Listening to Earth’s background hum, scientists mapped deep salt caverns that could store renewable energy at massive scale

Yunnan Province generates more renewable electricity than it can use. The fix may lie underground — in vast salt caverns that could store surplus energy as compressed air. But before engineers can safely pressurize those caverns, they need to know exactly what shape they are in.

The problem: the site sits on an urban fringe where road traffic, industrial machinery, and electromagnetic interference render conventional seismic surveys nearly unworkable. So researchers turned to the ground’s own background vibrations — the faint, constant hum the Earth produces on its own — to see what lies hundreds of feet below.

A renewable energy surplus with nowhere to go

Yunnan Province has built a substantial renewable energy portfolio — wind turbines, solar arrays, and hydropower dams whose combined installed capacity already surpasses the region’s peak electricity demand. The result is chronic curtailment: clean energy generated but never used, simply because the grid cannot absorb it all.

Large-scale storage is the obvious answer. Every megawatt-hour wasted represents both an economic loss and a missed opportunity to displace fossil generation elsewhere on the network.

Compressed Air Energy Storage, or CAES, ranks among the most promising technologies for storing electricity at bulk scale. The concept is straightforward: use surplus power to compress air underground, then release it later to drive turbines. Salt caverns are the preferred host — their extremely low permeability prevents air from seeping out, their creep behavior allows plastic deformation rather than cracking under pressure, and high solubility means large caverns can be carved out relatively cheaply through water injection, a process called solution mining.

Why mapping these caverns is harder than it sounds

Before any CAES project can proceed, engineers need precise data: cavern volume, geometry, and sealing integrity. These are not optional details — they underpin every stability and safety calculation that follows.

Conventional methods rely on active-source seismic surveys, which send controlled vibrations into the ground and record the echoes. The technique works well in quiet, open terrain. The Anning Basin site, roughly 16.8 miles from Anning City, is neither. Road traffic, industrial machinery, and electromagnetic interference all degrade signal-to-noise ratios, while rugged topography adds further complications. Previous evaluations elsewhere leaned heavily on 3D active-source data or well-logging — methods that could not deliver reliable results here.

Turning ambient noise into a subsurface map

The research team’s solution was to stop fighting the noise and start listening to it. They deployed Microtremor Array Measurements, known as MAMs, using the Spatial Autocorrelation method, or SPAC — which extracts Rayleigh wave dispersion curves from the Earth’s continuous background vibrations: traffic rumble, wind, ocean swell, and countless other sources, rather than generating artificial seismic signals.

The survey covered 31 lines totaling 76,000 feet, with all 1,188 measurement points georeferenced using Real-Time Kinematic GPS to a planar accuracy of ±8 mm. Terrain elevation varies by 328 to 656 feet across the site, so topographic corrections were applied to prevent geometric distortion. Repeat measurements at roughly 5 percent of survey points validated data quality, and the mean square relative error across the entire survey came in at ±2.96%.

What the ground revealed: Caverns with a split personality

The imaging resolved four salt caverns — designated An-1 through An-4 — to depths of up to 1,968 feet. Heights range from 147 to 623 feet, widths from 98 to 393 feet. Together, the four caverns represent an estimated total volume of approximately 15 million cubic feet.

Each cavern has a distinct two-part geometry. The lower section is broad and cylindrical, sitting within the J3an2 salt layer — a highly soluble halite formation that dissolves uniformly during solution mining. Above it, the upper section narrows into a cone extending into the J3an3 formation, which is interbedded with mudstone and gypsum. That conical shape forms as weaker roof layers collapse and settle into a self-supporting tapered arch, a geometry that improves geomechanical performance by redistributing overburden stress across the cavern crown.

Hidden channels and the sealing question

The survey also detected localized low-velocity anomalies in zones between wells An-1, An-2, and An-3 — pointing to hydraulic conduits formed by differential dissolution during brine extraction. Connected caverns do not behave as independent pressure vessels, so identifying these conduits early gives engineers the data needed to design safe gas containment systems.

Fracture zones in the overlying J3an3 formation reduce roof tensile strength. The deeper J3an2 halite exhibits viscoplastic behavior — flowing slowly under pressure and sealing micro-fractures in what researchers describe as “self-healing.” The passive nature of MAMs also carried a practical benefit: unlike active-source surveys, it produced no noise disturbance.

A quieter path to underground energy storage

The Yunnan study demonstrates that MAMs can replace or supplement active-source surveys when challenging environments make conventional methods impractical, delivering structural imaging to 1,968 feet depth without a single artificial seismic source.

For future work, the researchers recommend coupling MAMs with Wide-Field Electromagnetic and High-Density Resistivity methods — combining seismic velocity data with electromagnetic responses to better distinguish brine, sediment, and air-filled voids. For the Yunnan CAES project, the findings now provide the geophysical foundation needed for engineering design and safety evaluation. As renewable buildout accelerates globally, a method that converts urban noise into a subsurface imaging tool may prove relevant well beyond one basin in southwest China.