The Earth Observer--March/April, 1995 Vol. 7, No. 2

Summary of Workshop on Results from the GEOS-1 Five-Year Assimilation

I. Introduction

A three-day workshop on results from the Data Assimilation Office (DAO) five-year assimilation (Schubert et al. 1993) was held March 6-8 at the Goddard Space Flight Center. The primary objective of the workshop was to provide timely feedback from the data users concerning the strengths and weaknesses of Version 1 of the Goddard Earth Observing System (GEOS-1) assimilated products. A second objective was to assess user satisfaction with the current methods of data access and retrieval.

There were a total of 49 presentations, with about half (23) of the presentations coming from scientists outside of Goddard. The total attendance was about 120. The first two days were devoted to applications of the data. These included studies of the tropical circulation; geodynamics applications; consituent transport; momentum and energy diagnostics; precipitation and diabatic heating; hydrological modeling and moisture transport, cloud forcing and validation, various aspects of intraseasonal, seasonal, and interannual variability, ocean wind stress applications; and validation of surface fluxes. The last day included talks about several related efforts at the National Meteorological Center (NMC), the National Center for Atmospheric Research (NCAR), the Center for Ocean-Land-Atmospheres (COLA), the United States Navy, and the European Centre for Medium-Range Weather Forecasts (ECMWF). This was followed by talks from several members of the DAO on current and future development efforts. The workshop concluded with a general discussion on data quality, data access experiences, and suggestions for future development of the GEOS system.

The workshop findings highlighted a number of strengths and weaknesses of the GEOS-1 data assimilation products. The following is an attempt to summarize these findings and outline the development activities designed to address the deficiencies. First, however, we shall briefly review the mission of the DAO, and how the five-year assimilation and the workshop fit into the overall development activities of the DAO.

The DAO has as its primary mission the development of a global assimilation system suitable for ingesting the Earth Observing System (EOS) and other satellite and in situ observations to produce dynamically, physically, and chemically consistent, gridded high-level data products for studying the Earth System. Such a synthesis of the various observations, which have widely differing error characteristics and irregular temporal and spatial coverage, is deemed vital to enhance the utility of the EOS satellite observations. In addition to synthesizing the available observations, data assimilation provides numerous value-added products; these include estimates of quantities or processes not readily observable (e.g., vertical motion, surface fluxes, and latent heating), and objectively determined estimates of the errors of the final synthesized data products. The emphasis on climate applications, and the large range of potential applications of the data, make it imperative that the DAO obtain feedback on the quality and usefulness of the data products from a broad spectrum of users.

In recognition of the need for timely user feedback, Version 1 of the GEOS system was fixed or "frozen" in March 1993, and a committment was made to generate a five-year assimilation (March 1985 - February 1990) suitable for climate applications. The idea of a reanalysis of historical observations with a fixed data assimilation system was first suggested a number of years ago (Bengtsson and Shukla 1988) as a way of improving the utility of current meteorological observations for climate studies by eliminating the spurious climate signals often found in operational NWP analyses. In addition to the DAO effort, there are a number of other reanalyses currently underway at NMC (Kalnay and Jenne 1991), ECMWF (Bengtsson and Shukla 1988), and the United States Navy. While some limited intercomparisons between these products have already been made (and presented at the workshop), the next year should see an abundance of studies addressing the quality of the various products. The DAO reanalysis is now completed through the end of 1993. It is planned to have the entire 15 year period (1979-93) assimilated with the GEOS-1 system by the end of 1995.

