Hello everyone, at AWI we were having yet another look at the SST biases of the coupled climate simulations made with OpenIFS 43r3 - FESOM2, but also at ECE3(IFS36r4 - NEMO3.6), and the ECMWF IFS43r1 - NEMO3.6 HighresMIP contribution. All of these models show the well known warm biases over the SO among other places. First we calculated the Spearman's rank correlation coefficient of the SST bias and the magnitude of the wind stress, which turned out to be significant and positive, motivating further digging.



ECMWF IFS43r1 - NEMO3.6 HighresMIP SW cloud radiative forcing bias compared to OpenIFS 43r3 - FESOM2 ocean surface wind-stress:


In the higher latitudes (>30°) there seems to be a positive correlation, and in the lower latitudes a negative one. This is supported by the literature on the effect of surface roughness on surface albedo with varying latitudes. At low latitudes the rough surface decreases the solar incidence angle, leading to more reflection and higher albedo. At high latitudes the solar solar incidence angle is increased leading to lower albedo, as illustrated by the following graphic from [1]. This is underpinned by observed and computed values from [2].



We were then wondering if scaling of the ocean surface albedo with the wind speed might be too strong in the model. we had a look through the code and found a reference in surfrad_ctl_mod.F90 that such a scaling was added some 25 years ago:

PJANSSEN/JJMORCRETTE ECMWF 96/11/07 WIND DEPENDENT SEA ALBEDO

But the actual calculation in line 418:

ZAPTI1=MAX(0.037_JPRB/(1.1_JPRB*PMU0(JL)**1.4_JPRB+0.15_JPRB),REPALB)

does not include the wind intensity PWND. Only the cosine of the incidence angle PMU0 is considered. Was the formula added in 1996 moved elsewhere in the code? If so, where too?  Or has it been dropped entirely? That would be bad news I guess, not only because the effect would be missing, but also because including it again would make things even worse in terms of radiations biases.


We also found a second body of literature that is concerned with breaking waves rather than non-breaking ones. These waves produce whitecaps that increase the direct and diffuse albedo in regions where storm systems move through frequently [3]. We are thinking about adopting such a scheme, assuming it is not yet part of OpenIFS.


Literature:

 [1] https://doi.org/10.1029/JC090iC04p07313

 [2] https://doi.org/10.1175/1520-0469(1972)029<0959:AOTSS>2.0.CO;2

 [3] https://doi.org/10.5194/gmd-11-321-2018







2 Comments

  1. Unknown User (nagc)

    Hi Jan,

    I contacted someone in RD about this but I've learnt you've already been in touch with various people.  I think it would be useful to summarise that email chain here?

    Glenn

    1. Jan Streffing

      Hello Glenn, yes I will copy the email communication here:


      Robin Hogan:

      Hi all,
      I'm pretty sure that white-capping is not included anywhere else in the IFS, so that must have been removed at some point. The earliest cycle where I can find surfrad_ctl_mod.F90 is 30r1 (from 2003?) which has the Taylor parameterization for solar zenith angle dependence but no white-capping (as now).
      It should be easy to add a white-capping parameterization to surfrad_ctl_mod.F90 since you have the wind speed in the PWND variable (if this is the lowest model level then I guess it corresponds to a 0-20m layer so should be quite close to the 10-m wind speed on which many parameterizations are based). We would obviously want to keep the solar zenith angle dependence.
      A couple of further points:
       - You'll notice that the ocean albedo has no wavelength dependence, which is wrong - it is darker in the near-IR, and this ought to be added at some point.
       - Much of the SST warm bias in the Southern Ocean in 43r1/43r3 can be attributed to the clouds being too optically thin (not enough supercooled water), a problem that was substantially improved in 45r1.  So I doubt that ocean albedo is the main culprit.
      Regards,
      Robin.


      Jan Streffing:

      Hello Robin,

      thank you for your input. I think the whitecaps represent only a few per mill change in the net shortwave radiation, even in particularly windy regions. I would not have expected this to be included in IFS 43, since the effect is small and research in this field is quite recent. The other wind based effect, the modification of the solar zenith angle by surface roughness, seems to be larger. It is this effect that I presume was added in the 1996, since the research in this field was mostly done in the 70s and 80s. It is this effect that we were looking for, but did not find.

      We have already backported the IFS cy45r1 changes to the liquid supercooled water content into OpenIFS cy43r3, and indeed they gave us a substantial improvement of the Southern Ocean surface shortwave radiation and sea surface temperature bias from a zonal mean +3 down to +2°. We convened with Richard Forbes and he confirmed that we have gotten everything we can out of his supercooled water changes. The region is however still substantially too warm compared to observations.



      The simulations on the left was made with IFS 43r1 + NEMO3.6+LIM2 (TCO199+0.25°ocean) as part of https://doi.org/10.5194/gmd-2018-90, the run on the right is our latest tuning run with OpenIFS 43r3 + FESOM2 (TCO159+1°ocean).

      I take note of the point about using broadband instead of spectral ocean albedo. I think in contrast to the wind dependencies the spectral albedo might be a bit more challenging of a task and probably not something we would explore soon.

      Best regards, Jan



      Jean Bidlot:


      Hi,

      Thank you for including me in these emails.

      I have not had a chance to look at the sea state effects on the ocean albedo, but I am aware of Peter Janssen and Jean-Jacques Morcrette work in the mid 1990’s (ECMWF Tech memo 228)

      I would have thought that the whitecap effect on the albedo would have been present, albeit with a simple wind speed parameterisation. There are indeed a few around, some quire fancy, even including SST. As an ocean wave person, I have to say there is also a way to obtain the whitecap coverage from the wave model. It is an area of research obviously. But it could be tried.

      As for the sea state effect on the ocean roughness modifying the solar zenith, again it could be good to re-instate this again and test, especially in line with the recent work I have done with Peter Janssen on adding the impact of short gravity capillary waves in the wave model. We are now getting good mean square slope estimates, which could be used in the parameterisation of Peter and Jean-Jacques.

      So, Robin, I might need to be added to the albdo revision working group…

      Jean


      Sea water albedo might indeed be revisited at some point in the future, but we concluded that it is not high on our priority list at the moment.


      Cheers, Jan