Summary: This paper introduces a generalised parallelisation concept for lattice Boltzmann (LB) simulations. Our main goal was to combine two approaches to speed up LB computations: parallelisation combined with an efficient implementation and utilisation of non uniform tree-type Cartesian grids. A clear separation of data with different semantics (i.e., topology, geometry, physical quantities) to facilitate expandability for multi-physics applications is another key aspect of this work. In addition, an abstract rule set was developed which can handle e.g., specific user definable discretisation stencils. Finally we simulated a tandem cylinder flow to validate the method as well as the implementation in 2D.