Questions about CMIP6 anomaly forcing with CRUJRA datm: overlap stage (2015–2023), anomaly cycling, and DATM_YR_ALIGN/START

[xueyao]

Hi all,

I'm running a CTSM/CLM land-only case with CRUJRA atmospheric forcing and CMIP6 anomaly forcing for future projections. Setup:

• Model version: ctsm5.3
• Compset / resolution: SSP245_DATM%GSWP3v1_CLM50%BGC-CROP_SICE_SOCN_MOSART_SGLC_SWAV/f19_f19_mg17
• DATM mode: CRUJRA
• Anomaly forcing: anomaly_forcing = 'Anomaly.Forcing.cmip6.ssp245' in user_nl_datm
• Run period: [e.g., 2015–2100]

I have three questions:

1. Forcing during the overlap period (2015–2023) With DATM_YR_ALIGN=2015, DATM_YR_START=2015, DATM_YR_END=2023 and anomaly forcing enabled from the start of the run, what does the model actually use for 2015–2023?

• (a) Only the observed CRUJRA meteorology, with anomalies applied only after the CRUJRA period ends, or
• (b) CRUJRA meteorology plus the CMIP6 anomalies starting in 2015?

If (b), the observed climate change would be counted twice during 2015–2023. Is the recommended practice to run 2015–2023 without anomaly forcing and then turn it on from 2024 (e.g., via a branch or restart)?

2. How the anomaly forcing is applied over time Are the CMIP6 anomalies read as a time-varying monthly series (each model year uses its own anomaly year, e.g., 2050 uses the 2050 anomaly)? Or are they averaged or cycled over some period? Also, are the anomaly stream's year settings (yearFirst/yearLast/yearAlign) independent of DATM_YR_*? What baseline period are the anomalies computed against?

3. Must DATM_YR_ALIGN equal DATM_YR_START? I set DATM_YR_START=2009, DATM_YR_ALIGN=2015, DATM_YR_END=2023 and run 2024–2100:

• Which CRUJRA year is used for model year 2024: 2009, 2015, or something else?
• Does the background meteorology cycle over the full 2009–2023 range or only over 2015–2023?

My understanding is that DATM_YR_ALIGN is the model year that maps to DATM_YR_START. In that case model year 2015 would read CRUJRA 2009, and model year 2024 would read CRUJRA 2018, cycling over 2009–2023. Is that correct? Is there a recommended choice of cycling period for future runs?

Thanks very much for any clarification!

 

[oleson]

1. The model will use 2015 CRUJRA forcing with anomalies from 2015 applied, so (b). I expect the applied anomalies are small early on, so normal practice is to do the entire run with anomaly forcing.
2. Read as time-varying monthly time series. tintalgo is nearest so it would use the nearest monthly value at each model time step (no interpolation). The stream settings for years are independent of the DATM_YR_* An ncdump -h on the anomaly file indicates:

:title = "anomaly forcing data" ;
:note1 = "Anomaly/scale factors calculated relative to 2005.0-2025.0" ;
:history = "Tue Jun 28 2022: created by ./ssp_anomaly_forcing_smooth" ;
:gitdescribe = "branch_tags/ssp_af.n02_ctsm5.1.dev099" ;
:Source = "CMIP6 CESM simulations" ;
:Conventions = "CF-1.0" ;
:comment = "Monthly scale factors for given SSP scenario compared to a climatology based on data centered on 2015 over the range given in note1" ;
:number_of_ensemble_members = "3" ;

3. They don't have to equal each other. In the example you cite, I would expect model year 2024 would read CRUJRA 2018 and cycle over 2009-2023. The out-of-the-box cycling period for CRUJRA2024 appears to be 2001-2023.