Skip navigation.
Home

Schedule and progress

IncuSAR workplan

 

Gantt chart showing project schedule

 

Task list: 

  • Implement neural network inverse CMOD function to retrieve wind speed from given roughness (sigma0) and incidence and wind-look angle. This is to speed upthe Bayesian inversion scheme.
  • Compare SAR/model wind speed distribution versus observed frequency distributions.
  • Add more sources of wind into retrieval scheme:
    • Several models, e.g. NCEP, ECMWF and Hirlam. The wind information from these models is not independent (assimilated from same sources), but it should be tested if accuracy nevertheless can be improved by including more models.
    • Scatterometer winds, e.g. ASCAT which is often colocated closely in time with Envisat.
  • Test interpolation of wind to time of SAR image, instead of using only nearest model field. Should test sensitivity the time difference.
  • Test sensitivity of spatial averaging of SAR backscatter.
  • Develop inverse CMOD function to retrieve most probable (nearest) wind direction, given wind speed and roughness.
  • Splitting into radial and azimuthal components in wind retrieval scheme, instead of zonal and meridional (U and V) as used by default. Doppler velocity may then be used in cost function for radial component only.
  • Test if performance of certain CMOD-versions are better than others in Bayesian scheme. Kerbaol found that CMOD-IFR2 was more useful in Bayesian scheme because itis more sensitive to wind direction than CMOD-4, and hence can correctfor errors in model wind direction. He also found more improvement for higher incidence angles and cross winds, where sensitivity to wind direction is largest.
  • Test various cost functions for the Bayesian inversion: Maximum likelihood, maximum aposteriori, minimum mean square error.
  • Include Doppler velocity into wind retrieval scheme.
  • Include FFT wind streak directions into wind retrieval scheme. Could also include several possible FFT solutions.
  • Develop inverse CDOP function (fD -> wind) for both HH and VV polarization. Possible to get from ESA older (<2007/2008) ASAR WSM scenes with Doppler grid?
  • Should use CMOD developed on SAR data ("SARMOD") for better consistency between CMOD and CDOP model functions in inversion scheme. May be done in cooperation with CLS Radar Division.
  • Test bayesian scheme for dual pol SAR data.
AttachmentSize
Copy of Incusar_Gantt.PNG24.95 KB