II. Workshop Results

The results of the workshop are summarized below in terms of the strengths and weaknesses of the assimilated product that have been identified by the various investigators. We discuss also the relevant development efforts that address the deficiencies. To help assess the quality of the products, various technical reports have been prepared documenting the current version of the GEOS data assimilation system (DAS). The GCM is described in Takacs et. al. (1994), Suarez and Takacs (1995), and Molod et al. (1995). Takacs et al. (1994) also provide a complete listing and description of all the diagnostic quantities available from the data assimilation system. The analysis system is described in Pfaendtner et al. (1995). Some early results and a user guide to all the datasets available from the five-year assimilation are provided in Schubert et al. (1995). All documents are available on-line or in hard copy form. Further information about data access may be obtained by sending e-mail to data@dao.gsfc.nasa.gov.

The primary strength of the GEOS-1 assimilation system lies in its ability to capture many of the key climate variations associated with El Niño and La Niña events, monsoons, droughts and other low frequency variations. This appears to be, in part, due to the incorporation of an incremental analysis update procedure (Bloom et al. 1995), which virtually eliminates the shocks associated with data insertion, and allows the model to respond gradually to the observations. Also, the GCM's physical parameterizations appear to respond quite realistically to the low frequency variations in boundary forcing. The results on low frequency variability include the following:

• The global signature of low frequency variations in the atmospheric moisture field compares favorably with SSM/I data. This signal appears to be captured better in the GEOS-1 assimilation than either the NMC reanalysis or the ECMWF operational products.

• The tropical zonal winds and pressure fields have been successfully used to describe and analyze various aspects of the large-scale circulation such as the Madden-Julian Oscillation and westerly wind bursts. The large scale tropical divergent wind field captures the evolution of the Madden-Julian Oscillation quite well.

• Various El Niño signals are successfully captured in the GEOS assimilation. These include a large cloud forcing anomaly in the central equatorial Pacific associated with 1986/87 El Niño that compares favorably with ERBE data. Also, the tropical Pacific precipitation anomalies associated with the 1987/88 El Niño/La Niña event compare favorably with outgoing longwave radiation (OLR) observations. The monsoon precipitation anomalies associated with the 1987/88 El Niño/La Niña event are also consistent with OLR observations.

• The reduced precipitation during the 1988 drought and the much enhanced precipitation during the 1993 wet conditions over the United States compare favorably with surface observations. In general, the monthly precipitation anomalies compare favorably with station observations, though there are cases where this is not true.

• GEOS-1 captures the seasonal placement of upper tropospheric moisture patterns quite well. The spatial correlations between GEOS-1 fields and observations (TOVS brightness temperatures) are similar to those found with ECMWF analyses.

A number of shorter term fluctuations are also well represented in the assimilation. These are primarily associated with fluctuations in the zonal wind and/or the boundary layer winds and surface stresses. Over land, these results indicate that the performance of the GCM's planetary boundary layer (PBL) parameterization generates very realistic wind fields, since the GEOS-1 DAS assimilates few wind observations below 850 mb. Over the oceans, the results suggest that both the surface wind/pressure analysis and the PBL parameterization are performing well. The relevant results are:

• The GEOS-1 equatorial winds appear to successfully capture the subdaily atmospheric tidal variations.

• Comparisons with length of day measurements suggest that the GEOS-1 momentum variations are well captured for periods as short as 8-9 days. Comparisons between NMC and GEOS-1 reanalyses shows coherence on all time scales longer than 3 days, which is shorter (better) than has been typically obtained from operational analyses. This suggests the reanalyses are in closer agreement than operational series have been in the past for this quantity.

• The horizontal winds, convective cloud mass flux, detrainment information, and PBL depth are of sufficient quality to to be used with some success as input to a 3-D tropospheric transport model for the tracers freon F-11 and radon Rn-222. In particular, Northern Hemisphere synoptic events are well captured; however, interhemispheric exchange is not well modeled.

• GEOS-1 wind stress provides good estimates of ocean transport through the Florida Straits. The GEOS wind stress generally provides improvements over the operational ECMWF results, but tends to overestimate amplitudes beyond about 10 days.

• Variations in the low level winds over the Great Plains are quite realistic despite the lack of observations going into the GEOS-1 DAS below 850 mb.

