CAM standalone run without start_ymd

[janmejaisinghrawat@yahoo_com]

Hi all,
I am running a 2 year simulation on CAM 5.3 standalone. The default IC file is cami_1987-01-01_0.9x1.25_L26_c060703.nc.I made two simulations with two different namelists.In the first namelist, I didn't provide start_ymd and ncdata, I checked the log file of execution it took the default IC file cami_1987-01-01_0.9x1.25_L26_c060703.nc.In the second, I provided the start_ymd=19870101 and ncdata=cami_1987-01-01_0.9x1.25_L26_c060703.nc.Apart from this, there is no other difference in the namelist.When I compared the PRECT of the two simulations they were different. My question is why is that so? Also, which one is the correct simulation?

 

[rneale]

Iwould check out the atm_in namelist in your run directories to make absolutely sure they are the same. Just for sanity check out all the *_in namelist files for similarity.Rich

 

[rneale]

Iwould check out the atm_in namelist in your run directories to make absolutely sure they are the same. Just for sanity check out all the *_in namelist files for similarity.Rich

 

[janmejaisinghrawat@yahoo_com]

Thanks Rich for the reply,
I checked for any differences in the *_in files.I found that in case of the first namelist, i.e start_ymd and ncdata not provided. In the lnd_in the line " finidat= '/home/2014ast2289/cam/inputdata/lnd/clm2/initdata/clmi.BCN.2000-01-01_0.9x1.25_gx1v6_simyr2000_c100303.nc' " is present.While in the the case of second namelist, where start_ymd=19870101 and ncdata=cami_1987* is provided, this line is missing.I am fairly new to the model and don't know why the file clmi.BCN.2000-01-01_0.9x1.25_gx1v6_simyr2000_c100303.nc is used in the previous case but not latter.Can you please explain this? Thanks

 

[janmejaisinghrawat@yahoo_com]

Thanks Rich for the reply,
I checked for any differences in the *_in files.I found that in case of the first namelist, i.e start_ymd and ncdata not provided. In the lnd_in the line " finidat= '/home/2014ast2289/cam/inputdata/lnd/clm2/initdata/clmi.BCN.2000-01-01_0.9x1.25_gx1v6_simyr2000_c100303.nc' " is present.While in the the case of second namelist, where start_ymd=19870101 and ncdata=cami_1987* is provided, this line is missing.I am fairly new to the model and don't know why the file clmi.BCN.2000-01-01_0.9x1.25_gx1v6_simyr2000_c100303.nc is used in the previous case but not latter.Can you please explain this? Thanks

 

[rneale]

So that is the land initial condition file and is probably being set as default based on your ymd and ncdata setting.I presume you are using cam standalone scripts. Is there any way you could use CESM scripts with specific compsets (FAMIP) as the runs may behave more consistantly with your ymd changes. Why is it your are you not providing the CAM initial file in one run?

 

[rneale]

So that is the land initial condition file and is probably being set as default based on your ymd and ncdata setting.I presume you are using cam standalone scripts. Is there any way you could use CESM scripts with specific compsets (FAMIP) as the runs may behave more consistantly with your ymd changes. Why is it your are you not providing the CAM initial file in one run?

 

[janmejaisinghrawat@yahoo_com]

As I am new to the model, I have yet to work on CESM. Till now I have only used CAM standalone. So you are right I am using standalone script.At present it would be difficult for me to work on CESM as I don't any experience.As for the question why not provide CAM IC file, I wanted to see which file the model uses. As I thought it used the same default IC file which I provided through the second namelist.Hence I assumed that given the same IC file and same prescribed SST data, the two simulation would be same. Also I assumed in case of absent start_ymd the history files would be written from the date provided in the IC file. But it is not so.  I still am not able to understand why the difference is there. If I provide the execution log files, would it be any help to draw any conclusions. If so I will attach those. Thanks 

 

[janmejaisinghrawat@yahoo_com]

As I am new to the model, I have yet to work on CESM. Till now I have only used CAM standalone. So you are right I am using standalone script.At present it would be difficult for me to work on CESM as I don't any experience.As for the question why not provide CAM IC file, I wanted to see which file the model uses. As I thought it used the same default IC file which I provided through the second namelist.Hence I assumed that given the same IC file and same prescribed SST data, the two simulation would be same. Also I assumed in case of absent start_ymd the history files would be written from the date provided in the IC file. But it is not so.  I still am not able to understand why the difference is there. If I provide the execution log files, would it be any help to draw any conclusions. If so I will attach those. Thanks 

