In the process of modeling and simulation of bacterial growth, the calibration of a numerical model to the specific experimental setup is often necessary. This study addressed the question of whether it is possible to predict the growth of microorganisms based only on independent measurements without additional calibration. To this end, flow-through experiments with Escherichia coli (HB101 K12 pGLO) in a Hele–Shaw cell filled with quartz sand were simulated with a newly developed numerical code for reactive multiphase multicomponent flow based on an operator splitting approach. The hydraulic parameters of the sand were determined by multistep outflow experiments and the parameters of the microbial growth model from batch experiments. There was very good agreement between predicted and measured cell concentrations nondestructively determined from the fluorescence intensities of a green fluorescent protein produced by the bacterial strain. The correct description of the kinetics of gas–water phase exchange in the unsaturated zone and of cell attachment to the solid phase proved to be crucial. The operator splitting approach significantly reduced numerical dispersion in a convection-dominated test case. It allowed a flexible adaptation of the solution strategy to the specific problem.