Scientists Create 3D Image of Popocatépetl's Interior

By training an artificial intelligence (AI) system to identify small seismic signals generated inside the Popocatépetl volcano, researchers from the Institute of Geophysics (IGF) of the UNAM hope to complement the ambient noise tomography they currently perform of its interior.

Marco Calo, researcher in the Volcanology department of the IGF, and doctoral students Leonarda I. Esquivel Mendiola and Karina Bernal Manzanilla, explained that the current network of seismic instruments around the colossus allows them to obtain images of the interior of the volcano with unprecedented resolution.

The methods they use are similar to medical X-ray techniques. However, in this case, an image of the internal structure of Don Goyo is obtained from changes in the speeds of the seismic waves that travel through its interior.

Calo's team uses two different methods, the first is seismic interferometry, which uses ambient seismic noise, and the second is monitoring the signals of small tremors that occur within the volcano for which they use AI. The information collected together will serve to improve the models that explain how Popo works and to support Civil Protection in making more informed decisions and governments to better safeguard the population.

Previously, he continued, the seismicity linked mainly to the central part of the volcano was studied, but there was a lack of images of what the contrast with the surroundings could be, and this is what the students are doing.

The volcanologist arrived in Mexico almost a decade ago and helped develop the current seismic monitoring network, which was improved in 2018 at the National Center for Disaster Prevention (CENAPRED), and a year later he carried out the “installation of a network of eight additional stations” managed by the IGF volcanology department.

Esquivel Mendiola, a PhD student in Earth Sciences specializing in seismology, said that he has collaborated with Calo in various ways since he was studying for his master's degree. Currently, under his guidance and the collaboration of Arturo Iglesias Mendoza, he already has the first images of the internal structure of the volcano at a depth of up to 10 kilometers (five above sea level and five below sea level).

He said: "What we see now is an image of the entire structure of the volcano, each point indicates the seismicity and the colors, the anomalies, which refer to areas where seismic waves propagate more quickly. We have a complete image of the horizontal and deep structure of the volcano."

The researchers pointed out that, although there is a theory that they only have one magma chamber, the reality of what is observed in Popo is that it is not a room or a sphere, but that the material moves through different spaces or conduits that it finds between the fractures of the rocks. The material that feeds volcanoes is deep, but in the most superficial or shallow parts there are reservoirs or areas where it is stored.

Bernal Manzanilla,also a PhD student in Earth Sciences, specializing in seismology, trains an AI system to identify different types of tremors that occur within the volcano.

What I have been working on is a computer model or program that I am teaching to differentiate and recognize different types of earthquakes in the volcano and, based on that, we are going to create a new tomography, like Leonarda Esquivel's, to focus on the shallow area and have greater resolution and detail of the structures inside."

By comparing the results from January 2019 to March 2023 obtained with this system and the CENAPRED monitoring, experts have found that AI largely succeeds in identifying seismicity in official reports, even finding new events that may have been very difficult to identify with a simple manual review. This additional information will be useful to visualize in greater detail how the shallow part of the volcano behaves.

The researchers specified that until now there has been constant communication with CENAPRED and it is expected that in the future the work of the AI ​​system can be integrated, to contribute to the monitoring of the volcano. Implementing velocity models, calculated from the two methodologies to improve the location of seismic activity, and complementing manual reports with the help of AI are research challenges that help develop more efficient and reliable monitoring systems.

One of the main challenges facing IGF researchers is to improve the speed of data acquisition, since they are currently working with posthumous records, and although in the case of the implementation of AI, a month of data can be processed in one hour, the objective is to take advantage of it in real time.

Calo recalled that Popo is one of the most monitored volcanoes in the world due to its proximity to Mexico City, Puebla, and the State of Mexico. To do this, cutting-edge technology is available to review seismic, gas, acoustic, magnetic, ash, and deformation movements, and even monitoring is done with satellite images and fixed day and night cameras.