Reduce NorESM3 hydrological cycle by reducing drag over ocean

[mvertens]

One option to try to reduce the hydrological cycle would be to explore reducing the drag over ocean.

One option is to model what is done in the NOAA UFS via it's implementation in CMEPS. The NOAA UFS is doing this as an option (redrag) in the University of Arizona scheme. They do not modify the drag but rather the momentum roughness length over ocean.
The UFS CMEPS implementation (in $SRCROOT/components/cmeps/ufs) functions as a CCPP host model and the physics
is in the repository https://github.com/NCAR/ccpp-physics in the routine physics/SFC_Layer/UFS/sfc_diff.f

!! This file contains the surface roughness length formulation based on
!! the surface sublayer scheme in
!! Zeng and Dickinson (1998) \cite zeng_and_dickinson_1998.

and the implementation is as follows for the roughness length over ocean (the following are the key statements in the iteration loop):

              z0s_max=.317e-2_kp 
              .............................
              z0 = 0.01_kp * z0rl_wat(i)
               ............................
              if (redrag) then
                z0rl_wat(i) = 100.0_kp * max(min(z0,z0s_max),1.0e-7_kp)
              else
                z0rl_wat(i) = 100.0_kp * max(min(z0,0.1_kp), 1.e-7_kp)
              endif

In NorESM and CESM, the Large/Yeager, Large-diurnal, UA and COARE schemes are all in the file
$SRCROOT/components/cmeps/cesm/flux_atmocn/shr_flux_mod.F90.
I am not sure how we want to implement this type of reduction in COARE - since the scheme is considerably different. It might be good to get Thomas Toniazzo involved for advice.

I think it would be good at this point to also rewrite shr_flux_mod.F90 to actually be 4 separate modules - for Large-Yeager, Large-diurnal, COARE and UA.

@MichaelSchulzMETNO @adagj - @oyvindseland and I met with Thomas Toniazzo and have come up with the following plan.

• I will enable daily output of the roughness length and surface stress for a year using the latest simulation configuration.
• Once we have that - we can determine how we want to constrain the roughness length in the COARE scheme.

[adagj]

@JensBDebernard Can you please provide a description of what you suggested (this morning) to test?

[JensBDebernard]

My first thought was that it should be possible to use the 'Charnock' parameter for tuning the ice-ocean stress. Charnock (1955) suggested relation between surface stress and the roughness parameter that involves this parameter, with a typical value between 0.01 and 0.02. Almost all (the relevant for us) air-sea drag formulations use this in a way (except the Large & Pond / Large and Yeager) formulation).

However, looking into the parameterizations, and their development, there has been a shift to 1) larger drag in the 'intermediate' wind regime (10 - 25 m/s), but also a limitation of the roughness for very strong wind (hurricane strength). The latter has been necessary to allow the atmosphere model to develop these very strong storm conditions.
For neutral conditions, the main difference between the NOAA UFS model, the UA parameterization, and the COARE30 and COARE35 schemes are in the way the Charnock 'parameter' is set, or modeled. The UFS and UA uses fixed values (0.018 and 0.013, respectively), while COARE30 and COARE35 increases the value from 0.011 at 10 m/s to 0.018 at 18 m/s for the COARE30, and 0.027 at 19 m/s for COARE35. In addition, to allow for hurricane wind, the UFS impose an upper limit at the roughness length.
In principle, it is not difficult to impose a similar upper limit on the roughness in the COARE and UA schemes. We should, however, be aware that the rest of the flux package (heat and moisture), are somewhat tuned towards the specific neutral wind drag model.
We should also recognize that the COARE and UA schemes are not validated with many 'high' wind situations, and they are typically considered valid up to 20-25 m/s.

In the figure below, I have plotted z_0, and the neutral 10-meter drag coefficient Cd_10N for some of the schemes. I have also included the Pond and Yeager 2009 scheme, and a parameterization from Andreas et al 2012. The Pond and Yeager scheme stabilizes at fixed values for velocities above 33 m/s, and have lower drag for high winds than many of the others, but much higher drag coefficients for the low wind regime. The Andreas scheme is special because it is developed and validated with a lot of high wind data (up to 60 m/s). It flattens of to constant values (but higher that the UFS and P&Y schemes).

If the goal is to reduce the latent heat flux by reducing drag, I would think that it is more important to reduce the drag for intermediate winds (Trade wind region), than for the peak situations. The UA scheme has naturally somewhat lower drag than the COARE schemes due to a lower Charnock parameter.
How important are high wind events in the model, do we ever get up to strengths where a 'clipping' of the roughness really matters?

[MichaelSchulzMETNO]

Nice !

[oyvindseland]

Interesting. At the moment Mariana, Thomas and myself had put this work aside due to the complexity compared to possible gain.
At least part of the complexity is that the atmospheric calculations of moisture and momentum convergence is not consistent with the calculations in the coupler.

Then comes my general question. If you reduce the drag, can you not end up with a negative feedback as in: Reduced drag --> higher windspeed -> regaining the decrease in flux?

There may actually still be a reason for tuning the parameter and that is transport of heat to the Arctic.

I did a test where I divided z0 by 2 in the atmospheric formulation of convergence.
The conclusions may not be the same for changes in the couple formulations.

z0/2 in the atmosphere --> Very slight increase in windspeed and LHFX. Increase in global wind stress. Reduction in the tropics. I have not calculated whether change LHFLX have the same sign in the tropics and on global scale.

The only place with a strong impact was in the North Atlantic.
Reduction in u0 --> Increase in LHFX (higher positive bias)
Northward shift and strengthening of the North Atlantic jet --> warming of north-eastern Atlantic. (Regional reduction in temperature bias.)
So it looks like there may be a tradeoff between improving the fluxes in the tropical region and heat transport to the Arctic.

Diagnostics
https://ns9560k.web.sigma2.no/datalake/diagnostics/noresm/oyvinds/n1850.ne16_tn14.noresm3_0_beta03_ustar.20251013/

https://ns2345k.web.sigma2.no/datalake/oyvinds/ADF/plots/n1850.ne16_tn14.noresm3_0_beta03_ustar.20251013_1_10_vs_n1850.ne16_tn14.noresm3_0_beta03_rafsipmasstonum.20250930_1_10/website/index.html

[JensBDebernard]

When it comes to the question about the fluxes. As long as we use a consistent set of parameterizations, a reduced drag will also reduce the magnitude of latent and sensible heat fluxes, given the same temperature/humidity difference across the boundary. However, there are negative feedback such that the reduced drag roughness, gives a larger wind speed, which partly compensates for the reduction in the coefficient. Similar, a reduction in the exchange coefficients for heat and humidity will partly be compensated by larger gradients in heat and humidity between the surface and air. And, in principle, if all energy dissipation needed by the atmosphere is going through the surface stress, the atmosphere should be able to fully account for this. However, there are other processes dissipating energy also, so that should not happen.

As you say, it is very confusing with the calculations of fluxes both in the coupler, and in CAM, and I have no overview of how the inconsistency is handled. It might be a good thing to wait with a drag reduction all the time the C_0 changes seems to give a larger reduction in hydrological cycle over the ocean.

If we want to do tests with reduced drag, the figure above show that the UA scheme should be the easiest options.

Where in CAM did you try the reduction?

[oyvindseland]

The reduction was done in clubb_intr.90, specifically in the equation that calculates ustar
So the only connection was through ustar, hence the name of the experiment.