The following climate mean quantities are generally consistent with available verifying observations, and/or are consistent with or better than found in other analyses:

• The climate mean and seasonal evolution of the basic prognostic fields appear to be well captured in the GEOS analysis. Differences with ECMWF analyses over the Northern Hemisphere land masses are small. The largest differences occur over the tropics, and the Southern Hemisphere oceans, where observations are sparse and model bias is apparently playing a role (more on this below).

• The clear sky longwave flux and albedo are in good agreement with ERBE measurements.

• The general patterns of tropical convection and their seasonal evolution are consistent with available observations, but details of local maxima and amplitudes are not.

• GEOS -1 wind stress fields have been employed to force an ocean model in the North Pacific with some success, particularly in producing the subpolar circulation.

The greatest deficiencies in the GEOS-1 products are tied to biases in the humidity and cloud fields. There are several reasons for this. Moisture biases of the GCM are clearly playing a role, as well as exhibiting deficiencies in how the available moisture observations (currently only radiosonde) are being assimilated. One of the most disturbing aspects of the results is the manner in which the observations and model first guess appear to generate spurious feedbacks (Molod et al. 1995). A number of development activities are geared to addressing these deficiencies. For example, substantial improvements in the moisture field have been obtained with the assimilation of SSM/I observations. The introduction of downdrafts, a cloud water/ice scheme, improvements to the PBL (moist turbulence scheme, an improved mixed layer), further tuning of the convective parameterization, and the assimilation of relative humidity (instead of mixing ratio) should alleviate many of these problems. Improving the weaker-than-observed moisture gradients will likely require increased horizontal resolution. Some of the key moisture and cloud-related problems are:

• A much too wet upper troposphere (300 mb) over the Pacific Ocean compared with available observations. Also, the horizontal moisture gradients between very moist and very dry regions of the upper troposphere are too weak.

• The tropics and subtropics over the oceans are too dry compared with the vertically integrated moisture from SSM/I.

• Longwave and shortwave cloud radiative forcing (LCRF, SCRF) are overestimated over regions of deep convection in the ITCZ, especially during Northern summer.

• Middle-latitude LCRF and SCRF are weaker than ERBE especially over the storm tracks of both hemispheres. The vertical distribution of diabatic heating and vertical heat transport may also be too shallow in these regions (especially the North Atlantic).

• Low level coastal stratiform clouds are underestimated.

There are various problems with the precipitation, and near-surface temperature and humidity fields. Over land, these include substantial errors in the diurnal cycle. Some of these appear to be tied to the convective parameterization and should be remedied with the introduction of the changes outlined above. Improvements to the diurnal cycle and longer term impacts of soil moisture variations must await the introduction of a land surface model (currently being implemented). The known problems are as follows:

• Precipitation

• Summertime precipitation over eastern North America is overestimated.

• The amplitude of the diurnal cycle of the precipitation over the southeast U.S. is too large with little evidence of a nocturnal maximum over the Great Plains.

• Wintertime precipitation is too low over the Northern Rockies and along the southern coast of the United States.

• Over India, data in the gap between the western Ghats and the Bay of Bengal are missing.

• Too much rain occurs over continental Europe and northern Asia in July and too little over the Mediterranean during January.

• The assimilation produces too many days with small rain amounts and not enough days with no rain compared with station observations. The most intense rain amounts also appear to be underestimated. However, it was shown that this may not be the most meaningful comparison, since the observations represent point measurements while the values from the assimilation represent a grid square.

• The 1988 drought over the United States appears to extend too far into the summer with warmer-and-drier-than-observed conditions in July.

• The near-surface temperature is too cold over the northern United Sates and Canada during winter.

• Subtropical deserts are less reflective than in ERBE. The diurnal cycle of out-going longwave radiation over land shows significant phase shifts with respect to ERBE.

