Abstract: Aiming at the characteristics of quantity and fluctuation range on the gross error from global navigation satellite system (GNSS) deformation monitoring time series, we propose an improved 3σ method of gross error detection based on wavelet analysis. We use simulated data and actual engineering data, and compare the detection results with the traditional 3σ method and inter-quartile range (IQR) method respectively. The experimental results show that compared with the traditional 3σ method and IQR method, the improved 3σ method not only has better effect on the overall detection rate, but also has more obvious advantages in detecting 3σ to 5σ gross error. Thus, it is more suitable for the actual requirement of GNSS deformation monitoring project.