Difference between revisions of "GEOS-Chem nested grid simulations"

From Geos-chem
Jump to: navigation, search
(Transport fix)
(Transport fix)
Line 59: Line 59:
 
--[[User:Ccarouge|Ccarouge]] 16:35, 4 August 2010 (EDT)
 
--[[User:Ccarouge|Ccarouge]] 16:35, 4 August 2010 (EDT)
  
I applied this fix, and the RAM requirements were increased. This is problematic since I was operating near the maximum RAM of the node for nested sims before the fix. It went from 11 GB to over 16 GB (the maximum RAM).  I am using 148 tracers, which is why the RAM requirement is already high.  Is there a way to fix the transport issue without increasing the RAM?
+
'''''[mailto:jeffrey.pierce@dal.ca Jeff Pierce] replied:'''''
  
--[[User:Jeff Pierce]] 02:38, 9 August 2010 (EDT)
+
I applied this fix, and the RAM requirements were increased.  This is problematic since I was operating near the maximum RAM of the node for nested sims before the fix.  It went from 11 GB to over 16 GB (the maximum RAM).  I am using 148 tracers, which is why the RAM requirement is already high.  Is there a way to fix the transport issue without increasing the RAM?
  
 
=== 0.5x0.66 China nested grid simulation ===
 
=== 0.5x0.66 China nested grid simulation ===

Revision as of 18:49, 9 August 2010

This page contains some basic information pertaining to the GEOS-Chem nested grid simulations. For more detailed information about met data availability for the various geographic regions, please see our Available met data for nested grid simulations wiki page. We also invite you to visit the wiki page of the Regional Air Quality Working Group, which is responsible for coordinating nested-grid simulations within the GEOS-Chem user community.

Please also see our protocol for downloading and processing nested grid met data, which outlines the responsibilities for user groups using GEOS-Chem nested grid simulations.

Nested Grid Frequently Asked Questions

Can I perform a nested-grid simulation with GEOS-3 meteorology?

Yes. You may perform a GEOS-3 nested-grid simulation for one of the following regions:

You should be aware that GEOS-3 is now considered to be obsolete. Also, there is only a limited amount of GEOS-3 met data (2000-2001) available for the nested grid simulations.

Can I run a nested grid simulation with GEOS-5 meteorology?

Yes! With GEOS-Chem v8-01-02 and higher versions you may perform a nested grid simulation for the China/SE Asia region at the native GEOS-5 resolution of 0.5° x 0.666°. Also, GEOS-Chem v8-03-01 will contain several code updates for the North American and European nested-grid regions.

Boundary conditions

Dylan Millet wrote:

Do you know if the 0.5° x 0.667° nested simulation that's being developed uses a 4° x 5° or a 2° x 2.5° grid outside of the nested domain?"

Yuxuan Wang replied:

The nested simulation uses 4° x 5° simulation results only at the lateral boundaries of the nested domain. For example, the East Asia nested domain covers 10S to 55N and 70E to 150E, one needs to save 4x5 simulation results (tracer concentrations only) for each vertical layer, at a frequency of every 3 hours, at the following four boundaries:
lat = 10S, lon = 70E
lat = 10S, lon = 150E
lat = 55N, lon = 70E
lat = 55N, lon = 150E
One can easily change the code to have the nested-grid simulation use 2x2.5 simulation as boundary conditions, but the default has been to take 4x5 results since it's much faster to run a 4x5 simulation.

Steven Barrett replied:

Could you let me know the standard LL box and UR box of the 4x5 region used as boundary conditions for NA nested 1x1 GEOS-Chem simulations? I think it'll be LL = (9,26) and UR=(29,38), but the upper latitude extent of the NA region falls right on a grid cell boundary, so it's not clear to me which one would be the right one.

Philippe Le Sager replied:

Those indices are correct. They are default Fortran (starting at one for 1st cell), and correspond to cell within the domain.

