[RegCNET] sigma2p unpredictable interpolation errors

[Rebecca Porter]

Hello!

Been running into extensive sigma2p interpolation errors and am looking for
some help finding a reliable solution. I am running many different
configuration tests of RegCM, and then interpolating to pressure as the
vertical coordinate using sigma2p. I can't seem to find a reliable
methodology to follow for sigma2p implementation, as it seems like whether
sigma2p errors out is completely random, and the fix is not the same for
all configurations.

Whether this error occurs seems to not just depend on the configuration,
but also how many levels I am interpolating to. I've written about 2
examples and what debugging processes work or what don't work to get
sigma2p running. The below examples also represent two different
configurations of RegCM:

Example 1: sigma2p works perfectly fine if set for 11 plevs, but bumping to
12 plevs (based on the formatting suggested by regcm documentation),
sigma2p does not work. Subselecting to have less data using nco operator
"ncks" does not fix the problem.
Example 2: sigma2p errors for 12 plevs. I've tried subselecting the data
using "ncks" and that has resolved the problem for this example. I can even
run at higher plevs like 15-20 plevs.

Of course, there are also configurations where running sigma2p just works
the first time around and doesn't error out.

This is the error I receive:

`Program received signal SIGSEGV: Segmentation fault - invalid memory
reference.

Backtrace for this error:
#0 0x7fa8c141e62f in ???
#1 0x440f39 in ???
#2 0x403974 in ???
#3 0x7fa8c1e3eed5 in GOMP_parallel
at ../.././libgomp/parallel.c:178
#4 0x40bdbe in ???
#5 0x403388 in ???
#6 0x7fa8c140a864 in ???
#7 0x4033bd in ???
#8 0xffffffffffffffff in ???
Segmentation fault (core dumped)`

The forum will not let me attach the example netcdf files or namelist files
I would like to, but if anyone would like those I am happy to share over
email.

Thanks for the input!

Best,
Rebecca Porter
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ictp.it/pipermail/regcnet/attachments/20260318/55bbfc82/attachment.htm>