Hi Victoria,

did you get any (offline) help yet?

I can't say anything on the qarts side. also, some of my ARTS knowledge is
a bit rusty in the meanwhile. and it's been a while, too, since I
implemented and used jacobianDoit. But let's see, whether I can be of some

First, a cautionary note. I'm using current ARTS2.3 terminology unless
otherwise noted. Some variables and methods have changed names somewhere
along the way from 2.2 to current 2.3, which might lead to confusion when
not being considered/checked. so you might have to look up, what I
specifically refer to. This definitely is true for the
scat_data/meta(_/single/array) variables (what now is scat_data, once was
named scat_data_array. and what was scat_data then, now is
scat_data_single! easy to mix up!)!

Then, are you aware that there are newer methods to derive
scattering/all-sky Jacobians (in the dev version, 2.3) that use the
"normal" in-yCalc semi-analytical jacobian approach? check out
jacobianAddScatSpecies and the example in
controlfiles/artscomponents/wfuns/TestWfunsHybCloudy.arts (I see that in
this controlfile, there is also still a - commented out - setup for the
same Jac calculation using jacobianDoit. for comparison.).
I'm fairly confident, that this approach can be used through a current
qarts, since this approach was pushed for a lot by Patrick and I think, he
makes (at least made) use of it through qarts. (I'm not sure, whether
jacobianDoit can. Patrick might say more whether there might be an issue.).

Your scat_meta problem might be an issue for either approach. Both
approaches are, if I remember correctly, made with the assumption that you
use one of the ARTS-internal size distribution (PSD) parametrizations. I
think, we haven't considered the case that you just scale all scatt optical
properties by a constant (equivalent to scale the PSD with a
size-independent factor). Which should be the only exception when in the
calculation of the Jacobians you wouldn't need access to the PSD.
That is, in the Jacobian calculation the PSD is modified, which means we
need size-binned scat_data AND we need "meta" info for these data (which
particle size/mass does each specific scattering element represent?). (one
could probably circumvent the size-binning requirement by constructing
single-scatt-element hydrometeor species and using an arbitrary PSD
parametrization as dummy. but I'm not sure, the PSD methods accept
single-scatt-element species. if not, the easiest could be to add a
specific method for "one element PSDs". Which, however, would still need a
scat_meta since it needs to know how much mass that single element holds.
However, you'd have to do that yourself; I don't have time for ARTS devs in
the foreseeable future).
That for the explanation, why this info is needed for running the Jacobian

As for how to get the scat_meta into ARTS. The base rule is, scat_meta has
to have the same internal structure as scat_data (nowadays being an
ArrayOfArrayOfSingleScatteringData, that means same number of first level
array elements - those correspond to the hydrometeor species considered -
and same number of second-level array elements for each equivalent first
level element - corresponding to the individual scattering elements for
each hydrometeor species).

This is guaranteed when you use the dedicated scat_data/meta read method
(ScatSpeciesScatAndMetaRead). You can do that manually, too, using ReadXML,
but need to be careful to create the correct 2-level array structure, which
can be a bit tedious, I assume.

In ARTS2.2 (and some earlier 2.3), the same base rule applies, but both
variables are single-level array structures there, making the reading a bit
easier. But you will need an additional index array that tells ARTS which
scattering elements in scat_data (and scat_meta) belong to which
hydrometeor species. Not sure anymore, what was its name (scat_species?)
and how exactly that was supposed to get into ARTS consistently with
scat_data and scat_meta. Checking out the jacobianDoit controlfile example
in your ARTS version is the best way to find out, I think.

That's as much as I can diagnose from my memory of ARTS and the
online-accessible code & documentation. Hope it allows you to go on!
Best wishes,

On Tue, Jul 23, 2019 at 2:21 AM Victoria Sol Galligani <
victoriasgallig...@gmail.com> wrote:

> Hi everyone!
> Can anyone help me with What the status of the jacobianDoit calculations
> is?
> I'm trying to run jacobianDoit though qarts with my own scat data and pnd
> fields. So far I have never used scat_meta_data, but for jacobianDoit
> calculations I see that I should be using it. I have created my own
> scat_meta_data_single files, but I don't see what method generates the
> merged scat_meta_data that jacobianDoit needs.
> Thank you in advance for your help,
> Victoria
> _______________________________________________
> arts_users.mi mailing list
> arts_users.mi@lists.uni-hamburg.de
> https://mailman.rrz.uni-hamburg.de/mailman/listinfo/arts_users.mi

Jana Mendrok, Ph.D.
Deutscher Wetterdienst
Frankfurter Str. 135
63067 Offenbach am Main, Germany

Email: jana.mend...@dwd.de
Phone : +49 (0)69 8062 3139
arts_users.mi mailing list

Reply via email to