Scheduled Downtime
On Tuesday 24 October 2023 @ 5pm MT the forums will be in read only mode in preparation for the downtime. On Wednesday 25 October 2023 @ 5am MT, this website will be down for maintenance and expected to return online later in the morning.
Normal Operations
The forums are back online with normal operations. If you notice any issues or errors related to the forums, please reach out to help@ucar.edu

Questions about setup emission-driven simulations

starrystarrynighttt

LLD
New Member
Dear all,

I just installed CESM2_1_3 on our cluster and I am trying to set up some emission-driven simulations.

For example, here are two cases I found on the CESM2 SCIENTIFICALLY VALIDATED CONFIGURATIONS list (https://csegweb.cgd.ucar.edu/exp2-public/cgi-bin/expListPublic.cgi), which I would like to start with:

b.e21.B1850.f09_g17.CMIP6-esm-piControl.001
b.e21.BHIST_BPRP.f09_g17.CMIP6-esm-hist.001


My question is:
1) Is there an available COMPSET that can be used for emission-driven simulations?

For example, B1850_BPRPcmip6 has the same components with b.e21.B1850.f09_g17.CMIP6-esm-piControl.001, is the atmospheric CO2 concentration in B1850_BPRPcmip6 calculated from surface emissions flux (my understanding is that the atmospheric CO2 concentration is fixed here)?

2) Is there any document or record that can be referred to explaining how to modify the parameters for these simulations?

Thanks very much!

Best regards!
 

starrystarrynighttt

LLD
New Member
A quick following up question is: in the transient emission-driven simulations (e.g., b.e21.B1850.f09_g17.CMIP6-esm-piControl.001), how the carbon release or carbon sequestration due to plant functional type changes is considered? Are they been added or subtracted from the surface emission input data?
 

klindsay

CSEG and Liaisons
Staff member
starrystarrynighttt,

The compset used for b.e21.B1850.f09_g17.CMIP6-esm-piControl.001 was B1850_BPRPcmip6. The cmip6 suffix leads to more model output. It can be omitted, yielding compset B1850_BPRP for less output.

The compset used for b.e21.BHIST_BPRP.f09_g17.CMIP6-esm-hist.00[12] was BHIST_BPRPcmip6. There is an analogous compset BHIST_BPRP with less output.

In all 4 of these compsets atmospheric CO2 is a transported tracer/constituent in CAM, and is calculated prognostically. In the B1850_BPRP compset, the surface boundary condition for atmospheric CO2 consists of land-air and sea-air CO2 fluxes that are computed by the land and ocean models respectively and sent via the flux coupler to the atmosphere model. The BHIST_BPRP compset additionally has surface CO2 fossil fuel emissions and 3-dimensional time-varying aircraft emissions. These emissions are datasets that are read directly by the atmosphere model.

Carbon release due to land use and land cover change (LULCC) are included in the net land-air CO2 flux.

Could you be more specific about what parameters you are interested in knowing more about?

Hope that helps,
Keith Lindsay
 

starrystarrynighttt

LLD
New Member
Dear Lindsay,

Thanks very much for your clarification.

We would like to run simulations and replace surface planetary function types. For example, we would like to replace all trees with grass over the tropical land. At the same time, we want to consider the additional carbon emissions from these land-use changes. So part of the carbon currently stored in trees should be released to the atmosphere when been replaced.

Based on your responses, if the carbon release due to land use and land cover change (LULCC) is calculated by the land model and included in the net land-air CO2 flux, maybe we could just change the PFT fraction and let the model determine the CO2 concentration?

Or should we add these additional CO2 emissions to some input files (such as the CO2 fossil fuel emissions)?

Thanks again!
 

lawrencepj1

Peter Lawrence
New Member
Hi Keith and Starry Starry Night

The land use time series file would be the best way to address this if you were looking to start with 1850 initial conditions and then perform a full transition to grass from trees. The PFT transitions will be handled by the land use code with the carbon transferred to the atmosphere, wood product pools which then decay over 10 and 100 year time periods and litter and coarse woody debris pools that will decay to the atmosphere based on heterotrophic respiration turn over rates. This was already performed as a transient with prescribed atmospheric CO2 concentrations in LUMIP with the idealized deforestation experiments. This experiment deforested at 1% rate for 80 years but this could be accelerated to a single year if required.

