Variational data assimilation system combines the useful observation information and model dynamic constraint and improves the optimal analysis using adjoint method in order to provide correct and high quality initial conditions for the meteorological numerical weather prediction model. The observation error is assumed Gaussian in traditional variational assimilation method. While the gross error is inevitable in the actual observation data, which is not strictly agreed the hypothesis of Gaussian observation error. The observational term of cost function need to be modified by the variational quality control to deal with the non-Gaussian observation error, and the variational quality control is incorporated within the variational analysis itself. The variational quality control technique was implemented in the pre-operational global four dimension data assimilation system in this paper. Assimilation and forecast experiments were carried out with the conventional observation data. It was demonstrated that the variational quality control guaranteed the fully consistence of background and model dynamic to ameliorate the analysis results and improve the forecasting skill.
[1]
C.,et al.
Analysis methods for numerical weather prediction
,
2022
.
[2]
N. B. Ingleby,et al.
Treatment of Gross Errors Using Maximum Probability Theory
,
1992
.
[3]
D. Dee.
On-line Estimation of Error Covariance Parameters for Atmospheric Data Assimilation
,
1995
.
[4]
P. Courtier,et al.
A strategy for operational implementation of 4D‐Var, using an incremental approach
,
1994
.
[5]
Heikki Järvinen,et al.
Variational quality control
,
1999
.
[6]
Andrew C. Lorenc,et al.
Objective quality control of observations using Bayesian methods. Theory, and a practical implementation
,
1988
.
[7]
N. B. Ingleby,et al.
Bayesian quality control using multivariate normal distributions
,
1993
.
[8]
A. Lorenc.
A Global Three-Dimensional Multivariate Statistical Interpolation Scheme
,
1981
.
[9]
Jean-Noël Thépaut,et al.
The Global Observing System
,
2010
.