GRACE MONTHLY MASS GRIDS - LAND
NEW LAND GRID DATA VERSION (02/14/2014)
Current data version: [RL05.DSTvSCS1401]
Please download ALL MONTHS from these new solutions
and replace previous versions to work with a consistent time series
LAND DATA PROCESSING
- The land data are based on the RL05 spherical harmonics from CSR, JPL and GFZ
- The C20 (degree 2 order 0) coefficients are replaced with the solutions from Satellite Laser Ranging [Cheng et al., 2011]. The C20 values derived from GRACE observations have a larger uncertainty than the SLR-values.
- The degree 1 coefficients ( geocenter) are those derived by Swenson, Chambers, and Wahr (2008).
- A glacial isostatic adjustment (GIA) correction has been applied based in the model from Geruo A and J. Wahr (2013).
- A destriping filter has been applied to the data, to minimize the effect of an error whose telltale signal are N-S stripes in GRACE monthly maps.
- A 300 km wide gaussian filter has also been applied to the data (please note: make sure that the Gain Factor file is the correct 300km version!).
The spatial sampling of all grids is 1 degree in both latitude and longitude (approx. 111 km at the Equator). However, this does not mean that two neighboring grid cells are 'independent' because a spatial smoothing has been applied.
LAND GRID SCALING
Due to the sampling and post-processing of GRACE observations, surface mass variations at small spatial scales tend to be attenuated. Therefore, USERS SHOULD MULTIPLY THE GRCTellus LAND DATA BY THE PROVIDED SCALING GRID. The scaling grid is a set of scaling coefficients, one for each 1 degree bin of the land grids, and are intended to restore much of the energy removed by the destriping, gaussian, and degree 60 filters to the land grids. To use these scaling coefficients, the time series at one grid (1 degree bin) location must be multiplied by the scaling factor at the same 1 degree bin position. The netcdf file with gain factors is CLM4.SCALE_FACTOR.DS.G300KM.nc in the netcdf directory , and it must be applied to the GRACE grids in the same directory (an identical grid in ascii format can be found in the ascii directory) .
The scaling coefficients were computed by applying the same filters applied to the GRACE data to a numerical land-hydrology model (NCAR's CLM4). In a nutshell, the gain coefficient is the multiplicative factor that minimizes the difference between the model's smoothed and unfiltered monthly water storage variations at any geographic location. The coefficients are independent of the GRACE data proper, hence they are provided as a separate file. Furthermore, the gain factors tend to be dominated by the annual cycles of water storage variations, and may thus not be suitable to quantify trends from the GRCTellus land data. While the dependence of the gain factors on the specific land model used is generally small, please note that inter-annual trends in particular in hydrology models are very uncertain.
For a detailed description of the data processing, gain factor derivation & caveats, please see DATA PROCESSING and CAVEATS DESCRIPTION FOR LAND GRIDS (PDF, 3.68 MB).
NOT SUITABLE FOR CRYOSPHERIC STUDIES
The current GRCTellus Land grids are not suitable to accurately quantify ice mass changes over Greenland or Antarctica, or glaciers and ice caps. These regions require region-specific averaging kernels, as well as proper treatment of signal contamination from nearby land hydrology and adjusted GIA effects. We recommend the paper by Jacob et al. (Nature 2012, full citation below), for a thorough discussion of these aspects.
The units of the 'equivalent water thickness' grids are cm of water thickness. The units of the error grids are cm. The scaling factors are dimensionless. If each grid node is g(x,y,t) where x is longitude index, y is latitude index, t is time (month) index, and the scaling grid is s(x,y), then the time series is simply
g'(x,y,t) = g(x,y,t)*s(x,y)
These grids have 360 longitudes (0.5,1.5,2.5,...,359.5), and 180 latitudes (-89.5, -88.5, ..., -0.5, +0.5, ...+89.5). However, missing grid points are not included in the ascii files
The data are provided in
- NETCDF format, suitable for automatic ingestion into several software packages.
- ASCII, a plain text format (compressed with gzip)
- Error estimates due to the measurement and errors due to leakage are also provided, in separate files (ascii) or together with the scaling coefficient file (netcdf).
To compute error estimates for the scaled values, two additional grids are provided (as separate ASCII files or in the same netCDF file as the scaling coefficients).
