NCCS Computing Capabilities Enable Calculating a Multi-Reanalysis Ensemble


Reanalysis is a comprehensive record of how weather and climate are changing over time. Objectively combining up to tens of billions of observations with a computer model of the atmosphere generates a synthesized estimate of the Earth’s climate system.

The NCCS Collaborative REAnalysis Technical Environment (CREATE) has collected all of the major modern reanalyses in one site and prepared them for distribution to researchers around the world:

  • Modern-Era Retrospective analysis for Research and Applications (MERRA) and MERRA-2 — NASA
  • Climate Forecast System Reanalysis (CFSR) — National Oceanic and Atmospheric Administration (NOAA)
  • Japanese 55-year Reanalysis (JRA-55) — Japan Meteorological Agency
  • ECMWF Interim Reanalysis (ERA-Interim) — European Centre for Medium-Range Weather Forecasts (ECMWF)

CREATE’s objective is to repackage reanalysis data originally produced in a variety of formats and structures. Each dataset requires some customization using the NCCS ADAPT computing system to produce a standardized product for distribution through the Earth System Grid Federation (ESGF) and THREDDS, both hosted on the NCCS Dataportal.

The CREATE-V interface allows exploring data from all of the major modern reanalyses. Shown here is air temperature on September 19, 2018 from the MERRA-2 reanalysis.

Scientists often question which reanalysis is best for their own use. The NCCS CREATE team has prepared the Multi-Reanalysis Ensemble v2 (MRE2) average of key variables. Often MRE2 performs better than any single member, much like hurricane path ensembles outperform a single member.

An example in the figures below shows average precipitation compared to observations from the Global Precipitation Climatology Project (GPCP). The more blue the map, the better the representation. MRE2 (bottom right) is closer to the observations of precipitation than any one of the reanalyses.

Maps of the standard deviation of the difference between each reanalysis and observed precipitation from the Global Precipitation Climatology Project (GPCP). The darker blue depicts a better representation of the observed precipitation.

Preparing MRE2 requires NCCS ADAPT computing resources to regrid each reanalysis to a common grid and calculate the ensemble average and standard deviation.

Time history of the global standard deviation from the GPCP. The lower the number, the better the representation. Red is the Multi-Reanalysis Ensemble v2 (MRE2).

NCCS staff announced the MRE2 products at the 2018 AGU Fall Meeting in December 2018 and the 99th AMS Annual Meeting in January 2019.