[MITgcm-support] GM/Redi in adjoint runs

[Christopher L. P. Wolfe]

Hi Modelers,

I'm trying to set up some adjoint sensitivity runs using a coarse-resolution model which uses the GM/Redi package. I've had a few people tell me that I could into issues using GM/Redi in the adjoint, though I've also seen several papers (e.g., Ferreira and Marshall 2006 and Hill et al. 2004) which have used GM/Redi in the adjoint.

Does anyone have any advice for working with GM/Redi in the adjoint? I would be using GM in advective form and running sensitivities with timescales of decades. I'm especially interested in using the BVP form of GM (Ferrari et al. 2010). Has that routine ever been successfully adjointed? If not, how would one go about instrumenting the routine so that it could ingested by an adjoint compiler?

There was a question similar to this one posted in 2010, but I couldn't find a response in the mailing list archive.

Thanks in advance,
Christopher

-----------------------------------------------------------
Dr. Christopher L. P. Wolfe
Research Oceanographer
Climate, Atmospheric Science, and Physical Oceanography
Scripps Institution of Oceanography, UCSD
clwolfe at ucsd.edu                           858-534-4560
-----------------------------------------------------------






-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20120430/b78436b5/attachment.htm>