Response of Precipitation to Soil Moisture Constraints

in the NCEP Global Model Simulations

Earlier pilot study showed a wide disparity among general circulation models (GCMs) in the land-atmosphere coupling strength (LACS), the degree to which the atmosphere, especially precipitation, responds to anomalies in land surface state, in the boreal summer season.  A recent presentation by Kenneth Mitchell of NCEP/Environmental Modeling Center showed the LACS in the NCEP Global Forecast System (GFS) in the context of multi-model GLACE* approach.  Results demonstrated a broad disparity in the warm season precipitation response to the constrained land states across GCMs (Figure).  The NCEP GFS had weak warm season precipitation response to constrained land states, especially when only deep soil moisture was prescribed (see the figure below).  It seemed to emerge mostly from weak transpiration response to constrained deep soil moisture, indicating that other canopy stress factors (humidity, temperature, insolation) would not be negligible.  The differences among three GFS land surface model modes were small compared to the inter-atmospheric GCM differences.

*  GLACE: Global Land-Atmosphere Coupling Experiment, an inter-comparison study across a range of more than ten atmospheric general circulation models.

 

The figure shows sensitivities of Intra-ensemble variance for R (left) and S (right) ensemble simulations.  From top to bottom are GFS simulations with the land surface model configuration of NOAH, OSU and NOAHX, respectively.  The experiment design is described as follows. 

Atmospheric GCM:  GFS T62 L64 (as in NCEP operation of seasonal forecast)

Land Surface Model (LSM) configurations:

OSU:  OSU LSM (as in NCEP operation of seasonal forecast)

Noah:  Noah LSM

NoahX:  Noah LSM, initialized from Noah cycled Global Data Assimilation System (GDAS)

Simulation period:  1994/06/01 ~ 1994/08/31

Initial conditions: CPC AMIP, using OSU

Three ensemble experiments (each with 16 members):  W, R, S

W simulations:  W1: establish a time series of surface conditions;

                             W2-W16: repeat without writing out land variables

R simulations:  each ensemble member given the same time series of all land prognostic

                             states from W1

S simulations:  each ensemble member given the same deep soil moisture from W1

(Contact Kenneth Mitchell)