How to safely separate aerosol direct and indirect effects in coupled runs (BHIST, CESM2.1.3)?

[Lindsay_TJ]

Hi everyone,

We are working with CESM 2.1.3 and trying to separate the specific climate responses (e.g., changes in surface temperature, precipitation, and circulation) to the anthropogenic aerosol direct radiative effect (ARI) and indirect radiative effect (ACI). Now，we have completed a BHIST base run and a BHIST-NoAnthro run. I now want to separate the anthropogenic aerosol direct and indirect radiative effects.

To isolate the direct and indirect effects, my initial thought was to run two additional sensitivity experiments based on the ones above, but with the aerosol indirect effects turned off.
I found that I could fix the ice and cloud droplet number concentrations in user_nl_cam by setting:
micro_mg_nicons = .true.
micro_mg_nccons = .true.

However, since I am running fully coupled B compsets, I am highly concerned that abruptly modifying these microphysics parameters to constants will severely break the TOA energy balance. This would likely lead to a massive climatological drift (e.g., severe SST drift) and ruin the baseline mean state.

Is there a safe and standard experimental design in CESM2 (CAM6) to isolate the climate responses of aerosol direct and indirect effects without breaking the coupled energy balance?
Instead of using a hard-coded constant (nccons = .true.), is it possible to prescribe 3D monthly climatologies of cloud droplet and ice number concentrations into the microphysics scheme? If so, which namelist variables or code modifications are required?


Any advice on the best practices for this experimental design would be deeply appreciated!

Thank you,

 

[brianpm]

I agree that setting the ice/liquid numbers to constant in a B-case will cause significant drift.

I'm not sure that there's a totally satisfactory way to separate ARI and ACI in a coupled setting without having some caveats. Your suggested direction is close to what I was thinking, but since much of the microphysics calculation is determining the ice and liquid number (and mass), I don't think that is where you should inject prescribed fields. One potential method would be to follow the "cloud-locking" method that we've used previously. In this approach, you would have **instantaneous** 3D cloud fields provided to (**only**) the radiation scheme. These fields would need to come from one of your previous runs (and they'd need to be re-run to get the high frequency cloud fields). This fixes the cloud radiative effect, which removes the cloud-radiative feedbacks while preventing the model's climate from drifting. If your total aerosol effect is (BHIST - BHIST-NoAnthro), then I think you could think about it in terms of three simulations:

BHIST-NoAnthro - pre-industrial aerosol
BHIST - present day aerosol
BHIST-NALock - present day aerosol, but clouds locked from BHIST-NoAnthro

So then

ARI = BHIST-NALock - BHIST-NoAnthro
The cloud radiative effect is the same, but the aerosols differ, so the difference is due to the difference in aerosol (direct radiative effects). The caveat is that there will still be cloud-masking effects. (You could have diagnostic "clean sky" radiation calculations to try to isolate that.)

ACI = BHIST - BHIST-NALock
The aerosols are the same (kind of) but the radiation sees "dirty clouds" in BHIST and "clean clouds" in BHIST-NALock, so the difference is the impact of aerosols on the cloud radiative effect. I say kind of because there will be heating differences and the aerosols will interact with the predicted clouds (which in principle are going to be similar between the runs), so I think there will be a residual in the reconstruction of the total effect

TOTAL_AEROSOL = (BHIST-NALock - BHIST-NoAnthro) + (BHIST - BHIST-NALock) ≈ (BHIST - BHIST-NoAnthro)

We haven't done this for aerosol effects, but we've got a series of papers with CESM1 and CESM2 that use the cloud-locking approach:

• E. A. Middlemas, A. C. Clement, B. Medeiros, and B. Kirtman.
  Cloud radiative feedbacks and el ni˜no–southern oscillation. Journal of Climate, 32(15):4661–4680, 2019. doi: 10.1175/JCLI-D-18-0842.1.
• E. A. Middlemas, J. E. Kay, B. M. Medeiros, and E. A. Maroon.
  Quantifying the influence of cloud radiative feedbacks on arctic surface warming using cloud locking in an earth system model.
  Geophysical Research Letters, 47(15):e2020GL089207, 2020. doi: 10.1029/2020GL089207.
• J. J. Benedict, B. Medeiros, A. C. Clement, and J. G. Olson.
  Investigating the role of cloud-radiation interactions in subseasonal tropical disturbances.
  Geophysical Research Letters, 47(n/a):e2019GL086817, 2020. doi: 10.1029/2019GL086817.
  e2019GL086817 2019GL086817.
• Investigating the influence of cloud radiative effects on the extratropical storm tracks
  K. M. Grise, B. Medeiros, J. J. Benedict, and J. G. Olson
  Geophysical Research Letters, 46(13):7700-7707, 2019
  DOI: 10.1029/2019GL083542
• Investigating the impact of cloud-radiative feedbacks on tropical precipitation extremes
  B. Medeiros, A. C. Clement, J. J. Benedict, and B. Zhang
  npj Climate and Atmospheric Science, 4(1):18, 2021
  DOI: 10.1038/s41612-021-00174-x

(Note -- there are also examples of cloud-locking with other models, and with CAM4 ... see papers by Paulo Ceppi and Aiko Voigt for example, and some alternative versions are presented by Harrop et al. 2024, DOI: 10.5194/gmd-17-3111-2024)

The implementation isn't super easy. We've done it by adding a module the follows other prescribed tracers (e.g. ozone), with some customized output fields for RUN-1 and some code to inject them into the radiation fields for RUN-2. We have code mods that we can share with you (email me).

Another potentially useful approach is diagnostic, but relies on having the MODIS simulator outputs and using radiative kernels. And it is limited to shortwave effects right now. See:
A new method for diagnosing effective radiative forcing from aerosol--cloud interactions in climate models
B. M. Duran, C. J. Wall, N. J. Lutsko, T. Michibata, P.-L. Ma, Y. Qin, M. L. Duffy, B. Medeiros, and M. Debolskiy
Atmospheric Chemistry and Physics, 25(4):2123–2146, 2025
DOI: 10.5194/acp-25-2123-2025


If you devise another method, please let us know!