Question

[austinrwg]

Hi!

If you can help me with this problem, then I would very much appreciate it: which variable(s) will allow me to replicate this graph, as shown below.

Specifically, the FaIR 1.6.2 solid black line.

Thank you so much!

[lrennels]

Hi @austinrwg, to replicate the temperature impulse response graph above you will want the temperature anomaly variable from the FaIR climate model, which could for example use the variable global_temperature_norm in the TempNorm_1850to1900 component, or since it's just the delta you could use T from the temperature component. The former is normalized to preindustrial average over 1850 to 1900, the latter is raw from FaIR which runs from 1750, but since it's the impulse response they'll give the same results once taking the difference between baseline and pulse runs.

(:TempNorm_1850to1900, :global_temperature_norm), # Global surface temperature anomaly (K) from preinudstrial
# or
(:temperature, :T)

To replicate the graph you'd take the difference in temperature trajectories between the base run and the marginal run, which includes a pulse of carbon dioxide in the year 2030, and then normalize to pulse size and molecular mass as needed.

Give the substantial uncertainty in the FaIR climate module you'll want to run a Monte Carlo simulation to get your distribution and then take the mean.

[austinrwg]

Thank you.

Why do temperature anomalies differ slightly between GIVE and DSCIM?

Although they follow the same trajectory and their values are extremely close, the numbers are still slightly different.

[lrennels]

I'm not sure I can reliably answer without more information on the current version of the DSCIM implementation and where it's run from? If the differences are small, then a few answers may include

• uncertainty due to the MCS -- even if they are using the same 2,237 parameter sets derived from AR6 the means may not match exactly
• different specifications or versions of the FaIR model -- we use version 1.6.2 matched to the settings and uncertain parameters of AR6 but there are quite a few versions of FaIR available in its original Python form

[austinrwg]

Thank you.

I just have one clarification to ask for this sentence: "To replicate the graph you'd take the difference in temperature trajectories between the base run and the marginal run, which includes a pulse of carbon dioxide in the year 2030, and then normalize to pulse size and molecular mass as needed."

I changed pulse_size to 1, but what would I need to change about the molecular mass?

Having pulse_size = 1 seems to do the job to replicate the graph.

[lrennels]

Hi @austinrwg, it sounds like you're good to go then.

That comment was just that especially when comparing temperature impulse responses between models I find that the two units conversions always worth double checking are magnitude of pulse size and molecular mass.

For pulse size, you're right, the units for a pulse of carbon are Gt, so if you have pulse_size = 1 you should be fine. For HFCs, for example, the baseline pulse size is in units of Mt, and for most of our social cost work we use a small pulse size like pulse_size = 1e-4 so I just wanted to check what you were using.

And then sometimes those graphs are per Gt of carbon dioxide instead of carbon, so the 44/12 conversion can be needed to get the y-axis to be what you want.

[lrennels]

@austinrwg I'll close this issue for now, as it seems resolved, but feel free to open a new issue with further questions!

[austinrwg]

Thank you!

[bryanparthum]

Hi @austinrwg, please see here for replication code for that figure.

As for the difference between DSCIM and GIVE temperature anomalies, @lrennels hit the nail on the head with the pulse size and chemistry conversions. Please see the DSCIM User Manual for the relevant pulse sizes.