Hi Patrick,
Thanks a lot.
I am trying to retrieve ozone profiles considering the covariance matrix of
temperature and water vapour.
With 'good shape' I mean that it is around a number in the whole frequency
range (straight). Now I included a calculation of a baseline in each
iteration and it seems to improve the residual. I will also try to set
POLYFIT in Q.DEFINITIONS.m.
Thanks again.
Regards
Facundo


2015-12-06 10:05 GMT-03:00 Patrick Eriksson <[email protected]>:

> Hi Facundo,
>
> I am writing again because I did that I said in the previous mail and
>> retrieved ozone profiles are not bad, but the residual (yf-y) is not
>> around zero...It goes down to around 0.3 (the shape is good, but goes
>> down). It happens when I introduce the error related to T and H2O, and
>> also when I introduce only T or only H2O. Residual are very similar in
>> this 3 cases (T and H2O, T or H2O). When I do not include these errors
>> (only considering noise) it does not happen. I was looking for what is
>> wrong and I could not find it. Do you have any suggestion about it?
>>
>
> This is not easy to answer without having all the details at hand. Exactly
> how you calibrate the measurements, exact retrieval settings etc. Some
> questions:
>
> * What do you mean with good shape of the residual?
>
> * Do you make a non-linear retrieval? With T or H2O you might need to
> iterate to find the exact solution.
>
> * Do you include any POLYFIT? To include POLYFIT is quite standard, to
> reflect that your calibration newer is perfect. It is likely that the
> complete spectrum is wrong with roughly 1K just due to calibration issues.
>
>
> On the other hand, what is exactly the use of the variable 'xnorm'?
>>
>
> It is a feature to avoid numerical problems if your quantities have very
> different values. If you use Qpack, this is activated automatically, when
> could be needed.
>
> Bye,
>
> Patrick
>
_______________________________________________
qpack mailing list
[email protected]
https://www.sat.ltu.se/mailman/listinfo/qpack

Reply via email to