Modern-Era Retrospective analysis for Research and Applications, Version 2
MERRA-2 FAQ
Soil moisture is available from MERRA-2 in two different kinds of units in the Land Surface Diagnostics file Collections (M2T1NXLND, M2TMNXLND, and M2TUNXLND).
The first set of variables are ground wetness values (GWET*) in (dimensionless) units of relative saturation for different layer depths (more on that below). A value of 1 indicates a completely saturated soil, and a value of 0 indicates a completely water-free soil.
The second set of variables are soil moisture contents (*MC) in volumetric units of m3/m3, i.e., the volume of water within the volume of bulk soil (including all solid material, water, and air).
In both cases, the soil moisture variables are provided for three nested layers, the profile (PR) layer from the surface down to the bedrock, the 0-100 cm root zone (RZ) layer, and the 0-5 cm topmost or surface (SF) layer.
The layer depth associated with PRMC (and GWETPROF) is dzpr.
The layer depth associated with RZMC (and GWETROOT) is dzrz.
The layer depth associated with SFMC (and GWETTOP) is dzsf.
These layer depths can be found in the Constant Land-Surface Parameters file (M2CONXLND). For example, SFMC is the volumetric soil moisture content in the 0-5 cm surface layer, and GWETROOT is the relative saturation in the 0-100 cm root zone layer.
Volumetric moisture content (*MC) outputs can be converted from m3 m-3 into equivalent water depths (in length units) by multiplying the *MC value with the associated layer depth.
Finally, the Constant Land-Surface Parameters file (M2CONXLND) also includes grid-cell average soil porosity (poros) values in volumetric units m3 m-3. Note that the MERRA-2 model computes land surface states and fluxes on a cube-sphere grid that is sub-divided into tiles, whereas values in the output data products are provided on a regular 576-by-361 lat/long grid. For each computational unit (or "tile"), the following equations hold:
PRMC = poros*GWETPROF
RZMC = poros*GWETROOT
SFMC = poros*GWETTOP
These equations do not hold, however, for the gridded output on the lat/long output grid because the porosity and soil moisture values are not constant across each lat/long grid cell.
Snow depth (SNODP) is recorded as the snow depth within the snow-covered portion only. Snow mass (SNOMAS), on the other hand, is recorded relative to the entire grid cell area, including the snow-covered and snow-free portions.
The snow depth averaged across the entire grid cell (including the snow-covered and snow-free portions) can be computed by multiplying SNODP with FRSNO.
As fresh snow falls on a previously snow-free tile – a tile is the model's computational unit, several of which make up a grid cell – the snow is first accumulated within a small area of the tile such that the snow mass (or snow water equivalent) per unit area in the snow-covered portion of the tile is fixed at 26 kg/m2. This threshold is known in the model as WEMIN, or the minimum snow mass in the snow-covered area fraction. It is applied to safeguard against very thin layers of snow that would be numerically unstable.
As additional snowfall occurs, the snow-covered area fraction (FRSNO) is expanded while the snow mass within the snow-covered fraction is kept constant at 26 kg/m2. That is, SNOMAS (which is the snow mass averaged across the entire tile) increases gradually from 0 to 26 kg/m2 as FRSNO increases from 0 to 1 such that SNOMAS/FRSNO is constant (and equal to WEMIN). Eventually, FRSNO reaches 1 and SNOMAS increases beyond WEMIN.
The snow depth (SNODP), on the other hand, is given by SNOMAS/FRSNO/snow_density, where SNOMAS/FRSNO reflects the effect of piling the snow into the snow-covered fraction of the tile. If snowfall stops short of covering the entire tile with a SNOMAS of 26 kg/m2 or more, SNODP is therefore (nearly) constant regardless of how much snow fell. (SNODP is "nearly" constant because the modeled snow density varies with time due to snow aging and compaction, but these processes are very slow.)
Note also that the value of WEMIN was changed between the original MERRA reanalysis and MERRA-2. Some discussion on this topic can be found in the following papers:
Reichle, R. H., C. S. Draper, Q. Liu, M. Girotto, S. P. P. Mahanama, R. D. Koster, and G. J. M. De Lannoy (2017), Assessment of MERRA-2 land surface hydrology estimates, Journal of Climate, 30, 2937-2960, doi:10.1175/JCLI-D-16-0720.1.
Toure, A. M., R. H. Reichle, B. A. Forman, A. Getirana, and G. J. M. De Lannoy (2018), Assimilation of MODIS Snow Cover Fraction Observations into the NASA Catchment Land Surface Model, Remote Sensing, 10, 316, doi:10.3390/rs10020316.
