Decoupling atmospheric and multipath errors in NRTK landslide monitoring: A refined multipath modelling strategy

• A novel method is proposed to decouple GNSS multipath and atmospheric errors. • The method improves NRTK positioning accuracy by up to 19% over unfiltered solutions. • Performance remains robust across various VRS baseline lengths and rainfall scenarios. • Heavy rainfall amplifies both multipath and atmospheric errors in NRTK observations. Network Real-Time Kinematic (NRTK) high-precision deformation monitoring is affected by the superposition of multipath effects and residual atmospheric delay. Unmodeled atmospheric errors can contaminate multipath correction models and degrade positioning accuracy if not properly separated. This study proposes a refined multipath mitigation method. First, Complementary Ensemble Empirical Mode Decomposition (CEEMD) is applied to suppress high-frequency white noise. Multichannel Singular Spectrum Analysis (MSSA) is then used to jointly decompose the denoised residuals from multiple Orbital Repetition Times (ORTs) and generate the corresponding Reconstructed Components (RCs). We then evaluate the temporal coherence of RCs across channels, components that exhibit high temporal similarity across ORTs are identified as multipath-related. Using data from the Hong Kong SatRef network, including various Virtual Reference Station (VRS) baseline lengths and an ultra-short baseline scenario, we demonstrate significant improvements in positioning accuracy and robustness across baselines. It consistently improved horizontal positioning by 9% and vertical positioning by 6% compared to a conventional three-day SF, and by 15–19% relative to unfiltered solutions. Moreover, heavy rainfall, as an important landslide trigger, was observed to amplify both multipath and atmospheric errors, temporarily reducing NRTK positioning accuracy.