dynamic viscosity

[ziu1986]

Recently, I have found at least three different definitions of "dynamic viscosity" throughout different parts of the model:

• fallingaerosols.f90 (it cites RoY94 but is in fact the original formulation of Sutherland's Law but with different parameters) L560

  vsc_dyn_atm = 1.72e-5_r8 * ((tem(L)/273._r8)**1.5_r8) * 393._r8/(tem(L) + 120._r8)

• seasalt.f90 (does not cite anything, powerlaw relationship similar to Sutherland's law) L758-775

  visc0air = 1.75e-5_r8, & !// Viscosity of air at temperature T0visc
  T0visc = TK_0C, &        !// Temperature at which visocity=visc0 (K)  
  nviscair = 0.65_r8       !// Power law dependence viscosity on T
  !// Viscocity of air at ambient temperature
  viscair = visc0air * (TEM(L) / T0visc)**nviscair

  Later on in the same code we have it duplicated (L908-940)

  !// VISCOCITY OF AIR
  viscair = visc0air * (TEM / T0visc)**nviscair

• drydeposition_oslo.f90 (citing Prather, PBL using some wired linear fit through origin to describe the powerlaw relation, states "abs.visc.") L1489-1492

  !//   SFMU = 6.2d-8*SFCT    ! abs.visc. 6.2d-8*T (lin fit:-30C to +40C)
  SFNU  = (**6.2e-8_r8 * T2M** * T2M * R_AIR) / (100._r8 * PSFC)

1. There are far too many hard coded parameters and constants in the code... they should be tracked down and put into a "namelist".
2. I would really love to implement a xml-based namelisting and steering...
3. As can be seen from this example, there is much of duplication of code though out the model, there is a tremendous potential of line reduction...

[ziu1986]

Note: Sutherland's law

[ziu1986]

Note: Linear fit used in drydeposition_oslo.f90 dynamic_viscosity.pdf

[ziu1986]

Update: I have tested the range of z0 and the resulting Rb for both parametrizations of the dynamic viscosity. Well, I only ran the model for one entire day, though. The good news first: The difference in Rb is only about 1.4%. Meaning the average Rb using the "Prather"-fit is only about 1.4% smaller than it would have been using Sutherland's law. This is mainly due to the fact that Rb is generally, in about 91% (88%) of all cases, below the threshold of 10 sm^-1 (see dynamic_viscosity_diffs.pdf dynamic_viscosity_zo_Rb.pdf). Calculating the weighted mean for both Rb distributions, while setting all values which appear below the threshold to the threshold value:

<Rb_prather> = 1007.92 scm^-1 (=> vb_prather = 0.00099 cms^-1)
<Rb_sutherland> = 1021.80 scm^-1 (=> vb_sutherland = 0.00098 cms^-1)

Hence, the impact is rather small, but we should switch to a consistent dynamic viscosity through out the model (see above comments).