A Sensitivity Study of the November 25, 2011 GFS Dropout Using the GSI Hybrid EnKF vs GSI 3DVAR

NCEP has recently upgraded the analysis system with the Gridpoint Statistical Interpolation (GSI) Hybrid Ensemble Kalman Filter (EnKF) system that shows substantial improvement in the background field and 5-day Global Forecast System(GFS) forecast skill, and since this upgrade there has been a considerable reduction of both Northern Hemisphere (NH) and Southern Hemisphere (SH) dropouts.

A prominent 5-day NH dropout with a very low anomaly correlation (AC) score of 0.54 occurred with the forecast from 00Z 25 Nov. 2011 initial conditions (IC) when the GSI 3-dimensional variational (3DVAR) assimilation system was operational. The new GSI Hybrid EnKF system barely alleviated this dropout improving the AC score to 0.71, while the European Centre for Medium-Range Weather Forecasts (ECMWF) 5-day AC score was 0.82 which was below their average skill. The IC differences between the GFS and ECMWF analyses showed large height differences over the Central Pacific and Western CONUS, and the GFS had a weaker 500 hPa jet-streak in the Pacific. Large GFS forecast errors in amplitude and phase were concentrated near 500 hPa in the western CONUS at 24 and 48 hours and resulted in a short wave trough that propagated downstream compared to a cutoff low for the ECMWF forecast and the verifying analysis. The large concentrated short-range forecast errors suggested this case would be easy to solve, but until we conducted initializing the GSI IC using ECMWF analysis (called ECM runs) with the high resolution T574 GFS model, instead of our usual low resolution T382 model, there was little improvement in forecast skill indicating that resolution played an important role in improving the forecast skill.

NCEP's dropout team has been contributing towards understanding and alleviating GFS dropouts since 2008. A variety of diagnostic tools such as ECM experiments, conducting Observing System Experiments (OSEs), studying analysis differences between the GFS and ECMWF, and making research to operational (R2O) changes to quality control procedures for conventional and satellite observations have been used to investigate about two dozen NH and SH dropouts. We found a majority of the NH dropouts originate from higher latitudes or around the Central North Pacific Ocean. The SH dropouts occur along the baroclinic zone between 50S to 70S due to data quality issues and the assimilation system.