Testing IC4 Wave Attenuation Scheme with Updated Numerics in WW3

[ezhilsabareesh8]

We are currently testing the IC4 empirical wave attenuation scheme with the updated numerics in WW3. The detailed issue discussion can be found here: NOAA-EMC/WW3#1294.

The corresponding WW3 branch is here: ACCESS-NRI/WW3 – IC4_Numerics

The ACCESS-OM3 test branch is here: ACCESS-NRI/ACCESS-OM3 – IC4_Numerics

Related PR: ACCESS-NRI/ACCESS-OM3#152

Related Configuration: IC4_Numerics-dev-MCW_100km_jra_iaf

@NoahDay is currently testing different empirical IC4 wave attenuation schemes using this setup. Results and outputs will be updated in this issue as testing progresses.

Next Steps:

• Run baseline (IC4M2) experiments with the current IC4 setup.

• Test alternative empirical schemes like IC4M10 by modifying namelists_Global.nml.

• Collect outputs and document results here.

[NoahDay]

TLDR: I think we should properly disable IS2.

The first test I've run used IC4M2 (Meylan et al. GRL, 2014) and I attempted to disable IS2 scattering by setting IS2BACKSCAT=0.0 (see config here).

It was apparent immediately that wave heights were being attenuated without sufficient sea ice cover. Below is a figure of areal sea ice concentration, significant wave height, ice floe diameter, and mean wave direction (all output from WW3). The solid magenta line shows the 15% sea ice concentration edge, at which wave heights are already negligible. The point at which the wave attenuation begins corresponds with the PUNY ice edge (1e-12) after which wave heights decrease from around 1 m in height (dashed black line) to 1e-8 (dotted black line) before they reach the 15% ice edge (solid magenta line).

Legend (values and corresponding line styles):

• Ice concentration contours in magenta: [1e-12, 0.15, 0.8] = [":", "-", "--"]
• Significant wave height contours in black: [1e-8, 0.1, 1] = [":", "-", "--"]

This PUNY threshold corresponds to the "mask" which is used for the ice floe diameter (see bottom left panel of the figure). So I'm thinking that the large floe sizes (of ice that isn't really there) are also impacting the attenuation calculation.

In WW3, the ice source terms are calculated when the ice concentration is greater than 0 (e.g., here), meaning that despite there being practically no ice in these cells WW3 still calculates the impact on the wave energy. I will also note that there is a discrepancy between the NOAA code and the ACCESS-NRI fork, where IC_Numerics only impacts IC4 in our fork as can be seen here), and we also have a sub-cycling step (see here) which isn't present in the NOAA version.

Anyway, what I believe to be exacerbating the attenuation is the following step in w3srcemd.F90 where the attenuation term (ATT) is calculated. Since W3_IC4 and W3_IS2 are defined the attenuation term is essentially adding both of the exponents from IC4 (VDIC(IS)) and IS2 (VDIR2(IS)) together, which is further contributed to by setting IS2BACKSCAT = IS2PARS(2) = 0, hence adding VDIR(IS) to the attenuation rate. The big issue here is that attenuation from IS2 increases with increasing floe diameters, hence it is likely adding a lot of attenuation near the ice edge since the floe diameters are so large, yet there is practically no ice these cells.

#ifdef W3_IC4
          ATT=EXP(ICE*VDIC(IS)*DTG)
#endif

#ifdef W3_IS2
          ATT=ATT*EXP(ICE*VDIR2(IS)*DTG)
          IF (IS2PARS(2).EQ.0) THEN ! Reminder : IS2PARS(2) = IS2BACKSCAT
            !
            ! If there is not re-distribution in directions the scattering is just an attenuation
            !
            ATT=ATT*EXP((ICE*VDIR(IS))*DTG)
          END IF
#endif

So for the meantime, I think we should remove IS2 completely from the MCW config. After which I can look into whether we need to tweak the PUNY threshold for sea ice in CICE and the threshold of 0 in WW3 and then we can think about updating our WW3 fork.

[ezhilsabareesh8]

@NoahDay I’ve disabled the floe-size–dependent scattering term IS2 in this commit. This should make it easier to isolate the contribution of a_ice to attenuation, and also gives us a clearer pathway to test IC4M10 (where scattering is already incorporated into the attenuation model).

To use the latest build, update your config.yaml with:

modules:
  load:
    - access-om3/pr152-3

Also note that mod_def.ww3 and restart.ww3 will need to be regenerated by loading the access-om3/pr152-3 module and executing ww3_grid and ww3_strt, and the new filenames updated in config.yaml.

[NoahDay]

After @ezhilsabareesh8 made numerous updates (#771 #841) to the RYF configuration (771-MCW_100km_jra_ryf), we now have some reasonable looking model outputs!

Waves are propagating into the sea ice (from right to left with latitude on x-axis):

For reference, transects before the changes looked like this:

where none of the wave energy ($H_s$) would make it past a sea ice concentration of ~10^-3.