Effects of model chemistry and data biases on stratospheric ozone assimilation
1E.O. Hulburt Center for Space Research, Naval Research Laboratory, Washington, DC, USA
2Department of Physics and Astronomy, Dordt College, Sioux Center, IA, USA
3Sciences Applications International Corporation, Beltsville, MD, USA and Global Modeling and Assimilation Office, NASA Goddard Space Flight Center, Greenbelt, MD, USA
4Naval Research Laboratory, Monterey, CA, USA
Abstract. The innovations or observation minus forecast (O−F) residuals produced by a data assimilation system provide a convenient metric of evaluating global analyses. In this study, O–F statistics from the Global Ozone Assimilation Testing System (GOATS) are used to examine how ozone assimilation products and their associated O−F statistics depend on input data biases and ozone photochemistry parameterizations (OPP). All the GOATS results shown are based on a 6-h forecast and analysis cycle using observations from SBUV/2 (Solar Backscatter UltraViolet instrument-2) during September–October 2002. Results show that zonal mean ozone analyses are more independent of observation biases and drifts when using an OPP, while the mean ozone O−Fs are more sensitive to observation drifts when using an OPP. In addition, SD O−Fs (standard deviations) are reduced in the upper stratosphere when using an OPP due to a reduction of forecast model noise and to increased covariance between the forecast model and the observations. Experiments that changed the OPP reference state to match the observations by using an "adaptive" OPP scheme reduced the mean ozone O−Fs at the expense of zonal mean ozone analyses being more susceptible to data biases and drifts. Additional experiments showed that the upper boundary of the ozone DAS can affect the quality of the ozone analysis and therefore should be placed well above (at least a scale height) the region of interest.