Bug #400

Too low cloud effective radius with Martin parameterization.

Added by Jason Williams over 6 years ago. Updated over 6 years ago.

Target version:
Start date:
Due date:
% Done:


Estimated time:
24.00 h


When the online photolysis was coded the parameterization of McFarlane et al. (1992) was used and gave a wide range of potential r(eff) between 4-12um (See Figures 1a and 1b in the supplementary material in Williams et al. 2012). This was replaced in the code by the approach of Martin et al (1994) as in the C-IFS. Analysing the output in j-statistics shows that the new range is way too small (4-6um) i.e. introducing too much scattering at the top of the boundary layer!!
There seems to be a bug in the unit conversion for this as implemented in TM5. Should give a range between 4-16um.

This needs to be fixed as a matter of urgency considering it now is adopted across versions!

In future maybe tests of what comes out of any new update should be made rather than uploading untested code on the SVN.


#1 Updated by Twan van Noije over 6 years ago

I tried to figure out what is going wrong is the code. As usual, comments are missing, but with help of the IFS documentation, I believe I understand what the idea here is.

The basis equation is Eq. (11) from Martin et al. (JAS, 1994).

Here L is the mass of liquid water per unit volume of air, rho_w is the density of liquid water, and the unitless coefficient k is defined by Eq. (10).

k is expressed as a function of a size distribution parameter, using Eq. (23), which leads to Eq. (24).

The problem with the current code, is that the units seem to be wrong.

Ntot is the droplet number concentration in units cm^-3 and lwc in TM5 is given in kg water/kg air.

Compared to Eq. (24) from Martin et al., the ratio numerator/denom misses a factor rho_air/rho_w.

Including that factor, this ratio to the power 1/3 will give the effective radius in cm. Multiplication by 10^4 will give it in um. Currently, a factor 100 is used as a prefactor for the effective radius.

In short, an additional factor (10^4/100)*(rho_air/rho_w)^(1/3) needs to be included to obtain the effective droplet radius in um.

#2 Updated by Twan van Noije over 6 years ago

In the same subroutine in photolysis.F90, the calculation of the ice water concentration ciwc is also incorrect. It should be multiplied by a factor xmair to get it in g/m3.

#3 Updated by Twan van Noije over 6 years ago

I see two ways in which the calculation of the cloud droplet radius can be further improved:

1. Calculation of the cloud droplet number concentration (CDNC=Ntot) based on M7.
In the current code, Ntot is calculated from A, the aerosol number concentration in a certain size range (wet radius between 0.05 and 1.5 um) based on Martin et al. (1994). Instead, we could use A from M7.

An alternative would be to use the parameterization from Menon et al. (JAS, 2002), which calculates Ntot from the mass concentrations of sulfate, organic matter and sea salt. This scheme will possibly also be used in EC-Earth to calculate Ntot from the concentrations provided by TM5.

2. In the heterogeneous chemistry calculation, another definition of the mean radius is used. Currently, it is calculated in a rather arbitrary way by reducing the effective radius according to some table. It would make more sense to use the same assumptions regarding the size distribution as made in the photolysis routine, where a fixed value for the spectral dispersion d is assumed over land and sea, which is then used to relate the effective radius and the mean volume radius (through the dimensionless coefficient k).

#4 Updated by Jason Williams over 6 years ago

The global_eff_radius(i,j,l) array was given a value of 6.0 during initialization at every time-step which has been changed to 0.0 (no clouds therefore r_eff=0).

In the IFS formulation of the Martin parameterization, default CCN values of 40. and 900. were adopted for ocean and land cells, with an arbitary limit of 50% land cover for switching between CCN values. At the resolution of IFS this will not introduce the crude distribution observed at 3 x 2, the resolution adopted for EC_EARTH. Therefore a weighting component is introduced using the explicit land fraction values in each grid cell during the derivation of the grid cell r_eff. This will introduce more variability in the r_eff values per timestep.

#5 Updated by Twan van Noije over 6 years ago

In the photolysis routine, the extinction for ice particles is calculated based on Eq. (3.9a) from Fu (J. Climate, 1996), using coefficients valid for wavelengths between 330 and 360 nm. The corresponding formulas for the single-scattering albedo and the asymmetry factor are not used in the TM5 code. These are set to the values obtained for liquid water clouds, which are calculated based on Slingo (JAS, 1989) for wavelengths 250-690 nm. Moreover, the effect of the ice clouds on photolysis rates is not included if the liquid water content is close to zero.

Just wondering, why don't we calculate the ice optical properties and do a proper averaging?

#6 Updated by Jason Williams over 6 years ago

  • % Done changed from 0 to 100

#7 Updated by Jason Williams over 6 years ago

  • Status changed from New to Closed

Also available in: Atom PDF