Aerosol Optical Depth (AOD) Advisory for Users of CCM and CRM

The following describes the interpretation and usage of the input
parameter "tauvis", which sets the AOD in the CCM and CRM. 

************************************************************************
Description of problem:
************************************************************************
Standard CCM uses a globally uniform background aerosol (GUBA) with the
optical properties of pure sulfate. As of CCM3.6 and CRM 2.0, the
namelist parameter "tauvis" sets the GUBA burden (but tauvis is not
the optical depth, keep reading). tauvis is obviously related to
aerosol optical depth but tauvis is NOT the aerosol extinction optical
depth.  

The aerosol mass mixing ratio calculated from tauvis in routine
physics/aermmr() includes a factor 1/(1-omega*f) =
1/(1-omega*g*g). For the CCM sulfate aerosol in the visible, omega ~ 1.0
and g ~ 0.7 and this factor equals 1.9337. That means that the mixing
ratio in array aermmr is about twice as large as it would be
if tauvis represented the aerosol extinction optical depth. 

Let tau_ext be the true extinction optical depth (i.e., the optical
depth seen by the unscaled direct beam in radcsw()). Then the
relations between tau_ext and tauvis are tauvis=tau_ext*(1-omega*f)
and tau_ext=tauvis/(1-omega*f). This means tau_ext is 0.27 in the CCM.

************************************************************************
Symptoms:
************************************************************************
It is important to remember this factor of two (1.93) now that we have
reached the stage where prognostic sulfate and dust mixing ratios are
being combined with tauvis to obtain a total optical depth. In effect,
the GUBA one gets by naively setting tauvis has twice the extinction
efficiency as the prognostic sulfate.   

Another area to be careful is the sensitivity of the diffuse beam to
aerosol optical depth. The (unscaled) direct and diffuse components of
solar radiation are additional diagnostic variables provided by the
CRM. The downwelling diffuse flux in the CRM is twice as sensitive to 
aerosol when computed relative to tauvis rather than tau_ext. 
Comparing the CRM diffuse beam to my shortwave narrow band model
diffuse beam (and seeing a factor of two difference) is what
enlightened me as to the true meaning of tauvis.

************************************************************************
Origin of problem:
************************************************************************
The factor (1-omega*f) was originally placed in the CCM to increase
measured extinction optical depths to account for the fact that some
of the scattered photons were present in the direct beam, and thus
uncounted. It turns out that this method of correction makes two
assumptions, both of which are questionable (for the case of
aerosols):   

1. This factor relies on the forward scattering peak being smaller
than the instrument aperture (instrument apertures today are
generally less than 5 degrees full width). Sulfate aerosol is
typically smaller than one micron (area-weighted radius), and scatters
more isotropically than the cloud particles (e.g.,
sub-visible cirrus) for which the (1-omega*f) approximation is widely
(and appropriately) applied. Since the forward scattering peak is
wider for sulfate, fewer of the scattered photons are in the direct
beam. Figure 3.5 on p. 118 of Liou's (1992) book "Radiation and Cloud
Processes in the Atmosphere" (there's a copy in my office) illustrates
this concept. The "forward scattered peak" (where a fraction f=g*g=50%
of the scattered energy resides) for a sulfate aerosol may be many
tens of degrees wide. For example, both the Mie and Henyey-Greenstein
phase functions' "forward beam" for the CCM sulfate aerosol (g=0.7)
are about 50 degrees full width. 

2. This correction assumes that the data on which tauvis (0.14) is
based have not already been corrected for scattering into the direct
beam. Many, if not most, instrument mentors who release measurements
of aerosol optical depth nowadays already account for photons
scattered into the direct beam, when appropriate. Some have used 3D
Monte Carlo models to characterize this effect. 

Neither of these two assumptions should be made lightly for sulfate. 
Certainly none of this invalidates the concept or usefulness of
tauvis, either. The problem I have described is not what most would
call a bug. As currently implemented, tauvis describes real physical
phenomena, and is a useful tuning parameter for TOA and surface
energy budgets, but we need to stop calling it plain old "aerosol
optical depth".

************************************************************************
Courses of action:
************************************************************************
For the present, we have the question of nomenclature: What we call
tauvis in CCM3 is not the aerosol extinction optical depth. It is, as
near as I can tell, the aerosol extinction optical depth that does not
reside in the "forward scattered peak" of the assumed Henyey-Greenstein
phase function.

Since it is a namelist parameter in CCM3.6, tauvis has been made to be
explicitly adjustable from the input file in the new CCM3.6-based
CRM. With all the interest in aerosols and diffuse/direct
partitioning, many user's will probably take advantage of this new
parameter. We should provide them with a clear explanation/rationale
of what tauvis is so they understand what they are doing. This will
lessen the chances that the CRM will be misused in upcoming 1D model 
intercomparison studies. The new CRM prints the aerosol extinction
optical depth as a standard output. 

As far as CCM4 goes, the Climate Modeling Section advocates getting
rid of the (1-omega*f) factor in aermix() so we can change the 
namelist parameter from tauvis to something like tauext 
(whose physical meaning is tau_ext, and whose value is about 0.27).