Precision Intelligence: Solving for When Tectonic Plates Move Faster Than Your Accuracy Budget

Authored By: Wolfgang Schickler, Senior Engineering Fellow at Vantor

The Earth's surface moves. Tectonic plates drift up to 70 centimeters, or almost a meter, every ten years. Most geospatial intelligence (GEOINT) companies that collect imagery from space don’t have to think about this geological creep. The accuracy of their imagery isn’t impacted by these subtle shifts.

At Vantor, which operates on the leading edge of the accuracy curve, this geological fact matters very much. Our spatial foundation is built on satellite imagery with sub-5-meter accuracy, making it the most accurate global-scale foundation in the industry. And we’re on a journey to achieve sub-meter accuracy, pushing that gold standard even further.

Accuracy is critical to ensuring that images, imagery basemaps, and 3D datasets correctly represent real-world locations and enable precise measurements for a wide range of geospatial intelligence use cases. In this context, every centimeter difference counts.

Our engineering team recently deployed a solution that accounts for plate motion during bundle block adjustment (BBA), our photogrammetric process to increase the accuracy and align the imagery we collect. The capability is now live in production, correcting for geological drift that would otherwise slowly degrade the accuracy of imagery collected at 30-centimeter native resolution.

The accuracy threshold where geology matters

The BBA process refines satellite pointing accuracy by tying multiple images together and measuring them against known ground control points (GCP). GCPs have a survey epoch: the date when they were originally measured on the ground. As plates move, so do these ground control points. This introduces error between the GCP survey date and the image acquisition date that compounds over time.

For example, in Australia, the Indo-Australian Plate moves 6.9 centimeters per year, introducing 0.7 meters of positional discrepancy over a decade. In Los Angeles, where the Pacific Plate moves 4.9 centimeters per year relative to the North American Plate, the same ten-year span introduces 0.5 meters of error.

This image shows the movement of tectonic plates. Data source: © 2006 Pearson Prentice Hall, Inc.

This problem can even impact accuracy within a single satellite image. North of San Francisco, the San Andreas Fault divides the Pacific Plate from the North American Plate. A single WorldView satellite pass spanning this boundary captures terrain moving in measurably different direction.

Complex plate motion dynamic near San Andreas fault (left), causing plate motion of varying magnitude and direction within a single satellite image (right). Graphics source: Spatio-Temporal Mapping of Plate Boundary Faults in California Using Geodetic Imaging, by Andrea Donnellan, Ramon Arrowsmith, and Stephen DeLong. https://www.mdpi.com/2076-3263/7/1/15.

These discrepancies matter for real-world missions. Consider change detection, which is one of the most important GEOINT use cases today with the rise of geospatial AI. If a customer has a 10-year-old baseline dataset over Australia, analysis of new imagery may result in results that show a mix of real change and apparent change from plate drift. That’s a problem.

Engineering the­ right plate motion correction workflow for improved spatial accuracy

The industry has attempted to address this problem using a few different methods, but each of them have come with their own trade-offs:

• Pre-BBA correction: Adjust the attitude and ephemeris (satellite position and orientation) of individual images before BBA. This approach is easiest to implement but provides only approximate corrections. Since it applies one uniform correction per satellite pass, it cannot account for variances within a single image.
• Image-space correction within BBA: Adjust image measurements during BBA based on plate motion alone using the Global Strain Rate Model (GSRM), which quantifies the variation of tectonic motion. This approach handles varying motion within images while remaining relatively straightforward to implement.
• Observation-based correction within BBA: Introduce plate motion as stochastic variables (observations with uncertainty) within the BBA process itself. This approach is mathematically most precise and can handle varying motion but adds significant complexity to the adjustment equations and may lead to unstable results depending on control point distribution.

We developed a variation of the image-space correction approach that ensures corrections are made across both space and time, meaning customers can maintain accuracy across historical datasets. The process adjusts both GCPs and tie points for plate motion, which are features measured across multiple images to link satellite passes together (think: major landmarks or natural features). This approach gives us more flexible parameters for better outcomes.

Here’s the workflow:

1. Define an adjustment date, for example a customer-specified date for matching legacy data.
2. Using GSRM, correct all GCP measurements for each GCP separately from the date of the GCP survey to the adjustment date.
3. Correct all Tie Point image measurements for each image independently from the image acquisition date to the adjustment date.
4. Run the BBA with these corrected measurements.

With this approach, corrections happen in image space rather than ground space. This design choice simplifies the mathematics while maintaining precision across varying motion fields, since image measurements are what BBA processes.

Innovating at the accuracy frontier

The plate motion correction is now live in our production processes, ensuring our ability to maintain industry-leading accuracy for our satellite imagery as well as for broader spatial intelligence products like Vivid™ Mosaic basemaps and Vivid™ Terrain 3D data. It also helps ensure that our AI-powered insights are accurately detecting real changes on the ground.

The plate motion correction demonstrates Vantor’s continued commitment to forging the new frontier of spatial intelligence. As we’ve continued to improve our global accuracy, we’ve had to account for even the subtlest factors to maintain the quality our customers have come to expect.

And it gives us even more refined capabilities as we continue our march towards sub-meter accuracy at global scale.

About the author: Wolfgang Schickler is a Senior Engineering Fellow at Vantor, where he leads photogrammetric processing innovation for the company's satellite imagery foundation. His work focuses on accuracy refinement methods that bring us closer to delivering sub-meter accuracy from space-based sensors.