Robust hydrologic models require an accurate formulation of infiltration and soil water redistribution. The application of Richards' equation can provide the most accurate description of these processes but for some applications it can be computationally intensive and prone to numerical instability and convergence errors. A conceptual, physically based formulation like the Green–Ampt with Redistribution (GAR) can be an attractive alternative in many applications. Original GAR applications, however, showed significant errors in simulated surface water content for soils with high saturated hydraulic conductivity values, and an increasing surface water content deviation after subsequent redistributions during long simulations. A modified GAR method (MGAR) is proposed that provides improved infiltration and soil water redistribution predictions during uneven multistorm time series for a wide range of soils. An increased number of redistributing wetting fronts more accurately represents the naturally curvilinear soil water content profile during the redistribution phase. A redistribution coefficient decreases the surface soil water prediction during nonuniform precipitation series as a function of three variables: saturated hydraulic conductivity, redistribution number, and redistribution time for each storm event in the time series. Simulations of uneven multistorm precipitation time series using GAR and MGAR for 11 soil textural classifications were compared against Richards' equation. The MGAR markedly improved surface soil water predictions (coefficients of efficiency >0.935 and RMSE <0.011). The method also provided a good approximation of average water content for soil observation depths within the top 1 m, corresponding with the area of interest for many vadose zone modeling applications.