The accuracy of numerical models has been improved continuously and weather forecast and climate prediction rely on numerical models more and more. Numerical models are the equations which describe the motion and thermodynamic processes of the atmosphere. Statistical methods are also widely used to numerical models, becoming important components of numerical models. This paper reviews some new advances in the application of statistical method to numerical models. Firstly, application of statistical methods to data assimilation, ensemble forecast, parameterization of physical processes, statistical interpretation, extendedrange weather forecasts and model verification for numerical model are analyzed. Then, the latest application of Bayes statistics to numerical models is elaborated. Finally, several important research trends for future application of statistical methods to numerical models are put forward.