Ensemble-based observation impact estimates using the NCEP/GFS

Observation impact estimates within the NCEP global forecast system (GFS) have been calculated using the EnKF ensemble-based estimation method (Kalnay et al. 2012). This method is convenient for an ensemble-based data assimilation system since it does not use the adjoint of the forecast model nor the data assimilation, which are necessary for the adjoint-based observation sensitivity method. Since the hybrid 3DVAR and EnKF global data assimilation system became operational in May 2012 at NCEP, it is appropriate to apply this method to estimate the observation impacts.

All observation types assimilated in the operational global analysis are used in the EnKF (T254 with 80 member ensembles) and their impacts on the 24 hour forecast are evaluated using both moist and dry total energy norms. Spatial localization is applied on the observation impact by applying horizontal advection of the localization functions. Results from a one-month experiment will be shown in detail and we will discuss relative merits and characteristics of each observing system in the short-range forecasts in different regions of the world.