GEOS-5 met field reprocessing

From Geos-chem
Revision as of 21:05, 18 February 2009 by Bmy (Talk | contribs) (Fix in a6_read_mod.f)

Jump to: navigation, search


Data versions

GEOS-Chem uses the GEOS-5 operational data product, which is the product that starts approximately in 2004 and goes forward in time. Over the last couple of years there have been 3 different versions of GEOS-5 met data:

  1. GEOS-5.0.1 (produced prior to approx 1/1/2007)
  2. GEOS-5.1.0 (produced between 1/1/2007 to approx. 9/2007)
  3. GEOS-5.2.0 (produced after approx. 9/2007)

When we started preparing GEOS-Chem to ingest the GEOS-5 met data, the only available GEOS-5 data was the very oldest data product, GEOS-5.0.1. Therefore when we downloaded and regridded our met data years 2005-2007, these data were the GEOS-5.0.1 product.

Around the beginning of 2008, GEOS-5.0.1 was turned off and replaced by a newer version of the dataset, GEOS-5.1.0. Therefore, the GEOS-5 met data that we downloaded and regridded for the calendar year 2008 was the GEOS-5.1.0 product. It also turned out that GEOS-5.1.0 had several improvements in the convective fields (e.g. CMFMC, DTRAIN) used by GEOS-Chem with respect to GEOS-5.0.1. Therefore we were recommended to replace all of our GEOS-5.0.1 data (i.e. years 2005-2007) with the newer GEOS-5.1.0 data.

In September 2008, the GEOS-5.1.0 data set was switched off and replaced by a newer data set, GEOS-5.2.0. The GEOS-5.2.0 is now the stable operational dataset of GEOS-5. This is also the version that will be used to create the MERRA reanalysis that will extend for many years into the past. Thus, all GEOS-5 met data that we downloaded and regridded for GEOS-Chem on or after 9/1/2008 is the GEOS-5.2.0 data. For more information about the switchover to GEOS-5.2.0, please visit this link.

Optical depth error

After several years of GEOS-5 met data had been downloaded & regridded for GEOS-Chem, it was discovered that there was an error in the optical depth field. For detailed information please see this link.

Long story short: the optical depth fields provided to us by GMAO were mislabled. We had thought that these were grid-box optical depths but in reality they were in-cloud optical depths. We were able to make a quick fix in order to let GEOS-Chem users continue to use the existing GEOS-5 met data. However, the more robust solution was to reprocess the GEOS-5 met data so as to handle the optical depth regridding properly. (Hongyu Liu gave us a regridding algorithm for this).

Data regridding

As described above, we have already commenced to reprocess the entire GEOS-5 met data set to:

  1. To fix the optical depth error.
  2. To replace our 2005-2007 GEOS-5 data files from the older GEOS-5.0.1 data product with the more recent GEOS-5.1.0 data product (and thus take advantage of the better convection fields).
  3. To extract the GEOS-5.1.0 data product for the year 2004 (which had not been previously downloaded).

The GEOS-5 met data reprocessing is totally automated, but the limiting factor is the file transfer speeds. We therefore ask for your patience in this matter. At present we can process about a month of GEOS-5 met field files per business day.


As of February 18, 2009, we have reprocessed the following GEOS-5 met data:

  • 2008: entire year
  • 2004: May - December

Required source code modifications

In order to use the new GEOS-5 met data fields, we recommend that you upgrade to GEOS-Chem v8-01-04 as soon as it is released. Version v8-01-04 will have all of the necessary source code modifications that you need to use the newly reprocessed GEOS-5 met data.

However, if you are already using GEOS-Chem v8-01-02 or GEOS-Chem v8-01-03, these have some (but maybe not all) of the required changes. You may have to add some of the changes into these versions manually.

If you are using a version of GEOS-Chem older than v8-01-02, then you will have to upgrade.

File list

In include file define.h make sure that you have the IN_CLOUD_OD switch uncommented:

! NOTE: If you are using GEOS-5 met fields that were reprocessed to 
! correctly regrid the in-cloud optical depth and cloud fraction fields, 
! then be sure to uncomment the following line of code.  This will cause 
! FAST-J to interpret the optical depth correctly.  Leaving this option
! commented will cause a "quick fix" (i.e. multiplying the optical depth
! by the cloud fracton) to be applied, which should be a good enough fix
! in the meantime. (bmy, hyl, 10/24/08)

The source-code routines that depend on the IN_CLOUD_OD switch are:

  1. a3_read_mod.f
  2. a6_read_mod.f
  3. i6_read_mod.f
  4. fast_j.f
  5. lightning_nox_mod.f

In particular, if you are using v8-01-02 or v8-01-03, please double-check the following files and make sure you have these changes implemented. (These were based on bug reports from users after v8-01-02 and v8-01-03 had already been released:

Fix in a6_read_mod.f

We need to manually set the number of expected A6 fields in the file to be 21. Otherwise GEOS-Chem will not read in all of the fields properly. Please place the following #if defined block after the READ statement in subroutine OPEN_A6_FIELDS (in source code file a6_read_mod.f):

         ! The last 2 digits of the ident string
         ! is the # of fields contained in the file
         READ( IDENT(7:8), '(i2.2)' ) N_A6_FIELDS

#if   defined( GEOS_5 ) && defined( IN_CLOUD_OD ) 
         !%%% KLUDGE: set N_A6_FIELDS=21 when using the reprocessed
         !%%% GEOS-5 met.   This accounts for CMFMC (which doesn't seem
         !%%% to get counted) as well as for MOISTQ, which is an extra
         !%%% derived field.  (jmao, bmy, 2/12/09)
         N_A6_FIELDS = 21

Fix in lightning_nox_mod.f