Dear Patrick, I agree with you and first I started by trying to retrieve LWC.
But I have a high convergence problem in Qpack when I retrieve the liquid water content but I had to do some modifications in the oem.m routine due to different units in my background error covariance matrix and my jacobians. The problem is that the conversion from kg/kg to g/m³ depends on pressure, temperature and the liquid water content whereas the B matrix was computed from an ensemble assimilation system with several members during a large period of time with different LWC profiles.. I do not think I can convert my B matrix to g/m3 by selecting one profile of liquid water content, temperature or pressur randomly. The covariances depend on the atmospheric state. Thus, I decided to convert the jacobians for each profile from g/m³ to kg/kg which is not so complicated and I already did that with another 1D-Var algorithm. With my changes, Qpack runs but never converges. I probably made a mistake but I checked several times and I do not see where my problem is. Thus, I have already reviewed one my article and said that I cannot retrieve the liquid water content profile. Now it is the second review and the reviewer thinks it would be easier with the liquid water path. Maybe the rate of convergence with the liquid water path will be better as ground-based microwave radiometers are only sensitive to the integrated content and not the profile. Do you think I can try to change my retrieval to liquid water path or, for you, it is too complex ? Thank you again. Best regards, Pauline ----- Météo-France ----- Dr. Pauline Martinet Chercheur CNRM/GMEI/LISA [email protected] Fixe : +33 561079031 Site web: www.sites.google.com/site/martinetpauline31 ----- Mail original ----- De: "Patrick Eriksson" <[email protected]> À: "Pauline Martinet" <[email protected]>, [email protected] Envoyé: Lundi 7 Septembre 2015 17:08:49 Objet: Re: [Qpack] liquid water path retrieval with Qpack Dear Pauline, Just because the reviewer wrote LWP, I don't think you have to retrieve exactly LWP. Maybe the reviever meant "liquid water" when he/she wrote LWP!? In fact, the most efficient way the estimate LWP is the retrieve the LWC profile. If my memory is correct, Rodgers makes a point of this in his book. Let me also make the point that a LWP of e.g. 0.3 mm will have different impact on the temperature retrieval if the water is found between e.g. 0-1 km, or 5-6 km. What matters is in fact the LWC profile. Bye, Patrick On 09/07/15 11:54, Pauline Martinet wrote: > Dear everybody, > > Sorry for all my questions these last days, I have two projects with > Qpack and I have to deal with both of them at the same time. > > I am reviewing an article using Qpack but the reviewer asks me to > simultaneously retrieve the liquid water path (LWP) with the temperature. > > I would like to modify the routine oem.m of Qpack to minimize the cost > function with respect to the liquid water path and not the liquid water > content profile. > > I have some ideas in mind: when I am in the oem.m routine like: > > 1) Modify the B matrix to only have one value for LWP instead of > background errors for the whole liquid water content profile > 2) After the computation of the Jacobians for the liquid water content > profile, change it to only have one value of Jacobians (delta TB with > respect to LWP) > 3) after finding the xstep that minimizes the cost function, go back > from the LWP increment to an increment on the whole liquid water content > profile. > Then, the new x profile is found, no modification for the convergence > test should be done. > > I do not know if it is more complicated then that as I do not know well > how the whole Qpack is coded, I mainly know this oem.m routine. > > Does anyone have ever done such a modification ? If not, does anyone > could help me doing it alone ? > and do you think it would take a lot of time. I only have 4 weeks to > send back my reviews. If it sounds too complicated, I will try > to explain to the reviewer this is not possible with the current version > of Qpack. > > Thank you very much in advance ! > > Best regards, > > Pauline > > ----- Météo-France ----- > Dr. Pauline Martinet > Chercheur CNRM/GMEI/LISA > [email protected] > Fixe : +33 561079031 > Site web: www.sites.google.com/site/martinetpauline31 > <https://www.sites.google.com/site/martinetpauline31/> > > > > _______________________________________________ > qpack mailing list > [email protected] > https://www.sat.ltu.se/mailman/listinfo/qpack > _______________________________________________ qpack mailing list [email protected] https://www.sat.ltu.se/mailman/listinfo/qpack
