UOEXETER-CMIP-1-1-3: Volcanic optical properties and SO2 emissions

[shipengzhang]

Comparison of draft CMIP7 (cmip6plus) with CMIP6 (v2) volcanic forcing

Special thanks to Thomas Aubry for providing the CMIP6plus volcanic datasets.

1. Comparison of Prescribed Stratospheric Aerosol Properties (Extinction Profile at 550 nm)

The CMIP7 extinction dataset has improved significantly over the CMIP6 version. Specifically:

• Under-tropopause values: In CMIP7, extinction values below the tropopause are masked (during the satellite era) or set to zero (pre-satellite era). This is an enhancement over the CMIP6 extinction profile. CMIP6 has unreal values under the tropopause, which were extrapolated from the stratosphere.
• Background Stratospheric Aerosol Optical Depth (SAOD): CMIP6plus appears to have smaller background SAOD compared to CMIP6.
• Major Eruptions: The differences in SAOD during major volcanic eruptions between CMIP6 and CMIP7 vary on a case-by-case basis. However, several eruptions, such as Pinatubo, show larger SAOD in CMIP6plus.

2. Comparison of Volcanic SO₂ Emissions Between VolcanEESM and CMIP6plus

VolcanEESM (Volcanic Emissions for Earth System Models):

CMIP6plus:

Key differences between the VolcanEESM database and the CMIP6plus dataset include:

• Injection height variability: CMIP6plus exhibits greater variability in injection heights, while pre-satellite eruptions (before ~1970) are generally modeled with injections around 17–18 km altitude.
• Eruption locations: Some adjustments have been made to the locations of eruptions in CMIP6plus.
• Inclusion of missing eruptions: CMIP6plus includes eruptions that were missing from earlier versions of the dataset.
• Satellite era improvements: More eruptions are included in the post-1970 satellite era in CMIP6plus.

Request for Enhancement

In the CMIP6plus volcanic SO₂ emission dataset, it would be helpful to:

• Include volcano names for each eruption (currently, eruptions are only labeled by number).
• Include the Volcanic Explosivity Index (VEI) for each eruption to provide additional context about eruption intensity.

[johndunne13]

While I admittedly not an expert in volcano forcing, I am very concerned at the low background for mall volcano in the pre-satellite period in this dataset. The new dataset show much more high frequency small volcano activity than CMIP6 which is a huge advance, but gives the impression that a higher background should be applied in the pre-satellite era. While I appreciate the volcano community aspiration that the small volcano SAOD forcing should be provided as episodic, the high frequency apparent from the satellite period and multi-year timescale of the climate response suggest that lack of representation of this background will lead to spuriously high sensitivity to large volcanos. My recommendation would be to add a pre-satellite small volcano to give the same small volcano magnitude as the satellite period. I see this as particular concern with respect to the potential role of volcanos in contributing to the "1960's pothole" among CMIP6 models, and am concern that the current approach will induce a spurious additional cooling affect without inclusion of a "best guess" of the satellite-derived background small volcano activity in the decades before and after the pothole.

[thomasaubry]

Thanks for your comment John. Brief reply to make sure what we discussed is publicly available:

1. The lack of small eruptions pre-satellite era was very much present in any previous CMIP dataset so this is not a bias new to this dataset, there is simply more awareness about it now.
2. We are also concerned and are trying our best to fix it. The latest version of our dataset (1.2.0, incoming on ESGF) still has this bias but we have been working on a fix which will be deployed in a new version early January.

[vnaik60]

ping thomasaubry!

