Estimating the Impact of Observations in Hybrid GSI-EnKF Data Assimilation System Using Ensemble Based Method

The ensemble based observation impact estimate metric (Liu and Kalnay 2008) was implemented for the GSI-EnKF hybrid data assimilation system for GFS. The accuracy of the estimate was first accessed and methods to improve the estimate was proposed and tested. The impact of rawinsonde temperature observations on the forecast of wind and temperature was first estimated using the ensemble method and verified against the actual impact. For the 24 hour temperature forecast, the estimated impact of the rawinsonde temperature observation closely follows the actual error reduction for all the eight levels considered. For the 24 hour wind forecast, the estimated impact of the radiosonde temperature was found to be less precise when compared against its actual impact. The estimated impact was much less accurate at longer lead times as compared to short lead time when the fixed, distance based localization method was used. An adaptive localization scheme has been developed and the initial results have shown it improved the estimate. The same evaluation is being conducted for satellite radiance data and the results will be reported in the conference.