1. The errors given in CLM4.*.DS.G200KM.txt are in centimeters (same as the GRACE data).
2. The measurement errors have already been scaled so no further multiplication is necessary.
3. The leakage errors are residual errors after filtering and rescaling, such that the total error in Total Water Storage for a given grid pixel is:
total_err_pix = sqrt(leakage_err_pix^2+measurement_err_pix^2).
4. The errors in nearby pixels are correlated. Therefore, if the total error in a region of adjacent pixels is desired, this covariance needs to be considered. Here is pseudo-code to get the total leakage (lerr) and measurement (merr) errors for a region:
var_merr = 0. ; measurement error
var_lerr = 0. ; leakage error
betam = 300. ; km ~ measurement error de-correlation length
betal = 100. ; km ~ leakage error de-correlation length
for i=0, npix-1 do begin
for j=0, npix-1 do begin
dist = sqrt((lon[i]-lon[j])*cos(lat[i]))^2.+(lat[i]-lat[j])^2.) * (pi/180) * 6371. ; lon, lat in degs, dist in km
expdbm = exp(-(dist^2.)/(2.*betam^2.))
expdbl = exp(-(dist^2.)/(2.*betal^2.))
var_merr = var_merr + merr[i] * merr[j] * expdbm
var_lerr = var_lerr + lerr[i] * lerr[j] * expdbl
sigma_merr = sqrt(var_merr)/npix
sigma_lerr = sqrt(var_lerr)/npix
TIME AVERAGE REMOVED FROM MONTHLY SOLUTIONS
Each monthly GRCTellus grid represents the surface mass deviation for that month relative to the baseline average over Jan 2004 to Dec 2009. If you compare against other data or model, it is critical that anomalies relative to the same time-average are compared. This is simple to do: for example, if the new baseline is 2004-2006, average the GRCTellus grids over 1/2004 to 12/2006, and subtract this average grid from all other monthly grids.
TIME SPAN OF GRCTellus MONTHLY SOLUTIONS
'Monthly' is used somewhat loosely: please see the TABLE OF ACTUAL GRACE DATA DAYS
used for each 'monthly' solution. Note that from 2011 on, the GRACE instruments are periodically turned off due to battery management.
BROWSE IMAGES and NUMERIC DATA
The LAND gridded data and browse images are available here
ACKNOWLEDGEMENT and CITATION
When using any of these data, please acknowledge
GRACE land data (available at http://grace.jpl.nasa.gov) processing algorithms were provided by Sean Swenson, and supported by the NASA MEaSUREs Program;
Landerer F.W. and S. C. Swenson, Accuracy of scaled GRACE terrestrial water storage estimates. Water Resources Research, Vol 48, W04531, 11 PP, doi:10.1029/2011WR011453 2012.
Swenson, S. C. and J. Wahr, Post-processing removal of correlated
errors in GRACE data, Geophys. Res. Lett., 33, L08402,
If you encounter any problems with the data, please contact the person listed at bottom right.
ADDITIONAL REFERENCES used above:
Cheng, M. and Tapley, B.D.: Variations in the Earth's oblateness during the past 28 years, J. Geophys Res v109, B9, 2004
Jacob T., J. Wahr, W.T.Pfeffer, and S. Swenson, Recent contributions of glaciers and ice caps to sea level rise. Nature 2012. doi:/10.1038/nature10847
Swenson, S. C. and J. Wahr, Post-processing removal of correlated errors in GRACE data, Geophys. Res. Lett., 33, L08402, doi:10.1029/2005GL025285, 2006.
Swenson S.C , D. P. Chambers, and J. Wahr: Estimating geocenter variations from a combination of GRACE and ocean model output. J Geophys. Res.-Solid Earth, Vol 113, Issue: B8, Article B08410. 2008.
Wahr, J., M. Molenaar, and F. Bryan, Time-variability of the Earth's gravity field: Hydrological and oceanic effects and their possible detection using GRACE, J. Geophys. Res., 103, 32,20530,229, 1998.
LAST UPDATE: 2014-02-18 F. Landerer, Zheng Qu, V.Zlotnicki. We thank Holly Maness and Sean Swenson for their contributions.