The Earth Observer, November/December, 1995

Data Assimilation Configurations for TRMM and AM1

The Goddard Data Assimilation Office (DAO) plans to produce a hierarchy of products during the TRMM and EOS AM-1 eras. Some of these will be produced only as needed while others will be produced routinely. Although we're still in the definition phase of the specific products, the special configurations of the Goddard EOS Data Assimilation System (GEOS DAS) we expect to use are given below. In addition, following these special configurations is a brief description of the basic GEOS DAS followed by some hints on selecting a product. Feedback on these plans is welcome. Please contact Jim Stobie.

Special Configurations

First-Look Analysis (routinely produced)
The first-look analysis takes the basic GEOS DAS configuration and adds a special "fly-through" module. This module extracts incremental analysis update (IAU) data (see basic GEOS DAS below) at every model time step (approximately one minute) along a given satellite subtrack. The first-look analysis runs daily 12 to 24 hours after data time using primarily non-EOS data (see Table 1).

GCM Forecast/Simulation (provided as needed)
The GCM forecast/simulation looks like the basic GEOS DAS except that the objective analysis portion is turned off. The only outside data that enter the system are the boundary conditions such as sea surface temperature climatology. This configuration is used to produce 10-day forecasts for various NASA field experiments. It is also used to produce multi-year simulations that investigate the climatology of the GCM itself. Such investigations are needed to isolate real climate signals from artificial model climate signals.

Final-Platform Analysis (provided as needed)
The final platform analysis is just like the basic configuration except it brings the new EOS observations into the objective analysis. It is called a final "platform" analysis because it will be tailored to the observations from a given EOS platform such as AM-1 or CHEM-1. It will usually be run several months after data time to allow full processing of the EOS instrument retrievals.

Off-line Analysis (Produced routinely and as needed)
A typical off-line analysis uses information from a first-look or final-platform analysis as input to another assimilation system. This input is assimilated with EOS and/or other observations to produce special off-line analysis products. An example of this is the current DAO constituent assimilation effort that uses winds from the GEOS DAS to drive a tracer model for off-line N2O assimilation. The major distinction between an off-line analysis and a normal analysis is that the off-line products do not cycle back into the GEOS DAS. During the TRMM/AM-1 era off-line analyses of CO and O3 will be provided.

Pocket Analysis (produced as needed)
Pocket analyses are just like the final platform analyses except that selected instrument data are excluded from the assimilation. By excluding a given instrument type from the assimilation, its impact on the climatic signal of the overall system can be assessed. Pocket analyses are important to determine if certain short-lived instruments produce artificial climate signals within the GEOS DAS and, if so, how these artificial signals might be reduced.

Reanalysis (routinely produced)
These are our primary archive products. Reanalyses configurations look exactly like the corresponding original analyses. For example, a platform reanalysis looks just like a platform analysis. The major distinction is that during a given reanalysis, no changes are permitted to the assimilation algorithms. This is done to eliminate false climatic signals that sometimes result from intermittent algorithm changes. Approximately once every 4 years the DAO reanalyzes the past 20 years of data using the latest data assimilation techniques. In addition, shorter term reanalyses (several months to several years) are conducted periodically to satisfy specific scientific requirements.

Basic GEOS DAS

Data assimilation combines observations from many different instruments with a general circulation model (GCM) to produce gridded data sets (for a complete list of assimilation products see DAO homepage, http://dao.gsfc.nasa.gov/). Here's how the basic assimilation process works:

Cycling
The GEOS DAS operates on a fixed-length cycle (currently 6-hours). That is, it gathers observations in 6 hour blocks and uses them to correct the GCM's estimates for that same time period. Figure 1 shows a typical cycle.

Figure 1. Typical 6-hours cycle of the GEOS DAS. Processes are rectangles and data files are ovals. Times (UTC) are shown at the top of each process or data file. The vertical coordinate (pressure or sigma) is shown at the bottom or lower right corner. In the lower left corner, the GCM integrations are identified as either an incremental analysis update (IAU) or first guess (FG). Vertical interpolation steps between sigma and pressure levels are shown in the small rectangles.