--Bob Y. 13:51, 3 March 2010 (EST)

Setting up a 0.5x0.667 nested-grid simulation for GEOS-5

Transport fix

Lin Zhang reported a problem with the transport for nested simulations with GEOS-5 met fields. The problem was traced back to the fact that neither the latest pressure fixer nor the new advection have been set up to be used with nested grids. The problems in both were:

  • the use of polar caps for the first and last latitude bands of the nested domain, although there is no pole in the domain.
  • the use of periodicity both in longitude and latitude.

To solve the problems we modified both modules so that variables are updated only for an inner window of the domain, so that we do not need periodicity. The buffer zone for the pressure fixer is equal to 3 grid cells in all directions (defined in INIT_PJC_PFIX in GeosCore/pjc_pfix_geos5_window_mod.f). The buffer zone for the transport is the zone where the Boundary Conditions are fixed (as defined at the end of the NESTED MENU in input.geos). Note that the buffer zone for the advection must be larger than the buffer zone for the pressure-fixer.

The fix will be available in GEOS-Chem v8-03-02.

--Ccarouge 16:35, 4 August 2010 (EDT)

Jeff Pierce replied:

I applied this fix, and the RAM requirements were increased. This is problematic since I was operating near the maximum RAM of the node for nested sims before the fix. It went from 11 GB to over 16 GB (the maximum RAM). I am using 148 tracers, which is why the RAM requirement is already high. Is there a way to fix the transport issue without increasing the RAM?

0.5x0.66 China nested grid simulation

0.5x0.66 NA nested grid simulation

Setting up a 1x1 China nested-grid simulation for GEOS-3

NOTE: With the advent of the GEOS-5 meteorology, the GEOS-3 1x1 nested-grid simulation described below is now obsolete. We strongly recommend to use the GEOS-5 meteorology for your nested grid simulation.

--Bob Y. 14:00, 3 March 2010 (EST)

Here is a check list and input.geos file that you need to run the nested-grid China simulation (supplied by Yuxuan Wang):

define.h

These are the settings you need to make in the GEOS-Chem define.h file:

#define GEOS_3      'GEOS_3'
#define GRID1x1     'GRID1x1'
#define NESTED_CH   'NESTED_CH'
#define GRID30LEV   'GRID30LEV'

input.geos

These are the settings you need to make in the GEOS-Chem input.geos file:

Saving out 4 x 5 boundary conditions

You will have to run the global model at 4 x 5 resolution to save out the transport boundary conditions. Here are the relevant settings in input.geos:

%%% SIMULATION MENU %%% :
Start YYYYMMDD, HHMMSS  : 20010101 000000
End   YYYYMMDD, HHMMSS  : 20020101 000000
Run directory           : ./
Input restart file      : restart.YYYYMMDDhh
Make new restart file?  : F
Output restart file(s)  : restart.YYYYMMDDhh
Root data directory     : /data/ctm/GEOS_4x5/
 => GCAP       subdir   : AGRID/YYYY/MM/
 => GEOS-3     subdir   : GEOS_3/YYYY/MM/
 => GEOS-4     subdir   : GEOS_4_v4/YYYY/MM/
 => GEOS-5     subdir   : GEOS_5/YYYY/MM/
Dir w/ 1x1 emissions etc: /data/ctm/GEOS_1x1/
Temporary directory     : TEMP/
Unzip met fields?       : T
Wait for met fields?    : T
Use variable tropopause : N
Global offsets I0, J0   : 0 0

 ... 

%%% TRANSPORT MENU %%%  :
Turn on Transport       : T
 => Use Flux Correction?: F
 => Fill Negative Values: T 
 => IORD, JORD, KORD    : 3  3  7
Transport Timestep [min]: 30  
Use strat O3/NOy BC's   : T

...

%%% CONVECTION MENU %%% :
Turn on Cloud Conv?     : T
Turn on PBL Mixing?     : T
Convect Timestep (min)  : 30

...

