Atm Chem Modeling Group 

TRACE–P Robotic Automatic Processing System (TRAPS)

Last Updated October 14, 2003

Atm Chem Modeling Group 

TRACE-P Products

NOTE: The TRAPS system was shut down on April 13, 2001, at the end of the TRACE-P mission. 

TRAPS forecast products are no longer available at this website, but have been archived to tape. If you need to access these products, please contact Bob Yantosca bmy@io.harvard.edu for assistance.

A second-generation build of TRAPS (named IAPS) was used to successfully provide daily forecasts of tagged CO tracers during the ITCT 2k2 mission (April-May 2002).

At present, we are preparing to resurrect TRAPS for providing forecasts for the ITCT 2k4 mission, which will take place in 2004 along the East Coast of the United States.  Here we go again!

What is TRAPS?

The TRACE-P Robotic Automatic Processing System (TRAPS for short) is the automated forecast system for the GEOS–Chem model of tropospheric chemistry.  TRAPS will be used to generate 4-day "look-ahead" chemical forecasts in support of the TRACE-P aircraft mission during Spring 2001.

NASA's Data Assimilation Office (DAO) will provide Harvard University with GEOS-Terra meteorological data during the TRACE-P mission.  This data consists of fields (such as winds, pressures, cloud mass fluxes, etc.) needed to drive the GEOS–Chem model. DAO will produce two types of meteorological data during TRACE-P: 1-day "first-look" assimilations (based on daily worldwide observations), and 5-day "look-ahead" forecast data (i.e., forward runs initialized from the 1-day assimilations).

Each day, Harvard will receive one assimilation data set (starting at 0 GMT) and two 5-day forecasts (starting at 0 GMT and 12 GMT) from DAO.  However, these data sets will not be ready for downloading until approximately 12 AM EST and 9 AM EST.  Additional processing of the DAO meteorological data must also be done by Harvard before it can be fed into the GEOS–Chem model.  The sheer volume of the incoming data (~1 GB/day), combined with the fact that much of the data processing has to be done during the overnight hours, necessitated that an automated data processing system be installed at Harvard.

What TRAPS does

The TRAPS package consists of Unix shell scripts, Fortran 90 code, IDL code, and the GEOS–Chem model source code.  The shell scripts (written in tcsh) act as "drivers" and invoke the various operations.  TRAPS operates in two cycles: a 0 GMT Forecast Cycle and an Assimilation/12 GMT Forecast Cycle.  The operations for each of the TRAPS cycles are listed below.

0 GMT Forecast Cycle

Step Code Operation
1 Shell script

Transfer 0 GMT forecast data (2 x 2.5 resolution) from DAO to Harvard via FTP. This consists of 5-days of GEOS-Terra "look-ahead" data on a 2 x 2.5 grid, starting at 0 GMT of the current date.

2 Fortran

Extract the fields needed to drive the GEOS–Chem model.

3 GEOS–Chem

Forward run the GEOS–Chem model with the 0 GMT forecast data at 2 x 2.5 resolution.  The smulation includes 5 Tagged CO tracers.  The model will be initialized with a restart file generated by the previous 12 GMT cycle.  Generate output files containing timeseries of tagged CO tracer concentrations.

4 IDL

Generate movies (animatable GIFs) of GEOS–Chem tagged CO tracer timeseries from the 0 GMT forecast run.

5 Shell Script

Place movies of tagged CO tracer timeseries on an FTP site so that field personnel can access them.

6 Shell Script

Transfer binary data files containing tagged CO timeseries concentrations to CAVE at NASA/Langley.  The CAVE will produce 3-D animations from this model output.

Assimilation/12 GMT Forecast Cycle

Step Code Operation
1 Shell script

Transfer assimilation data from DAO to Harvard via FTP.  This consists of 1-day of GEOS-Terra "first-look" assimilation data on a 1 x 1 grid, starting at 0 GMT of the previous date.

2 Shell script

Transfer 12 GMT forecast data (2 x 2.5 resolution) from DAO to Harvard via FTP.  This consists of 5-days of GEOS-Terra "look-ahead" data on a 2 x 2.5 grid, starting at 12 GMT of the current date.

3 Fortran

Extract necessary fields from the 1 x 1 assimilation data.  Regrid from 1 x 1 grid to 2 x 2.5 grid.

4 GEOS–Chem

Forward run the GEOS–Chem model with the 2 x 2.5 assimilation data for 1 day.  Generate a restart file which will be used to initialize the 12 GMT forecast run.

5 Fortran

Extract necessary fields from the 12 GMT forecast data.  This data is already on the 2 x 2.5 grid so no further regridding is necessary.

6 GEOS–Chem

Forward run the GEOS–Chem model for 5 days, using the 12 GMT forecast data.  Initialize the model using the restart file generated from the assimilation run.  Generate a new restart file, which will be used to initialize the next day's 0 GMT forecast run.  Generate output files containing timeseries of tagged CO tracer concentrations.

7 IDL

Generate movies (animatable GIFs) of GEOS–Chem tagged CO tracer timeseries from the 0 GMT forecast run.

8 Shell Script

Place movies of tagged CO tracer timeseries on an FTP site so that field personnel can access them.

9 Shell Script

Transfer binary data files containing tagged CO timeseries concentrations to CAVE at NASA/Langley.  The CAVE will produce 3-D animations from this model output.

TRAPS People

TRAPS was developed at Harvard University by scientific programmer Bob Yantosca and post-doctoral research fellow Isabelle Bey, under the guidance of TRACE-P project scientist Daniel Jacob.  Gi-Kong Kim and Rob Lucchesi, our colleagues at DAO, are responsible for generating the GEOS-Terra assimilation and forecast data that TRAPS will process and feed into the GEOS–Chem model.

Harvard and DAO collaborated in a "dry-run" test of the TRAPS system during the week of November 13, 2000.  DAO generated meteorological data files and stored them on an FTP server where Harvard could download them.  After a few initial missteps, TRAPS was routinely able to execute each of the individual operations for the 0 GMT Forecast Cycle and the Assimilation/12 GMT Forecast Cycle without human intervention.  We therefore have a high degree of confidence in our ability to generate quality CO forecasts for mission flight planning during TRACE-P.

During the months of January and February, 2001, minor upgrades have been made to the TRAPS system.  Harvard has upgraded its disk storage capabilities significantly in anticipation of the TRACE-P mission, which will require modifications to the TRAPS directory structure.  TRAPS can now switch to a backup FTP server if DAO's primary FTP server should fail during the mission.  Also, the IDL code used to generate animations from model output was significantly enhanced.

http://www.as.harvard.edu/ctm/geos/geos_traps.html