A number of other problems have been identified. For example, the model's zonal wind bias introduces a bias in the assimilation in data sparse regions. Current model experiments suggest that much of the westerly bias (and related cooling at high latitudes) can be eliminated with the introduction of gravity wave drag. Problems associated with noise at the poles have been addressed with the generalization of the dynamics module to allow rotation of the poles. Comparisons with NMC results suggest that increased vertical resolution in the PBL should improve the representation of the low level winds. These, and other problems, are noted in the following list of deficiencies:

• The Hadley cell appears to be too weak during Northern winter (based on comparisons with ECMWF and NMC).

• There is a zonal mean bias at 40S-50S in sea level pressure and the zonal wind.

• The northward component of the summertime low level wind over the Great Plains appears to be underestimated.

• Various quantities including the temperature and sea level pressure show evidence of noise near the poles.

• The magnitude of the wind stress appears to be underestimated over the North Pacific.

III. Data Access and Retrieval

Feedback on experience with data access and retrieval was rather limited. One of the major findings, however, was that most users were accessing the data directly from the NCCS Unitree system, instead of from the Goddard DAAC. A large number of users have also obtained the monthly means from the local DAO server. Most users were satisfied with the access from the NCCS Unitree system which is geared to high speed near-online (robotically-controlled) data access. Complaints were generally associated with data organization and occasional glitches with the retrieval system. Those accessing the monthly means from the DAO server were also generally happy with this type of access, and especially the online documentation; however, the large sizes of the files gave some users problems due to their limited local storage capabilities. For those people obtaining the data from the Goddard DAAC, the experiences were somewhat mixed. Some of the early users had difficulties obtaining the data in a timely manner. The problems were primarily associated with "growing pains" of the DAAC, and the more recent users appear to be satisfied with DAAC performance. The Goddard DAAC received very high marks for responsiveness to user concerns and suggestions. From a more general perspective, there is a concern that the mode of access provided by the DAACs (search and order) is not satisfactory for many users accustomed to high-speed interactive access.

The major complaints centered on the large file sizes of the DAO output and the lack of a subsetting capability. This is an important issue which the DAACs and data providers must address. The options range from providing a more efficient organization of the files (including smaller file sizes), to providing several popular versions (organizations) of the data, to providing an on-demand subsetting capability. The first option may not satisfy enough of the users, while the last option would very likely quickly overwhelm the current resources of the DAACs.

Acknowledgments: Funding was provided by NASA Headquarters through the support of Kenneth Bergman. Jean Rosenberg (workshop coordinator) provided excellent logistical and technical support. Travel arrangements were made by Denise Dunn of the Universities Space Research Association.

References

Bengtsson, L., and J. Shukla, 1988: Integration of space and in situ observations to study global climate change. Bull. Amer. Meteor. Soc., 69, 1130-1143.

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

Kalnay, E., and R. Jenne, 1991: Summary of the NMC/NCAR reanalysis workshop of April 1991. Bull. Amer. Meteor. Soc., 72, 1897-1904.

Molod, A., H.M. Helfand and L.L. Takacs, 1995: The climate of the GEOS-1 GCM and its impact on the GEOS-1 data assimilation system. J. Climate, 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. No. 104606, volume 4, Goddard Space Flight Center, Greenbelt, MD 20771.

Schubert, S. D., R. B. Rood and J. Pfaendtner, 1993: An assimilated data set for Earth science applications. Bull. Amer. Meteor. Soc., 74, 2331-2342.

Schubert, S., 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. No. 104606, volume 6, Goddard Space Flight Center, Greenbelt, MD 20771.

Suarez, M.J., and L.L. Takacs, 1995: Documentation of the Aries-GEOS Dynamical Core, Version 2. NASA Tech. Memo. 104606, volume 5, Goddard Space Flight Center, Greenbelt, MD 20771.

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. No. 104606, volume 1, Goddard Space Flight Center, Greenbelt, MD 20771.

[Table of Contents] [Previous] [Next]