%%% NESTED GRID MENU %%%:  
Save TPCORE 4x5 BC's    : T
TPCORE 4x5 BC directory : /as/home/ctm/yxw/bc/bc_ozone_std/
4x5 BC timestep [min]   : 180
LL box of 4x5 BC region :  51  21
UR box of 4x5 BC region :  67  37
1x1 offsets I0_W, J0_W  :   3   3

1x1 nested grid simulation

Then once you have saved out the 4 x 5 boundary conditions you can do the 1 x 1 nested grid simulation. Here are the appropriate settings in input.geos:

%%% SIMULATION MENU %%% :
Start YYYYMMDD, HHMMSS  : 20010301 000000
End   YYYYMMDD, HHMMSS  : 20010302 000000
Run directory           : ./
Input restart file      : restart.YYYYMMDDhh
Make new restart file?  : F
Output restart file(s)  : restart.YYYYMMDDhh
Root data directory     : /data/geos/GEOS_1x1_CH/
 => GCAP       subdir   : AGRID/YYYY/MM/
 => GEOS-3     subdir   : GEOS_3/YYYY/MM/
 => GEOS-4     subdir   : GEOS_4_v4/YYYY/MM/
 => GEOS-5     subdir   : GEOS_5/YYYY/MM/
Dir w/ 1x1 emissions etc: /data/ctm/GEOS_1x1/
Temporary directory     : TEMP/
Unzip met fields?       : T
Wait for met fields?    : T
Use variable tropopause : F
Global offsets I0, J0   : 250 79       # need to be changed for NA; index refer to 1x1 resolution

 ...

%%% TRANSPORT MENU %%%  :
Turn on Transport       : T
 => Use Flux Correction?: F
 => Fill Negative Values: T 
 => IORD, JORD, KORD    : 3  3  7
Transport Timestep [min]: 10  
Use strat O3/NOy BC's   : T

...

%%% CONVECTION MENU %%% :
Turn on Cloud Conv?     : T
Turn on PBL Mixing?     : T
Convect Timestep (min)  : 10

... 

%%% NESTED GRID MENU %%% 
Save TPCORE 4x5 BC's    : T
TPCORE 4x5 BC directory : /as/home/strat/yxw/bc/  # change to be your own directory
4x5 BC timestep [min]   : 180
LL box of 4x5 BC region :  51  21                 # need to be changed for NA; index refer to 4x5 resolution
UR box of 4x5 BC region :  67  37                 # need to be changed for NA; index refer to 4x5 resolution
1x1 offsets I0_W, J0_W  :   3   3                 # do not change for NA

NOTES to 1x1 nested-grid simulation for GEOS-3

General

  1. If not NESTED_CH (East Asia window) or NESTED_NA (north America window), need to first obtain
    • The 1x1 meteorological field for the nested domain;
    • Tll the input files (mostly emissions) at 1x1 resolution for the nested domain.
  2. There needs to be two run directories: one for the 4x5 global run and one for the 1x1 nested run
  3. Only one code directory is necessary. Code version v7-01-02 and higher can be compiled for both the global run and the nested run
  4. Run the 4x5 global simulation first. With correct switches in the 4x5 input.geos file, boundary conditions will
    • be saved out automatically into the designated directory and will be read later by the nested run.
    • The code automatically takes care of formatting.

Code

  1. Change grid size in define.h
  2. Change scaling factor in lightning_nox_mod.f (if not Chinese window)
    • search for 1x1 in the file to locate the lines of code where changes should be made
  3. Change IGLOB, JGLOB in CMN_SIZE (if not Chinese window)
    • IGLOB and JGLOB refers to the size of the nested window measured at 1x1 resolution
  4. Change PARAMETER( NLAND=2681, NPULSE=3 ) in commsoil.h (if not Chinese window)
    • changes are already made in the code; just uncomment the NA part
  5. Change datadir for O3 column in toms_mod.f --- if not CHINA or NA windows
    • need to create TOMS_O3col_2001.geos file at 1x1 resolution for the nested region
    • use the IDL code: regrid_4to1_O3_column.pro (note that line 33 and 34 needs to be modified for NA)