Opened 10 years ago

Closed 8 years ago

#424 closed defect (fixed)

Problem with ropp_fm_bg2ro_2d

Reported by: Ian Culverwell Owned by: Ian Culverwell
Priority: major Milestone: 9.0
Component: ROPP (all) Version: 8.0
Keywords: Cc:

Description

Axel von Engeln (EUM) reports

I am just trying to run the ROPP with the 55 case profiles
generated in a DMI/RUAG/ECMWF study. For that I converted those 
background profiles into ROPP format and then ran it through ROPP8, 
either on the default 300 vertical levels, or the 247 ones. On the 
247 I get this for occultation 18 (index starts at 1):

FATAL ERROR (from set_obs_levels_refrac_247):  GPH iteration failed to converge in 100 iterations

The default run I do is:

ropp_fm_bg2ro_1d WOP_Study_Cases.nc -f -comp -new_op -o Output.nc

When adding the -247L levels in, I get the error above. When
removing "-comp -new_op", I already get the error at profile 3.

Any idea why that is? Or are you not having these issues with your
ROPP version?

I replied

This is not too surprising. The forward model generates
refractivities and bending angles on the corresponding 247L, and if 
you want the latter to match the standard 247 impact heights, then 
you need to do a bit of iterating on the refractivity levels.  I 
had some problems when developing this, but it finally seemed to 
work OK.  I'll look into it.

One thought: his levels are monotonically descending, ECMWF-style, not ascending as I (probably) assumed. This might have something to do with it.

Attachments (2)

WOP_Study_Cases.nc (260.8 KB ) - added by Ian Culverwell 10 years ago.
WOP_Study_Cases.nc
FM_levs.pdf (72.5 KB ) - added by Ian Culverwell 9 years ago.
FM_levs.pdf

Download all attachments as: .zip

Change history (9)

by Ian Culverwell, 10 years ago

Attachment: WOP_Study_Cases.nc added

WOP_Study_Cases.nc

comment:1 by Ian Culverwell, 10 years ago

Note that a longer term solution might be to allow ropp_fm_bg2ro_1d to produce either refracs or bangles, and therebye break the correspodence between the levels of both, which is causing the problems here.

comment:2 by Ian Culverwell, 10 years ago

The title of this problem should of course read Problem with ropp_fm_bg2ro_1d.

comment:3 by Ian Culverwell, 10 years ago

I've reopened #314 to address this. It now includes a possible solution.

comment:4 by Ian Culverwell, 9 years ago

Some of the nasty subtleties about this are illustrated in FM_levs.pdf (attached).

by Ian Culverwell, 9 years ago

Attachment: FM_levs.pdf added

FM_levs.pdf

comment:5 by Ian Culverwell, 9 years ago

First set of changes in support of this have been made at r4872.

comment:6 by Ian Culverwell, 8 years ago

Tested code thoroughly. ropp_fm/config directory added at r4876, and further cosmetic changes made at r4878. ROPP FM user guide updated at r4879.

comment:7 by Ian Culverwell, 8 years ago

Resolution: fixed
Status: newclosed

Satisfied with these changes, so closing ticket.

Note: See TracTickets for help on using tickets.