Abstract: Similarities exist in spatial patterns over time, such as in serial landscapes of an area at different times, serial expressions of an area in difference scale maps, spatial patterns in different areas and so on. These similarities often refer to spatial attribute (i.e., location, shape, size) and non-spatial attribute data (i.e., type, quality). This paper presents a "flowing water model" to evaluate similarities between different spatial patterns. Both global and local situation of landscape features are considered in this method. This method uses flow overlay areas for time points to describe the landscape patterns instead of landscape indexes. To some extent it avoids the uncertainty that one value may describe more than one type of landscape indexe and disturbances in spatial correlation. We explored the similarities of Beijing traffic-network backbones from 1938 to 2014, revealing the tendency of the traffic-network pattern and change in this pattern over time.