RESOLUTION
The tropospheric version of the OI analysis scheme being used for the control assimilation has been carried out at a horizontal resolution of 2 degree latitude by 2.5 degree longitude at 14 upper-air pressure levels (20, 30, 50, 70, 100, 150, 200, 250, 300, 400, 500, 700, 850, 1000 mb) and at sea level. The analysis increments are computed every 6 hours using observations from a +/- 3 hour data window centered on the analysis times (00, 06, 12, and 18 UTC). The innovation vector (observation minus background forecast) used as input to the OI is computed using a single forecast valid at the analysis time.
THE ASSIMILATION
The OI scheme is multivariate in geopotential height and winds and employs a damped cosine function for the horizontal correlation of model prediction error. The height-wind cross-correlation model is geostrophic and scaled to zero at the equator. The multivariate surface analysis scheme over the oceans adopts an Ekman balance for the pressure-wind analysis. The moisture analysis for mixing ratio employs only rawinsonde moisture data. Although, it should be noted, that all TOGA COARE upper air moisture data was withheld from the assimilation. All grid point analyses are done using up to 75 nearby observations from within a radius of 1600 km.
The assimilation system does not include an initialization scheme and relies
on the damping properties of a Matsuno time differencing scheme to control
initial imbalances generated by the insertion of observations. However, the
initial imbalances and spinup have been greatly reduced over earlier versions
by the introduction of an incremental analysis update (IAU) procedure
[Bloom et al., 1991]. As shown in the figure of
the IAU procedure,
standard OI analysis increments are computed at the analysis times (0, 6, 12, 18Z). The increments are then inserted gradually into the AGCM by rerunning the forecast and adding a fraction of the increment at each model time step. Over the 6 hour period centered at the analysis time the full effect of the increment is realized. The assimilation thus effectively consists of a continuous AGCM forecast with additional heat, momentum, moisture and mass source terms updated every 6 hours from observations. It is the output of this IAU procedure that makes up the dataset provided here. An important difference between the IAU scheme and the usual Newtonian nudging procedure is that the IAU forcing terms are held constant over the insertion period, while in Newtonian nudging they are proportional to the difference between a target analysis and the instantaneous current model state.
