GSWP-2 Multi-Model Analysis

The multi-model analysis (MMA) from GSWP-2 is the first global land-surface analysis of its kind, including state variables, fluxes, and other land surface properties. This is in many respects a land-surface analog to the atmospheric reanalyses in that it represents a best estimate of conditions over an extended period of the recent past. Using the results of multiple land surface models provides a model-independent result that is superior the results of any individual model, and typically as good as or better than the best model at each point and time. Of particular value, this multi-model analysis includes uncertainty estimates for all of the fields, based on the inter-model spread.

The MMA covers the period from January 1986 through December 1995, at a spatial resolution of 1° by 1° on a regular latitude-longitude grid, including all land surface points north of 60°S according to the land-sea mask of ISLSCP Initiative II.

Not all of the 13 models necessarily contribute to the calculation of any particular variable. See the data inclusion chart for a list of which models contributed to which variables.

There are three categories of model output data provided in the GSWP-2 MMA:

Monthly mean fields - These include all of the relevant standard land surface model output fields from GSWP-2 as listed by ALMA (120 time steps) as both means among all models, and the inter-model standard deviations.
Climatological monthly fields - These are the 10-year mean values of the standard ALMA fields for each month (12 time steps).
Daily state variables - These are daily mean values for soil moisture and temperature at 6 levels.

Documentation of the common input data sets is also provided here:

Model input data sets - These are global 1° grids of land surface parameters and meteorological forcing data used to specify initial conditions and boundary conditions (forcing) for the participating land surface models.
Documentation of the forcing data is available as COLA Technical Report #159.

.png Images

Images of the monthly and climatological MMA fields are also provided for browsing (click on the variable name for a table of images):

Caveats about the data

In addition to the disclaimer regarding this analysis, there are some specific issues with regard to the fields that users should be aware of:

NormDeepSoilMoist has some bogus points, particularly over the deserts of Arabia and North Africa, that are due to the lack of variability in time of deep soil moisture affecting the normalization proceedure (these quantities should average to zero over the 10-year period, but do not at these locations).
SurfStor and EWater each have only one model (SiBUC) contributing non-zero values, and are probably not representative of a multi-model extimate.
WaterTableD refers to the depth to the saturated zone in LSSs with TOPModel or similar groundwater interactions, and is not analogous to actual water table depths, especially outside of very humid regions.
Many of the snow-related variables (e.g., SWE, SnowDepth) show an unusually large range of values over permanent ice points (e.g., parts of Greenland) which are largely the result of the behavior of a single model.
High precipitation amounts are found in some high-latitude coastal areas (e.g., Norway, southeastern Alaska) particularly for snow. This is a result of a combination of factors that led to an overzealous wind correction for gauge undercatch in these regions.

Additional caveats are discussed in the text of the accompanying technical report.