JP6.10
Upper-air forecast improvement utilizing radiance bias correction algorithms that account for forecast model bias
Young-Joon Kim, NRL, Monterey, CA; and W. F. Campbell and S. D. Swadley
Forecast skill in the stratosphere can be improved by increasing both the quantity and the quality of satellite data assimilated. Data assimilation requires accurate, unbiased model forecasts (i.e., the model “background”). If the model stratosphere is not very accurate, valuable observations can be rejected by the data assimilation system when the innovation (observation-background) magnitude exceeds a prescribed limit. If the limit is relaxed excessively, significant errors can be introduced, and the forecast skill can be degraded. Innovation limits must be set properly, which is difficult to do objectively because most data assimilation systems are bias-blind, designed only to correct random error rather than systematic error.
As a first step, this study investigates the sensitivity of the analysis to the NOGAPS (Navy Operational Global Atmospheric Prediction System) satellite data outlier check. A series of experiments are performed by varying the prescribed error limits for the highest-altitude AMSU-A channel (i.e., ch14) used in the operational NOGAPS as well as the tolerance factors for the AMSU-A quality control and NAVDAS (NRL Atmospheric Variational Data Assimilation System). An experimental version of the NOGAPS / NAVDAS T239L42 system with its top at 0.1 hPa is used for the experiment. Cycling data assimilation, initialized at 2006-12-20-00Z, is run through 2007-01-31-18Z, and is analyzed at selected times. Preliminary results show improved simulation of the middle atmosphere and corresponding forecast skill with some combinations of the limiting parameters. The improvement is due to use of more abundant AMSU-A radiance data, which help correct the model bias with observations of good quality.
Joint Poster Session 6, Improvements to NWP and Short-term Forecasting
Wednesday, 14 January 2009, 2:30 PM-4:00 PM, Hall 5
Previous paper Next paper