Abstract: The authors analyzed spatial and temporal characteristics of traffic flow under congested traffic based on detection data on urban expressway.First,autocorrelation function method was adopted to analyze the stationary of traffic flow parameter time series;Second,the spatial characteristics,which included the impact between the upstream and downstream sections,were analyzed by using detection data;At last,a model was developed to convert non-stationary speed time series to stationary time series based on G-P algorithm.The empirical research shows that traffic flow and occupancy are stationary time series,speed is non-stationary time series under congested traffic.The speed at upstream section could impact speed at downstream section through the transfer effect,the speed at downstream section could impact on speed at upstream section through the feedback effect.The speed can be converted to stationary time series by phase space structure with time delay of 2 min and embedding dimension of 4.