The case directory for this simulation is available on cheyenne at:

/glade/work/cmip6/cases/LUMIP/b.e21.B1850.f09_g17.CMIP6-deforest-globe.001

the user_nl_clm changes to do this are:

flanduse_timeseries = '/glade/work/cmip6/cases/LUMIP/b.e21.B1850.f09_g17.CMIP6-deforest-globe.001/landuse.timeseries_0.9x1.25_hist_78pfts_CMIP6_simyr1850-2015_c170824.20mil_top30_tdf.nc'
use_init_interp = .true.

best
Peter
 

starrystarrynighttt

LLD
New Member
Dear Peter,

Thanks very much for your responses, which sounds very encouraging to us.

The case you referred to seems good and we also want to do some additional simulations based on different deforestation strategies.

So my understanding is that "fsurdat" sets the initial condition of PFTs and then "flanduse_timeseries" describes the time transition of PFTs. Basically I just need to create an input *.nc file myself, which includes the desired time series of PFT distributions, and set it using "flanduse_timeseries", then the model will do the rest for me.

Sounds really great and thanks again for your explanations!

Best regards
 

lawrencepj1

Peter Lawrence
New Member
Hi Starry Starry Night

Yes exactly. The historical and SSP simulations all have the fsurdat set for 1850 then flanduse_timeseries files derived from the LUMIP/CMIP6 time series data. Those can be found under the DECK and ScenarioMIP directories:

/glade/work/cmip6/cases/DECK
/glade/work/cmip6/cases/ScenarioMIP

Peter
 

QINKONG

QINQIN KONG
Member
Hi Keith and Starry Starry Night

The land use time series file would be the best way to address this if you were looking to start with 1850 initial conditions and then perform a full transition to grass from trees. The PFT transitions will be handled by the land use code with the carbon transferred to the atmosphere, wood product pools which then decay over 10 and 100 year time periods and litter and coarse woody debris pools that will decay to the atmosphere based on heterotrophic respiration turn over rates. This was already performed as a transient with prescribed atmospheric CO2 concentrations in LUMIP with the idealized deforestation experiments. This experiment deforested at 1% rate for 80 years but this could be accelerated to a single year if required.

The case directory for this simulation is available on cheyenne at:

/glade/work/cmip6/cases/LUMIP/b.e21.B1850.f09_g17.CMIP6-deforest-globe.001

the user_nl_clm changes to do this are:

flanduse_timeseries = '/glade/work/cmip6/cases/LUMIP/b.e21.B1850.f09_g17.CMIP6-deforest-globe.001/landuse.timeseries_0.9x1.25_hist_78pfts_CMIP6_simyr1850-2015_c170824.20mil_top30_tdf.nc'
use_init_interp = .true.

best
Peter
Hi Peter,
First thanks for your answer which is very helpful to me. I'm trying to do a idealized tropical deforestation experiment as my climate modeling course project. But I still have some confusions. My plan is as follows:
(1) start with a ETEST control run.
(2) modify the fsurfdata file of the control run, basically changing the 'PCT_NAT_PFT' value of tropical forest into zero, and add its original value to C4 grass.
(3) Save this modified fsurfdata file. Create a new ETEST case, but modify the user_nl_clm to let fsurfdata point to this modified file.
(4) Run both the control run and this experiment run for 100 years (first 50years spin-up, and last 50 years for comparison).

