Premium
Horizontal structure of height‐forecast errors over the Southern part of South America
Author(s) -
Vera Carolina
Publication year - 1994
Publication title -
quarterly journal of the royal meteorological society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.744
H-Index - 143
eISSN - 1477-870X
pISSN - 0035-9009
DOI - 10.1002/qj.49712051911
Subject(s) - autocorrelation , bessel function , mathematics , series (stratigraphy) , exponential function , data assimilation , standard deviation , range (aeronautics) , fourier series , truncation (statistics) , statistics , meteorology , geography , mathematical analysis , geology , paleontology , materials science , composite material
A study of the height‐forecast error autocovariances was made for a simple intermittent data‐assimilation system (SADI) applied over the southern part of South America. The main objective of this study was to determine from the data the height‐forecast error autocorrelation, and to provide new estimations of the forecast‐ and observation‐error standard deviations. In performing this derivation several significant differences were found between the estimates obtained with a simple forecast model in an area which is data sparse, and those derived for data‐rich regions and advanced forecast models. The forecast‐error correlations were fitted to the Fourier—Bessel series and also to a negative square exponential (NSE) function at the six model levels. The fitted Fourier—Bessel series was able to represent more accurately the wavelength range resolved by the analysis scheme than the fitted NSE function. New estimations of forecast and observation errors have been obtained from the Fourier—Bessel expansion. As the representative errors are much larger because of the coarse resolution of the network and the low‐order truncation of the fit, the observation errors are greater than those used for the northern hemisphere analysis systems. Results also show that the forecast errors are dominated by the synoptic‐scale components in all the levels considered. An evaluation of the impact of using the new set of horizontal covariances was done and it mainly showed that the analysis done with the fitted Fourier‐Bessel series better represents the data than both the original SADI system and an analysis system using a fitted NSE function.