Dataset Description

The NA-CORDEX data collection comprises output from different regional climate models (RCMs) at various spatial resolutions driven by the ERA-Interim historical reanalysis and by different global climate models (GCMs) using different representative concentration pathways (RCPs) for the future composition of the atmosphere. The various combinations of RCM, GCM, RCP, and spatial resolution that have been simulated are shown in the Simulation Matrix.

The NA-CORDEX data collection archives output from these simulations for a number of variables of interest to end users, especially those who work with climate change impacts. The full list of variables is available on the Variable List page.

Each variable is stored at the daily timescale. The collection also includes monthly, seasonal, and annual averages of each variable, as well as monthly and seasonal climatologies (multi-year averages). In addition, the archive also includes sub-daily precipitation at the highest frequency available.

The model outputs are stored on the native grid used for simulation, which has a consistent spatial resolution across the simulation domain. In addition, the archive also includes a version of each dataset that has been interpolated to a common lat-lon grid at roughly equivalent spatial resolution. The interpolation has been performed using the ESMF "Patch" algorithm.

The archive also includes bias-corrected version of each GCM-driven dataset. These datasets have been adjusted using a multivariate quantile-mapping method (Cannon's MBCn algorithm) against a gridded daily observational dataset. Most impacts users find bias-corrected data more useful than the raw model outputs. See below for more details.

Dataset Facets

Each data file in the NA-CORDEX collection is named "[var].[scen].[gcm].[rcm].[freq].[grid].[bc].nc". The components vary as follows:

  • var: Variable
    See the variable list for names of variables. We have published all essential and high-priority variables, as well as some aspirational variables.

  • scen: Scenario
    ERA-Interim runs are given a scenario named "eval" (for evaluation) for the sake of regularity. The historical ("hist") scenario runs 1950-2005; future (RCP-based) scenarios run 2006-2100.

  • gcm: GCM (driver)
    The source of the boundary conditions for the RCM. Usually a GCM, but includes the ERA-Interim reanalysis.

  • rcm: Regional Climate Model

  • freq: Temporal frequency
    Sub-daily data is available only for precipitation for certain RCMs. "ymon" and "yseas" are multi-year climatological averages.

  • grid: Grid and spatial resolution
    The grid names follow the CORDEX convention for domain naming. The NAM-44, NAM-22 and NAM-11 indicate 0.44°, 0.22°, and 0.11° native rotated-pole grids, respectively; the "i" suffix indicates that the data has been interpolated to a common half/quarter/eighth-degree lat-lon grid. RegCM4 and WRF simulations with a grid value of 44 or 22 are actually using a 50-km or 25-km projected grid, which is roughly equivalent in resolution to the 0.44° and 0.22° grids.

  • bc: Bias-correction
    "Raw" data is uncorrected model output. "MBCn" data has been bias-corrected using Cannon's MBCn algorithm against the gridMET or Daymet gridded observational datasets. The gridMET dataset uses climatically aided interpolation to blend data from PRISM with data from the NLDAS-2 reanalysis; it includes surface winds, but only covers CONUS. The Daymet dataset interpolates and extrapolates GHCND station data using purely statistical methods; it does not include surface winds but covers the entire United States, Canada, and Mexico. Both datasets include tmin, tmax, precip, specific humidity, and incoming solar radiation. We also derive relative humidity and tmean (the average of tmin and tmax) from other variables.
    NOTE: A previous bias-correction ("mbcn-METDATA") has been deprecated; it used an older version of gridMET and had a sign error in the surface winds.