Is the above workflow correct? I do have some concerns about it:
(1) When I looked at the fsurfdata file of the control run, there are many variables within it, like "MONTHLY_LAI" (leaf area index), and "MONTHLY_SAI" (stem area index). Do I need to change these variable as well for my deforestation experiment? Or only change 'PCT_NAT_PFT' would be sufficient?
(2) You mentioned that land use code will transfer carbon to atmosphere from the land use change. I can imagine this for a transient run. However, for the highly idealized case here (remove all tropical forest all at once and run a startup run), can the carbon from deforestation be transferred to atmosphere? Or the model will never know the existence of tropical forest?
(3) If I also want to try the transient land use but with a different deforestation scenario compared with that of LUMIP, should I just modify the flanduse_timeseries you give?
(4) why do we need "use_init_interp = .true" in the case above?
(5) Do we also need: "do_transient_pfts = .true" ?

Thanks a lot!
Best,
-Qin
 

lawrencepj1

Peter Lawrence
New Member
Hi Peter,
First thanks for your answer which is very helpful to me. I'm trying to do a idealized tropical deforestation experiment as my climate modeling course project. But I still have some confusions. My plan is as follows:
(1) start with a ETEST control run.
(2) modify the fsurfdata file of the control run, basically changing the 'PCT_NAT_PFT' value of tropical forest into zero, and add its original value to C4 grass.
(3) Save this modified fsurfdata file. Create a new ETEST case, but modify the user_nl_clm to let fsurfdata point to this modified file.
(4) Run both the control run and this experiment run for 100 years (first 50years spin-up, and last 50 years for comparison).

Is the above workflow correct? I do have some concerns about it:
(1) When I looked at the fsurfdata file of the control run, there are many variables within it, like "MONTHLY_LAI" (leaf area index), and "MONTHLY_SAI" (stem area index). Do I need to change these variable as well for my deforestation experiment? Or only change 'PCT_NAT_PFT' would be sufficient?
(2) You mentioned that land use code will transfer carbon to atmosphere from the land use change. I can imagine this for a transient run. However, for the highly idealized case here (remove all tropical forest all at once and run a startup run), can the carbon from deforestation be transferred to atmosphere? Or the model will never know the existence of tropical forest?
(3) If I also want to try the transient land use but with a different deforestation scenario compared with that of LUMIP, should I just modify the flanduse_timeseries you give?
(4) why do we need "use_init_interp = .true" in the case above?
(5) Do we also need: "do_transient_pfts = .true" ?

Thanks a lot!
Best,
-Qin
Hi Qin

Yes this seems like a perfectly good workflow to me. To be more specific however the intention of the experiment and the hypothesis you are testing needs to be captured in your model set up. The CLM model can be run in various configurations from Satellite Phenology (SP) mode to Biogeochemical (BGC) to BGC with the crop model on (BGCCROP). If you are only concerned with the surface energy and moisture fluxes you can run these experiments in SP mode. If you are concerned about carbon fluxes you will need one of the two BGC modes. The model has supported initial condition files which have been spun up for water, energy, carbon and nitrogen. These are important as the determine the starting state of the model. Starting at a non-equilibrium state will cause the model to drift which is going to compromise your experiment.

If your experimental design is to do a simple comparison of the land surface state with two alternative vegetation descriptions you can do what we call are time slice equilibrium comparisons. In this case you have the model start with the two alternative fsurfdata files and namelist option "use_init_interp = .true" will take the initial conditions from the first vegetation distribution and extrapolate it to the modified vegetation distribution. If these are spun up then they should be very close to spun up assuming there are similar vegetation types geographically located with similar climate.

The transient simulations are very different questions. I am going to point you to two papers that cover the nature of these types of simulations.


In these cases you have human activities that change the land use and land cover change trajectory with the model performing fluxes of carbon to the atmosphere. Depending on the CESM configuration this can result in changes to CO2 in the atmosphere, ocean and land.

