Repeated unmanned aerial vehicle laser scanning (ULS) surveys allow for the derivation of differential digital elevation model (DoD), facilitating the detection of geomorphic changes. However, insufficient understanding of methods for estimating DoD uncertainty constrained the implementation of the above research. In this study, four methods for DoD uncertainty estimation, including the direct uncertainty derivation method (DUD), the uniform-error thresholding method (UET), the variance-overcome thresholding method (VOT), and the errorsource thresholding method (EST), were applied based on ULS data acquired from gullies and slopes in a small catchment of the Chinese Loess Plateau. Two strategies (subtraction and exclusion) were employed to remove the derived DoD uncertainties. The volumetric changes were validated with field investigation and terrestrial laser scanning (TLS) surveys. Results showed that the DUD method preformed superiorly compared to other methods on the road, which quantified the highest percentage (62.64 %) of gully volumetric change with field survey as a reference. While the VOT showed the worst accuracy in estimating erosion gully volume on the road. The DUD method using the subtraction strategy and VOT method showed comparable performance and outperformed other methods on the sloping field. The EST method exhibited mediocre levels of performance for estimating volumetric change on both road and sloping field. In summary, our study provided a reference for the application of the DoD uncertainty estimation method in soil erosion monitoring.