Abstract:The role of data assimilation is to provide a superior initial field for application in numerical models by means of applying observation data.The three-dimensional variation method is widely used in operational forecasting,due to its rather low computational costs and strong assimilation ability.An important component of the three-dimensional variation method is background error covariance,which can directly affect data assimilation and numerical prediction results.Due to the fact that the background error covariance is very substantial,in practical application the analysis increment is often used to solve the minimization problem of the three-dimensional variational cost function.Different control variables have various assimilation and prediction effects,thus selecting appropriate control variables is an important prerequisite for building a reasonable assimilation system.At present,the two most frequently used dynamical control variable schemes are the stream function potential function(ψ-χ) scheme,and horizontal wind components(U-V) scheme.These are considered to be respectively suitable for large-scale assimilation and convective scale assimilation.However,neither the specific characteristics of the two schemes in typhoon scale assimilation nor their impact on typhoon forecasting have yet to be investigated.In this study,based on the WRF-3DVarDA system,we analyze the statistical characteristics of the background error covariance of the two different control variable schemes.Next,we carry out an ideal test of single point observation assimilation and 20 groups of continuous cycle assimilation and prediction experiments in regard to five typhoon events.The analysis results of the background error covariance statistical characteristics of the two schemes show that the background error is smaller and the length scale is larger in the ψ-χ scheme,while more obvious local features and larger variance are found in the U-V scheme.In addition,the results of a single point observation test confirm the statistical characteristics of the two schemes,and also reveals that ψ-χ as a control variable can lead to a false negative increment.The results of batch experiments show that in general the U-V scheme is superior to the ψ-χ scheme,in terms of typhoon track error,maximum wind speed error and minimum sea level pressure error.The diagnostic analysis results show that the greater number of reasonable thermal and dynamic structural features in the U-V scheme are the source of improvement in the typhoon forecast.