SODA logo

Data Structure, Access

Downloading the reanalysis files

To download: go to the pages of links to the individual ensemble members:

SODA3.3.0NO MERRA2* Large-Yeager 1980-2015
SODA3.3.1YES MERRA2* Large-Yeager 1980-2015
SODA3.3.2 YES MERRA2*COARE41980-2015
SODA3.4.0NO ERA-I*Large-Yeager 1980-2015
SODA3.4.1YES ERA-I*Large-Yeager 1980-2015
SODA3.4.2YES ERA-I*COARE41980-2015
SODA3.5.1YES ERA20C*COARE41980-2010
SODA3.6.1YES CORE2Large-Yeager 1980-2009
SODA3.7.0NO JRA-55*COARE41980-2013
SODA3.7.2YES JRA-55*COARE41980-2013
SODA3.8.1YES 20CRv2*Large-Yeager 1980-2013
SODA3.9.1YES OAfluxLarge-Yeager 1980-2015
SODA3.10.1YES CFSRRLarge-Yeager 1980-2015
SODA3.11.0NO DFS5.2COARE41980-2015
SODA3.11.2YES DFS5.2COARE41980-2015
(* indicates there has been flux-correction. Flux correction has not been applied to reanalyses such as CORE2 and DFS5.2, which are themselves corrections of other atmospheric reanalyses. Experiment members marked in grey are not yet available)

File naming conventions

Files are given names of the form:
  • 3.X indicates which meteorological forcing set used:
    • 3.X=3.3 indicates MERRA2
    • 3.X=3.4 indicates ERA_int ,
    • 3.X=3.5 indicates ERA20C ,
    • 3.X=3.6 indicates CORE2 ,
    • 3.X=3.7 indicates JRA-55 ,
    • 3.X=3.8 indicates 20CRv2 ,
    • 3.X=3.11 indicates DFS5.2 ,
  • x, the experiment number, indicates whether assimilation is included and the nature of the forcing:
    • x=0 indicates the simulation
    • x=1init indicates assimilation reanalysis prior to flux bias correction using Large-Yeager bulk formula
    • x=1 indicates assimilation reanalysis with flux bias correction using Large-Yeager bulk formula
    • x=2 indicates assimilation reanalysis generally with flux bias correction using COARE4 bulk formula (COARE2 and DFS5.2 forcings are uncorrected)
  • Z indicates the time averaging and sampling interval
    • Z=mn indicates calendar monthly average.
    • Z=5dy indicates 5-day average.
    • Z=10dy indicates 10-day interval (for increments).
  • V indicates which set of variables is included, and on what vertical grid.
    • V=ocean indicates that the file constains 3D fields of temperature, salinity, and velocity (specified at 50 levels), and a series of 2D fields, including such variables as surface stress, mixed layer depth, and SSH.
    • V=isopycn indicates that the file contains temperature and salinity remapped to constant potential density surfaces.
    • V=ice indicates the file contains a series of sea ice-related fields.
    • V=transp indicates the file contains 3D vector volume transport (only available on the original grid).
    • V=increm indicates that the file constaint the 3D temperature and salinity increments (e.g. [H-1(To-HTf)]) computed by the data assimilation.
  • G indicates the horizontal grid on which the data is presented.
    • G=reg the reanalysis has been remapped onto a uniform 1/2x1/2 Mercator coordinate horizontal grid with 720x330 grid points, and with the first point at (74.75S, 0.25E).
    • G=or the reanalysis has been provided on the original model grid with 1440x1070 grid points in the horizontal and with the first point at (86.55S, 279.875W).
    • G=1x1 indicates a field mapped onto a uniform 1x1 degree Mercator coordinate grid.
  • YYYY indicates the date.
    • YYYY = year (4 digits) for monthly files (so each file contains 12 month time samples).
    • YYYY = year_mo_dy (4y,2m,2d digits) for 5dy files (so each file contains a single 5dy time sample).

    Native and remapped grids

    The native interlaced horizonal velocity and conserved tracer (e.g. temperature and salinity) grids form a tripolar Arakawa-B grid, varying from 0.1°x0.25° at high latitude to 1/4°x1/4° in the tropics (quasi-isotropic grid spacing increases from ~11.7km at 65 latitude to ~28.0km at the Equator, 1440x1070 grid points). The (1.5Gb sized) topography map (created by Whit Anderson of GFDL) is here. The topography file contains a complete definition of the model grid (including locations of the temperature, salinity, and velocity points), the sizes and angles of the grid quadralateral boxes (they are not cubes because of the distortion required by the displaced north pole), and the depths of the bottom cells (for better resolution of the topography the model employs partial bottom cells). There are a few errors in this topography at high latitude and in the Indonesian throughflow which were introduced by errors in the original topography data set and also exist in the GFDL CM2.5. The topography has also undergone slight massaging, for example to open up Nares Strait (pointed out to us by Yury Vikhliaev).

    The netcdf v4 variable ordering is: {time, depth, latitude, longitude} written out surface to bottom, south to north (velocity begins at 79.968°S while tracers begin at 80.021°S), and west to east (velocity begins at 279.75°E while tracers begin at 279.875°E).

