An approach for Reference Point Selection in SBAS-InSAR Time Series Slope Deformation Monitoring

Imagine standing on a hillside, watching the earth beneath your feet shift imperceptibly over months or years. To the human eye, the ground appears still, but to the instruments of modern science, it is a landscape in constant, slow motion. Measuring this movement is vital for safety, especially near dams or steep slopes where a sudden slide could be catastrophic. Scientists use a technique called Interferometric Synthetic Aperture Radar, or InSAR, to track these shifts from space. Satellites bounce radar signals off the ground and measure the time it takes for the echo to return. By comparing these signals over time, they can detect if the ground has moved even a few millimeters. However, there is a fundamental catch: the satellite cannot tell where the ground is "still" and where it is "moving" on its own. It can only measure how much one spot has moved relative to another spot. To make sense of the data, scientists must pick a single point on the ground to act as a fixed anchor, a reference point that they assume is not moving at all. If they pick a spot that is actually shifting, even slightly, the entire map of movement becomes distorted, showing false slides or hiding real dangers.

For decades, researchers have struggled with this choice. The standard method has been to look for areas that appear very clear and steady in the radar images, often relying on a single point that looks stable. This approach, however, is prone to human error and bias. One scientist might pick a rock that looks solid, while another picks a different one, leading to conflicting reports about the same landslide. In some cases, the chosen anchor point might be subtly sinking or rising, which would make the rest of the ground appear to be moving in the opposite direction. This uncertainty has made it difficult to trust satellite data for critical engineering decisions, such as monitoring the stability of tailings dams, which hold back massive volumes of mining waste.

A team of researchers at The University of Melbourne has developed a new way to solve this problem, moving away from guessing and toward a rigorous, automated process. Instead of picking just one lucky pixel from a map, their method scans the entire area to find multiple zones that appear stable. They then apply a series of strict filters to these zones, checking not just how clear the radar image is, but also how well the data connects across the landscape and whether the area is large enough to be trustworthy. Think of it as checking a group of people to see who is standing perfectly still; you wouldn't just pick the first person who looks calm, you would observe a group, check if they are all standing together in a stable formation, and then average their positions to find the true center of stillness. By combining the data from several of these stable zones, the researchers can cancel out local noise, such as atmospheric interference or small, isolated shifts, to find a single, highly reliable reference point.

The researchers tested this new framework on the Cadia Dam in New South Wales, Australia, a site that experienced a significant failure in March 2018. Before the dam collapsed, the ground had been slowly deforming, a warning sign that was difficult to interpret with previous methods. The team applied their multi-step selection process to the satellite data covering the dam and its surroundings. They identified three distinct areas that met all their stability criteria and used a mathematical averaging technique to determine the single best reference point from among them. When they compared the results of their new method against physical instruments called prisms that were installed on the dam and measured the movement directly, the match was striking. The satellite data, anchored by their new method, aligned closely with the ground measurements, showing an average error of just 2.24 millimeters. This level of accuracy represented a fifty percent reduction in error compared to the worst-case scenarios found when using the standard, default methods of selecting a reference point.

The study went further to prove that their approach was not just a lucky guess for this specific location. The researchers systematically removed parts of their new method to see which steps were truly necessary. They found that the most critical steps were the ones that checked the overall shape and connectivity of the stable areas, ensuring that the chosen points were not isolated islands of data but part of a coherent, stable landscape. They also tested how the method held up when the amount of time covered by the satellite data was shortened. They discovered that while the method could find stable points even with less data, the accuracy of the speed at which the ground was moving improved significantly when the observation period was longer than one year. This confirmed that while the new selection process is robust, the length of time the satellite watches the ground remains a key factor in getting the full picture.

By replacing subjective choices with a structured, multi-parameter framework, this research offers a clearer path for using satellite technology to monitor earth movements. The work demonstrates that with the right anchor point, InSAR can capture complex ground behaviors with a precision that rivals physical instruments on the ground. This is particularly important for infrastructure safety, where the difference between a false alarm and a missed warning can be the difference between a routine inspection and a disaster. The study confirms that the proposed method effectively removes the risk of human error in selecting a reference point, providing engineers and geologists with a more reliable tool to watch over the slow, silent shifts of the earth.