MERRA-2 implementation details
Overview
MERRA-2 is the newest reanalysis meteorological data product from NASA/GMAO. It is created with the NASA/GMAO GEOS Data Assimilation System (aka GEOS-DAS), version 5.12.4. Overall, the structure of MERRA-2 is very similar to the GMAO "forward processing" data product, which we refer to as GEOS-FP. Therefore, within GEOS-Chem we can (for the most part) treat MERRA-2 in the same way as we do for the GEOS_FP.
For more information about MERRA-2, please see these resources:
- MERRA-2 page on the GEOS-Chem wiki
- List of MERRA-2 met fields used for GEOS-Chem
- Version history of GMAO met data products
Comparing MERRA-2 to other GMAO met data products
The table below compares many features of MERRA-2 with other GMAO met data products:
| Feature | MERRA-2 | GEOS-FP | MERRA | GEOS-5 |
|---|---|---|---|---|
| Native vertical grid |
|
|
|
|
| Native horizontal grid |
|
|
|
|
| Time-invariant fields |
|
|
|
|
| Time archiving: surface fields |
|
|
|
|
| Time archiving: 3D fields |
|
|
|
|
| Timestamps | ||||
| Precipitation fields |
|
|
|
|
| Temperature |
|
|
|
|
| Relative humidity |
|
|
|
|
| Specific humidity |
|
|
|
|
| Vertical pressure velocity |
|
|
|
|
| Ertel potential vorticity |
|
|
|
|
| Snow mass |
|
|
|
|
| Sea level pressure |
|
|
|
|
| Surface pressure |
|
|
|
|
| Tropopause pressure |
|
|
|
|
| GEOS-Chem convection routine |
|
|
|
|
| GEOS-Chem wet deposition routine |
|
|
|
|
| GEOS-Chem transport routine |
|
|
|
|
| File format read by GEOS-Chem |
|
|
|
|
--Bob Y. (talk) 16:41, 17 August 2015 (UTC)
Timeline for integration into GEOS-Chem
Initial modifications
As of August 2015, we have we have modified GEOS-Chem (as described below) so that it can be driven by the new NASA/GMAO MERRA-2 reanalysis data product. These source code updates were validated with the 1-month benchmark simulation v11-01c and approved on 14 Sept 2015.
Currently, several of the GEOS-Chem specialty simulations are already compatible with MERRA-2. These include:
- CO2 simulation
- CH4 simulation
- Hg simulation (but some tuning factors for nested grids are still needed)
- POPs simulation
- Rn-Pb-Be simulation
- Tagged O3 simulation
--Bob Y. (talk) 14:54, 18 August 2015 (UTC)
UPDATE: GEOS-Chem v11-01 and newer versions can now be driven by the MERRA-2 reanalysis field data product.
Modifications for the nested grid simulations
There were several modifications needed to run nested grid simulations with MERRA-2. The source code updates for running MERRA-2 on the 0.5x0.625 nested NA and nested CH grids were included in v11-01e (approved 04 Jan 2016).
The nested CH grid for MERRA-2 is defined as 10S to 55N and 60E to 150E. It therefore includes China, Southeast Asia, Japan, and India. In order to make it transparent to users that this domain has been expanded, the nested CH grid has been renamed to the nested AS grid for MERRA-2. This update was included in v11-01h (approved 11 Oct 2016).
--Melissa Sulprizio (talk) 17:13, 14 December 2015 (UTC)
Modifications for the full-chemistry simulations
The various "full-chemistry simulations" (i.e. those using the benchmark, UCX, SOA, and tropchem chemistry mechanisms) use emissions (e.g. lightning, dust) that have scale factors that must be tuned specifically for each met field product. The GCST has computed the dust and lightning scale factors using MERRA-2 met fields for 2009–2014. These updates were included in v11-01f (approved 16 Apr 2016).
When the entire MERRA-2 coverage period (1980–present) has been processed, the lightning and dust scale factors for MERRA-2 will be adjusted as needed.
