McCormack 2015 ozone parametrization does not include ozone hole

[DWesl]

What is wrong?

Catalytic reactions on polar stratospheric clouds lead to ozone depletion in the Antarctic spring. The model used to derive the parameters of the current parametrization does not include the heterogeneous surface chemistry that drives the ozone hole. Instead of decreasing, global total column ozone increases during this period, while Antarctic total column ozone is nearly constant.

What should have happened?

Ozone mixing ratios should decrease, with some places reaching 0 kg/kg near maximum depletion around 40 mb/18 km.

Literature on loss rates:

• Hassler et al. (2011) has profiles of ozone loss rates
• Johnson et al. (2023) and Schofield et al. (2015) have more data on maximum ozone loss rates
• Lecouffe et al. (2022) has information that can be used for spatial extent

What machines are impacted?

All or N/A, Hera

Steps to reproduce

1. Set up an experiment with a forecast including late August to late September (35 days starting 15 August should work)
2. Compare output mixing ratio to NOAA South Pole Ozonesonde data or total column ozone data, or to NASA's satellite-derived polar cap ozone, or to GDAS estimates
   • The World Ozone and Ultraviolet Radiation Data Centre and Network for the Detection of Atmospheric Composition Change might have more relevant data

Additional information

This issue is present in each of twenty years of C192 runs Kate did a few months ago.

Do you have a proposed solution?

The layer of ozone loss is 12-20 km, according to NOAA GML, or 15-150 mb, according to Hassler et al. (2011).

The shape of the ozone loss profile is roughly given by $\log\left(\frac{p}{15\ \mathrm{mb}}\right) \log^2\left(\frac{150\ \mathrm{mb}}{p}\right)$, although $\sqrt{p - 15\ \mathrm{mb}}\ (150\ \mathrm{mb} - p)^2$ also works.

Hassler et al. (2011), Johnson et al. (2023), and Schofield et al. (2015) give the peak ozone loss rate at around 60 nmol/mol/day, or around 6 nmol/mol/sunlit hour. The latter value allows for a dependence on the amount of sunlight available.

Given the structure of the parametrization file (fix/am/ozprdlos*.f90), the reference/base state/climatological net production ($(P-L)_0$) values for August through October would need to change poleward of -60 or so. There is not enough information to tell if the other parameters would need to change.

Alternate methods:

• Re-derive the parameters in their entirety, using a model that includes stratospheric heterogeneous chemistry (some CESM configurations do this)
• Tweak the parameter file so the ozone parametrization instead nudges strongly toward climatology (timescales < 1 week instead of > 2 months)
• Derive new net ozone production by tweaking the parameter file to nudge strongly (timescales < 12 hours) toward observed/GDAS zonal-mean ozone, save the tendencies, and average over multiple years of such experiments
• Derive the parameters as regression coefficients (instead of Taylor series coefficients) of GDAS non-advective ozone tendencies against temperature, ozone mixing ratio, and ozone partial column

[DWesl]

I zipped up the Fortran files I used to create an updated parameter file with the Antarctic ozone hole. The program is designed to be run from the global-workflow root, and will write the new parameter file to the root as well.

In that ZIP archive is additionally the resulting parameter file, in cast that is easier to use. The new parameter file has additional time steps in late August and September instead of just the middle of each month. I haven't checked if that's helpful.

ozprdlos_2015_new_sbuvO3_tclm15_nuchem.zip

[DWesl]

That version (pressure dependence given by $\log(p - p{top}) \log^2(p{bottom} - p)$ ) puts the ozone hole too high in the atmosphere (~30 hPa instead of ~40 hPa), so it does not decrease the magnitude of the bias at 30 hPa. Changing the pressure dependence of the ozone loss to $\sqrt{p - p{top}} (p{bottom}-p)^2$ may help with that. (As noted above, $p{top} = 15\ \text{hPa}$ and $p{bottom} = 150\ \text{hPa}$ )