TL;DR
This paper introduces a novel Azimuth-Dependent Elevation Weight (ADEW) model to improve the accuracy of real-time deformation monitoring using multi-GNSS in challenging environments. By adaptively adjusting satellite observation weights based on azimuth and elevation, the ADEW model significantly mitigates multipath errors, leading to centimeter-level positioning accuracy.
What Problem Does This Paper Address?
Real-time deformation monitoring, particularly for geological hazards like landslides, is crucial for disaster prevention. Global Navigation Satellite Systems (GNSS) are a powerful tool for this, but their accuracy can be severely degraded in complex environments. These environments, common in landslide-prone areas, feature mountains, dense vegetation, and human structures that cause signal diffraction and multipath errors. Traditional methods often struggle to effectively handle these errors, leading to unreliable positioning and potential misinterpretations of deformation. Existing models like the fixed cut-off elevation (TFC) or the azimuth-dependent elevation mask (ADEM) have limitations, such as failing to account for varying environmental conditions or introducing breaks in the data.
Methods
The paper proposes a two-tiered approach:
- Azimuth-Dependent Elevation Mask (ADEM): This model moves beyond a simple geoid-based cut-off elevation. Instead, it creates a physical elevation mask that is dependent on the satellite's azimuth. This is achieved by analyzing the minimum elevation angles (Emin) of satellite trajectories over a 24-hour period to define the physical horizon around the antenna. This mask is then used to adjust the elevation angle used in the stochastic model.
- Azimuth-Dependent Elevation Weight (ADEW): Building upon ADEM, the ADEW model introduces an adaptive weighting strategy. It uses a projection transformation to map satellite elevation angles to a virtual range, allowing for a continuous variance estimation. This approach dynamically adjusts the weight of observations based on their elevation and azimuth, specifically down-weighting observations that are likely affected by multipath errors due to terrain or obstacles. The core idea is to use a stochastic model where the variance of observations is modified based on a virtual elevation angle (E') derived from the measured elevation (E) and the ADEM mask.
The overall processing flow utilizes a real-time kinematic (RTK) platform, incorporating double-difference (DD) observations, least-squares estimation, and the Lambda method for integer ambiguity resolution.
Datasets
The study utilized GNSS data collected from two sites in Shaanxi, China, known for frequent landslide occurrences.
- Site (a): NingQiang city, with a rover station (NQ02) located in a complex environment with ridges and trees. Data was collected on DOY 021, 045, and 072 in 2018.
- Site (b): HanZhong city, with a rover station (BM02) in a mountainous area. Data was collected on DOY 040 in 2018.
The experiments involved multi-GNSS observations, including BDS, GPS, and GLONASS. For comparison, the paper also references data used for the Sidereal Filter (SF) model.
Key Insights and Ideas
- Environment-Specific Masking is Crucial: The study highlights that a one-size-fits-all elevation cut-off is insufficient. The physical environment around a GNSS antenna significantly impacts signal quality, and an adaptive, azimuth-dependent approach is necessary.
- ADEW Offers Continuous Weighting: Unlike ADEM, which might introduce discontinuities, ADEW's projection transformation and adaptive weighting provide a more robust and continuous way to handle varying multipath conditions.
- Physical Coefficients Have Longevity: The study suggests that the derived ADEM coefficients can be applicable for extended periods (even over a month), reducing the need for frequent recalibration, which is a significant practical advantage.
- Multi-GNSS Synergy: The use of multi-GNSS (BDS, GPS, GLONASS) enhances the ability to create a comprehensive physical elevation mask, as more satellites provide a richer set of trajectory data.
Key Results
The ADEW model demonstrated significant improvements over existing methods:
- ADEM vs. TFC: The ADEM model improved horizontal accuracy by 18.91% and vertical accuracy by 34.93% compared to the traditional TFC model. The integer ambiguity fixing success rate also increased from 63.4% to 87.8%.
- ADEW vs. ADEM: The ADEW model further improved upon ADEM, achieving an additional 21.9% improvement in horizontal accuracy and 29.8% in vertical accuracy.
- ADEW vs. TFC (Multi-GNSS): For multi-GNSS data, ADEW showed mean RMS improvements of 34.69% (horizontal) and 31.58% (vertical) compared to TFC.
- ADEW vs. TFC (Site b): At the second site, ADEW achieved improvements of 18.10% (horizontal) and 9.21% (vertical) over TFC.
- Reduced Coordinate Bias: The RMS of coordinate bias was consistently lower with ADEW across different datasets and days.
Limitations and Future Work
- Universal Algorithm: The authors acknowledge that the ADEW model relies on specific coefficients for each ADEM. Developing a universal algorithm to eliminate effects in real-time without site-specific calibration is a goal for future work.
- Dilution of Precision (DOP): The paper mentions considering the dilution of precision (DOP) of GNSS satellites in future research.
- Real-time Implementation: While the platform is designed for real-time processing, the full implications of real-time implementation and potential computational burdens are areas for continued exploration.
Summary updated