[MITgcm-support] Negative concentrations in pTracer

[Roland Young]

Hi,

I'm using a number of atmospheric tracers carried around by the pTracers 
package. These typically begin as a thin layer over a handful of model 
levels, with the other levels set to zero. I have found that when 
tracers are advected vertically (tracer diffusion coefficients are zero) 
the concentration in the cells where initially there is no tracer can go 
negative. The effect is small but noticeable - typically 7-8 orders of 
magnitude smaller than the initial concentration. For example, 
schematically:

Initial condition

space
0.0d0
0.0d0
0.0d0
1.0d0
1.0d0
1.0d0
1.0d0
0.0d0
0.0d0
0.0d0
ground

After 1 timestep

space
  0.0d0
  0.0d0
-2.0d8
  1.0d0
  1.0d0
  1.0d0
  1.0d0
  3.0d-8
  0.0d0
  0.0d0
ground

After 2 timesteps

space
  0.0d0
  3.0d-15
-1.0d8
  1.0d0
  1.0d0
  1.0d0
  1.0d0
  2.0d-8
-4.0d-16
  0.0d0
ground

I was wondering why it's going negative and whether there is an option 
that will keep these values positive? I'm currently using 
pTRACERS_advScheme=77 as that's what is used by most of the verification 
examples, with mom_vecinv for momentum. I tried pTRACERS_advScheme=33 
but it made little difference. Also PTRACERS_ImplVertAdv=.FALSE.; with 
it set to TRUE it is worse (in that case the negative concentration can 
be of comparable magnitude to the initial postive one).

I'm wondering why values in pTracers can go negative at all - surely 
this is unphysical?

Best wishes,

Roland