Project Description

Reanalysis Scientists are interested in reproducing the success of the Coupled Model Intercomparison Project (CMIP5) to study reanalysis differences and uncertainties to improve reanalyses. Also, reanalysis data, essentially a re-forecast of past weather using the latest forecast models, allow interdisciplinary scientists to compare their datasets (e.g. biodiversity, water planning, wind power) with 30 or more years of gridded climate data. Both research efforts require large datasets of monthly and hourly data, formatted identically to facilitate comparisons. The Climate Model Data Services (CDS) collaborated with the five major reanalysis projects to collect this data and present it through their range of services: Distribution, Visualization, Analytics, and Knowledge, resulting in the Collaborative REAnalysis Technical Environment – Intercomparison Project (CREATE-IP). Data was processed up to 2020 for most reanalyses data products. As of 2022, we recommend using ESMValTool available from DLR to process data after 2020.

We have identified five areas where our contributions could benefit the research community’s reanalysis intercomparison efforts. These areas include the following:

1. Co-Location and improved methods of discovery — Researchers need a one-stop shopping place to go to find the best reanalysis to use, metadata information and search tools to identify variables of interest, and multiple interfaces to distribution, visualization, and analytics tools to support a range of researcher expertise. Collecting the full set of reanalysis data in one place with regular updates will save researchers time and resources and significantly reduce the time spent finding and gathering data.
2. Common data formats — Collaborating with the reanalysis centers to provide a common interface to reanalysis data is an important step.  There are a number of sites that document the available reanalyses (e.g. http://www.reanalyses.org/) and each of the major reanalysis centers provides access to their data in various formats with differing organization. For example, the ECMWF-interim data provides the fluxes 4 times/day and the data must be summed to produce the proper monthly average. CFSR data is distributed in grib format with an OPeNDAP interface but the organization of the diagnostic and prognostic variables is not the same as other OPeNDAP environments. This emphasizes the need for a common interface with data in a standardized format.


3. Access to internal metrics — Each reanalysis produces a set of internal metrics referred to as observations minus forecast (O-F) and observations minus analysis (O-A) and are used to help understand how much information comes from the observations and how much comes from the model. These metrics are not routinely shared among the reanalysis centers and not generally formally published (Gregow, 2014). The metrics are referred to as innovations and are essentially the corrections applied to the model to bring it back in line with the observations. They are also useful in determining observational biases. 
   
   Providing the research community with these metrics from various reanalyses may help with understanding why climate models produce certain systematic errors. A simple preliminary example of the concept can be found in Phillips et al. (2004) where they initialized a climate model with reanalysis data and studied the error progression to understand initial systemic errors. If the internal metrics had been available from several reanalyses, it is possible that considerable more insight would have been obtained about the climate model biases.


4. Uncertainty quantification — A further value of reanalysis intercomparison is outlined in Gregow (2014). One of the key arguments for such a project is identification of reanalysis uncertainty. There is a basic need to enable users to “conduct and interpret intercomparisons tailored to their own applications.” This approach, although more complicated than providing users with a category of appropriate use of each field (Kalnay et al., 1996) will, in the long run, be a much more valuable method of assessing the appropriate use of reanalysis for both climate scientists and interdisciplinary researchers.


5. Broader use by other communities — Due to the ability to take observations from a wealth of sources and generate a gridded data set of climate variables over 30 years or more, reanalysis presents a unique opportunity to compare known climate conditions to related scientific phenomena.  Not only does this provide the opportunity to validate climate models, it also allows scientists from other disciplines to incorporate reanalysis data into investigations of subjects such as biodiversity, water supply, including droughts and floods, agriculture, wind power, and pandemics.  The majority of these interdisciplinary scientists are not familiar with climate data analysis tools or file formats (e.g. netcdf) and will benefit from data services that provide access through protocols that integrate with their existing tools, for example, using ArcGIS with georeferenced data. 

To support the above areas, CREATE-IP collected all the available reanalysis data in one place, formated it following the standards set out by the Earth System Grid Federation (ESGF), and provided improved access to these data via multiple services. CREATE-IP supports improvements in reanalysis research by providing the ability to study the differences and similarities between the existing reanalyses. 

References:

Gregow, H. (2014). COordinating Earth observation data validation for RE-analysis for CLIMAte ServiceS Procedure for comparing reanalyses , and comparing reanalyses to assimilated observations and CDRs, (July), 1–45. http://www.coreclimax.eu/sites/coreclimax.itc.nl/files/documents/Deliverables/WP_Reports/Deliverable-D553-CORECLIMAX.pdf

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., … Joseph, D. (1996). The NCEP/NCAR 40-Year Reanalysis Project. Bulletin of the American Meteorological Society, 77, 437–471. doi: http://dx.doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2

Phillips, T., Potter, G., Williamson, D., Cederwall, R., Boyle, J., Fiorino, M., … Yio, J. (2004). EVALUATING PARAMETERIZATIONS IN GENERAL CIRCULATION MODELS Climate Simulation Meets Weather Prediction BY. Bulletin of the American Meteorological Society, 85(December), 1903–1915. doi: http://dx.doi.org/10.1175/BAMS-85-12-1903