Abstract Quantitative characterisation of landscapes through derivation of landscape indices helps to define landscape composition, but provides limited insight into configuration attributes. This research demonstrates how the combined use of landscape indices and graph theory (and its associated analytical tools) can produce more substantive insights into the influence of landscape structure upon landscape dynamics relative to using them individually. Landcover datasets are used to delineate and analyse the dynamics of landscape systems for contrasting locations on the Qinghai-Tibetan Plateau: a grassland system near Maduo, and a human-tended agricultural system around the city of Xining. The “natural” grassland system is shown to be very dynamic, characterized by ever-changing relationships between its components, while the human influenced landscape is configurationally less dynamic with more persistent patterns. Prospectively, combined approaches to analysis of landscape composition and configuration provide more leverage in efforts to explain landscape dynamics (e.g. appraisals of connectivity and fragmentation), guiding both research and management/monitoring applications.