With urban population rapidly growing, traffic congestion is becoming a major problem. In this paper, a framework is proposed for identifying the spatial congested partitions in a dynamic urban road network and for monitoring the temporal changes in their locations and structure. To that end, a given road network is transformed into a suitable graph representation, an initial partitioning based on a spectral clustering approach is performed, and then the partitions continue to be updated incrementally on the basis of the newly obtained traffic data at each new time point. The congested partitions are then identified on the basis of traffic measures (e.g., volume and green time utilization) available from the traffic signal control system. Experiments with the proposed method are conducted with real historical traffic data collected from the 493 signalized traffic sites in Melbourne, Victoria, Australia, with a total of 1,444 road segments and 581 intersection points. Experimental results show that large-scale urban traffic networks undergo many rapid but regular and frequent traffic patterns, which often go unnoticed by the traffic network operators. Tracking these kinds of changes in real time by means of the proposed framework can improve the reaction time of the traffic management team and result in less congestion.