Urban traffic congestion is a prevalent concern, carrying significant implications for economic efficiency, environmental integrity, and societal welfare. Unlike traditional approaches which favour infrastructure enhancements for traffic congestion mitigation, contemporary solutions emphasize the implementation of traffic management and control strategies. This attention has given rise to diverse and adaptive traffic control and management strategies, emphasizing the importance of comprehending the system’s state, whether at the level of links, intersections, corridors, or the overall network. In recent years, attention has been shifted towards the network-level traffic flow models due to the emergence of control strategies based on the notion of the Network Macroscopic Fundamental Diagram (NMFD). This evolution has transitioned from analyzing traffic congestion in heterogeneous single-reservoir systems to more uniformly distributed multi-reservoir systems. Despite significant efforts in this direction, it has been found that the prediction accuracy of network-level traffic flow models in large-scale urban networks requires extensive calibration efforts, leading to less accurate simulation of traffic states in such systems. This thesis is an attempt towards the direction to provide a better representation of traffic state evolution in such complex systems by taking advantage of real-time sparse data from the network. The initial step towards a better understanding of traffic states in urban networks lies in introducing a framework for NMFD estimation, with a particular focus on reduced NMFD estimation due to limited data availability. Although it is well-established that the NMFD of a network is influenced by various factors, encompassing trafficrelated and topological characteristics, existing literature predominantly directs its efforts towards estimating the reduced NMFD through the utilization of traffic-related features. More importantly, prevailing methodologies often prioritize the loading period in reduced NMFD estimation, neglecting the crucial significance of the unloading period. These shortcomings have been addressed in the proposed method in this thesis, in which the simultaneous influences of traffic and topological characteristics are evaluated to estimate the reduced NMFD. Unlike existing approaches, the proposed method also takes into consideration the effect of the unloading period, thus offering a more accurate representation of the network performance during this period. Previous works on traffic dynamics simulation in heterogeneous urban networks have evolved from relying on significant assumptions about prior trip knowledge to utilizing either path flow distribution methods or state estimation techniques that take into account the presence of loop detectors on the periphery of regions. In this study, however, the focus is shifted towards identifying the evolution of Turning Rate (TR) at macroscopic nodes along the perimeter of the reservoirs with the aim to leverage the outputs to comprehend flow exchanges in multi-reservoir networks. In contrast to previous approaches, the proposed data-driven TR estimation method offers networkwide estimations at intersections capable of detecting both low and high-frequency variations. This approach allows us to better identify outflow/transfer flow ratios across multi-reservoir networks by integrating TR estimates with local penetration rate estimates using sparse Floating Car Data (FCD). Moreover, our results showcase the potential to directly estimate the outflow-NMFD of the network using sparse FCD. To address potential errors in model predictions, a model-driven state estimator is introduced which comprises two steps: (i) prediction step and (ii) update step. In the former step, the accumulation-based model regulates the system dynamics by considering estimated outflow-NMFD and real-time estimates of transfer flow ratios based on FCD. In the latter step, the sparse LDD is integrated with model predictions to minimize the model’s prediction errors and provide more accurate estimates of traffic states (i.e. accumulation and transfer flows).