Improved GSI Forward Model for Surface Observations on Coastlines

The Gridpoint Statistical Interpolation (GSI) Data Assimilation System has been developed as analysis component for both global (GFS) and regional (NAM, RAP/HRRR) numerical weather forecast systems in NOAA/NCEP. GSI can analyze many types of the observations with well-developed forward model, such as soundings, aircraft observations, satellite radiance, and GPS refractivity/bending angle. But in most of the GSI applications, including RAP/HRRR, the same forward model that interpolates background fields to observation location only based on the distance among observation and the grid point is used for upper air and surface observations.  This forward model for surface observations can induce large representativeness error when surface conditions are inhomogeneous, such as on coastlines. In this study, an improved forward model for surface observations is developed to analyze the observations on coastlines with GSI. Both ideal and real cases were conducted to test and verify the improved forward model under different seasons and resolutions. Results shows that the improved forward model can distinguish the characteristic of coastal observations automatically, and then improve the analysis along the coast area.