Macroscopic modelling of road traffic flow is far from complete, different models exhibit strengths when used in varying situations. Continuum models, based on fluid dynamics are accurate and robust in describing traffic on a single road. Knowledge-based models, derived from heuristics based on either statistical methods or Artificial Intelligence techniques and are efficient at describing traffic processes at intersections. Neither of the existing approaches is separately able to capture effectively traffic dynamics in road networks. The thesis Introduces a hybrid macroscopic approach, combining continuum methods and knowledge-based models. It Is implemented using three different forecasting methods, namely neural network, random walk and SARIMA models each coupled with the Lighthill-Whitham and Richards continuum model. Results from numerical experiments confirm the promising features of the introduced approach in describing effectively traffic dynamics in road networks. The developed models are theoretically rigorous, numerically reliable, computationally efficient and suitable for real world applications.