local NN. Difficult to have a global NN that works globally even accounting for vegetation
- Important to filter out frozen soils
- RMS of 2dB is still very large compared to the signal. The measurement noise is 0.2dB. It would be useful to have RMS closer to the measurement noise. At a given location in the temporal domain, the main driving factor is SM. Spatial approach give more information on vegetation. Subsurface scattering: issue everywhere. The best variable to predict is the coarse factor.
- Also need to account for slope and incidence angle computation
- Trend in backscatter data related to land cover changes
- ASCAT is also affected by RFI issues
Meeting notes:
Slides: