Hi,
I was looking into cppdefs.h and I found this:
#undef M2CLM_NUDGING /* Nudging of 2D momentum climatology */
#undef M3CLM_NUDGING /* Nudging of 3D momentum climatology */
#undef TCLM_NUDGING /* Nudging of tracer climatology */
#undef ZCLM_NUDGING /* Nudging of SSH climatology */
#undef NUDGING_SSH /* nudging of SSH observations */
#undef NUDGING_SST /* nudging of SST observations */
#undef NUDGING_T /* nudging of tracer observations */
#undef NUDGING_UVsur /* nudging of surface current observations */
#undef NUDGING_UV /* nudging of horizontal currents observations */
So there are two kinds of Nudging. Can anyone please tell me what is the actual use of Nudging and what is the difference between the two kinds of Nudging mentioned above
Thanks.
Goutam
Understanding Nudging
Some are nudging to a climatology, some are nudging to observations. The difference would be whether it is pointwise information you have or a full 2D or 3D field of values to nudge to.
The difference is also in the purpose - the climatology nudging is usually to keep things from drifting too badly, or keeping boundaries well-behaved. Nudging to observations is more like a primitive data assimilation scheme.
The difference is also in the purpose - the climatology nudging is usually to keep things from drifting too badly, or keeping boundaries well-behaved. Nudging to observations is more like a primitive data assimilation scheme.
on the dependence of nudging coefficients
Hi all,
I have a run forced by monthly wind values for a specific year. The domain has 4 open boundaries. Initially, I was using 3 days for "inflow nudging" (the model is restored to Levitus climatology). However, after 1.5 years of run (started from rest) it blows up. After sanity-checking, there was nothing else for me to do except to change the nudging values (I changed it within the range 1 to 2.5 days). It worked, that is, the model kept running until it blew up again when I, once more, changed the nudging coeffients.
However, this unstable behavior made me wonder about the dependence of a run on nudging values:
1) Is there an "optimal" nudging coefficient for each region? If so, is there a "method" to find it other than trial and error?
2 ) In the litterature, I found that a "couple of days" would be reasonable. So, if I stay within this range (1 to 2.5 days) are my results 'robust' (are the values obtained with a given value comparable to another?)
3) Should I consider this unstable behavior "normal"?
Thanks
Paulo
I have a run forced by monthly wind values for a specific year. The domain has 4 open boundaries. Initially, I was using 3 days for "inflow nudging" (the model is restored to Levitus climatology). However, after 1.5 years of run (started from rest) it blows up. After sanity-checking, there was nothing else for me to do except to change the nudging values (I changed it within the range 1 to 2.5 days). It worked, that is, the model kept running until it blew up again when I, once more, changed the nudging coeffients.
However, this unstable behavior made me wonder about the dependence of a run on nudging values:
1) Is there an "optimal" nudging coefficient for each region? If so, is there a "method" to find it other than trial and error?
2 ) In the litterature, I found that a "couple of days" would be reasonable. So, if I stay within this range (1 to 2.5 days) are my results 'robust' (are the values obtained with a given value comparable to another?)
3) Should I consider this unstable behavior "normal"?
Thanks
Paulo
That sounds incredibly annoying to me. In my experience, we had trouble with these boundary conditions until we moved from seasonal/monthly Levitus boundary information to something more frequent from a larger model domain. There are more and more options all the time for these larger models - what is your domain of interest and is there some particular year or years you'd like to model? Our latest Northeast Pacific used boundary conditions from a 47-year global POP simulation.
Thanks Kate. and yes, It is annoying....
when you say "more frequent" you mean like daily/weekly climatology? In this case, wouldn't the model be so strongly constrained to the boundaries (my domain is a 20x20 degree grid at 1/12 degree resolution) that it wouldn't be free enough to "develop" its own dynamics?
Anyway, I'll use a so-called eddy-permitting climatology and compare the differences.
thanks again,
Paulo
when you say "more frequent" you mean like daily/weekly climatology? In this case, wouldn't the model be so strongly constrained to the boundaries (my domain is a 20x20 degree grid at 1/12 degree resolution) that it wouldn't be free enough to "develop" its own dynamics?
Anyway, I'll use a so-called eddy-permitting climatology and compare the differences.
thanks again,
Paulo
More frequent is every four days. We are doing a hindcast for specific years, not exactly trying to generate a climatology. Have you looked at the trouble you are getting? Do you get eddies in the corners that won't go out or what? Do you have increased horizontal viscosity/diffusivity around the edges?
what I note is that there's a tendency for things to get "squeezed" at the western boundary of the domain which would suggest that they do not propagate properly (and yes, I cheked everything and all boundaries are open) or that, as the variables leave the domain, they take too long to "nudge" to the climatogy (?). Btw, I haven't seen eddies that won't go out in the corners.
The increased viscosity is done with a sponge layer around the edges.
The increased viscosity is done with a sponge layer around the edges.