Objective This study investigates the response characteristics of water storage at gully heads to rainfall in the black soil region of Northeast China， reveals the variation patterns of water storage at gully heads before and after rainfall， and establishes a quantitative prediction method for its changes before and after rainfall. Methods Four gully heads with similar developmental stages and scales in three study areas （Jiusan， Binxian and Jiutai） were selected as research objects. Differences in soil hydraulic parameters， decay coefficients of water storage， and variation patterns of water storage before and after rainfall were investigated through transient water release and imbibition method （TRIM）， numerical fitting， and prediction model construction. Results Obvious hysteresis effects were observed in soil-water characteristic curves （SWCC） of soils at gully heads， with the intensity ranking as gully 1 ≈ gully 2 < gully 3 < gully 4. After rainfall， water storage decayed exponentially over time， showing rapid decline in the initial stage and slower decline in the later stage. The decay coefficient （K_d） of water storage decreased progressively with increasing soil depth. Model simulations showed that water storage changes at 85.7 % of the monitoring sites were best predicted using 1 h maximum rainfall intensity （I_{1 h}） as the rainfall input parameter， whereas 14.3 % of the sites were better simulated using 2 h average maximum continuous rainfall intensity （I_{2 h}） as input. Therefore， in predicting soil water storage changes， I_{1 h} could be selected as the rainfall input parameter for sites with good infiltration performance， such as gully heads or severely eroded sites. However， I_{2 h} should be selected as the rainfall input parameter for sites with poor infiltration performance or strong water redistribution. Conclusion The evolution of water storage at gully heads after rainfall in the black soil region of Northeast China can be effectively characterized using exponential decay patterns and short-duration rainfall indices. The constructed model can accurately and quantitatively predict short-duration water storage changes at gully heads at the event scale， which provides methodological support for prediction and early warning of gully erosion processes.