The key characteristics of noise information are spatial propagation and distance attenuation. Traditional noise maps are dominated by two-dimensional and static representations, similar to the heat maps. The researches in the past were focused on the ways to get data or evaluating the accuracy of data, resulting that noise map visualization and noise symbols were less studied. The traditional drawing methods are usually used like color grading which makes noise maps have simple symbolic form such as flat color blocks corresponding to different values. It is difficult to show its spatial features. Thus, this paper proposed a new expression method of noise maps in three-dimensional dynamic visualization for the spatial distribution and intensity change of noise. Based on the Bertin symbol parameters, the new method integrates the characteristics of size, color in the visual variables, and the rate and order of change in the dynamic parameters, and it can convey noise information dynamically by using height-variable and color-changing square bars as noise symbols. Three-dimensional dynamic noise maps can present relevant information on different scales, and they contain corresponding symbol sizes and positioning arrangements in different map levels in order to show the multi-granularity of the maps. So as to represent the attenuation and addition of noise during propagation, gave some examples of a single-point sound source and multi-point sound sources with graphics and data, and multi-level information expression was briefly exemplified. In this paper, we simulated and expressed the road traffic noise in a certain period and area of Huangpu District, Shanghai. In addition, noise data was calculated from traffic noise models and the taxi GPS trajectory data. The specific noise map was deployed in Web browser, using the base map provided by Mapbox. As expected, the three-dimensional dynamic expression expands the performance dimension of noise data. Compared with the flat color symbols used in traditional two-dimensional noise maps, multi-parameter stereo symbols can not only accurately describe the distribution and intensity of noise, but also improve the audience's interest and concentration from the visual perception. Simultaneously, compared with the static maps which can only show the final results, the three-dimensional dynamic noise maps are able to display the processes of information changes, which could help users find subtle anomalies and provide more accurate references for noise prevention and control. Furthermore, three-dimensional dynamic noise maps proposed in this paper can provide a better visual reference for the analysis of spatial rules such as hot spots distribution and transmission trends of noise.