Estimation of Observation Impact using the Adjoint of the GSI Analysis at NCEP

An adjoint-based procedure for estimating the impact of observations on the Gridpoint Statistical Interpolation (GSI) analysis scheme is described and the influence of observations to poor GFS forecasts (called dropouts) is examined. An exact adjoint of the GSI, developed by the NASA Global Modeling and Assimilation Office (GMAO), is setup and tested at the National Centers for Environmental Prediction (NCEP). Since the adjoint of the GFS model is not currently available, this study focuses on the sensitivity of various observing systems based on the analyzed increments of temperature and zonal wind. Results show that specific observing systems play a major role in cycled GFS forecast skill, and improvements in the GSI analysis account for a ~5% reduction in poor GFS forecasts over the last five years. Using the adjoint of the GSI, we intend to target specific observation types for improvement in the data quality control used in the NCEP global data assimilation system.