Creating initial conditions for CAM5

[jayant_pendharkar@cptec_inpe_br]

 Dear Olson I am a user of cesm1.2.2. I am interested in simulating global precipitation for Sept, 2016 - Jan, 2017 with CAM-SE 1deg set up. I am following the step (#3) in https://bb.cgd.ucar.edu/node/1001566 for generating the ICs. 1. I spun up the land from 2010/01 - 2016/02. Now I've land ICs and atmospheric startup file for 2016/03. 2. I regridded the ERA-Interim data for 2016-03-01:00:00:00 to SE 1 deg. I did not fully understand the next step:For spin up of atmosphere, 6-hourly runs are suggested for one model month replacing with ECMWF values after every 6-hourly interval. Does it mean that I replace the state variables with ERA-Interim for 2016-03-01:00 and stop it after first 6-hours (with inithist='6-hourly'). Then again replace with ERA-Interim 2016-03-01:06 values and run again for next 6 hours and so on. Do this until 2016-03-31:18. a. Is my understanding correct?b. why this exercise has to be done for 1 month (= [31 days x 4 timesteps] times)?c. What way will it effect if I just do this for the first timestep i.e, 2016-03-01:00 and let it run until jan 2017?

 

[olson]

> I did not fully understand the next step:For spin up of atmosphere, 6-hourly runs are
> suggested for one model month replacing with ECMWF values after every 6-hourly interval.
> Does it mean that I replace the state variables with ERA-Interim for 2016-03-01:00 and
> stop it after first 6-hours (with inithist='6-hourly'). Then again replace with ERA-Interim
> 2016-03-01:06 values and run again for next 6 hours and so on. Do this until 2016-03-31:18.



> a. Is my understanding correct?

   Yes, that is correct

> b. why this exercise has to be done for 1 month (= [31 days x 4 timesteps] times)?

  We found, when first developing this strategy, that it took the model (particularly the land model) about 1 month
  to equilibrate to the ECMWF analysis (or any foreign analysis) environment.  Since you are doing global precip investigations, it seems land equilibrium would be important


> c. What way will it effect if I just do this for the first timestep i.e, 2016-03-01:00 and let it run until jan 2017?

    If your intent is to overwrite the model state with ECMWF state fields just once and then let the model
    run for a year, it will quickly lose the memory of the ECMWF state and drift back to its original internal
    equilibrium state anyway.  In which case there would be no value in spinning up the model with
    ECWMF anyway unless there's something about your research I'm not understanding.

     Our strategy is to never let the model state drift too far from the ECMWF state.  Therefore, after the 1-month
   spinup, we only run forecasts out 5-30 days.  And we make these forecasts every day, each time replacing the
   initial state with ECMWF.  So we build an *ensemble* of forecasts and evaluate them in the aggregate to see
   how the model drifts from *reality*

 

[olson]

> I did not fully understand the next step:For spin up of atmosphere, 6-hourly runs are
> suggested for one model month replacing with ECMWF values after every 6-hourly interval.
> Does it mean that I replace the state variables with ERA-Interim for 2016-03-01:00 and
> stop it after first 6-hours (with inithist='6-hourly'). Then again replace with ERA-Interim
> 2016-03-01:06 values and run again for next 6 hours and so on. Do this until 2016-03-31:18.



> a. Is my understanding correct?

   Yes, that is correct

> b. why this exercise has to be done for 1 month (= [31 days x 4 timesteps] times)?

  We found, when first developing this strategy, that it took the model (particularly the land model) about 1 month
  to equilibrate to the ECMWF analysis (or any foreign analysis) environment.  Since you are doing global precip investigations, it seems land equilibrium would be important


> c. What way will it effect if I just do this for the first timestep i.e, 2016-03-01:00 and let it run until jan 2017?

    If your intent is to overwrite the model state with ECMWF state fields just once and then let the model
    run for a year, it will quickly lose the memory of the ECMWF state and drift back to its original internal
    equilibrium state anyway.  In which case there would be no value in spinning up the model with
    ECWMF anyway unless there's something about your research I'm not understanding.

     Our strategy is to never let the model state drift too far from the ECMWF state.  Therefore, after the 1-month
   spinup, we only run forecasts out 5-30 days.  And we make these forecasts every day, each time replacing the
   initial state with ECMWF.  So we build an *ensemble* of forecasts and evaluate them in the aggregate to see
   how the model drifts from *reality*

 

[olson]

> I did not fully understand the next step:For spin up of atmosphere, 6-hourly runs are
> suggested for one model month replacing with ECMWF values after every 6-hourly interval.
> Does it mean that I replace the state variables with ERA-Interim for 2016-03-01:00 and
> stop it after first 6-hours (with inithist='6-hourly'). Then again replace with ERA-Interim
> 2016-03-01:06 values and run again for next 6 hours and so on. Do this until 2016-03-31:18.