Click on "Read more" to see the aerosol sizes used in MERRA2 (and in the current version of GEOS/GOCART).
Using fields from the 2D tavg1_2d_aer_Nx collection, the concentration of particulate matter can be computed using the following formula: PM2.5 = DUSMASS25 + OCSMASS+ BCSMASS + SSSMASS25 + SO4SMASS* (132.14/96.06)
Sulfate requires a multiplication factor since the species tracer in MERRA-2 is the sulfate ion. Unlike with MERRAero, a multiplicative factor is not included for OCSMASS to convert organic carbon to organic matter since this is already handled within the model. For users of GEOS FP, please note that this equation is not applicable to FP as MERRA-2 does not include nitrate aerosol.
For additional details, see Section 3.1 of https://gmao.gsfc.nasa.gov/pubs/docs/Collow1489.pdf
References:
Buchard, V., Randles, C. A., DA SILVA, A. M., Darmenov, A., Colarco, P. R., Govindaraju, R., et al. (2017). The MERRA-2 Aerosol Reanalysis, 1980 Onward. Part II: Evaluation and Case Studies. Journal of Climate, 30(17), 6851–6872. http://doi.org/10.1175/JCLI-D-16-0613.1
Buchard, V., DA SILVA, A. M., Randles, C. A., Colarco, P., Ferrare, R., Hair, J., et al. (2016). Evaluation of the surface PM2.5 in Version 1 of the NASA MERRA Aerosol Reanalysis over the United States. Atmospheric Environment, 125, 100–111. http://doi.org/10.1016/j.atmosenv.2015.11.004
Provençal, S., Buchard, V., DA SILVA, A. M., & Leduc, R. (2017). Evaluation of PM2. 5 Surface Concentrations Simulated by Version 1 of NASA's MERRA Aerosol Reanalysis over Israel and Taiwan. Aerosol and Air Quality
Provençal, S., Buchard, V., DA SILVA, A. M., & Leduc, R. (2016). Evaluation of PM surface concentrations simulated by Version 1 of NASA's MERRA Aerosol Reanalysis over Europe. Atmospheric Pollution …, 8(2), 374–382. http://doi.org/10.1016/j.apr.2016.10.009
Unlike PM2.5 where the dust and sea-salt contributions are included in the 2D aer_Nx collections, there is no readily available PM1/PM10 diagnostic in MERRA-2. However, P1/PM10 concentrations can be calculated from the aerosol mass mixing ratios in the inst3_3d_aer_Nv collection. Start with the aerosol mass mixing ratios in lowest model layer 72 (recall: MERRA-2 vertical layers are arranged top-down) and compute the particulate matter concentration according to these formulas:
PM1 = (1.375*SO4 + BCphobic + BCphilic + OCphobic + OCphilic + 0.7 * DU001 + SS001 + SS002) * AIRDENS
PM10 = (1.375*SO4 + BCphobic + BCphilic + OCphobic + OCphilic + DU001 + DU002 + DU003 + 0.74 * DU004
+ SS001 + SS002 + SS003 + SS004) * AIRDENS
Unlike with MERRAero, a multiplicative factor is not included for OCphobic or OCphilic to convert organic carbon to organic matter since this is already handled within the model.
References:
Buchard, V., Randles, C. A., DA SILVA, A. M., Darmenov, A., Colarco, P. R., Govindaraju, R., et al. (2017). The MERRA-2 Aerosol Reanalysis, 1980 Onward. Part II: Evaluation and Case Studies. Journal of Climate, 30(17), 6851–6872. http://doi.org/10.1175/JCLI-D-16-0613.1
Buchard, V., DA SILVA, A. M., Randles, C. A., Colarco, P., Ferrare, R., Hair, J., et al. (2016). Evaluation of the surface PM2.5 in Version 1 of the NASA MERRA Aerosol Reanalysis over the United States. Atmospheric Environment, 125, 100–111. http://doi.org/10.1016/j.atmosenv.2015.11.004
Provençal, S., Buchard, V., DA SILVA, A. M., & Leduc, R. (2017). Evaluation of PM2. 5 Surface Concentrations Simulated by Version 1 of NASA's MERRA Aerosol Reanalysis over Israel and Taiwan. Aerosol and Air Quality
Provençal, S., Buchard, V., DA SILVA, A. M., & Leduc, R. (2016). Evaluation of PM surface concentrations simulated by Version 1 of NASA's MERRA Aerosol Reanalysis over Europe. Atmospheric Pollution …, 8(2), 374–382. http://doi.org/10.1016/j.apr.2016.10.009
Please refer to the Supplementary Description of Differences between ANA and ASM Data Collections [PDF].