I hope that helps.
best
Peter
 

QINKONG

QINQIN KONG
Member
Hi Qin

Yes this seems like a perfectly good workflow to me. To be more specific however the intention of the experiment and the hypothesis you are testing needs to be captured in your model set up. The CLM model can be run in various configurations from Satellite Phenology (SP) mode to Biogeochemical (BGC) to BGC with the crop model on (BGCCROP). If you are only concerned with the surface energy and moisture fluxes you can run these experiments in SP mode. If you are concerned about carbon fluxes you will need one of the two BGC modes. The model has supported initial condition files which have been spun up for water, energy, carbon and nitrogen. These are important as the determine the starting state of the model. Starting at a non-equilibrium state will cause the model to drift which is going to compromise your experiment.

If your experimental design is to do a simple comparison of the land surface state with two alternative vegetation descriptions you can do what we call are time slice equilibrium comparisons. In this case you have the model start with the two alternative fsurfdata files and namelist option "use_init_interp = .true" will take the initial conditions from the first vegetation distribution and extrapolate it to the modified vegetation distribution. If these are spun up then they should be very close to spun up assuming there are similar vegetation types geographically located with similar climate.

The transient simulations are very different questions. I am going to point you to two papers that cover the nature of these types of simulations.


In these cases you have human activities that change the land use and land cover change trajectory with the model performing fluxes of carbon to the atmosphere. Depending on the CESM configuration this can result in changes to CO2 in the atmosphere, ocean and land.

I hope that helps.
best
Peter
Hi Peter. Thanks for the reply. It's very helpful! You answer makes me realize that "use_init_interp = .true" will interpolate the default initial data to match the changed pft distribution (I initially thought "use_init_interp = .true" is aimed to deal with mismatch in grid).

I checked that both my control run and deforest run use the SP mode, co2_type=diagnostic for clm. The atm history file shows that co2 level is a fixed seasonal cycle. So, I guess it means in my case, all the carbon in the deforested tropical forest just disappear.

In terms of the aim of this course project, you are right that I want to do a simple comparison of the land surface state between two alternative vegetation descriptions, and slice equilibrium comparison should be what I'm going for. I want to look at changes in some surface variables, like air temperature, precipitation, energy and moisture flux. Given this aim, is making the deforest carbon simply disappear acceptable (for example carbon release to atmosphere will affect temperature but maybe very slightly)?

Thanks.

best,
-Qin
 

lawrencepj1

Peter Lawrence
New Member
Hi Qin

The SP option of CLM does not have a carbon cycle as it only has photosynthesis and prescribed LAI and SAI, so will not do what you are looking for with that configuration. The BGC version will have carbon pools and a carbon cycle. By doing the alternative vegetation distribution with BGC you will have a different base carbon content in terms of wood in the trees compared to the grasses you replace them with and different carbon uptake. This will not cover other elements of land use and land cover change but should be more than enough for your project.

best
Peter
Hi Peter. Thanks for the reply. It's very helpful! You answer makes me realize that "use_init_interp = .true" will interpolate the default initial data to match the changed pft distribution (I initially thought "use_init_interp = .true" is aimed to deal with mismatch in grid).

I checked that both my control run and deforest run use the SP mode, co2_type=diagnostic for clm. The atm history file shows that co2 level is a fixed seasonal cycle. So, I guess it means in my case, all the carbon in the deforested tropical forest just disappear.

In terms of the aim of this course project, you are right that I want to do a simple comparison of the land surface state between two alternative vegetation descriptions, and slice equilibrium comparison should be what I'm going for. I want to look at changes in some surface variables, like air temperature, precipitation, energy and moisture flux. Given this aim, is making the deforest carbon simply disappear acceptable (for example carbon release to atmosphere will affect temperature but maybe very slightly)?

Thanks.

best,
-Qin
 

QINKONG

QINQIN KONG
Member
Hi Qin

The SP option of CLM does not have a carbon cycle as it only has photosynthesis and prescribed LAI and SAI, so will not do what you are looking for with that configuration. The BGC version will have carbon pools and a carbon cycle. By doing the alternative vegetation distribution with BGC you will have a different base carbon content in terms of wood in the trees compared to the grasses you replace them with and different carbon uptake. This will not cover other elements of land use and land cover change but should be more than enough for your project.

best
Peter
Yeah! Thanks for the answer. It's clear!
 
Top