Ocean surface winds and pressure forecasts are important for coastal and offshore marine weather guidance. In particular, ocean winds are most essential for driving global wave and ocean circulation forecast models. This paper evaluates performance of model forecasts of winds and pressures over the oceans from 2000 to 2003 at the National Centers for Environmental Prediction (NCEP). During the last three years, there has been a great increase in the amounts of surface data and area coverages of ocean wind observations from several operational satellites. These surface wind data include DMSP/SSMI wind speeds (since 1987 to present time), ERS 1-2 scatterometer winds (during 1998 to 2001), and most recently, QuikSCAT wind data (since 2001 to the present time). NCEP global forecasts of ocean winds and pressures are based on the Global Forecast System (GFS), while regional forecasts are based on the ETA forecast model. During this period, the two operational forecast models together with their atmospheric analysis schemes have also been undergoing various changes and improvements, especially in spatial resolutions and model physics. That the increase in use of various satellite surface wind data in the atmospheric analyses and data assimilation, together with global and regional model changes and improvements, which undoubtedly have affected the performance of NCEP model forecasts of winds and pressures over the oceans, is briefly reviewed.

To evaluate the model performance, NCEP's global and regional forecast model forecasts of surface wind and pressure fields are compared with buoy observations. Depending on their platform locations, buoys are classified into deep ocean buoys and near shore buoys, with 50 km from the coastline being the demarcation boundary between the two groups. These buoy locations are further geographically divided into five regions: East Coast of the U. S., West Coast of the U. S., Gulf of Mexico, North Sea, and TOGA regions. Buoy winds are adjusted to the standard 10 meters level above the ocean surface where model winds are forecasted. Time series of bias and RMS error statistics are presented for onshore and offshore buoys, for the five geographical regions, as well as at several individual buoy locations. To further investigate error characteristics of models' first guess forecasts (i.e., model background errors), EOF analysis is applied to time series of model background errors of surface wind speeds and sea level pressures at several selected buoy locations. The mode-amplitude structures of model background errors for the NCEP global and regional forecast models are analyzed and presented, and their potential applications to atmospheric analyses of surface winds and pressure fields are discussed.