Using a ground temperature map as an initial condition (restart file)

I don't have specific guidance on this since I'm not aware of anyone having done it before. But yes I would think you'd have to modify T_SOISNO and T_GRND at a minimum. And possibly H2OSOI_LIQ and H2OSOI_ICE of that top layer if the temperature is inconsistent with the liquid water/ice composition of that layer. It might be that the model will try to freeze that water or melt that ice if there is an inconsistency and cause some spurious fluxes you might want to avoid. These are all column level variables so you'd want to restrict your changes to just soil columns (e.g., exclude urban columns, etc.)

oleson said:

I don't have specific guidance on this since I'm not aware of anyone having done it before. But yes I would think you'd have to modify T_SOISNO and T_GRND at a minimum. And possibly H2OSOI_LIQ and H2OSOI_ICE of that top layer if the temperature is inconsistent with the liquid water/ice composition of that layer. It might be that the model will try to freeze that water or melt that ice if there is an inconsistency and cause some spurious fluxes you might want to avoid. These are all column level variables so you'd want to restrict your changes to just soil columns (e.g., exclude urban columns, etc.)

Click to expand...

Dear Keith,
Thank you so much for your response. I will give it a try. But before I test that, I should mention that I discussed this idea with another colleague, and they told me that there is another, easier, better way. Here is what they said:

"you should use the average of your GST (over North America, (which is the mean of 1300–1700 CE)) and find another temperature time-series from that period of time (lets call it mem2) and calculate the mean of that too, then figure out the bias between these two means (Δt) and whatever the Δt is should be applied to the mem2 timeseries. This way, you reconstructed the temperature (TBOT) from GST estimates, and you can run your offline model using this new reconstructed temperature for a spin-up and transient run."

But there is a problem with their suggestion, which I think you can confirm too. For a land-only model (offline simulation) to run, we need those seven boundary conditions, and temperature is only one of them. I have seen that precipitation, humidity, and even longwave radiation show the same pattern as TBOT in their time series. So only by imposing a bias (Δt) into TBOT we can not simply run the offline model and call it a day, right?

I guess you have come across similar ideas where they wanted to apply bias to the TBOT boundary conditions. Do they impose that bias on other boundary conditions too? If yes, how? Because let's say Δt is 0.2 deg C, then this value can not simply be applied to humidity or precipitation; it needs to be converted into a rational bias related to each boundary condition, I suppose.

Thank you for your help in advance.

The forcing temperature TBOT is of course different from the actual ground surface temperature, so changing TBOT by the bias you mention wouldn't necessarily result in the same change in ground surface temperature. But it could be considered a sensitivity study method.
We use an anomaly forcing method to run land-only simulations for future scenarios where the anomalies are calculated as departures from present day using output from fully coupled future scenario simulations (e.g., SSPs). Variables include TBOT, RAIN/SNOW, FSDS, etc. See:

Thank you very much for the clarifications.
So my idea, as you mentioned, would be a TBOT anomaly sensitivity experiment, not a complete reconstruction of the 1300–1700 atmosphere.

I need to confirm something here:
When we use ISSP compsets, it automatically expects the DATM files to be from GSWP3 or the available DATM that the model itself downloads, so when we assign the start and end years of DATM forcing in the beginning (I did try this before by running SSP scenarios, which did set it to 1995-2014 for my case before), it applies the SSP anomaly to that DATM data (also applies the domain too: for example in user_datm.streams.txt.Anomaly.Forcing.Precip we need to point to the anomaly and the domain files: af.allvars.CESM.SSP2-4.5.2015-2100_c20220628.nc and domain.permafrostRCN_P2.c2013.nc)

But here I have some issues. First, my map of GST is for North America only, and I don't know if CLM can run only for north america or not (on the ctsm user guide, I only found regional single point documentation and the link for clm_userdat is broken). Secondly, my mem2 forcings (boundary conditions) that I want to impose the bias are for 1300- 1700 CE, and they have a different configuration from the default DATM model downloads, so if I follow the anomaly forcing steps, I need to change every detail regarding the domain, resolution, etc. So it would be easier to manually impose the bias into the TBOT and run it without the anomaly-forcing setup method.
But still, I am not sure if we could run it only for North America or not, if not, do you think it would be rational to just impose the bias on my DATM files only for North America and set the bias for other cells to zero? Is that a norm in simulation setups you've encountered?

I assume these are the related links for regional cases. However, I didn't find anyone creating North American files for a regional case.