Underground gas storage can alter pressure deep below ground, raising an important question for operators and nearby communities: does the associated earthquake response remain stable as injection and withdrawal cycles continue, or does it evolve over time?
A new peer-reviewed study of the Serajeh underground gas storage site in central Iran suggests the answer can be surprisingly dynamic. Researchers identified 29,781 local earthquake detections, relocated a subset using multiple seismic stations, examined earthquake source mechanisms and compared the seismic record with operational gas data. They found evidence of a delayed association between injection and seismicity, followed by a pronounced drop in earthquake activity around April 2015, almost four years after storage operations began. Seismicity then remained lower even though cyclic gas operations continued.
The findings do not establish a simple one-to-one causal relationship between every injection cycle and every earthquake. Instead, they point to an evolving interaction between industrial operations and a pre-existing fault-reservoir system. That distinction matters for seismic hazard assessment because a storage facility’s earthquake behaviour may change over years rather than follow a fixed response to the amount of gas injected.
Tracking almost 30,000 earthquake signals
The study, published in Scientific Reports on 2 October 2026, focused on the Serajeh underground gas storage facility. The storage interval lies approximately 2.4 to 3.0 kilometres below the surface. To reconstruct local seismicity in greater detail than a conventional earthquake catalogue alone could provide, the researchers combined deep-learning-based earthquake detection with multi-station relocation, centroid moment tensor analysis and records of gas injection and withdrawal.
The deep-learning workflow detected 29,781 earthquake signals at a nearby seismic station. The researchers then used independently relocated earthquakes to test whether the time difference between arriving S and P seismic waves at that station could serve as a practical proxy for source distance. This allowed the much larger single-station detection record to contribute information about how local activity changed through time.
Relocated hypocentres formed a compact cluster around the Serajeh structure. Most were about 3 to 8 kilometres deep. This means much of the detected activity occurred below the 2.4 to 3.0 kilometre storage interval, although the earthquake cluster overlapped the reservoir area and extended beneath it.
That spatial pattern is important. It suggests the seismic response cannot be understood simply as small fractures occurring only inside the storage formation. The authors instead interpret the observations in the context of a pre-existing fault-reservoir system whose response may be influenced by changing subsurface stresses and pressures.
The strongest injection relationship appeared with a delay
When monthly gas injection was compared with seismicity, the cross-correlation reached a distinct maximum at an offset of approximately eight months. In other words, the strongest statistical alignment was not simultaneous. Earthquake activity appeared most closely aligned with injection after a substantial delay.
The researchers did not treat that eight-month value as an unquestionable physical constant. They tested the relationship using phase-randomised ReshuffleCorr analysis, which reproduced the feature, but additional tests designed to preserve aspects of the time series structure showed that both the global statistical strength and the exact timing depended on how temporal variability was handled.
This is a useful caution against an overly simple interpretation. A correlation peak can identify a potentially meaningful timescale, but it does not by itself prove that injected gas directly caused earthquakes exactly eight months later. Subsurface systems contain multiple pathways for pressure diffusion, stress transfer and fault response, while both operations and earthquake occurrence vary over time.
A major change appeared around April 2015
The clearest long-term change was a substantial decline in seismicity centred on April 2015. An objective change-point analysis identified this shift nearly four years after underground storage operations started.
Importantly, the decline was not merely an artefact of including the smallest detections, which can be especially sensitive to changes in monitoring conditions or detection performance. The pattern remained visible when the researchers applied a more conservative magnitude threshold.
After the change, seismicity stayed at a lower level despite continued cyclic injection and withdrawal. This makes the Serajeh record particularly informative. If earthquake occurrence responded to storage operations in a stable and proportional way, continued cycles might be expected to reproduce broadly similar seismic activity. Instead, the system appears to have evolved.
Several physical processes could potentially contribute to such evolution, including changes in effective stress, pressure redistribution and progressive adjustment on existing faults. The study does not isolate one definitive mechanism, so these possibilities should not be treated as proven explanations. Its stronger conclusion is observational: the temporal relationship between operations and local seismicity changed substantially over the monitoring period.
Faulting style remained broadly consistent
The researchers also examined focal mechanisms for selected earthquakes. These mechanisms describe the geometry and style of fault movement responsible for seismic events.
Across pre-operational, operational and later periods, the analysed earthquakes were dominated by strike-slip to oblique source mechanisms. Strike-slip motion involves rock masses moving mainly sideways past one another, while oblique mechanisms combine components of different faulting styles.
The persistence of broadly similar mechanisms while the rate of seismicity changed supports the interpretation that operations interacted with an existing geological structure rather than continually generating an entirely new style of faulting. It also reinforces why reservoir depth alone is not enough to define the zone that should be monitored.
Why this matters for underground storage
Underground gas storage is used to balance differences between gas supply and demand. Gas can be injected when demand is lower and withdrawn when it rises. Similar subsurface storage concepts are also relevant to emerging energy systems, including geological storage of other fluids.
The Serajeh findings show why seismic hazard assessment benefits from long monitoring windows. A short observation period could capture only the relatively active phase and imply that the response will persist, or capture only the quieter phase and underestimate earlier behaviour. Neither snapshot would represent the full evolution documented in this study.
The delayed statistical association also means that monitoring strategies should not necessarily expect seismicity to track operational changes immediately. Integrating high-resolution earthquake detection with accurately relocated events, focal mechanisms and operational records can provide a more complete picture than any one dataset alone.
For regulators and operators, the practical implication is not that cyclic gas storage inevitably produces dangerous earthquakes. The study does not make that claim. Rather, it shows that a site’s seismic response can be time-dependent and structurally complex, making continuous site-specific assessment more informative than assuming a fixed relationship between injection and earthquake activity.
Important limitations
The authors emphasise uncertainty in the timing and strength of the injection-seismicity relationship. Although the approximately eight-month cross-correlation feature survived one randomisation approach, structure-preserving tests showed that its statistical strength and precise timing were sensitive to the treatment of temporal variability.
The large detection catalogue also contains many single-station events. S to P arrival-time differences can provide a useful first-order distance proxy, and the researchers validated that relationship against relocated earthquakes, but single-station detections do not provide the same spatial precision as events constrained by multiple stations.
Finally, this is a detailed study of one underground gas storage system. Reservoir geology, fault networks, operational histories and background tectonic conditions differ between sites. The observed decline at Serajeh therefore should not be assumed to occur on the same schedule, or at all, at other storage facilities.
What the study does provide is unusually detailed evidence that induced or operation-associated seismicity can evolve over multi-year timescales. Nearly 30,000 detections reveal a system that did not simply respond in the same way year after year.
Source Information
Study: Temporal evolution and decline of seismicity associated with cyclic operations at the Serajeh underground gas storage, Iran
Authors: Ali Songhori, Habib Rahimi, Mohammadreza Jamalreyhani, Fenglin Niu, Badrodin Karampour, Mostafa Ebrahimi Moghaddam, Reza Zeynaddini-Meymand and Abdolreza Ghods
Journal: Scientific Reports
Published: 2 October 2026
DOI: 10.1038/s41598-026-74195-y
