P1.32
The Real Time Mesoscale Analysis System. On-going system improvements and challenges
Manuel De Pondeca, SAIC and NOAA/NWS/NCEP/EMC, Camp Springs, MD; and R. J. Purser, S. Y. Park, G. S. Manikin, D. F. Parrish, and G. DiMego
The Real Time Mesoscale Analysis (RTMA) System is essentially a 2DVar system for the high spatial and temporal resolution analysis of surface observations on National Digital Forecast Database (NDFD) grids. The current resolutions used in space and time are 5km and one hour, respectively. A distinct characteristic of the 2DVar algorithm of the RTMA is the use of background error covariance shapes that follow the contour lines of the underlying topography to a controlled degree. For each analyzed field, the system also computes an estimate of the corresponding analysis error. In addition to the 2DVar analysis, the RTMA also interpolates Stage-II precipitation and GOES sounder cloud cover data to the NDFD grids. A thorough description of the RTMA, which was developed at the National Centers for Environmental Predictions (NCEP) and the Earth System Research Laboratory (ESRL) is given in an accompanying paper. The present work addresses the efforts which are taking place at NCEP aimed at improving various aspects of the RTMA system. Specifically, progress in the following areas is discussed: (i) background error model parameter estimation with the help of cross-validation, whereby the subsets of cross-validating data are derived with the help of a space filling Hilbert curve; (ii) use of a variational quality control scheme for the observations; (iii) use of alternate structure functions of background error correlations, with emphasis on covariances mapped to the background potential temperature and wind field; (iv) estimation of the analysis error covariance matrix via the the Lanczos algorithm for solving large scale eigenproblems in its connection with the pre-conditioned conjugate gradient method of the 2DVar scheme; and (v) use of a modified, computationally more effective formulation of the recursive filter that implements the anisotropic covariance model of background errors.
Poster Session 1, Monday Poster Viewing
Monday, 25 June 2007, 4:35 PM-6:30 PM, Summit C
Previous paper Next paper