Hi George (and all), Great that you are asking. Gives a pressure to fix things that should have been fixed earlier. At start of qpack2 I tried to avoid to include stuff that demanded knowledge on the order inside the state vector (x), but such variables have been added step by step and there has not been a complete solution to the issue.
In short, I have now made two commits that should fix this. The changes are at hand in both atmlab v2.2 and v2.3. As Ole Martin wrote, you can now get the complete A matrix (selected by the string 'Afull'). In addition, I have just added that when A or any other variable matching the complete x is selected, you also get two structures denoted as jq and ji. These structures are described on page 9 in latest version of qpack2.pdf (found in folder atmlab/documentation/qpack2). The positions that Hugh gave you seem to be correct. However, from now on that information is at hand by ji. For the order between the retrieval quantities, see jq. The order in jq and ji is the same. To be clear, L2.A is the complete averaging kernel matrix, while e.g. L2.species1_A is the part around the diagonal of the complete A matching the retrieval quantity of concern. Bye, Patrick On 2015-01-09 11:24, Ford, George P. wrote: > So to clarify, the avk I have been using up to now are those for the > species named in that L2 structure only (L2(1).wind_u_A), but that > using the method you described gives avk for all contributions to the > retrieval? > > A collaborator at Edinburgh (Dr Pumphrey) broke the state vector into > these constituents for me, am I to assume the resulting avk will have > much the same arrangement? % 1-48: Species % 49:72: One wind % 73-96: > the other wind % 96-102: The baseline. > > Many thanks for your help, > > George > > > > -----Original Message----- From: Ole Martin Christensen > [mailto:[email protected]] Sent: 08 January 2015 16:14 > To: Ford, George P. Cc: [email protected] Subject: RE: Qpack: > Retrieval using multiple species > > Hi, > > I'm sorry, but I realized that qpack does not give you the complete > AVK-matrix even if you follow my instructions from last mail. To get > the full averaging kernel matrix you need to print out the Jacobian- > and gain matrix and multiply them together: > > Q.L2_EXTRA = { 'J', 'G' }; > > AVK = G*J; > > This gives you the complete averaging kernel matrix and you can then > analyse how much the different species contribute to your retrial of > the winds at different altitudes. Note that there is currently no > indication from qpack in where in this matrix the different variables > are stored, so some experimentation from your side is probably needed > to find the correct entries. > > PS: Patrick added the possibility to get the full AVK to the > development version (2.3) of atmlab today, so if you are using this > version you can get the matrix by adding the variable 'Afull' to the > L2_extra field. > > PPS: If you want to experiment I can suggest setting the retrieval > grid for the wind field to a different length than those of the > species to more easily separate between the variables. > > Regards > > Ole Martin ________________________________________ From: > [email protected] [[email protected]] on behalf > of Ole Martin Christensen [[email protected]] Sent: > Monday, January 05, 2015 1:38 PM To: Ford, George P. Cc: > [email protected] Subject: Re: [Qpack] Qpack: Retrieval using > multiple species > > I might be misunderstanding you, but the averaging kernels you get > from Qpack in the species1_A or wind_u_A, only contain the part of > the averaging kernel matrix related to the variable retrieved. This > means that the wind_u_A tells you the relation between the true wind > and retrieved wind, and as such is independent on the number of > species you retrieve. > > To get the full AVK matrix, you can include 'A' in the L2_EXTRA > field. > > Q.L2_EXTRA = {'A'}; > > I hope this helps > > Ole Martin ________________________________________ From: > [email protected] [[email protected]] on behalf > of Ford, George P. [[email protected]] Sent: Monday, January 05, 2015 > 12:21 PM To: Ole Martin Christensen Cc: [email protected] Subject: > Re: [Qpack] Qpack: Retrieval using multiple species > > I think I'm doing that already (I have 4 or 5 species in my model > atmosphere) but the resulting averaging kernel for wind (L2.wind_u_A) > suggests it's only using the first species despite '.RETRIEVE' and > '.L2' being set to true for both CO and Ozone. My issue might be with > getting the useful diagnostic information out. > > > -----Original Message----- From: Ole Martin Christensen > [mailto:[email protected]] Sent: 05 January 2015 11:13 > To: Ford, George P.; [email protected] Subject: RE: Qpack: > Retrieval using multiple species > > Hi, > > you can add a second (or more) species to Qpack by expanding the > Q.ABS_SPECIES field in the following way: > > Q.ABS_SPECIES(1).TAG = { 'CO' }; . . Q.ABS_SPECIES(2).TAG = > { 'O3' }; . . > > And so forth. Note that the spectral line for each species must also > be added to your linefile as well as to your atmospheric fields > (Q.ABS_SPECIES(2).ATMDATA). To retrieve the species, the covariance > matrix, retrieval grids, etc. must be set separately for each > species. > > For a complete example check out the qpack2_demo2.m file in > atmlab/demos > > Regards > > Ole Martin Christensen ________________________________________ From: > [email protected] [[email protected]] on behalf > of Ford, George P. [[email protected]] Sent: Monday, January 05, 2015 > 12:03 PM To: [email protected] Subject: [Qpack] Qpack: Retrieval > using multiple species > > Hello Qpack community, and a happy new year. > > I am currently running wind retrievals on simulated Ozone and CO > lines with some success. I was wondering if anyone knew how to run > the retrieval on two species simultaneously? I have been able to use > double lines but only of the same species. > > Any pointers would be much appreciated. > > Regards, > > > --------------- Dr George Ford Atmospheric Scientist British > Antarctic Survey > > E: [email protected] T: +44(0)1223 22 1395 > > > > This message (and any attachments) is for the recipient only. NERC is > subject to the Freedom of Information Act 2000 and the contents of > this email and any reply you make may be disclosed by NERC unless it > is exempt from release under the Act. Any material supplied to NERC > may be stored in an electronic records management system. > _______________________________________________ qpack mailing list > [email protected] https://www.sat.ltu.se/mailman/listinfo/qpack > > This message (and any attachments) is for the recipient only. NERC is > subject to the Freedom of Information Act 2000 and the contents of > this email and any reply you make may be disclosed by NERC unless it > is exempt from release under the Act. Any material supplied to NERC > may be stored in an electronic records management system. > _______________________________________________ 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 > > This message (and any attachments) is for the recipient only. NERC is > subject to the Freedom of Information Act 2000 and the contents of > this email and any reply you make may be disclosed by NERC unless it > is exempt from release under the Act. Any material supplied to NERC > may be stored in an electronic records management system. > _______________________________________________ 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
