Alpine deciduous shrubs play an important role in local water cycle for ecosystems of Qinghai‐Tibet Plateau, but rainfall partitioning of the deciduous shrubs has not been fully understood due to significant variation in spatial coverage and leaf storage capacity, especially the lack of appropriate model to describe relevant processes. To assess these processes, field experiments were performed to determine vegetation and atmospheric parameters for the deciduous Potentilla fruticosa shrub in the growing season of 2012. Based on the experimental data, a variable parameters Gash model specifically for deciduous shrub rainfall partitioning was adapted from the revised Gash model to simulate rainfall partitioning. Results showed that the interception, throughfall, and stemflow accounted for 21.44%, 29.26%, and 49.30% of total rainfall in the shrub patches, respectively. Nearly half of interception (46.1%) was lost through evaporation from saturated leaf canopy during rainfall events. The performance of the original model on deciduous shrub canopy was improved by utilizing a variable parameters Gash model that adopted event‐based changeable meteorological and canopy parameters. The rainfall partitioning pattern in the shrub patch was very sensitive to changes in canopy coverage, evaporation, and rainfall intensity, in which great varieties can add large potential errors to the revised Gash model simulations. To accurately model rainfall partitioning, it is reasonable to adopt variable parameters.