Problem in radiation physics

[mirin]

I'm running 1870-control using FV 1x1.25 with ocean gx1v3. After 32 years I encounter NaN in trcems.F90, in computation of emissivity for CH4 band (ech4). Input to this calculation are CH4 path length (uch4), pressure factor for CH4 (bch4), emissivity Planck factor (emplnk(3,:)) and pressure squared h2o path length (up2). I suspect problem is in func(uch4(i,k),betac)), which is affected by uch4 and bch4. Can somebody suggest reasonable ranges of the various inputs, so that I could add diagnostic tests? Unfortunately I cannot use interactive debugger due to 5-executables. Thanks.

 

[mirin]

It turns out that the N2O mass mixing ratio goes to zero at the top level (k=1). I suppose it most likely goes negative and then gets reset to zero in qneg3. The problem is that the radiation physics requires a positive value. The real problem is why it goes to zero in the first place. Would it work to force a small positive value, or is that postponing the inevitable?