Mesoscale predictability and background error convariance estimation through ensemble forecasting

Abstract

Over the past decade, ensemble forecasting has emerged as a powerful tool for numerical weather prediction. Not only does it produce the best estimate of the state of the atmosphere, it also could quantify the uncertainties associated with the best estimate and the predictability of an event. This additional information could be used to provide a more accurate estimate of the background error covariance matrix for use in data assimilation. This preliminary study explores the use of ensemble forecasts to gain information on predictability and the background error covariance matrix using data from the 24-25 January 2000 snowstorm that occurred along the East Coast of the United States. Adding randomly distributed uncorrelated initial perturbations to this data produced four sets of ensemble forecasts. The error growth characteristics estimated from these ensemble forecasts were found to be quantitatively similar to previous estimates that were achieved by perturbing the individual observations or by adding monochromatic wave disturbances. Analyses show that the isotropic random perturbations initially dissipated everywhere except for in the conditionally unstable regions. Error from the ensemble forecasts quickly developed horizontal and vertical structure that was associated with the cyclogenesis of the system. This error growth was a nonlinear process that was dictated by moist dynamics but not by the initial magnitude of the perturbations added to create the ensemble. The ensemble forecasts were also used to estimate the background error covariance matrix. These results show that initially uncorrelated errors develop a strong covariance or relationship between and among variables especially in the regions of cyclogenesis and convection. The covariances' development was controlled by the underlying governing dynamics, among which moist processes play a fundamental role. These covariances (correlations) indicate that observation/estimation of one state variable could provide valuable information on the state estimation of another state variable which is key to data assimilation. This paper demonstrates that further study on using ensemble forecasts to estimate the flow-dependent background error covariance matrix is warranted.