A one-dimensional (1D) – two-dimensional (2D) coupled numerical model for shallow-water flows is developed, inspired by a large river-lake network. The model includes an improved method of coupling subdomains as part of the water stage prediction-correction (WSPC) approach, exploiting the characteristic curves of the shallow-water equations. Compared to other coupling methods, the one presented here results in better computational efficiency, convergence and stability. The coupling method uses parameters with a clear physical meaning, which can be set directly without appealing to empirical values. With the help of the WSPC approach, individual submodels are solved separately with their coupling parameters being consistent with each other at the coupling boundaries. The coupling method proposed herein allows large time-stepping for the coupled model. Practical approaches to achieve large time steps are discussed and analysed by comparing two different solution algorithms for the 2D submodel. The proposed method and model are examined by applying to a real-life case in China, and the results indicate their good validity, practicality and computational superiority.