> a. Is my understanding correct?

   Yes, that is correct

> b. why this exercise has to be done for 1 month (= [31 days x 4 timesteps] times)?

  We found, when first developing this strategy, that it took the model (particularly the land model) about 1 month
  to equilibrate to the ECMWF analysis (or any foreign analysis) environment.  Since you are doing global precip investigations, it seems land equilibrium would be important


> c. What way will it effect if I just do this for the first timestep i.e, 2016-03-01:00 and let it run until jan 2017?

    If your intent is to overwrite the model state with ECMWF state fields just once and then let the model
    run for a year, it will quickly lose the memory of the ECMWF state and drift back to its original internal
    equilibrium state anyway.  In which case there would be no value in spinning up the model with
    ECWMF anyway unless there's something about your research I'm not understanding.

     Our strategy is to never let the model state drift too far from the ECMWF state.  Therefore, after the 1-month
   spinup, we only run forecasts out 5-30 days.  And we make these forecasts every day, each time replacing the
   initial state with ECMWF.  So we build an *ensemble* of forecasts and evaluate them in the aggregate to see
   how the model drifts from *reality*

 

[jayant_pendharkar@cptec_inpe_br]

Thank you. I have few queries:1. Which parameters to look for in order to know whether the land model is spun up or is stabilized?2. In order to make 6-hourly runs, a branch run needs to be set up? or I can include "inithist='6-hourly'" in user_nl_cam and resubmit it?3. I downloaded the ERA Interim data in pressure levels. I will then interpolate the data at *.cam.i.*.nc pressure levels and then regrid to SE - 1 deg. Is it the right way or I need to take care of something else?4. I am interested in performing the same process (land and atm spin up) at CAM-SE 2 deg and 0.25 dec. Can I use these files that are generated for 1 deg and interpolate at the required resolution?

 

[jayant_pendharkar@cptec_inpe_br]

Thank you. I have few queries:1. Which parameters to look for in order to know whether the land model is spun up or is stabilized?2. In order to make 6-hourly runs, a branch run needs to be set up? or I can include "inithist='6-hourly'" in user_nl_cam and resubmit it?3. I downloaded the ERA Interim data in pressure levels. I will then interpolate the data at *.cam.i.*.nc pressure levels and then regrid to SE - 1 deg. Is it the right way or I need to take care of something else?4. I am interested in performing the same process (land and atm spin up) at CAM-SE 2 deg and 0.25 dec. Can I use these files that are generated for 1 deg and interpolate at the required resolution?

 

[jayant_pendharkar@cptec_inpe_br]

Thank you. I have few queries:1. Which parameters to look for in order to know whether the land model is spun up or is stabilized?2. In order to make 6-hourly runs, a branch run needs to be set up? or I can include "inithist='6-hourly'" in user_nl_cam and resubmit it?3. I downloaded the ERA Interim data in pressure levels. I will then interpolate the data at *.cam.i.*.nc pressure levels and then regrid to SE - 1 deg. Is it the right way or I need to take care of something else?4. I am interested in performing the same process (land and atm spin up) at CAM-SE 2 deg and 0.25 dec. Can I use these files that are generated for 1 deg and interpolate at the required resolution?

 

[olson]

  1. It has been so long since we looked at those that I don't remember.  Certainly land surface T, possibly
     some measure(s) of soil moisture

  2.  Yes, you need to make a branch after you overwrite the *.cam.r.* file with the interpolated analyses

  3.  Pressure level analyses have a lot of data loss.  It is best to use the analyses on the ERA-I native
      vertical grid.  You can find a repository of model-level data here:

      https://rda.ucar.edu/datasets/ds627.0/

  4.  It is best to interpolate using the original ERA-I analyses not your 1-deg interpolated data

 

[olson]

  1. It has been so long since we looked at those that I don't remember.  Certainly land surface T, possibly
     some measure(s) of soil moisture

  2.  Yes, you need to make a branch after you overwrite the *.cam.r.* file with the interpolated analyses

  3.  Pressure level analyses have a lot of data loss.  It is best to use the analyses on the ERA-I native
      vertical grid.  You can find a repository of model-level data here:

      https://rda.ucar.edu/datasets/ds627.0/

  4.  It is best to interpolate using the original ERA-I analyses not your 1-deg interpolated data

 

[olson]

  1. It has been so long since we looked at those that I don't remember.  Certainly land surface T, possibly
     some measure(s) of soil moisture

  2.  Yes, you need to make a branch after you overwrite the *.cam.r.* file with the interpolated analyses

  3.  Pressure level analyses have a lot of data loss.  It is best to use the analyses on the ERA-I native
      vertical grid.  You can find a repository of model-level data here:

      https://rda.ucar.edu/datasets/ds627.0/

  4.  It is best to interpolate using the original ERA-I analyses not your 1-deg interpolated data

 

[jayant_pendharkar@cptec_inpe_br]

Apologies for a naive question. I am downloading the model level data from the site you suggested. It has hybrid coordinates in the vertical. Is there a way or a tool to interpolate the variables to the cam start up file grid (pressure levels)?