First Guess
Each cycle begins with a restart file from the previous cycle. Using this for its initial conditions and boundary conditions from other sources, e.g.. sea surface temperature, terrain elevation, etc., the GCM integrates 3 hours into the future to produce the first guess.

Objective Analysis Observations (EOS and non-EOS)
are gathered for the 6 hours surrounding the valid time of the first guess. For example, if the first guess valid time is 06 UTC, then observations from 03 to 09 UTC are used. An objective analysis system compares these observations with the first guess and produces a set of gridded corrections. These corrections are called analysis increments. One assimilation technique is to apply these increments all at once to the first guess, producing the analysis in Figure 1.

IAU
The GEOS DAS uses another method called the incremental analysis update (IAU). Rather than putting the increments in all at once at 06 UTC, IAU goes back and reruns the GCM from 03 to 09 UTC, gradually inserting the analysis increments at each model time step. This has several benefits including the ability to produce assimilations at much higher temporal resolution. Thus, even though data are gathered in 6 hour blocks, each single-level product is provided every 3 hours through the IAU process. Multi-level (pressure and sigma) assimilations are still archived every 6 hours, but are based on the same IAU process. Furthermore, a first-look analysis (see below) uses IAU to provide selected fields at very high temporal resolution along the satellite subtrack for use by TRMM and AM-1 instrument retrievals. The same boundary conditions that went into the first guess are also used for the IAU.

Mandatory Pressure vs. Sigma Levels The World Meteorological Organization (WMO) has established the following mandatory pressure levels:

The objective analysis is done at these mandatory pressure levels, while the GCM operates on model sigma levels. The GCM sigma levels are based on the following formula:

sigma= ( p - pt ) / (ps - pt )

where: p = pressure of the sigma level,

pt = pressure at the top of the GCM,

ps = surface pressure.

The GEOS DAS GCM uses approximately 70 sigma levels from the surface to 0.01 mb. Interpolation steps between sigma and pressure levels are indicated by the small rectangles in Figure 1.

Continuing the Cycle
The next 6-hour segment begins by creating a new restart file at 09 UTC and then extending the GCM integration 3 more hours to 12 UTC. The first guess (FG) integration is done without analysis increments.

Selecting a Product

The GEOS DAS provides the user with many product options. The assimilations on pressure surfaces are compatible with a wide variety of other data sets since they are on the WMO mandatory pressure levels. However, for the user who demands higher vertical resolution or wants to avoid the final sigma-to-pressure interpolation, the assimilations on sigma surfaces are the best choice. The boundary conditions, observations, and increments are available for users who want to delve deeper into the sources of the assimilated products. Finally, the first guess and analysis products are more for internal monitoring of the GEOS DAS performance and are not recommended for detailed climatic studies.

REFERENCES

Bloom, S. C., L.L. Takacs, A. M. da Silva, and D. Ledvina, 1995: Data assimilation using incremental analysis updates. Mon, Wea, Rev., submitted.

Pfaendtner, J., S. Bloom, D. Lamich, M. Seablom, M. Sienkiewicz, J. Stobie, and A. da Silva, 1995: Documentation of the Goddard Earth Observing System (GEOS) Data Assimilation System - Version 1,NASA Tech Memo. 104606, Vol. 4, NASA, Goddard Space Flight Center, Greenbelt, MD, 44 pp.

Schubert, S. D., C.-K. Park, C.-Y Wu, W. Higgins, Y. Kondratyeva, A. Molod, L. Takacs, M. Seablom, and R. Rood, 1995: A multiyear assimilation with the GEOS-1 system: Overview and results. NASA Tech. Memo. 104606, Vol. 6, NASA, Goddard Space Flight Center, Greenbelt, MD, 207 pp.

Schubert, S. D., R. B. Rood, and J. Pfaendtner, 1993: An assimilated dataset for earth science applications. Bul. Amer. Meteor. Soc., 74, 2331-2342.

Takacs, L.L., A. Molod, and T. Wang, 1994: Documentation of the Goddard Earth Observing System (GEOS) General Circulation Model‹Version 1, NASA Tech. Memo. 104606, Vol. 1, NASA, Goddard Space Flight Center, Greenbelt, MD, 100 pp.

[Table of Contents] [Previous] [Next]