New ancillaries for future continuation run, Halos, NaNs in error term in BiCGstab

Hi CMS Helpdesk:
I have two ARCHER2 suites (u-cm922, u-cm923) that are continuation future runs of previous suites that ended on 20150101. They have different ancillaries for sea ice, SSTs, ozone, and aerosols than the previous suite. They both initially crashed with halo errors.

I have tried running the suites with a startdump from the previous suites’ runs from 20140101 instead of 20150101, thinking that maybe the 20140101 startdump might work better, since the 20150101 was the final time step of the prior suites. But this attempt still gave the halo errors.

I tried to follow the advice at
making sure the extra diagnostic printing was enabled.
And for u-cm922, I set l trap uv =.true., which led to an ‘NaNs in error term in BiCGstab’ error.
There is online advice for that too, like reducing the time step. But I don’t know the best way to go now. Maybe I have something set up wrong?
Can you suggest anything?

By Halo error, I meant "(North/South halos)/Halos too small for
advection ", as on that MetOffice web page :

Here is an example error message that I got.

[0] ???!!!???!!!???!!!???!!!???!!!       ERROR
[0] ?  Error code: 15
[0] ?  Error from routine: LOCATE_HDPS
[0] ?  Error message: North/South halos too small for advection.
[0] ?        See the following URL for more information:
[0] ?
[0] ?  Error from processor: 234
[0] ?  Error number: 64


Malcolm R. suggests: "Did you change the GHG and other forcings too? In your
u-cm922 do you have in the rose-suite.conf a line like:

UM_OPT_KEYS='global_ghg_cmip6_ssp585_2014_2100 cmip6_solar ens1_2015 noclimmean n512 ios'

this points to the ghg, solar for the future years (files in app/um/opt)."

Indeed these weren’t in the suite. So I will try to fix that.

The two future suites with the properly-configured future GHG’s have been running atmos_main for about 25 minutes now, which is much better than before.