The impact of satellite water vapor wind vector (WVWV) observations on the prediction of the development of a mid-Pacific-ocean cyclone during NORPEX (NORth Pacific Experiment, 14 January - 27 February 1998) is assessed using a 4D-Var approach in which a non-hydrostatic version of the Penn State/NCAR nonhydrostatic mesoscale model version 5 serves as a strong constraint. The satellite wind observations are retrieved through an automated tracking algorithm using water vapor (WV), visible (VIS), and infrared (IR) imagery from operational geostationary meteorological satellites GMS-5 and GOES-9 over the North Pacific basin. For the case studied, the amount of satellite wind data are much greater in the upper troposphere than in the lower troposphere.

The 4D-Var assimilation of the satellite wind observations is carried out on a single domain with 90-km horizontal resolution. Incorporation of satellite wind observations was found to increase the cyclonic zonal wind shear and the cross-front temperature gradient associated with the simulated cyclone. However, the improvement in the intensity of the simulated cyclone measured by the central sea-level pressure is marginal using the same assimilation model. Increasing the forecast model resolution by nesting a 30-km resolution domain yields a more significant impact of WVWV on the cyclone intensity prediction. The GMS-5 satellite winds (upstream data) have more influence on the quality of the cyclone development than the GOES-9 satellite winds (downstream data). An adjoint sensitivity study confirms that the most sensitive region is located upstream of the cyclone, and that the cyclone is more sensitive to the lower rather than the upper atmosphere. Therefore, it is anticipated that larger impacts on cyclone prediction in the mid-Pacific Ocean will occur when a greater or equal amount of satellite wind observations are made available for the lower troposphere as are available for the upper levels.