--Melissa Sulprizio (talk) 21:08, 5 February 2016 (UTC)
Required coding changes in GEOS-Chem
As noted in the section above, the similarities between MERRA-2 and GEOS-FP (especially of the vertical grids) means that in most cases we can just treat MERRA-2 in the same way as for GEOS-FP. In many areas of the code, the coding changes are straightforward and simply involve an extension of C-preprocessor statements such as:
#if defined( GEOS_5 ) || defined( MERRA ) || defined( GEOS_FP )
to
#if defined( GEOS_5 ) || defined( MERRA ) || defined( GEOS_FP ) || defined( MERRA2 )
In other areas of GEOS-Chem the changes are more involved. Below we provide a description of the areas of GEOS-Chem that were modified for MERRA-2:
Top-level directory
The following modifications for MERRA-2 were made in the root GEOS-Chem directory:
| File | Modifications made |
|---|---|
| Makefile_header.mk |
|
--Bob Y. (talk) 18:28, 12 August 2015 (UTC)
Headers directory
The following modifications for MERRA-2 were made in the Headers/ directory:
| File | Modifications made |
|---|---|
| CMN_SIZE_mod.F |
|
| gigc_input_opt_mod.F90 |
|
| gigc_state_met_mod.F90 | Multiple instances:
|
--Bob Y. (talk) 18:28, 12 August 2015 (UTC)
GeosUtil directory
The following modifications for MERRA-2 were made in the GeosUtil/ directory:
| File | Modifications made |
|---|---|
| bpch2_mod.F | In routine GET_MODELNAME:
In routine GET_NAME_EXT:
In routine GET_NAME_EXT_2D:
In routine GET_RES_EXT:
|
| pressure_mod.F | In routines GET_PEDGE and GET_PEDGE_FULLGRID:
In routine INIT_PRESSURE:
|
| time_mod.F | In routine GET_A1_TIME:
|
| transfer_mod.F | In routine INIT_TRANSFER:
In several routines:
|
--Bob Y. (talk) 20:33, 12 August 2015 (UTC)
GeosCore directory
The following modifications for MERRA-2 were made in the GeosCore/ directory:
| File | Modifications made |
|---|---|
| calcrate.F | Multiple instances:
|
| carbon_mod.F | Multiple instances:
|
| cldice_HbrHOBr_mod.F | Multiple instances:
|
| convection_mod.F | In routine DO_CONVECTION:
In routine DO_MERRA_CONVECTION:
|
| dao_mod.F | In routine COPY_I6_FIELDS:
In routine INTERP:
|
| depo_mercury_mod.F | In routine ADD_Hg2_SNOWPACK
|
| dao_mod.F | In routine COPY_I6_FIELDS:
In routine INTERP:
|
| diag3.F | Multiple instances:
|
| dust_dead_mod.F | In routine DST_MBL:
|
| fast_jx_mod.F | In routine FAST_JX:
|
| gamap_mod.F | In routine INIT_TRACERINFO:
|
| global_ch4_mod.F | In routines WETLAND_EMIS and READ_COPROD:
|
| input_mod.F | In routine READ_SIMULATION_MENU:
In routine READ_CONVECTION_MENU:
In routine VALIDATE_DIRECTORIES:
|
| land_mercury_mod.F |
In routines LAND_MERCURY_FLUX:
In routines SOILEMIS:
|
| main.F | In main.F:
In included routine READ_INITIAL_MET_FIELDS:
In included routine READ_MET_FIELDS:
|
| mercury_mod.F | In routine CHEMMERCURY:
|
| merra2_read_mod.F90 |
|
| ocean_mercury_mod.F | Multiple instances:
|
| planeflight_mod.F | In routine PLANEFLIGHT
|
| strat_chem_mod.F90 | Multiple instances:
In routine DO_SYNOZ:
|
| sulfate_mod.F | In routine CHEM_SO2:
In routine INIT_SULFATE:
In routine SRCSFC30 (TOMAS-only):
|
| tagged_ox_mod.F | In routine GET_REGIONAL_POX:
|
| tpcore_bc_mod.F: | Multiple instances:
|
| vdiff_mod.F90 | In routine VDIFFDR:
|
| wetscav_mod.F | In routine DO_WETDEP:
In routine DO_WASHOUT_ONLY::
In routine MAKE_QQ:
In routine RAINOUT:
|
--Bob Y. (talk) 18:29, 12 August 2015 (UTC)
HEMCO/Core directory
The following modifications for MERRA-2 were made in the HEMCO/Core directory:
| File | Modifications made |
|---|---|
| hco_chartools_mod.F90 | At the top of the module:
|
| hco_interp_mod.F90 | In routine MODLEV_INTERPOLATE:
|
--Bob Y. (talk) 18:28, 12 August 2015 (UTC)
HEMCO/Extensions directory
The following modifications for MERRA-2 were made in the HEMCO/Extensions directory:
| File | Modifications made |
|---|---|
| hcox_ch4wetland_mod.F90 | In routine WETLAND_EMIS:
|
| hcox_dustdead_mod.F | In routine HCOX_DUSTDEAD_GETTUN:
|
| hcox_lightnox_mod.F | In routine GET_OTD_LIS_SCALE:
|
| hcox_megan_mod.F | In routine GET_GAMMA_SM:
|
| hcox_tomas_dustdead_mod.F | In routine HCOX_TOMAS_DUSTDEAD_GETTUN:
|
--Bob Y. (talk) 18:53, 12 August 2015 (UTC)
Previous issues that are now resolved
Default timestep for MERRA-2 nested grid simulations
The original update was included in v11-02f (approved 17 May 2018).