At the end of the month, monthly files are created, and the whole month is run through quality checking. On approval, the data are released to the GES DISC. So, each new month is available approximately between the 15th and 20th of the next month.
However, if there is any disruption to the input observation stream or computing services, delays may occur.
The MERRA-2 File Specification Document provides extensive information on the collections of variables, units and data files.
MERRA's land parameterization is Randy Koster's Catchment model, but other surfaces, such as inland water, ocean surface and glaciers are also accounted for as sub-grid tiles. In the LND collection of variables, all the data are derived from the land model, and are not weighted according to the land fraction at that grid point. This data is provided to better compute land budgets for soil water and land energy.
The data in FLX, RAD or any other collection of variables represent the gridbox average of all the different tiles weighted by their fractional cover. This is where you would use evaporation to compute the atmospheric energy balance. The important distinction here is that LND is land only, while all other collections are representative of the whole grid box.
Fractional land cover in GEOS and MERRA is discussed more here: https://gmao.gsfc.nasa.gov/reanalysis/MERRA/land_fractions.php
The GEOS data assimilation system used to produce MERRA does not (or did not at the time of production) extrapolate data to pressure levels greater than the surface pressure. These grid points are marked by undefined values. The result is that area averages that include these points will not be representative compared to other data sets without additional screening. Time averages, such as monthly means, may also have substantial differences at the edges of topography. The lowest model level data and surface data are available so that users can produce their own extrapolation. A page discussing this issue is available. See https://gmao.gsfc.nasa.gov/reanalysis/MERRA/pressure_surface.php
The choice for a more complete derivation and discussion is a micrometeorology text book, for example, "Boundary Layer Meteorology" by Roland Stull.
Briefly, elements of the Earth's surface, grass, shrubs, crops, trees and buildings, all cause some friction and perturbation to the wind profile. The displacement height (or depth, or zero-plan displacement) accounts for their effect in the calculation of the surface layer log wind profile. The displacement height is the height at which the log wind profile projects the wind to be zero for purposes of computing the surface later turbulent fluxes. At heights less than displacement, different physical processes and theory take over from the log profile methods. For practical purposes, MERRA 2m and 10m output are intended to compare with screen level meteorology stations.
From land based surface meteorology stations, only surface pressure is assimilated. Radiosonde stations may contribute to the lower level analysis (T, Qv, U, V). Likewise, commercial aircraft can provide lower level data on ascent and descent (T, U, V). There are also wind profilers (U,V). Over ocean, ships and buoys may provide PS, T, Qv, U and V. See the MERRA-2 Observations Tech Memo for more details.
13. Why does downward shortwave (SWGDN) output differ between the radiation (RAD) and land forcing (LFO) collections?
Output of the surface incoming shortwave flux (SWGDN) is available in two sets of file collections: (1) tavg[*]_2d_lfo_Nx Land Surface Forcings (M2T[*]NXLFO, hereinafter LFO) and (2) tavg[*]_2d_rad_Nx Radiation Diagnostics (M2T[*]NXRAD, hereinafter RAD), but their values can differ. The SWGDN values in RAD reflect the flux averaged across all surfaces (ocean, land, landice, and lake), whereas the values in LFO reflect the flux over land only. This makes SWGDN in LFO appropriate for use in conjunction with the net shortwave land (SWLAND) flux, which is also available in LFO and similarly restricted to land only.
In the MERRA-2 model, SWGDN is first calculated on the model’s native cube-sphere grid, and all tiles (or sub-grid computational elements) within a given cube-sphere grid cell see the same SWGDN value, regardless of surface type. To write the RAD output, SWGDN is simply interpolated from the model’s native cube-sphere grid to the regular latitude-longitude grid of the MERRA-2 file collections. To write the LFO output, SWGDN is first masked to land tiles and then interpolated to the regular latitude-longitude grid. Along coastlines, lake shores, or land-landice boundaries, this different interpolation approach results in SWGDN values in LFO that differ from those in RAD.
For grid cells without land (FRLAND=0), SWGDN in LFO is set to the MERRA-2 no-data-value (1e+15).
Where the entire grid cell consists of land (i.e., FRLAND=1), small differences in SWGDN between LFO and RAD occur because a lossy compression is applied to RAD but not LFO.
In the MERRA-2 model, additional small differences in SWGDN between LFO and RAD are caused by small inconsistencies in the radiation calculations between the “Radiation” and “Surface” model components. More recent GEOS model versions no longer have these inconsistencies in the radiation calculations.
Seasonally-varying climatologies were used for leaf area index and greenness fraction as the data available over the MERRA-2 period were not considered reliable enough to capture the relevant interannual variability. The approach to producing those climatologies can be found here, pages 33-35.