|
An Inverse Model for the PBL The inverse model evolved from a 1980 suggestion at the NSCAT group meeting by Brown that we could use the scatterometer surface wind to generate a geostrophic wind simply by inverting the PBL Similarity model. This was done [Ocean Surface Pressure Fields from Satellite Sensed Winds, R.A. Brown and G. Levy, Mon. Wea. Rev., 114, pp 2197-2206, 1986.] and derived pressure fields from VG agreed with numerical analyses within 1-2 mb. When applied to the Southern Hemisphere, there was considerable difference with the ECMWF analyses (up to 20% of the storms were missed by the numerical model) [Southern Hemisphere synoptic weather from a satellite scatterometer, G. Levy and R.A. Brown, Monthly Weather Review, 119, 2803-2813, 1991]. Subsequently, ECMWF increased resolution and performance to 5% storms missed. As the numerical models are able to incorporate scatterometer data into the analysis, location of storms and fronts has gotten better. However, as long as the Southern Hemisphere and Tropics lack data for initialization of the numerical models, the satellite scatterometer data is expected to produce superior surface pressure fields. Since the nonlinear equilibrium model contains OLE explicitly, it will produce different pressure fields than the K-Theory PBL models used in the numerical models. The inverse PBL model is used to calculate the pressure fields given at this web site. It has been modified by Jerome Patoux to include the tropics (Patoux, Foster & Brown, 2002). Thus continuous global pressure fields are available from satellite scatterometer data. Currently the inverse model is in version 3 by Jérôme Patoux (UWPBL 3.0). UWPBL 3.0 technical introduction In 1997, Brown suggested that given the success of the inverse model in reproducing surface wind fields, we should try to derive a direct Pressure Model Function (PMF) by correlating the backscatter signal, so with pressure gradients directly. This could be done simply by replacing the existing model function process (for U10) with an input of pressure gradient vector instead of U10. When done, the model function behaved as well as that for U10. |