The final recommendation will be included GEOS-Chem 12.0.0.
Bram Maasakkers wrote:
- Melissa Sulprizio and I have been working on the methane simulation with MERRA-2 and the nested North America grid at 0.5x0.625. During our initial simulations, we found very high values in the stratosphere, originating at the edges of the domain and showing a checkerboard-like pattern in the domain. It appears this is caused by a violation of the CFL condition in the stratosphere due to strong stratospheric winds in MERRA-2. When the condition is violated at the boundaries, it leads to an out-of-bounds error but even preventing an out-of-bounds error there by capping the array index doesn’t prevent the high concentrations from originating. It seems that this previously occurred for the GEOS-5 simulation but it did not seem to compromise simulation results significantly. For our purposes of comparing total column measurements of GOSAT to model output, it is a larger problem. We found it also occurs in the full-chem simulation, I’m attaching a plot of column ozone here to illustrate, the high values are only present in the stratospheric layers.
- I think that a variable timestep could be implemented to prevent this from happening. For now, we recommend using a 5/10 transport/chemistry timestep with MERRA-2 nested grid simulations instead of 10/20.
--Bob Yantosca (talk) 16:01, 17 May 2018 (UTC)
Ke Li wrote:
- The default time steps for nested tropchem simulation using MERRA2 are 5 mins for transport and 10 mins for chemistry. It took me about 2 days to run 1 months of simulation at the cost of 24 cpu. Now I change time steps to 10 mins for transport and 20 mins for chemistry, and it costs 1.2 days to run 1 months of simulation at the cost of 24 cpu. The 5/20mins time step for a test simulation over Asia (Tropchem, MERRA-2, 24 CPUs) still costs about two days for one month of simulation. The cost time is estimated from that the test simulation has used 28.5 hours for 17.5 days of simulation, though it hasn’t finished.
Lin Zhang wrote:
- I am fine to use 10/20 mins time step as default which does not seem to affect the tropospheric simulation of air pollutants. We used to run GEOS-5 at 0.5 degree resolution with 10/20 mins, and tropospheric simulations were working well.
- We can also put a stop sign for CH4 and CO2 simulations to warn users changing time steps to 5/10 mins.
Yuxuan Wang wrote:
- I agree with Lin and would also opt for a 10/20 mins time step as the default for full-chemistry nested-grid simulation. The reason is exactly as Daniel said. The nested-grid model users are predominantly using the full-chemistry option and concerns with tropospheric or surface concentration. The specialty nested-grid simulations, such as tagged CO, methane and CO2, used to require a separate development and model tuning, as least during the past before the HEMCO era. As a continuation of that ‘tradition’, the idea of putting a note or stop sign for the CH4 and CO2 nested-grid code should work well.
--Melissa Sulprizio (talk) 12:50, 30 July 2018 (UTC)
Unresolved issues
PBL height
Chris Holmes wrote:
I’ve documented many cases in which the reported PBLH in MERRA2 does not align with the mixing height that an observer would diagnose from the MERRA2 vertical profiles of potential temperature and humidity. In my group we are diagnosing our own PBL heights for use in GEOS-Chem.
--Melissa Sulprizio (talk) 20:25, 30 May 2017 (UTC)