Adding an SST perturbation

[a_sengupta@unsw_edu_au]

Hi,
I'm trying to add a SST perturbation to the ocean (running in the fully coupled configuration). I'd like the perturbation to build over the period of a day say and then be left to see how the system evolves.
I have tried adding a flux term in the tracer_update module, something like

if ((iday.eq.8).and.(k.eq.1)) then
FT(:,:,1) = FT(:,:,1) + Tpert(:,:)
endif

where Tpert contains the spatially varying perturbation that leads to a ~0.3C anomally over a day.
I have compared this against a control run. The code produces the correct perturbation but after a few extra days/weeks noise builds up over the entire global domain that swamps the original perturbation.
I have tried reducing the size of the perturbation (by several orders of magnitude) but the magnitude of the noise is essentially unchanged. If I set Tpert(:,:)=0 then the control and perturbation runs match each other (so the problem is not elsewhere).

Can anyone suggest what the problem might be. It seems particularly odd that the magnitude of the perturbation has no bearing on the size of the numerical noise. OR can anyone suggest a better way of implementing such a perturbation.

Out of ideas on this end - so would be very grateful for any ideas.
Many thanks
Alex

 

[olson]

Alex,

You haven't stumbled upon a "problem" but rather a basic feature of solving the
Navier-Stokes equations, namely the addition of any non-zero perturbation to a control
simulation will forever alter the outcome of the series of deterministic forecasts
that proceed from the perturbation ("butterfly flapping its wings"). Within a few
weeks of the initial perturbation, the simulation looks nothing like its control
counterpart except in a statistical climate sense (the perturbed and control *climates*
should still look the same but the "forecast" details will look competely different).

It is possible to study perturbations in highly "idealized" scenarios but the setup
of the scenario would depend upon precisely what you are studying and I don't have
the background in that area to answer further. Hopefully others can step in to help.