If a road continues to sink, then simply laying an extra layer of asphalt will not help. It is important to know if the surface has been damaged or if there is movement within the ground. The same applies to a family with a crack running across their wall. They need someone to figure out what happened and explain what should happen next. Satellites can be used as part of the investigation. Radar observations over several years can show ground movement across neighborhoods, roads and industrial sites. Engineers can compare these measurements with groundwater records, soil investigations and construction activity. Artificial intelligence can help analyze the increasing amount of information.
The useful question is how this combination could help Pakistan investigate problems earlier, before another repair bill or a dangerous failure.
What Researchers Found in Rawalpindi
There is already local research to build on.
The study published in Frontiers in Earth Science (2024) investigated subsidence rates based on satellite data from Rawalpindi and Islamabad between 2015 and 2022. It discovered that the older, densely populated area of Rawalpindi had a subsidence rate of approximately 50 mm per year. This was most likely due to large amounts of groundwater extraction. The effect of local geology was also investigated. The measurement concerned the area and period studied; it should not be applied to every neighborhood in the twin cities.
Subsidence is when the ground surface is moving downward. A major concern for buildings and buried services is whether nearby locations settle at different rates. Such a pipe or foundation across these points is liable to strain. Just because movement is happening does not mean a specific building is unsafe. It gives engineers a reason to look into it and a record to compare what they find.
How Radar Measures a Change
It’s called InSAR, short for interferometric synthetic aperture radar. A satellite transmits radar signals to the Earth and measures the reflected signals. Specialists compare repeated observations of the same area to estimate changes in the distance between the satellite and reflecting surfaces. The method is able to detect motion of a few millimeters over large areas with proper observations and processing. Radar also works at night and through cloud cover.
But that doesn’t mean a satellite takes a picture of every foundation moving by a millimeter. Measurement sensitivity and the size of objects visible in an image are different things. There is another detail that gets lost in descriptions of the technology. Radar generally looks at the ground from an angle. It measures motion toward or away from the satellite. It does not directly measure vertical sinking. Additional observations or well-justified assumptions are required to distinguish the directions of motion.
This is important when you are reading a map. For a displacement value to have any impact on an engineering decision, the user needs to be aware of the direction it points and how reliable it is.
Why Engineers Still Need Measurements on the Ground
A satellite record can show where movement has changed. The change needs more evidence to account for it. Engineers might look at groundwater pumping, nearby excavation, drainage, or the materials under a structure. They must also verify whether any of the apparent motion is due to measurement errors.
Ground instruments remain important. Piezometers measure the pressure of water. Inclinometers measure motion beneath the surface. Boreholes give us the underground layers, and surveys and positioning instruments are independent checks. It would be misleading to call these methods slow or outdated. Some make repeated or continuous measurements. The main drawback is that they observe specific points, while satellite observations can be used to study the area around.
In 2024, a case study used InSAR to monitor mine waste dams. This involved comparing data from satellite observations with ground-truth observations and analyzing 5 failure cases. It was concluded that whilst InSAR can be a useful tool to help identify areas of risk and target further investigation, a multi-method approach should be used to monitor.
The point of combining them is to answer questions neither can settle alone.
What AI Adds to the Investigation
Many datasets can be produced by a monitoring program, including displacement, rainfall, groundwater, and maintenance data. This can be difficult to analyze manually. Artificial intelligence can be applied to locate areas that deviate from a trend. A predictive model can be used to predict the displacement over a set time period based on an observed dataset. Extra data can be obtained. This may include a digital elevation model describing the topography, geological maps showing the materials, water pressure data, and groundwater data.
However, these datasets may not all be describing the same area in the same level of detail. The geological map may cover a large area such as a district, whilst the engineering may only require information about a section of embankment. The groundwater data may not be a direct measurement under the building but could be from a nearby well. Care should be taken when combining the datasets. More data does not always mean the models will be more reliable.
