The ensemble forecast technique is a practical solution to the uncertainty problem of numerical weather prediction.At present,researchers around the world tend to focus on the Regional Ensemble Forecast(REF),which aims at the improvement of regional high impact weather forecast.As various uncertainty resources exist for meso-scale and small-scale weather phenomena,regional model simulation is a very complicated issue,thus how to generate effective perturbations for REF is a hot topic involving many technical difficulties.In the present paper,the progress of REF research is reviewed in terms of initial condition perturbation,model perturbation and lateral boundary condition perturbation,and the trends of methods related to these aspects are also presented.The results presented show the following:for initial condition perturbation,the mainstream methods include dynamical downscaling,using traditional methods developed from Global Ensemble to generate perturbations for REF,as well as some methods specifically designed for REF.All of these methods are characterized by some advantages and some shortcomings,as downscaling a lack of small scale and other components leads to the generation of insufficient large-scale uncertainty information.In addition,research on the REF initial condition perturbation has only begun to explore more effective methods such as blending,which consider not only sufficiently small-scale uncertainty,but also sufficiently large-scale uncertainty.Finally,the inconsistency problem between initial state and lateral boundary can also be ameliorated.Model perturbation is another important aspect for REF.This technique mostly perturbs model physics,such as multi-physics combination and stochastically perturb model physics.It has been reported that the multi-physics combination is quite simple and can effectively improve the ensemble spread of REF,while using the stochastic method to perturb model physics has greater scientific significance,thus this type of perturbation method has become the trend of model physics perturbations.Furthermore,multi-model combination is another practical method of model perturbation.Related studies have been carried out,the results of which show that this method possesses stronger skill than a single model ensemble,especially when a multi-model ensemble is applied to meso-scale severe weather forecast,such as with a Tropical Cyclone.As REF systems are constructed based on the regional models,the uncertainties originating from lateral boundary conditions cannot be ignored.Lateral boundary condition perturbation schemes mainly use large-scale ensembles,such as Global Ensemble Forecast Products,to provide different lateral boundary conditions for REF.Studies have proven that this method can achieve the goal of perturbing the lateral boundary condition of REF,and lateral boundary condition perturbation is found to aid in amplifying the ensemble spread of REF for long-range forecast lead times.In addition,the ensemble forecast skill can also benefit from lateral boundary perturbation.Although the perturbation techniques for REF have already led to some fruitful achievements,much work is still needed,and all of the methods related to initial condition perturbations,model perturbations and lateral boundary condition perturbations are still under development.It can be predicted that,as the REF perturbation methods continue to improve,the REF will become increasingly more effective,and will play a more important role in operational numerical weather prediction centers.