Atmospheric Reanalyses – Recent Progress and Prospects for the Future

Created by gilbert.p.comp… on - Updated on 08/09/2016 11:46

Atmospheric Reanalyses – Recent Progress and Prospects for the Future.

A Report from a Technical Workshop, April 2010


Michele M. Rienecker, Dick Dee, Jack Woollen, Gilbert P. Compo, Kazutoshi Onogi, Ron Gelaro, Michael G. Bosilovich, Arlindo da Silva, Steven Pawson, Siegfried Schubert, Max Suarez, Dale Barker, Hirotaka Kamahori, Robert Kistler, and Suranjana Saha


In April 2010, developers representing each of the major reanalysis centers met at Goddard Space Flight Center to discuss technical issues – system advances and lessons learned – associated with recent and ongoing atmospheric reanalyses and plans for the future. The meeting included overviews of each center’s development efforts, a discussion of the issues in observations, models and data assimilation, and, finally, identification of priorities for future directions and potential areas of collaboration. This report summarizes the deliberations and recommendations from the meeting as well as some advances since the workshop.


Summary of Recommendations

From Rienecker et al. (2012)

Target areas for improvements for the next generation of atmospheric reanalyses include:

• The hydrological cycle

• The quality of the reanalyses in the stratosphere

• The quality of the reanalyses over the polar regions

• Representation of surface fluxes

• Observational bias corrections and/or cross-calibration across platforms

• Estimates of uncertainty in the analyses, and

• Reductions of spurious trends and jumps associated with the changing observing system.


Several recommendations were made regarding areas for coordination between reanalysis centers in order to prepare for the next reanalyses:

• Preparation and sharing of lists of anomalous behavior or features to help identify how common anomalies are across the various reanalyses.

• Examination of data utilization, including QC decisions, innovation statistics, bias corrections, outcomes of data selection algorithms, cloud detection outcomes, etc.

• Identification of joint experiments to be conducted to elucidate issues found to be in common in different reanalyses.

• Sharing of results from jointly designed sensitivity experiments.

• Coordination of input observations and ancillary data and centralization of the serving of these observations where possible.

• Expansion of ACRE’s efforts for contributing surface observations to 20CR to contributing to all future reanalysis efforts, possibly acting as a data coordinator and provider of surface data for all future reanalyses in collaboration with working groups of GCOS and WCRP.

• Development of innovative diagnostics and metrics to help quantify observational issues, the quality and also agreement of the reanalyses.

The workshop recommended that a mechanism be established for the timely exchange of information about the quality of the reanalyses, results of experiments, and plans for future developments. This idea was quickly embraced with the establishment of However, further progress is needed in the utilization of such a capability to enhance communications between reanalysis groups.

Finally, consistent with the Arkin et al. (2003) workshop report, the workshop participants recommended extending the reanalysis record for as long as possible, to include the 1970s for reanalyses focused on the satellite era, and to go back at least to 1850 with those reanalyses using sparse observations.


Arkin, P., E. Kalnay, J. Laver, S. Schubert, and K. Trenberth, 2003: Ongoing analysis of the climate system: A workshop report. NASA, NOAA, and NSF, 48 pp.  Link to Full Report.

Rienecker, M.M., D. Dee, J. Woollen, G.P. Compo, K. Onogi, R. Gelaro, M.G. Bosilovich, A. da Silva, S. Pawson, S. Schubert, M. Suarez, D. Barker, H. Kamahori, R. Kistler, and S. Saha, 2012: Atmospheric Reanalyses—Recent Progress and Prospects for the Future. A Report from a Technical Workshop, April 2010. NASA Technical Report Series on Global Modeling and Data Assimilation, NASA TM–2012-104606, Vol. 29, 56 pp. Link to Full Report

Dick Dee (not verified)

Tue, 07/30/2013 - 03:50

Hello again.. The 6-hourly temperature and wind analyses from ERA-Interim are 'snapshots' i.e. instantaneous values. You are probably aware however that they are consistent with the relatively coarse time-space resolution of the global model. Winds in particular represent model grid-cell averages and thus do not account for local small-scale variability. Near the surface the winds are consistent with the model's representation of topography which of course is much smoother than the real topography.

Please excuse me if this is posted to the incorrect area and please point me to a different location, if necessary.

I have a question about the ERA-Interim temporal resolution of the temperature and wind reanalyses. What do the 6-hourly values represent? For example, is the 00Z temperature a reanalysis of the instantaneous temperature at 00Z, or is it a 6-hour average? If it is a 6-hour average, is the average from 21Z to 03Z or 06Z-12Z or 00Z-06Z? I think it represents an instantaneous temperature, but I cannot find literature to support this assumption. Also, please comment on the wind parameters. Does the reanalysis represent instantaneous wind components or averages over 6 hours?

I'm sure this is published somewhere, but I have read many documents on the ECMWF website and a few papers, but I have been unable to ascertain the answer. Thanks kindly.

Dear icystorm, From the data FAQ: 8. What is the difference between analyses, forecasts and accumulated forecasts? ECMWF data can be split into 3 main categories: analyses, instantaneous forecasts and accumulated forecasts. Analyses are produced by combining short-range forecast data with observations to produce the best fit to both. The data are available a few times per day. Instantaneous forecast data are produced by the forecast model, starting from an analysis, and are available at various forecast steps (hours) from the analysis date/time. (Note, forecasts are not initiated from all analyses.) These data are relevant to a particular verifying date/time (analysis date/time plus step). Accumulated forecast parameters are accumulated from the beginning of the forecast. You can divide values by the length of the forecast step to calculate averages over the accumulation period. Some parameters are only analysed (eg. model bathymetry), some are only forecast (e.g. radiative fluxes) and some are both analysed and forecast (e.g. temperature, winds and pressure). If you look at the GRIB records, the internal metadata should also answer your question directly. best wishes, gil compo (University of Colorado/CIRES and NOAA/Earth System Research Laboratory/Physical Sciences Division)

Add new comment

This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.