Assuming we have a road embankment. We have observed that the behavior has changed after heavy rainfall using satellite data. However, if our field instruments have also observed increased water pressure, this can provide us with evidence to take a closer look. Engineers will still need to determine whether the change in behavior poses a risk to the road and what remediation should be carried out.
Detecting a Warning Is Different From Predicting Failure
A July 2026 paper in Scientific Reports examined this approach at the Cadia mine-waste storage facility in Australia, which failed in March 2018. Researchers combined satellite observations, deep learning, and a model of the behavior of the ground and dam materials.
Their analysis identified unusual movement associated with the area that failed. The visibility of the signals depended on the satellite observations and the direction of movement being examined.
That is promising, but the researchers were examining a known failure. An operating service faces a harder task: deciding which changes deserve attention before it knows the outcome.
An unusual reading could reflect construction activity, ordinary movement, or an error. Some dangerous failures may not produce a detectable warning pattern. The 2024 dam study specifically cautioned that many failure mechanisms might not show acceleration detectable by Sentinel-1 in time for a failure prediction.
For that reason, “predicting structural failures before they happen” is too broad a promise. A more credible aim is to improve monitoring, identify concerning changes, and direct inspections. Forecasts can support that work where their reliability has been demonstrated.
Where Pakistan Could Begin
A useful pilot would start with a limited engineering problem: a district with documented subsidence, for example, or a road corridor with recurring settlement. The team would need usable satellite records, information about the exposed infrastructure, and field measurements for comparison. Engineers should help define which changes matter before a model is developed.
Testing should use later observations that were unavailable during development. Otherwise, a system can appear convincing because it has already been influenced by the period it claims to predict. The results should answer practical questions. How often did the system raise unnecessary alarms? Which concerning changes did it miss? Did it give an inspection team enough time to respond? Did combining the data improve decisions compared with existing practice?
The budget must cover processing, field visits, instruments and staff. Obtaining imagery is only one part of running a service. A successful pilot would demonstrate a useful improvement in maintenance or investigation, rather than simply producing an attractive dashboard.
What the Public Needs to Know
A clear explanation should be provided in the ground movement map about what areas couldn’t be assessed.
Those living in the area will want to know how extensive the assessment is. It’s much easier to notice movement in a whole area than just looking at one house. In this case, you’re not going to be able to get a reading from a satellite and just know the foundations and structure of the building aren’t an issue.
If there is a warning issued by the authorities, then they should provide an explanation of what has been assessed, how up-to-date the information is, what isn’t certain, and whether an inspection will take place. They should avoid using technical colors that people may not have seen before to give them an idea if it is safe to use the road or stay in their building.
Responsibility falls on the organizations running the service.
A Call to Action
In Pakistan, they can use satellite radar along with ground measurements and AI to see where they need to carry out inspections. This can show movement that may not have been obvious from individual repairs or readings. Whilst there will need to be someone to actually do something upon receiving the warning. Depending on the location, this could mean investigating a leaking pipeline, checking if groundwater is being pumped, or looking at the soil under the road. The warning should be sent to those with authority and resources to investigate.
By creating a map of ground movement, it can help those involved to make repairs or do other work based on the findings to reduce the risk.
If you want to submit your articles and/or research papers, please visit the Submissions page.
To stay updated with the latest jobs, CSS news, internships, scholarships, and current affairs articles, join our Community Forum!
Find our FREE quizzes and MCQs here!
The views and opinions expressed in this article/paper are the author’s own and do not necessarily reflect the editorial position of Paradigm Shift.
Inayat Arshad is an AI researcher and engineer based in Islamabad, holding a BS in Computer and Information Sciences from PIEAS. Her work focuses on low-resource language technologies, where she develops open-source tools and foundational datasets to bridge critical digital divides for underserved communities, particularly Urdu-speaking populations. Additionally, she specializes in multilingual AI safety and explainable machine learning, working extensively to make modern AI systems equitable, transparent, and trustworthy. Outside of engineering, she is an avid reader of classical literature and existential philosophy.
