Hi all,

Just made a commit with two changes that affect Qpack. One extension, 
and one that could require a small modification.

1. For both theoretical and practical reasons, I have changed the L2 
output for ABS_SPECIES to follow the selected retrieval unit. This 
affects the rel and logrel options. The output was before converted 
automatically to VMR. This conversion is now moved to the function 
qp2_rel2vmr.
The theoretical justification is to return data that are fully 
consistent with the assumptions of the retrieval. This point is most 
critical for the logrel case, where the returned data theoretically have 
a normal distribution while converted to VMR it would be log-normal. 
However, using logrel can cause substaintail non-linearity in the 
retrieval problem and it is likely that the errora not follow normal 
distributions.

2. The setting ABS_SPECIES.L2_RANGE is introduced, that allows to 
fine-tune L2 data. This simplifies to make use of the fact that 
retrieved profiles are treated as constant outside the end points. For 
example, with rel you can retrieve a constant relative change of the 
profile above some pressure (which then should be the minimum pressure 
of the retrieval grid).
See further qinfo(@qarts,'ABS_SPECIES')

Bye,

Patrick


-------- Original Message --------
Subject: [s-c-rt] r6966 - in atmlab/trunk: . arts geophysics 
retrieval/qpack2
Date: Sun, 22 May 2011 15:54:12 +0200
From: [email protected] <[email protected]>
To: [email protected] <[email protected]>


             * atmlab-1-0-129

             * Important for Qpack users. L2 output for ABS_SPECIES now 
follows
             the selected unit strictly. More information in qpack2.pdf and
             email to qpack mailing list.

             * retrieval/qpack2/qpack2.tex: Modified to reflect changes 
below.
             The information for L2 'ptz' was old.

             * retrieval/qpack2/qp2_rel2vmr.m: Created.

             * arts/arts_regrid.m (extract_grids): Introduced qname, to 
improve
             error message.

             * arts/qarts.m (qarts): Added L2_RANGE as sub-field to 
ABS_SPECIES
             and T.

             * retrieval/qpack2/qp2_l2.m (qp2_l2): For absorption 
species, the
             conversion to VMR for 'rel' and 'logrel' is removed. That 
is, the
             output follwos now throughout the retrieval unit. This is 
also the
             case for errors estimates.
             Added handling of L2_RANGE (and this part of the function 
revised).

             * arts/arts_x2QR.m (arts_x2QR): "Extrapolation" was done 
for VMR
             independently of retrieval unit selected. This follows now the
             retrieval unit. For example, if "rel" is selected and the end
             value is 1.1, the VMRs on this side of the grid are scaled with
             1.1. Before the end VMR value was used out to the atmospheric
             limit.

             * arts/arts_oem_init.m (arts_oem_init): Improved header. 
Revised
             the set of variables that are stored as part of R.

             * geophysics/nd2vmr.m: Spelling mistake in header.

_______________________________________________
qpack mailing list
[email protected]
https://www.sat.ltu.se/mailman/listinfo/qpack

Reply via email to