Welcome to the new DiscussCESM forum!
We are still working on the website migration, so you may experience downtime during this process.

Existing users, please reset your password before logging in here: https://xenforo.cgd.ucar.edu/cesm/index.php?lost-password/

How the initial dataset can be generated in detail?


I want to generate a special initial dataset from ECMWF reanalysis dataset. I downloaded U, V, T, HGT, MSLP, RHUM variables and I used definesurf to create a surface dataset containing SGH, LANDFRAC, PHIS. PS variable can also be diagnosed from the variables above. However, I still can not figure out how to generate TSICE, TS1,TS2,TS3,TS4, SNOWHICE, LANDM and CWAT. Can these data be downloaded from somewhere and interpolated to my speicial grids? I really appreciate your great help! Thank you!
Hi Xiexin,

I am also trying to generate my new initial condition from the same ECMWF data sets. I am not been able to find these remaining data sets. If you know it from some where they your help will be highly appreciable. I wonder how you have started making you initial conditions.



You don't specify which model you're using - but most recent versions of CAM
create initial conditions files by default (once/year on January 1). You could
run CAM at some standard resolution and then interpolate the fields you need to
your grid from this file. (The time-invariant field, LANDM should be generated
independently using software in the models/atm/cam/tools/definesurf directory).

But this would only create your startup file.
Since these fields aren't in balance with the ECMWF analyses, we spin up the
model by running it for 6 hours and dumping out initial files every 6
hours (namelist variable: inithist='6-hourly'). Replace the U,V,T, etc with
ECMWF values and then run for another 6 hours. You need to include the
resultant land initial file as well as atmospheric file in each subsequent
6-hour run. We spin up in this manner for at least one model month to get the
land spup up. To spin up the land more completely we may start 5 years earlier
than when we want to actually use ECMWF analyses and make a 5-year run with
AMIP sst's. (set inithist='monthly' or inithist='yearly' to generate initial
files). Then you will have spun-up land and atm initial files to begin the
ECMWF spin-up with.
Hi Olson,

I am using CAM3.0 and have run this model for 5 years and got 5 initial conditions (for every 1st January of each year). I have use the default T42 resolution for this run. Actually i want to use this model for some seasonal prediction experiments. I need to force model with initial conditions at each month as you know for seasonal prediction experiments initial condition are very crucial. As per you instruction I am rerunning this model again. So that i can get many initial condition generated by model. Then i will replace variables i have from ERA40 data sets as well as created by definesurf routine. Missing variables (T1 T2...) will be automatically added in my initial conditions from model it self. thats what i understood from you answer. Am i right?. Can you further define the methodology, so that i can avoid any mistake. Also can you refer me to any written document/paper for making initial condition in this way?

I am waiting for your answer as i am stuck in the stage where I have to generate my own initial condition.



Yes you are correct that the model will automatically place fields such as T1, T2, etc. on the IC file.

To replace the state variables with ERA-40, you may follow the methodology for
interpolations as described at the ECMWF web site. Particular attention should be paid
to vertical interpolations. Empirical formulae are used to account for differences in topography
between CAM and ERA-40 when performing interpolations near the surface:



I am having similar type of problem with CAM 3.0 generated initial condition file. I am doing ensemble run so I generated different initial condition file by doing control run for 12 years. But in the generated initial condition files PHIS, SGH, LANDM, LANDFRAC are missing. So when I am trying to use these generated initial condition as the initial condition of the model run I am getting error like "PHIS missing" . Can anybody please suggest what can I do to fix this error. I am stuck at this point for a long time any suggestion will be helpful.



Are you sure of the version of the model you are using to generate the initial conditions file? CAM3.0
does output PHIS, SGH, and LANDM_COSLAT on the IC file but later versions of the model such
as CAM3.1 do not. Versions such as CAM3.1 rely on a "bnd_topo" file to supply these additional
Thank you so much for the reply. Could you please tell me where can I get this bnd_topo file?
I am using bnd_topo in my namelist so I was wrong about my model version. How can I include these necessary fields in IC file so that I can use those IC file as Initial condition for starting my model run.



Do you know which version of the model you're running exactly? If you have bnd_topo in your namelist then you shouldn't need fields like PHIS in the IC file for a version like CAM3.1. Did you use bnd_topo
in your control run? Whatever file you used for bnd_topo has the necessary fields (e.g., PHIS) already. I'm not sure why you're getting the "PHIS missing" error message.
I am using CAM 3.1 p2 .I think I have figured it out. I did not use bnd_topo in control run. But as the supplied IC file have those 4 fields it did not make any difference. But while starting ensemble run with model generated IC file I need to use bnd_topo as bnd_topo have the missing fields.
Thanks Jerry.
Hi Absu,

If you remember this discussion, you said that you performed ensemble runs using model generated initials conditions. Did you use ECMWF data for replacing variables in these initial files. I am trying to sort out same kind of runs but its not working. Can you share some of your experience how you made ensemble runs.

Hi Olson,

I am tuning CAM4 for seasonal forecast and I had a discussion with you in this forum on the same post. If you can read my first post about how to created initial conditions for CAM4 form NCEP/ECMWF daily data. You asked me to run the model for your desired time and model will generate monthly initial conditions by its own and then replace state variables in the files.

I did the same and initially to check I just run the model for one year using prescribe SST and model has generated initial conditions. I then start replacing variables U,V T... in these files but it is not working. I am stuck here and will be very thank full to you if you could help me in detail to make initial conditions. I have both NCEP and ECMWF data and I want to use NCEP daily file. I first want to make some hindcast experiments and need monthly initial conditions form Januray 1980 to December 2000. Using these initial conditions at each month I will run the model for next three month to account seasonal climatology in model.

Looking forward to your reply.

I mean when I change variable in the files then this file is not stable. I started by just change temperature in the file only for the check if replacing one variables is ok. But when I run the model with this initial condition model is not able to read that file.



Did you plot maps of Temperature to see if it looked correct on your modified file? If you're
still having trouble, contact me at olson@ucar.edu. I could take a look copies of your
file - before and after you modified it.
Hi Siraj,I am a newer to CAM and have some questions. The CAM's files has different coordinate(p=A*Po+B*Ps) with NCEP(pressure/sigma) and ECWMF(P=A+B*Ps( 1013.250 hPa )). How did U change the U,V,T etc. using the NCEP/ECWMF data.Thanks a lotJeff
I was wondering how the atmospheric condition this would make, assuming small atmspheric change and just adjusted land condition? Does the ECWMF data have to be put in every 6 hour within the one month or just take the 1month outcome later?