 

[jayant_pendharkar@cptec_inpe_br]

Apologies for a naive question. I am downloading the model level data from the site you suggested. It has hybrid coordinates in the vertical. Is there a way or a tool to interpolate the variables to the cam start up file grid (pressure levels)?

 

[jayant_pendharkar@cptec_inpe_br]

Apologies for a naive question. I am downloading the model level data from the site you suggested. It has hybrid coordinates in the vertical. Is there a way or a tool to interpolate the variables to the cam start up file grid (pressure levels)?

 

[olson]

  Since both CAM and the ERA-I fields have hybrid coordinates, you
 will need to compute the 3-dimensional pressure fields for both.
 In CAM you have hyam, hybm, hyai, hybi for the hybrid coefficients.
 In ERA-I, the equivalent coefficients are named, lv_HYBL2_a, lv_HYBL2_b,
 lv_HYBL_i3_a, and lv_HYBL_i3_b

   The formula for computing pressure (Pa) on model levels is:

    P(i,j,k) = hybm(k)*Ps(i,j) + hyam(k)*P0

   Same at model interfaces:

    P(i,j,k) = hybi(k)*Ps(i,j) + hyai(k)*P0

   where Ps is the 2-dimensional surface pressure and P0 is the constant 100000. Pa

     Presumably you will only be interpolating to model levels, not
  interfaces.

 

[olson]

  Since both CAM and the ERA-I fields have hybrid coordinates, you
 will need to compute the 3-dimensional pressure fields for both.
 In CAM you have hyam, hybm, hyai, hybi for the hybrid coefficients.
 In ERA-I, the equivalent coefficients are named, lv_HYBL2_a, lv_HYBL2_b,
 lv_HYBL_i3_a, and lv_HYBL_i3_b

   The formula for computing pressure (Pa) on model levels is:

    P(i,j,k) = hybm(k)*Ps(i,j) + hyam(k)*P0

   Same at model interfaces:

    P(i,j,k) = hybi(k)*Ps(i,j) + hyai(k)*P0

   where Ps is the 2-dimensional surface pressure and P0 is the constant 100000. Pa

     Presumably you will only be interpolating to model levels, not
  interfaces.

 

[olson]

  Since both CAM and the ERA-I fields have hybrid coordinates, you
 will need to compute the 3-dimensional pressure fields for both.
 In CAM you have hyam, hybm, hyai, hybi for the hybrid coefficients.
 In ERA-I, the equivalent coefficients are named, lv_HYBL2_a, lv_HYBL2_b,
 lv_HYBL_i3_a, and lv_HYBL_i3_b

   The formula for computing pressure (Pa) on model levels is:

    P(i,j,k) = hybm(k)*Ps(i,j) + hyam(k)*P0

   Same at model interfaces:

    P(i,j,k) = hybi(k)*Ps(i,j) + hyai(k)*P0

   where Ps is the 2-dimensional surface pressure and P0 is the constant 100000. Pa

     Presumably you will only be interpolating to model levels, not
  interfaces.

 

[jayant_pendharkar@cptec_inpe_br]

Thank you for your inputs. I could download the ERA-I model level data for an entire month and all four time steps (0,6,12,18). However, the data has 60 model levels and cam startup file has 30 levels. Is there some kind of remapping to be done? Apologies agan if I did not understand your explanation.

 

[jayant_pendharkar@cptec_inpe_br]

Thank you for your inputs. I could download the ERA-I model level data for an entire month and all four time steps (0,6,12,18). However, the data has 60 model levels and cam startup file has 30 levels. Is there some kind of remapping to be done? Apologies agan if I did not understand your explanation.

 

[jayant_pendharkar@cptec_inpe_br]

Thank you for your inputs. I could download the ERA-I model level data for an entire month and all four time steps (0,6,12,18). However, the data has 60 model levels and cam startup file has 30 levels. Is there some kind of remapping to be done? Apologies agan if I did not understand your explanation.

 

[olson]

  I will send you a tarball of one of our regridding routines next week...