Differences in NEMO when coupled with OASIS (even with inactive second model)

Versions v4.2.x
,
1

Hi everyone,

I would like to ask a question concerning coupled simulations using OASIS

I am running some tests to compare two simulations using the eORCA025 configuration of NEMO v4.2.1 with XIOS. The only difference between the two runs is that one is uncoupled, while the other is coupled with OASIS.

In principle, both simulations should produce the same results, since the information exchanged from the second model does not affect any NEMO calculations. However, I am observing small difference between the two runs (less than two orders of magnitude smaller than typical values of temperature, salinity, velocity, etc.), especially near the coasts.

I did not specify the second model here, because my main question is whether the mere act of coupling NEMO with OASIS (even if the other component model does nothing) can introduce small differences in the NEMO outputs

Thanks in advance for any experiences you might share!

Marcela

2

I don’t really understand what kind of coupling you are doing… By default, switching on the coupling changes everything!

Using or not OASIS in NEMO requires that you recompile the code and you are not guarantee to get the same results if you use two different executables. Agressive compilations options can explain such differences.
You could try very strict compilations options such as -O0 -fp-model strict
You could also try all the debugging compilations options (see for example arch/CNRS/arch-X64_IRENE_DEBUG.fcm) to see if there is a problem somewhere such as the use of an undefined variable.

3

Hi,

Thank you for your reply.

I understand that my message might have been confusing since I didn’t mention a specific model.

However, your response was very helpful, and I’ll look further into the impact of different compilation options.

Thank you again for the suggestion!

4

Hi Marcela,
It doesn’t matter what model you use, when coupling is activated most basic things like surface forcing is entirely different, it might also change the forcing frequency so minor to major differences are to be expected.

5

Hi!,

Thank you for your reply,

So if I understand, this means that, for example, the surface forcing would also will be difference, even if the second model does not modify this variable or does not require it, right?