ensemble transform Kalman filter (ETKF)

• Main Authors: Published Online In Wiley Interscience, G. N. Petersen, A. J. Thorpe A
• Other Authors: The Pennsylvania State University CiteSeerX Archives
• Format: Text
• Language: English
• Subjects: KEY WORDS; North Atlantic
• Online Access: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.379.7875; http://enkf.nersc.no/Publications/pet07a.pdf

ABSTRACT: The spatial characteristics of ensemble transform Kalman filter (ETKF) sensitive area predictions (SAPs) are explored using ensemble forecasts from the European Centre for Medium-Range Weather Forecasts for the period of the 2003 North Atlantic THORPEX Regional Campaign. The ensemble size necessary for a robust sensitive area prediction is found to be surprisingly small: a 10-member ensemble is capable of replicating approximately the same sensitive area structure as a 50-member ensemble. This result is corroborated by the fact that the leading eigenvector of the ensemble perturbations explains over 70 % of the ensemble variance and possesses a nearly identical spatial structure regardless of the ensemble size. The structures of the SAPs were found to vary with the lead-time between the ensemble initialization and the adaptive observing time, indicating the necessity of using as recent an ensemble as possible in ensemble-based sensitive area predictions. The ETKF SAPs exhibit similar structures at different levels in the atmosphere and there is no indication of a vertical tilt. A relationship is found between the SAPs and the zonal wind, horizontal temperature gradient and the Eady index, indicating that the ETKF identifies regions with significant gradients in the mass-momentum field as regions of large initial error or large error growth. Copyright © 2007 Royal Meteorological Society