Impact of Mesoscale Satellite Wind Data on Numerical Model Simulations.

The initialization of a comprehensive mesoscale numerical prediction model is investigated using actual high-spatial-resolution satellite wind observations. Attention is focused on the impacts of this mesoscale data as part of an initial specification for a dynamical numerical model and on the use of diagnostic methods to assess the forecast impacts. Comparisons of the satellite wind data to special 1800 GMT rawinsonde observations provide some confidence in satellite winds and thus in the mesoscale features defined by these winds. Four objective analysis methods are tested on these data to determine sensitivity to analysis procedure. It is possible to obtain reasonable gridded data sets with the data containing large spatial variation in density. Four six-hour forecast experiments are made with the Drexel University/NCAR mesoscale prediction model with 35 km horizontal resolution to examine the assimilation characteristics of the mesoscale wind data. Both initially imposed and gradually imposed insertion techniques are tested. The initial mass field in the test experiments is the same as in the control forecast. Numerous diagnostic methods are made for assessing forecast impact of the inserted wind data. The primary impact of the inserted data is to alter the amplitude of mesoscale features developed in the control. The gradually imposed insertion technique in contrast to the initially imposed insertion produces the larger local extrema and an extra spurious external mode pressure oscillation. The divergence signal for the inserted data is lost after 1/2 h for the smaller mesoscale component and by 3 h for the larger mesoscale component. The vorticity signal persists for the entire 6 h period for both scale components. Assimilation of the satellite wind data into the model proceeds according to recognizable adjustment processes with time scales of 3 h and 1/2 h depending on motion scales. Changes in the 1 to 3 h precipitation forecasts due to the insetion data exist and are not evident after 3 h.