Fix tracer types for correct answer and faster throughput

[tangq]

Some explicit tracers were wrongly defined as "implicit" type in the mechanism file. This causes 1) wrong ansowers for those should be solved with the explicit solver; 2) slow throughput by adding unnecessary tracers in the implicit array. This commit fixes these issues by defining the following tracers as the explicity type.

We will need to propagate similar changes to other NGD mechanism files later.

• DMS, SO2, H2SO4 Fix MAM5 stratospheric sulfate solutions

• CO, C2H6, C3H8, CH3COCH3 These tracers have slow reaction rates, so solve them explicitly

• MAM tracers (i.e., *_a[1-5], expect SOA ones) Have no reactions. They are in the mechanism file only for the purpose of creating them.

• SOA tracers to be discussed

[BFB] for F20TR_chemUCI-Linozv3

[tangq]

@keziming , can you confirm if S-tracers (DMS, SO2, H2SO4) are defined as implicit, will they be solved correctly for MAM5 stratospheric aerosols?

@mkshriva , I haven't changed the SOA tracers. They appear to be two types: 1) soa_a[1-3] and 2) SOAG*. I am not sure how fast these SOA reactions are. If they are slow (compared to the 30 minutes timestep), they probably can be solved with the explicit scheme, which will be faster. What are the lifetimes of these species?

[mkshriva]

@tangq SOAG gases have a reaction lifetime of ~2 hours. But these vary with OH concentrations. The reaction timescales for SOAG might approach model timestep of 30 minutes in upper troposphere where OH is higher than surface. One could try moving them to explicit and test the results with current benchmark run before making a decision.

Particle phase soa_a* react by photolysis as you can see..

[hwangacme]

@mkshriva Were SOA and gas precursors always included for implicit solver in all previous runs?

[wlin7]

Merged to NGD_v3atm.

[wlin7]

Note: thought to be an NBFB PR. But the test results with compset F20TR_chemUCI-Linozv3 are BFB.