Reduction of Middle-Atmospheric Forecast Bias through Improvement in Satellite Radiance Quality Control

Abstract

Abstract This article discusses a practical problem faced in operational atmospheric forecasting and data assimilation, and efforts to improve forecast quality through the choice of quality control parameters. The need to utilize as much data as possible must be carefully balanced against the need to reject observations deemed erroneous because they are far from the background value. Alleviation of forecast bias in the middle atmosphere for a global atmospheric prediction system is attempted via improvement of the quality control and bias correction of the satellite radiance data; in particular, the sensitivity of the analysis to the satellite radiance outlier check parameters for the Naval Research Laboratory’s three-dimensional variational data assimilation system [Naval Research Laboratory Atmospheric Variational Data Assimilation System (NAVDAS)] is investigated. A series of forecast experiments are performed with an extended-top (0.04 hPa or ∼65 km) version of the U.S. Navy’s Operational Global Atmospheric Prediction System (NOGAPS) for the month of January 2007. The experiments vary the prescribed radiance observation error variance for the Advanced Microwave Sounding Unit-A (AMSU-A) and the tolerance factors for the AMSU-A and NAVDAS quality control processes. The biases of geopotential height, temperature, and wind in the middle atmosphere are significantly reduced when the observation error limit for the highest-altitude AMSU-A channel (i.e., 14) is relaxed from 0.95 to 3 K and the tolerance factors for the AMSU-A and NAVDAS quality control processes are relaxed from 3 to 4. The improvement is due to assimilation of more high quality AMSU-A radiance data from the highest-peaking channel.