 

[eaton]

The default for start_ymd is 101.  Starting in year 0 is a common convention when doing a climatological run.  The default ncdata file for the cam4 1deg FV run is cami_1987-01-01_0.9x1.25_L26_c060703.nc.  This file is assigned the date 101 (in namelist_defaults_cam.xml) and is suitable for doing a present day climatology run which is what the build-namelist script sets up by default.  Also by default the initial file for CLM is set to clmi.BCN.2000-01-01_0.9x1.25_gx1v6_simyr2000_c100303.nc which represents a spun up state using present day climatology boundary conditions.  This file is also assigned the date 101.  What happened when you set start_ymd=19870101 is that first of all you needed to explicitly set the ncdata file because that date doesn't match the date in namelist_defaults_cam.xml and so build-namelist will stop with an error because it must provide a value for ncdata in order for the model to run.  On the other hand when build-namelist didn't find a value for finidat that matched your specified date it happily produced a lnd_in file with no setting for finidat.  That's because CLM has the capability of starting from 'arbitrary initial conditions' and does not require that finidat be set.  This will produce a very different simulation than the one starting from a spun up finidat file.As Rich points out if you want to do climate simulations that represent a specific climatology or historical period then the use of the CESM scripts is what sets this up correctly.  On the other hand if a present day climatology is sufficient then that's what the build-namelist script sets up by default.  But when you change something like start_ymd then you no longer have a default case and you need to be careful, as you were in checking the precip results.  When using stand alone scripts it is good practice to look at namelist files produced by default, and compare to what is produced when you make any changes to the input given to build-namelist. 

 

[eaton]

The default for start_ymd is 101.  Starting in year 0 is a common convention when doing a climatological run.  The default ncdata file for the cam4 1deg FV run is cami_1987-01-01_0.9x1.25_L26_c060703.nc.  This file is assigned the date 101 (in namelist_defaults_cam.xml) and is suitable for doing a present day climatology run which is what the build-namelist script sets up by default.  Also by default the initial file for CLM is set to clmi.BCN.2000-01-01_0.9x1.25_gx1v6_simyr2000_c100303.nc which represents a spun up state using present day climatology boundary conditions.  This file is also assigned the date 101.  What happened when you set start_ymd=19870101 is that first of all you needed to explicitly set the ncdata file because that date doesn't match the date in namelist_defaults_cam.xml and so build-namelist will stop with an error because it must provide a value for ncdata in order for the model to run.  On the other hand when build-namelist didn't find a value for finidat that matched your specified date it happily produced a lnd_in file with no setting for finidat.  That's because CLM has the capability of starting from 'arbitrary initial conditions' and does not require that finidat be set.  This will produce a very different simulation than the one starting from a spun up finidat file.As Rich points out if you want to do climate simulations that represent a specific climatology or historical period then the use of the CESM scripts is what sets this up correctly.  On the other hand if a present day climatology is sufficient then that's what the build-namelist script sets up by default.  But when you change something like start_ymd then you no longer have a default case and you need to be careful, as you were in checking the precip results.  When using stand alone scripts it is good practice to look at namelist files produced by default, and compare to what is produced when you make any changes to the input given to build-namelist. 

 

[janmejaisinghrawat@yahoo_com]

Thanks Eaton for your reply,It was very helpful. My aim is to simulate years 1987 and 1988 and compare the precipitation produced by model to the observations to check how well the model simulates these years.From your anwer, I conclude that using CESM script is the correct way. Am I right?If I was to use CAM standalone, what is the correct method that serves the intended purpose? Thanks

 

[janmejaisinghrawat@yahoo_com]

Thanks Eaton for your reply,It was very helpful. My aim is to simulate years 1987 and 1988 and compare the precipitation produced by model to the observations to check how well the model simulates these years.From your anwer, I conclude that using CESM script is the correct way. Am I right?If I was to use CAM standalone, what is the correct method that serves the intended purpose? Thanks