Resending this as I used the old qpack mailing list. Sorry for this.
Note that the new list is: [email protected]
--
Hi Rolf, hi all,
Rolf: Congratulations to the new position. I cc the qpack mailing list
as this could be of general interest.
If we start by just discussing oem.m:
* Yes, the function is written in such way that it should be possible to
couple it to another forward model.
* Note that Bob Sica and Alexander Haefele have done exactly the above
for some other lidar measurements. See:
---
Applied Optics website:
http://www.opticsinfobase.org/ao/abstract.cfm?URI=ao-54-8-1872&origin=search
The paper is Open Access, if you can't retrieve it at the above link
try: http://drp.rjsi.ca/1lkj0/G6dBktZZ
---
* Since a few weeks there exists another option! In an ongoing master
thesis project, Simon Pfreundschuh is creating a generic OEM library in
C++, which is also part of ARTS v2.3. This library has more features
than oem.m and should be much more efficient for larger inversions. I
assume that Simon will make some PR for the invlib library later, If you
curious already now, contact Simon (see cc).
To be clear, it should be possible to link his library to all/many (?)
forward models.
Regarding Qpack:
* At some point I will switch from using oem.m, to run OEM inside of
ARTS (by Simon's invlib library).
* I have never thought about using the complete Qpack together with
another forward model, but that should be relatively straightforward.
Some hacking of qpack.m should suffice (in addition to set up the
interface to oem.m), but would be more nice to solve it a general manner
(i.e. specifying the forward model to use by some user settings). If
anyone would be interested in this option, contact me and I will take
closer look at it.
Bye,
Patrick
On 05/13/16 11:55, Rolf Rüfenacht wrote:
Hi Patrick
A few weeks ago I started a PostDoc at IAP Kühlungsborn where I gave a
talk last week. The people here are very interested in applying optimal
estimation to lidar observations. Until now I have just been using
ARTS/QPACK by simply specifying Q-structures and letting QPACK do the
rest of the job. When wanting to run QPACK with another forward model,
would it be possible to just use oem.m and specify an alternative
function to arts_oem.m containing the new forward model? Or is arts used
at other places or are there other things to consider?
Kind regards from the other side of the Baltic Sea
Rolf
_______________________________________________
qpack.mi mailing list
[email protected]
https://mailman.rrz.uni-hamburg.de/mailman/listinfo/qpack.mi
