[Objective] To verify the applicability of the method that combined self-organizing map neural network (SOM) and dynamic index methods in the analysis of interactions between land use types and structures in a certain period, and to analyze the evolution of land use types and structures from multiple scales in order to provide reference for the sustainable development and utilization of urban land resources.[Methods] Based on the land use data of Huangshi City, Hubei Province in 2005, 2010 and 2015, we used the single dynamic degree of land use to analyze the change and transition characteristics of land use types at the city level, and constructed a SOM model to express the spatial distribution of land use structures at the township-level, and explored the overall land use evolution with a method integrated K-means clustering and comprehensive dynamic degree of land use.[Results] ① The area of build-up land increased obviously, and they were transformed from cultivated land, woodland and unused land. The area of cultivated land was continuously decreased, as a result of conversions into construction land, mining land and bare land. The area of unutilized land reduced substantially. ② The transformation among types of land use structure was mainly from cultivated land to urban/cultivated land. ③ The central, western, and southern towns had experienced their slight changes in land use; while the towns in the northeast changed significantly, so did its structure.[Conclusion] The method in combination with SOM and land use dynamic index is suitable for comprehensive analysis about the spatial and temporal evolution of land use.