CISL’s RDA chosen for new WCRP reanalysis intercomparison

By Brian Bevirt
08/30/2015 - 12:00am

Masatomo Fujiwara of Hokkaido University in Sapporo, Japan leads a program to understand differences between the world’s leading global atmospheric reanalysis datasets, and his team determined that almost all of the datasets for this intercomparison project are available in CISL’s Research Data Archive (RDA). The SPARC Reanalysis Intercomparison Project (S-RIP) is coordinated by one of the core projects of the World Climate Research Programme (WCRP), named Stratosphere-troposphere Processes And their Role in Climate (SPARC).

“Reanalysis datasets with varied data formats and different access protocols are archived at various reanalysis centers around the world, and some older reanalysis datasets are no longer archived at some centers,” explains project lead Masatomo Fujiwara. “The NCAR RDA houses key climate datasets and is extremely useful for worldwide data users to obtain both older and newer reanalysis datasets.”

The S-RIP project is comparing reanalysis data sets, investigating the causes of differences among reanalyses, helping researchers use reanalysis products in scientific studies, and working to improve future reanalysis products by collaborating with reanalysis centers. S-RIP collaborators have selected the RDA as the primary data source for this project: of the 12 reanalysis datasets being used, researchers are directed to the RDA for 11 of them.

The RDA’s reanalyses are produced in high resolution by the world-leading meteorological institutions ECMWF, JMA, and NOAA. CISL staff recomputes these reanalysis products onto latitude-longitude grids to make them more convenient for climate and weather research. Another advantage to the S-RIP project is that the NCAR RDA offers a standardized suite of user-friendly services for high-throughput data access and retrieval. “We provide this archive and its services to support the NCAR and UCAR University research communities,” said Steve Worley (CISL Data Support Section Manager), “and the information is available to researchers everywhere via the Internet.” Added DSS database engineer Doug Schuster, “Researchers with login credentials for NCAR supercomputers have a speed advantage, however. They can access RDA data directly from spinning disks on GLADE, so they can bypass the data download step in their data analysis workflow.”

Because they are estimates, and because they are done with different assimilation systems on different computers using various models and resolutions, it is beneficial to cross-evaluate reanalyses in projects such as S-RIP. This multi-year WCRP effort involves nearly 70 collaborators from around the world. In addition to peer-reviewed journal papers by several groups of the collaborators, an interim report is planned for 2015 that covers basic intercomparisons. The full S-RIP report is planned to be completed in 2018. This effort will comprehensively compare reanalysis data sets for key diagnostics to help understand the causes of differences. The results will be used to provide guidance on appropriate usage of various reanalysis products in scientific studies. In addition, the reanalysis community will benefit from coordinated user feedback, which can lead to improvements in the next generation of reanalysis products.

Temperature response to ENSOThe annual temperature response to El Niño Southern Oscillation (ENSO) for each of nine reanalyses (named above each plot), obtained using multiple linear regression. The regression considers volcanic eruptions, ENSO, the Quasi-Biennial Oscillation, and the 11-year solar cycle as well as seasonal variations. The units shown by the contour lines are in K per standard deviation of the ENSO index, then multiplied by the difference between the warmest and the coldest ENSO events. The solid curves are for positive values, and the dotted curves are for negative values. So, the tropical troposphere shows warming signals, while tropical stratosphere mostly shows cooling signals. [Note that some reanalyses cover the region up to the 0.1 hPa level (about 65 km), and some cover only up to the 10 hPa level (about 32 km).] Shaded regions in the plots denote statistical significance at the 95% (lighter) and 99% (darker) levels, and these are the prime regions to be compared and contrasted.

The tropical signals are highly statistically significant and remarkably consistent between the reanalysis datasets. The primary signals consist of anomalously positive temperatures in the tropical troposphere of up to 1 K and anomalously negative temperatures in the tropical stratosphere of down to -1.5 K.

—Figure from Mitchell, D.M., L.J. Gray, M. Fujiwara, T. Hibino, J.A. Anstey, W. Ebisuzaki, Y. Harada, C. Long, S. Misios, P.A. Stott, and D. Tan (2015): Signatures of naturally induced variability in the atmosphere using multiple reanalysis datasets. Quarterly Journal of the Royal Meteorological Society, doi: 10.1002/qj.2492.