In water repellent soils, infiltration following dry periods will typically be limited to narrow pathways which enlarge gradually through winter to produce seasonal patterns of progressive water repellent breakdown. Simple, one-dimensional hydrological models, which assume moisture is horizontally uniform, will not produce representative results in soils where preferential flow dominates, but may produce good representations of moisture dynamics at late stages of the wet season, where declining water repellency has allowed pathways to spread to their maximum extent, producing flow which is close to homogeneous in nature. We propose a new metric, the Mean Modified Response or MMR, to quantify intermediate stages of seasonal water repellent breakdown in terms of the discrepancy between field data and a calibrated one-dimensional model representing the same soil in a hydrophilic state. The utility of this metric is demonstrated using four years of soil moisture sensor data collected at a woodland site in Perth, Australia with a highly water repellent A-horizon. Individual rain events were simulated using data from an on-site rain gauge. MMR results show strong seasonal trends in all years of study, comparable to those revealed by an older metric, the Effective Cross Section, which provides a measure of flow heterogeneity. However, the new MMR metric is particularly useful for identifying variations in soil moisture responses by depth. We show that the highly water repellent surface layer diverts moisture preferentially to deeper layers to produce increasing moisture responses at depth, in patterns which sharply contrast with model predictions. This effect is shown to decrease through winter as surface repellency breaks down, but may be highly significant in conserving moisture against evaporative loss during dry periods. Results of the MMR analysis suggest that soil was most successful in diverting flow to deeper layers in periods where significant rain events were separated by dry periods of at least a week, but less successful where rain events were either highly isolated or closely spaced. We conclude that comparison to the 1D model presents a useful tool in demonstrating how patterns of infiltration are altered under water repellent conditions.