In this work, we use the next sub-volume method (NSM) to investigate the possibility of using the compartment-based ("on-lattice") model to simulate water radiolysis. We, first, start with a brief description of the reaction-diffusion master equation (RDME) in a spatially discretized simulation volume ("mesh"), which is divided into sub-volumes (or "voxels"). We then discuss the choice of voxel size and merging technique of a given mesh, along with the evolution of the system using the hierarchical algorithm for the RDME ("hRDME"). Since the compartment-based model cannot describe high concentration species of early radiation-induced spurs, we propose a combination of the particle-based step-by-step ("SBS") Brownian dynamics model and the compartment-based model ("SBS-RDME model") for the simulation. We, finally, use the particle-based SBS Brownian dynamics model of Geant4-DNA as a reference to test the model implementation through several benchmarks. We find that the compartment-based model can efficiently simulate the system with a large number of species and for longer timescales, beyond the microsecond, with a reasonable computing time. Our aim in developing this model is to study the production and evolution of reactive oxygen species generated under irradiation with different dose rate conditions, such as in FLASH and conventional radiotherapy.