UFS_UTILS
UFS_UTILS copied to clipboard
ufs-community
fill negative tracers in chgres
Some (All?) old output of FV3GFS has some negative tracer values. These values get interpolated, but not filled, when using chgres and are read into the model as initial conditions. The input of negative tracers can perpetuate the problem of negative tracers and so on and so forth. We need to get rid of them. Clipping is generally considered a conservation no-go, but I'm much more worried about it in the forecast model than I am in preparing initial conditions. I'm certainly open to other methods if somebody has another idea.
@KevinViner-NOAA How is this impacting your work? Could we try clipping as a first solution?
@GeorgeGayno-NOAA I noticed the negatives in some RRFS cold starts I was running. Right now, given sharp gradients in tracer fields, the vertical remapping in the model can produce negatives. Knowing that this is the case, there is code implemented to try to fill the negatives with an elaborate (but conservative) borrowing scheme. However, sometimes this still fails to fill the negatives. Perhaps my issue is premature, but my thinking is that we don't want the initial conditions to be introducing negatives, especially in a (hopefully) near future where the model doesn't create them. I think clipping would be a great start.
@GeorgeGayno-NOAA I noticed the negatives in some RRFS cold starts I was running. Right now, given sharp gradients in tracer fields, the vertical remapping in the model can produce negatives. Knowing that this is the case, there is code implemented to try to fill the negatives with an elaborate (but conservative) borrowing scheme. However, sometimes this still fails to fill the negatives. Perhaps my issue is premature, but my thinking is that we don't want the initial conditions to be introducing negatives, especially in a (hopefully) near future where the model doesn't create them. I think clipping would be a great start.
Do you have an example case I can look at (i.e., the chgres script/config file used, and the input data)?
@KevinViner-NOAA - there has been no recent activity on this issue. Do you still want it worked? If not, I will close.