What’s more nerve-wracking than climbing an active volcano? Climbing it with kilos of delicate equipment, knowing that a single eruption could send both scientists and sensors hurtling downhill. Over the past five years, researchers from Mexico’s National Autonomous University have been doing just that on Popocatépetl, one of Earth’s most active volcanoes, driven by a mission to peer into its unseen depths. The result: the first-ever three-dimensional interior map of the entire volcanic edifice, built from seismic data processed with AI.
Popocatépetl rises 17,883 feet above central Mexico, sitting within reach of some 25 million people. Yet, unlike many other high-risk volcanoes, it had never been imaged in sufficient detail to reveal its magma storage zones. Earlier attempts more than a decade ago had produced contradictory results and lacked the resolution to show “how the volcanic edifice was being built,” as project leader Marco Calò explained. This time, the team expanded the monitoring network from 12 to 22 seismographs encircling the volcano an increase critical for resolving deep structural features.
Each of the seismographs captures ground vibrations at 100 samples per second, detecting the constant seismic chatter of magma on the move, shifting rock, and escaping gases. The data piles up, but PhD student Karina Bernal modified machine learning algorithms originally developed for other volcanoes to classify the unique tremor signatures at Popocatépetl. “I taught the machine about the different types of tremors there are in El Popo,” she said, letting the AI categorize the signals into classes with a speed and accuracy impossible to attain by hand.
This AI-driven seismic tomography is basically a medical CT scan, for a mountain. Seismic waves travel faster through solid rock and more slowly through either molten or softened material, enabling the researchers to infer composition, temperature, and state at depth. The resulting model reaches to 18 km below the crater and exposes a complex system: multiple magma pools stacked at various levels, separated by layers of rock, with the largest concentration toward the southeast flank. Complementary studies suggest a mushroomshaped magmatic system connected by a narrow conduit in which shallow zones are influenced by degassing that raises magma viscosity and deeper zones are pressurized by overlying rock.
Such detail is important for hazard forecasting. The activity of Popocatépetl usually develops a dome over its vent, which then collapses during eruptions to send ash and pyroclastic flows downslope. Since its reawakening in 1994, it has produced daily emissions of gas and ash, with major dome collapses in years like 2019 and 2023. Now, AI-enhanced mapping provides a baseline for tracking the changes in magma distribution that are key to anticipating shifts in eruptive style.
The expanded seismic network also enhances early-warning capabilities. While just three stations can trigger emergency alerts, a denser array is key to grasping the processes behind those alerts. This parallels advances elsewhere in volcano monitoring, where a few judiciously placed broadband seismometers have provided high-resolution subsurface images through the analysis of wavefields scattered from distant earthquakes.
The Popocatépetl expeditions also carry instruments besides those that analyze seismic waves to examine volcanic gas emissions-important because gas composition and flux can signal changes in magma depth and pressure. Although the volcano emits more greenhouse gases than almost any other in Mexico, its output is still dwarfed by the anthropogenic emissions of nearby Mexico City.
The list of engineering challenges is impressive. Stations have to withstand extreme conditions: high-altitude cold, corrosive gases, ash fall, and even interference from wildlife. Batteries can fail, rats chew through wires, and explosions can obliterate equipment. Yet when working, they provide steady streams of data that, with the help of AI, can be processed in a matter of hours rather than months, enabling near real-time updates to the internal model.
The 3D map is by no means an endpoint; it is a starting point because it raises more questions, like why seismic tremors are more frequent on the magma-rich southeast side, and opens possibilities for four-dimensional models that capture changes over time. As Calò succinctly put it, “We have a natural laboratory here… very important to be able to understand and give residents detailed, trustworthy information about what is happening inside the volcano.”
To science and technology aficionados, the mapping of Popocatépetl represents an important milestone-a melding of field endurance, sensor network design, and AI-powered geophysics that shifts the paradigm on how humanity sees, and prepares for, the forces moving beneath one of the world’s most watched volcanoes.