    The vertical grid consists of 50 levels at telescoping depths (z* coordinate, z* increases downward). Horizontal velocity and tracers are specified at z* = {5.03355m, 15.10065, 25.21935, 35.35845, 45.57635, 55.86325, 66.26175, 76.80285, 87.57695, 98.62325, 110.0962, 122.1067, 134.9086, 148.7466, 164.0538, 181.3125, 201.2630, 224.7773, 253.0681, 287.5508, 330.0078, 382.3651, 446.7263, 524.9824, 618.7031, 728.6921, 854.9935, 996.7153, 1152.376, 1319.997, 1497.562, 1683.057, 1874.788, 2071.252, 2271.323, 2474.043, 2678.757, 2884.898, 3092.117, 3300.086, 3508.633, 3717.567, 3926.813, 4136.251, 4345.864, 4555.566, 4765.369, 4975.209, 5185.111, 5395.023}.

    General information about the grids as well as other model characteristics are available in the document: Code downloads are available through and through github.

    Regridded 1/2°x1/2°x50lev 3D ocean files

    Regridding from the original staggered ~1/4°x1/4° tripolar horizontal grid described above onto a single uniform 1/2°x1/2°x50lev Mercator grid is done using the GFDL Flexible Modeling System (FMS) 'regrid' routine with conservative remapping for temperature and salinity and nonconservative remapping for velocity. 3D arrays are thus 720x330x50 ≈ 12Mb with the starting point at: 0.5E, 74.75S, 5.03355m depth. Units are MKS. Projection onto isopycnal surfaces is carried out using the linear interpolation routines contained in GRADS. Many files are converted to calendar monthly averages, which is carried out post run-time either using GRADS routines or NCO routines. The ocean files [written in 32-bit netcdf v4 classic] contain seven two dimensional fields:

    • ssh [= sea level + Pa/(rho0*g)]: MOM5 computes sea level from a complete conservation of mass equation (MOM5_elements.pdf section 38.2). This means sea level includes both steric (changes in column density) effects and eustatic (changes in column mass) effects. Currently SODA3 does not add mass due to continental ice melt, and does not include the impact of self-gravitation. Also note that the ocean bottom elevation in MOM5 is fixed and thus we do not include the effects of glacial isostatic adjustment or ocean loading.
    • {mlt,mlp,mls}: mixed layer depth calculated either as the depth where temperature differs from surface (z=5m) temperature 0.2K, or as the depth where density exceeds surface density by 0.03 kg/m3 , or where salinity differs from surface salinity by 0.01 psu.
    • anompb: bottom pressure minus change in hydrostatic pressure from z=0 to z=-H assuming a constant density rh0 [dbar=100hPa]
    • {taux,tauy}: wind stress calculated from bulk formulas in the model, e.g. accounting for ocean motion. [N/m^2]

    The ocean files also contain six three dimensional fields specified at the depths listed above.

    • temp: potential temperature [°C]
    • salt: SODA3 treats salinity as practical salinity (PSS-78) as expressed in Practical Salinity Units [psu]. The model uses preformed salinity as its salinity state variable and uses the TEOS10 equation of state. Absolute salinity is, on average, 0.4% larger than practical salinity.
    • {u,v,w}: horizontal and vertical components of velicity. w is computed diagnostically from conservation of mass and is evaluated on a different grid. [m/s]
    • prho: potential density computed from potential temperature, practical salinity, and pressure using the TEOS-10 formulas. [kg/m^3]

    2D sea ice files

    The sea ice files [written in netcdf (NetCDF3)] contain eight two-dimensional fields. Further description of the Sea Ice Simulator and its variables is provided at:

    • hi: sea ice thickness [m ice]
    • mi: sea ice mass [kg/m^2]
    • hs: snow thickness [m snow]
    • {cn1,cn2,cn3,cn4,cn5}: sea ice concentration [0:1] in five ice thickness classes

    3D ocean mass transport files

    The mass transport files [written in netcdf (NetCDF3) on the original grid] contain two two-dimensional fields and two three-dimensional fields

    • {mxvint,myvint}: (2D) total column vector horizontal mass transport across the face of a grid cell vertical column [10^9 kg/s ~ Sv]
    • {mx,my}: (3D) vector horizontal mass transport across the face of each grid cell [10^9 kg/s ~ Sv]

    Citing the data

    When using the data set in a paper, please provide the SODA3 ensemble member (e.g. SODA3.3), the date/site the data were downloaded from, and cite the following reference:

    Carton, J.A., G.A. Chepurin, and L. Chen, 2018: An updated reanalysis of ocean climate using the Simple Ocean Data Assimilation version 3 (SODA3), manuscript in preparation.


Jackett, D.R., T.J. McDougall, R. Feistel, D.G. Wright, S.M. Griffies, 2006: Algorithms for density, potential temperature, conservative temperature and the freezing temperature of seawater J. Atmos. Ocean. Technol., 23,1709-1728.