arXiv:2601.12878v2 Announce Type: replace Abstract: We propose a hierarchical splitting approach that constructs splitting methods for N-split differential equations by iteratively applying splitting methods for two-split systems along a splitting tree. Since assigning methods of order p to all nodes of the tree yields a hierarchical splitting method of order p, the order theory for two-split systems carries over to any number of partitions. We derive explicit formulas for the leading-order error terms, investigate self-adjointness, and show that state-of-the-art composition approaches are contained in the resulting class as special cases. Augmenting the approach with multiple time stepping yields a framework at the intersection of geometric numerical integration and multirate integration, for which we characterize the computational order of a multirate integrator and introduce a reweighting of the multirate factors that reduces computational overhead, but can lead to an order reduction when applied within composition techniques. We further discuss principles for designing splitting trees, with particular attention to separable Hamiltonian systems and port-Hamiltonian systems. Numerical experiments for both classes demonstrate that the hierarchical splitting approach yields efficient methods for N-split systems. We find that the approach is particularly efficient when commutator-based splitting methods become available and the partitions evolve on distinct time scales.