[tilmes]

@thomasaubry We are planning to compare the provided SO2 eruptive emissions in CESM with chemistry to see how different our aerosol distribution will look compared to our default emissions used for CMIP6. Here are some questions about the provided dataset:

1. The dataset provides boundaries for lon lat, altitude and time, see plot below. For lon and lat, we are wondering what these ranges (errorbars) represent. Are we supposed to consider point injections for each volcano (on the lon/lat provided) or include information of the bounds?
2. For time, we see that the eruptions are always giving for one full day. Previous considerations considered 6-hourly injections. For this dataset, to be consistent, should we consider that the provided injection amount is spread over 24 hours?
3. For altitude, again, there is a range provided. Do we consider point injections at one altitude level or spread the injections equally over the given altitude range? For example, for Mt Pinatubo, the altitude range is much larger than in CMIP6.

Thanks Simone

[vnaik60]

Thanks Simone. Tom is on vacation until the middle of this month. He will be picking up the forcings work after he gets back so will have answers to our questions and updated dataset with bug fixes.
Regarding your question 2, my understanding is that it is left to the modelers to decide how to distribute the injections over the course of the day. Mike chose to emit over the 6 h period between 12:00 UTC and 18:00 UTC on each day of eruption (the Schmidt and Neely dataset also provided emissions for the day). We followed the same option in Chloe's work. I wonder if the same assumption was used in UKESM. I also wonder how much the timing of the eruption in the day matters for climate simulations. Perhaps we (modelers) can coordinate on that.

[thomasaubry]

Thanks Simone! Documentation will come soon (by early Jan 2025) but in the meantime I'm replying here to ensure information is publicly available.
Note to anyone reading this reply: this is only relevant if your model is emission-driven rather than run with prescribed stratospheric aerosol optical properties. We welcome comments/alternative suggestions on all three points below.

1 - Injection lat/lon: Please ignore the bounds variable in the emission file, e.g. "lat_bnds". They are dummy bounds that had to or were automatically provided for cf compliant. Variables "lon" and "lat" are the recommended longitude and latitude of injection, and we recommend a column injection strategy i.e. inject all the SO2 in the model column containing the specified volcano lat/lon.

2- For time, we recommend a 24 hour injection. Typical duration of volcanic pulses will be on the order of hours rather than tens of hours, but for most eruption time is unknown so 24 hour is a "neutral" choice. It is also consistent with the temporal resolution of the satellite dataset used.

3- For altitude, again, ignore injection bound. There is a new "depth" variable in version 1.2.0 which should soon be landing on ESGF. We recommend that modelling group vertically distribute SO2 injections in the column using a Gaussian profile centered on the recommended injection height ("height" variable) and with a Gaussian width equal to the "depth" variable. If you are trying to work with version 1.1.3, I suggest you fix the depth to an ad-hoc value, e.g. 10% of the height. For modelling groups who want to implement a simpler vertical injection profile, scale the characteristic depth of it using our "depth" variable and center it on "height", and please document choices you make if they differ from our recommendation (documentation coming by early January).

[mdeushiMRI]

I am Makoto Deushi, working at Meteorological Research Institute (MRI)/Japan Meteorological Agency (JMA).
I compared the CMIP6Plus surface area density (sad) data with that of CMIP6.

CMIP6Plus sad: sad_input4MIPs_aerosolProperties_CMIP_UOEXETER-CMIP-1-1-3_gnz_175001-202312.nc
CMIP6 sad: multiple_input4MIPs_aerosolProperties_CMIP_IACETH-SAGE3lambda-3-0-0_gn_185001_201412.nc

The attached figure is the time series of CMIP6Plus (red) and CMIP6 (black) sad at 2.5N, 19.5 km. Unit is um2/cm3.
As seen in the figure, the magnitude of the CMIP6Plus sad is smaller than the CMIP6 sad by 1~2 order.
I would appreciate it if you would check whether the CMIP6Plus sad data is correct.

[vnaik60]

Thanks for looking at this dataset, Makoto. The issue you you note will be addressed by Tom and his team.

[thomasaubry]

Hello Makoto, thanks for finding and reported this issue! It came from a bug in the code which has now been fixed. Version 1.2.0, which is ready and undergoing validation before being uploaded to ESGF, should not have this issue anymore.