TL;DR
This paper introduces a Small Baseline Subset (SBAS) Offset Tracking method to precisely monitor very slow landslides in challenging environments like the Three Gorges Region, which are difficult for traditional InSAR. The method achieves centimeter-level accuracy, outperforming standard offset tracking in dense vegetation and steep terrain, and shows good agreement with GPS data.
What Problem Does This Paper Address?
Landslides are a major natural hazard, and monitoring their movement is crucial for disaster prevention. Traditional methods like GPS are expensive and difficult to deploy widely, especially in remote areas. Interferometric Synthetic Aperture Radar (InSAR) is a powerful remote sensing tool for deformation monitoring, but it faces significant challenges in areas with:
- Dense vegetation: This causes rapid loss of phase coherence, making it hard to get reliable measurements.
- Steep slopes and rugged topography: These lead to geometric distortions and can exceed InSAR's maximum detectable displacement gradient (DDG).
- Very slow displacements: Landslides moving at rates of millimeters to centimeters per year can be obscured by noise or errors in InSAR processing.
- High humidity and frequent cloud cover: These can limit the effectiveness of optical remote sensing.
The Three Gorges Region in China exemplifies these challenges, making it difficult to monitor its numerous landslides effectively. Existing offset tracking methods, while better at handling decorrelation and large displacements than InSAR, can still suffer from sparse measurements, especially when using a single master image.
Methods
The core innovation here is the Small Baseline Subset (SBAS) Offset Tracking approach. This method builds upon standard sub-pixel offset tracking but incorporates principles from SBAS InSAR to improve measurement density and reliability. Here's how it works:
- Sub-pixel Offset Tracking: This technique analyzes the intensity and amplitude of SAR images to find 2D shifts (offsets) between corresponding pixels in master and slave images. By using cross-correlation and oversampling, it achieves sub-pixel accuracy.
- Small Baseline Subset (SBAS) Construction: Instead of using all possible image pairs, this method strategically selects pairs with small temporal and spatial baselines. This minimizes temporal decorrelation (due to changes over time) and geometric decorrelation (due to different viewing angles), which are major sources of error. This creates a network of "small baseline subsets."
- Residual Offset Removal: Before deriving displacement, the method removes residual offsets caused by orbit inaccuracies and co-registration errors. This is done by fitting a 2D polynomial to offsets measured in stable areas. This step is critical for accurately measuring very slow displacements.
- Singular Value Decomposition (SVD): The 2D offset measurements from the connected pairs are then inverted using SVD to derive the time series of displacement for each acquisition.
- Quality Masking: Pixels with high root mean square error (RMSE) are discarded to ensure the reliability of the final measurements.
This SBAS approach aims to leverage scatterers that might have temporary medium correlation, thereby increasing the density of reliable measurements compared to traditional single-master offset tracking.
Datasets
- TerraSAR-X Staring Spotlight (TSX-ST) data: A stack of high-resolution SAR images acquired over the Tanjiahe landslide area from February 2015 to February 2016. The data was acquired with an 11-day interval.
- SRTM 1 arc-second global DEM: Used for modeling and correcting topographic distortions.
- GPS measurements: Historical GPS data from monitoring stations in the study area were used for validation.
Key Insights and Ideas
- SBAS is crucial for challenging environments: The paper strongly suggests that the SBAS approach is essential for achieving dense and reliable measurements in areas with dense vegetation and steep slopes, where traditional InSAR struggles.
- Residual offset removal is key for slow movements: For very slow landslides (centimeter-level annual rates), removing residual offsets due to orbit and co-registration errors is not just beneficial but crucial to avoid obscuring the actual displacement.
- Temporal baseline is the dominant factor for measurement density: The study found that the number of reliable measurements decreases exponentially with the temporal baseline, highlighting the importance of selecting appropriate temporal constraints for the SBAS network.
- Reservoir drawdown as a primary trigger: In the Three Gorges Region, the study identifies the rapid drawdown of the Three Gorges Reservoir water level as the major triggering factor for landslide movement, with rainfall playing a less significant role during the observation period.
Key Results
- Precision: The SBAS offset tracking approach achieved an average precision of 2–3 cm for landslide displacement measurements.
- Measurement Density: The SBAS method provided a measurement density more than 15 times greater than conventional offset tracking in the study area.
- Validation: The results showed good agreement with published GPS measurements, with RMSE values between offset time series and GPS data being statistically insignificant.
- Landslide Behavior: The analysis revealed a consistent temporal evolution of the landslide, with the highest displacement rates observed in April-August, and a spatial distribution showing greater deformation in the upper part of the slope.
- Triggering Factor: Reservoir drawdown was identified as the primary driver of landslide activity.
Limitations and Future Work
- Data Availability: The study notes that the availability of SAR data with shorter revisit cycles and higher resolution would further improve DInSAR and offset tracking capabilities, especially for addressing coherence loss in vegetated areas.
- 3D Displacement: The current offset tracking method measures displacement projected onto the slant range plane. To obtain full 3D displacement, data from at least two different orbits (descending and ascending) are necessary. The authors suggest that using data from both orbit types would improve estimation robustness.
- Complementary Techniques: While SBAS offset tracking is effective, the authors acknowledge that time series InSAR can still be valuable for detecting smaller deformation magnitudes, especially during less active periods. They propose that these techniques are complementary.
- Future Data: The paper implicitly suggests that future work should leverage upcoming satellite missions with improved revisit times and resolutions to further enhance landslide monitoring capabilities.
Summary updated