When a volcano erupts, the danger doesn’t just come from the blast. Rivers of lava carve through the landscape, swallowing roads and homes, while pyroclastic flows—scorching avalanches of gas, ash, and rock—race down the slopes at over 60 miles per hour, incinerating everything in their path. With temperatures hotter than 1,500°F, these flows are one of the deadliest forces on Earth.

Predicting where lava will flow and how pyroclastic surges will spread is crucial for protecting lives, but modeling these hazards is no simple task. Scientists rely on simulations to anticipate how an eruption might unfold, but these models require thousands of calculations—factoring in terrain, eruption size, and environmental conditions.

A Bottleneck in Volcanic Hazard Research

This was the challenge faced by a Colombian scientist studying volcanic hazards. Renette Jones-Ivey, a researcher at the University at Buffalo, was brought in to run set up workflows on Titan2D, a specialized software developed at the University at Buffalo to model geophysical mass flows like lava, pyroclastic density currents, and debris avalanches. “When I first started working on this, I had to run jobs manually,” Jones-Ivey said. “I would submit them, wait for completion, then start the next batch. It was slow and complicated, but Pegasus changed everything, automating the entire process.”

Pegasus, developed at USC Viterbi’s Information Sciences Institute (ISI), is a workflow management system that allows researchers to run large-scale computational tasks across high-performance computing clusters and cloud environments. 

Jones-Ivey quickly realized that the existing Pegasus configuration (using Amazon AWS Batch) couldn’t handle the massive volume of jobs needed. “There were certain problems I was running into, like I couldn’t get it to run over 100 jobs—it would get an error,” she said. “So I reached out to Karan Vahi for guidance.”

For the next month or two, Jones-Ivey worked closely with Karan Vahi, a senior computer scientist at ISI, and one of the lead developers of Pegasus. Together, they troubleshot the issue, adjusted the workflow, and reconfigured the system to handle the large-scale processing required. With Pegasus orchestrating the simulations, the research team was finally able to process the thousands of jobs needed for volcanic hazard modeling.

The Evolution of Pegasus

Pegasus was created to help scientists manage the growing complexity of computational research, a challenge that ISI Principal Scientist Ewa Deelman and her team recognized more than two decades ago. Deelman, who began developing Pegasus as an NSF-funded research project in 2001, saw that as scientific projects became increasingly data-intensive, researchers needed a way to efficiently schedule, track, and execute computations across distributed systems.

“Scientists shouldn’t have to spend their time troubleshooting computational workflows,” said Vahi, a senior computer scientist at ISI and Pegasus’ lead developer. “Pegasus was built to handle that complexity—automating job scheduling, managing data movement, and ensuring reliability across computing environments.” Vahi worked closely with Jones-Ivey to fine-tune her workflow, ensuring Pegasus was optimized for her needs.

Simulating Volcanic Hazards at Scale

Volcanic hazard assessments rely on probabilistic modeling, meaning scientists must run thousands of simulations to account for uncertainty in eruption size, terrain characteristics, and environmental conditions. This study used the specialized volcanology software Titan2D, which models volcanic hazards by combining physics-based simulations with digital elevation models, allowing researchers to predict how volcanic flows interact with natural landscapes.

But running Titan2D at scale isn’t as simple as pressing “go.” Each simulation has multiple steps, and results from one stage often feed into the next. Scientists must also track which jobs have completed, re-run failed simulations, and ensure data is correctly transferred between computing nodes. Pegasus automates all of these steps, ensuring that the entire workflow runs efficiently and without interruptions.

Beyond Volcanology: A Tool for Scientific Discovery

Although Pegasus was originally designed for fields like astronomy, bioinformatics, earthquake science, and gravitational-wave physics, its applications have expanded far beyond its initial scope. Today, it is used in climate modeling, medical imaging, and even cybersecurity. Deelman and her team continue to refine the system, with ISI computer scientists Mats Rynge and Vahi playing key roles in its ongoing development.

“I very much appreciate it when scientists like Renette share their experiences and challenges when running their workflows with Pegasus,” said Deelman. “This way we can make sure we improve our software so that the entire user community can benefit from it.” Vahi noted that Pegasus is one of the few open-source software projects developed at ISI that is actively downloaded and used by researchers worldwide. “Pegasus isn’t just about running workflows,” Vahi said. “It’s about enabling researchers to push the boundaries of what’s possible—whether that’s predicting volcanic eruptions, tracking climate change, or studying the cosmos.”

Pegasus is funded by the U.S. National Science Foundation under grant number 2138286.

Published on March 11th, 2025

Last updated on March 11th, 2025