This study was conducted on a stony‐soil hillslope that severely eroded due to land use change from bamboo to tea. First, three approaches were used to extract soil hydraulic properties and then inputted into the Denitrification–Decomposition model for simulations. The Denitrification–Decomposition model defaults and ROSETTA pedotransfer function were traditional approaches without considering the rock fragment impacts on soil hydraulic properties, while the dual‐porosity function considered rock fragment impacts since it was calibrated with observed soil water retention data. Results showed that dual‐porosity functions gave the most accurate or physically sound results in simulating the nitrous oxide (N2O) emissions and leachate nitrate nitrogen (NO3？‐N) concentrations, with the root‐mean‐square errors as 0.0043–0.0107 kg N/ha and 4.39–9.09 mg/L, respectively. This suggested that rock fragments could impact the soil N cycle by affecting soil hydraulic properties. Second, the model simulations in the first step were further corrected by the volumetric rock fragment contents. This correction significantly reduced the N2O emissions and NO3？‐N leaching by 25–43% and improved the simulations. This suggested that high rock fragment contents could reduce the N emission and leaching by decreasing the volume of soils participated in N cycle. Last, under the climate change, the NO3？‐N leaching were overestimated without considering the rock fragments, while simulation using regular sets of Denitrification–Decomposition was coincide to the real N2O emissions without physical meanings. Our findings suggested that rock fragments should be considered in soil N cycling studies in the stony‐soil areas.