Effects of large grid size discretization on coverage prediction using the ParFlow method

The ParFlow approach solves wave equations in the discrete time and space domain. ParFlow implementation has been designed for the coverage prediction for mobile networks in urban environments. In order to speed up the iteration process and save the computational memory, ParFlow uses a larger grid size than required by its constraint. This paper studies the effect of large grid size on the simulation results, in order to quantify the maximal allowed grid size for a satisfying coverage prediction. Applying a larger grid size mainly influences the simulation results in two ways: (1) it lowers the computation frequency which causes the changes of propagation attenuation rate, interfering pattern, and waveguiding effect between the building walls; (2) it distorts the building boundaries which leads to the change of scattering and artificial propagation blocking. Simple case studies and simulations enable us to propose criteria for adequate grid size which guarantees the required prediction quality. Furthermore the proposed criteria have been applied to the coverage prediction for part of the city of Munich in Germany. Compared to the measurements, the behavior of predictions to the grid size confirmed the proposed grid size criteria in this paper.