[MITgcm-support] LabSea model collapse

[尚丞明]

Dear MITgcm Community,




In "data" runtime file, I did not set ''diffKzT" and "diffKzS" when run the model. However, I have added values to the "data" in the attachment.


Sincerely,

Chengming




-----原始邮件-----
发件人:尚丞明 <22320250156689 at stu.xmu.edu.cn>
发送时间:2026-03-18 22:28:15 (星期三)
收件人: mitgcm-support at mitgcm.org
主题: LabSea model collapse



Dear MITgcm Community,




(Dimitris Menemenlis suggested I move this discussion from GitHub Issue to the mailing list.)




I am currently running a high-resolution (3.5 km, 340*468*43) regional model for the Labrador Sea. I am using the EXF and KPP modules. My sea surface forcing field is provided by directly specifying net heat flux (NHF), shortwave radiation, wind stress, and significant freshwater runoff along the coast.Specifically, I placed the runoff separately into the SFLUX channel.

Problem Description:
My model runs fine in winter and spring, but crashes in early autumn (September) with a SURF_ADJUSTMENT error (specifically, cg2d solution divergence).




Diagnostic Analysis:
By examining the output before the crash, I discovered a "heat trapping" issue: In summer, the sea surface temperature (SST) reaches unrealistically high values (approximately 30°C). The attached figure displays the 5-day mean temperature distribution: the first row shows the meridional inflow profile at the eastern boundary, while rows 2 through 4 present the temperature at the surface, 15 meters, and 60 meters, respectively. Please pay close attention to the colorbar.

I have attached the crucial configuration and runtime files. In the "output.txt" file, I retained only the last 100 "cg2d" outputs, as there were simply too many of them.




MITgcm version is checkpoint69e.




I eagerly await your suggestions and guidance!







Sincerely,

Chengming
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20260318/9e2052cc/attachment.html>