Lahars are a type of volcanic debris flow that occurs when water from rain or other sources mixes with volcanic debris sitting on the flanks of volcanoes. The often violent flows can damage infrastructure and threaten human lives. This project’s original proposal aimed to use infrasound (a form of sound at frequencies below human hearing range) as a way to gain insights of sound generation by lahars. As a complementary objective, the proposed study seeked to compare infrasound-derived lahar data with real-time rainfall distribution. This complementary objective was developed by the team at Michigan Technological University.

During the first field campaign in April-May 2021, the researchers installed a network of rain gauges on one of the active lahar channels on Volcan de Fuego, known as Las Lajas. A seismometer to measure ground vibrations generated by lahar activity was also installed. This deployment complemented the efforts by collaborators at Boise State University who installed infrasound sensors and video cameras for lahar monitoring, and the current surveillance operations by the local monitoring institute, INSIVUMEH. By the end of the first lahar season, it became evident that seismometers had the potential to provide valuable data on lahar initiation and propagation.

Over the following years, we employed seismometers and rain gauges to monitor lahars on Las Lajas and another lahar channel called Ceniza. Seismic data was spatially limited (few stations) prior to our deployment, but allowed us to retrieve a catalog of lahars for these two channels on Volcan de Fuego. This catalog was enhanced with observations from seismometers installed during this project. With this catalog, we are now able to describe what characteristics are common in seismic waves generated by lahar activity on Volcan de Fuego. Moreover, this allows us to track the changes these lahars incur as they propagate downstream.

With the rapid development of artificial intelligence-based (AI) techniques and their power for analyzing large data sets, we pivoted to incorporating AI techniques into an automated detection system into our objectives. To put this into scale, one of our seismometers running for a year and collecting 200 samples every second was capable of generating over 50 billion data points measuring lahar-generated ground vibrations (seismic waves) in three dimensions. While seismic data used in this project was obtained from seismometers with different specifications, these datasets all had the potential to be used in machine learning model (ML) training. With these, we developed LaharML, a machine learning-based algorithm that identifies portions of a seismic signal that were generated by lahar activity. This algorithm is in the process of being deployed into VICTOR, an NSF-funded platform that provides cyberinfrastructure for computational volcanology tools. These seismic datasets also made possible the development of a seismically-derived estimate of the size of lahars from ground rotation (caused by the mass of these flows moving along the channel and depressing the ground adjacent to the channel). In addition, students have used these datasets to get training in advanced processing techniques of seismic signals through this project.

Lahars on Volcan de Fuego are almost exclusively triggered by precipitation, so rain gauges are a critical component of this study. Our rain gauge network focuses on the same active lahar channels targeted by our seismic deployment: Las Lajas and Ceniza. We also included rainfall data from the summit of neighboring Volcan de Agua to constrain high elevation data at a volcano that is not erupting. On Volcan de Fuego, high elevation rain data from ground stations is virtually impossible to obtain given the dense vegetation and the persistent eruptive activity affecting the areas above the treeline. For this reason, we complemented our ground-based rainfall monitoring with radar measurements. The equipment, a portable micro-rain radar, was provided in collaboration with the National Center for Atmospheric Research (NCAR), a research center sponsored by the NSF. The radar allowed us to obtain first-ever remote measurements of precipitation in the high reaches of the Ceniza channel. Additional ML approaches will enable us to identify the common characteristics of rainfall that are associated with lahar activity. Rainfall data will also be instrumental to our ongoing efforts to model flow processes.

Overall, these project results have great potential to improve mitigation strategies in the form of alert systems and better hazard assessments. While this could be impactful to the communities surrounding Volcan de Fuego who are exposed to this type of hazard, we also expect these results and techniques to be widely applicable to other volcanic contexts and to non-volcanic debris flows as well.

Last Modified: 11/29/2024
Modified by: Gregory P Waite