[MITgcm-support] more KPP blues, maybe related?

[Martin Losch]

Hi,

another question about KPP (Patrick, I think I asked you something 
similar half a year ago, already, but we didn't get anywhere):
when I run global_ocean.90x40x15 with kpp everything seems stable, as 
long as I use the same input heatflux (that is ncep_qnet.bin), as in 
the repository, range(-224,441)W/sqm, without any short wave heat flux. 
But  I should use shortwave radiation dsolr (also from NCEP), 
range(-435,0)W/sqm (so only heat GAIN by the ocean). I then should use 
instead of qnet something like qres=qnet-dsolr, right?
But qres then has the range(-4,573)W/sqm, so basically only heat LOSS 
by the ocean. So if I use something like
  surfQFile='ncep_qres.bin',
  surfQswFile='ncep_dsolr.bin',
in the data file I should have the same net heat flux as with only
surfQfile='ncep_qnet.bin',
right? But somehow the model blows up after 20 timesteps (dt=2days), 
because the entire southern ocean is convecting all the way to the 
bottom! (kpp-boundary layer depth equal total water depth)

Do you have an idea what is going on?

Martin