search

FIDUCEO / FCDR_HIRS

Other
1 stars 2 forks source link

separate fully random effect from "filter wheel wobble" effect and propagate this through CURUC #224

Open gerritholl opened 6 years ago

gerritholl commented 6 years ago

In the CURUC recipes we need to estimate correlation matrices with associated uncertainties. Previously I added the "filter wheel" correlations (spatially and spectrally) to the fully random effect, but that is not right. However, I do need some uncertainty estimate for the structured "filter wheel" effect (correlated error). One way would be to take the difference between Allan deviation and standard deviation. In this case I need to be careful because the Allan deviation can sometimes be larger than the standard deviation. Would need to check how often that happens and decide what to do in this case. Related to colours of noise, see #63.

gerritholl commented 6 years ago

And then the result of this should be propagated into the CURUC recipes.

Perhaps until then I can assign the measured inter-channel correlation to the independent effects?

gerritholl commented 6 years ago

Closely related to #21, possibly a duplicate?