Search

Global navigation satellite system (GNSS) positioning technology is widely used in landslide monitoring due to its unique real-time and high-precision advantages. However, in complex landslide environments with severe occlusion, multipath error becomes one of the main error sources affecting GNSS real-time kinematic high-precision positioning, as it cannot be eliminated by differential correction. Traditional sidereal filtering methods cannot retain deformation information, which affects the accuracy of landslide early warning. To mitigate multipath error and preserve landslide deformation information, an improved sidereal filtering method based on mutation points and landslide evolution stage recognition is proposed. The method considers the landslide evolution stage, establishes a dataset for sidereal filtering, matches the multi-day correlation data with the sliding section-related data, and constructs a multipath model using averaging. Experimental results show that, compared with the traditional sidereal filtering and conventional segmentation methods, the proposed method reduces the coordinate deviation (QE) in the 3D direction by 25.02% and 22.13% and reduces the fitting deviation (QR) by 4.01% and 36.21% during the stabilization stage. During the critical sliding stage, QE decreases by 48.54% and 23.8%, and QR decreases by 60.12% and 42.54%, respectively. The proposed method has the best smoothness and power spectral density indicators in both stages. It has also been successfully applied to the Heifangtai landslide in Gansu Province, providing more stable and reliable data than traditional methods, avoiding delayed and false alarms, and issuing a red warning 3 h before the landslide failure, demonstrating its practical value. [ABSTRACT FROM AUTHOR]