aerosol: give radmod the particle radius the namelist asked for

[Sohex]

apart, the aerosol particle radius, is declared twice -- in aeromod and in radmod, both defaulting to 50 nm -- and only aeromod's is in a namelist. aero_ini reads aero_nl into aeromod's copy and use-associates l_aerorad and aerofile from radmod, but not apart, so radmod's copy keeps its compiled default however the run is configured.

The two halves of the shortwave optical depth then disagree about the particle. aerocore calls mmr2n(mmr, apart, rhop, ...) with aeromod's value, so the number density nrho is right for the particle the run asked for. radmod's aeroprof then computes

daerod = nrho * PI * apart**2 * qex1 * zdz

with radmod's value. For fixed mass nrho goes as 1/r³ and the optical depth as nrho·r², so a consistent calculation goes as 1/r; with the copies disagreeing the model computes nrho(r_true) * r_radmod², and

τ_computed / τ_intended = (50e-9 / r_true)²

Both defaults are 50 nm, so nothing happens until a run sets apart — but a run that sets it is exactly a run that meant to. At 500 nm the radiation sees 1% of the intended optical depth; at 1 µm, 0.25%.

Why it is invisible

radmod also declares namelist/aero_nl/ including its own apart, and nothing in radmod ever opens or reads that namelist — aero_ini is the only reader and it reads aeromod's group. So radmod looks wired up and is not. That declaration is removed here, because leaving it is what makes the defect impossible to see by reading.

The fix

Two lines of mechanism: aero_ini copies its namelist value into radmod's, and radini broadcasts it beside the l_aerorad broadcast already there. The two variables stay separate because radmod cannot use aeromod — aeromod already uses radmod, and make_plasim compiles it second.

Compiles at T21/8. No behaviour change for any run that leaves apart at its default.