Highlights
- Researchers used more than 200 drone videos, lava samples and numerical modeling to reconstruct how lava flow changed as it moved toward the ocean.
- Near the vent, the lava was nearly 80% gas bubbles by volume; as it traveled, gas escaped, crystals grew and the lava became much more resistant to flow.
- Bubble behavior, not just bubble abundance, strongly influenced how fast and how far the lava traveled.
- The findings suggest that accounting for changing bubble behavior can greatly improve lava-flow predictions.
The lava that buried entire neighborhoods during the 2018 Kīlauea eruption was composed of nearly 80% gas bubbles near its source. A recent study shows that those bubbles played a central role in controlling how fast and far the lava traveled, and that lava flow models need to account for bubbles to more accurately forecast where lava will stop.
One of Hawaii’s most destructive volcanic events in two centuries, the eruption lasted from May to September 2018, covering 13.7 square miles (35.5 sq. km)—an area more than half the size of Manhattan—and destroying more than 700 structures. The most prolific lava source was fissure 8, known as Ahu’ailā’au, which produced a fast-moving, river-like flow that reached the coast in five days.
The study, published in the Journal of Volcanology and Geothermal Research, draws on more than 200 drone videos, lava samples and numerical modeling to track how the fissure 8 flow changed in speed, internal makeup and viscosity as it traveled eight miles (13 kilometers) to the ocean. By treating lava as a mixture of liquid rock, gas bubbles and crystals, the study reveals how interactions among those components shaped the flow’s behavior.
“The findings show that getting the bubble dynamics wrong can throw off flow predictions dramatically,” said coauthor Einat Lev, an associate research professor at Lamont-Doherty Earth Observatory, which is part of the Columbia Climate School. “To make accurate predictions, you need the right description of how the lava changes as it moves.”
Reading the Lava Flow
Lava flow predictions shape where hazard zones are drawn and how communities respond to an eruption. “If there’s a lava flow heading toward your neighborhood, you want to know where it’s going to stop and how much time you have,” said Lev. “And if you define the hazard zone based on how far lava is going to go and you get it wrong, you might be under-insuring entire communities.”
To examine how the flow evolved, the researchers analyzed drone videos from 21 sites along the channel captured during the eruption. They used a technique called particle image velocimetry, which tracks pixel movement between video frames, to extract surface velocity measurements and then combined these with channel geometry and slope data to estimate the lava’s viscosity at each site.
They found that the lava moved at roughly 36 feet per second near the vent, with a viscosity similar to thick motor oil. Nearly seven miles (11 kilometers) downstream, the lava’s velocity had slowed to less than two feet per second, and its viscosity had increased substantially. The thickening process was gradual at first, then it accelerated with distance.
To understand why, the team used CT scans and microscopy to analyze lava samples, measuring changes in gas and crystal content at different points along the flow. The lava was extremely bubbly near the vent, with 79 to 88% gas by volume. As the lava traveled and gas escaped, its bubble content decreased to between 16% and 26% by 7.7 miles (12.5 kilometers) from the vent.
The bubbles’ behavior—not just their presence—shaped how the lava flowed. Larger bubbles deformed when the flow was strong enough to overcome surface tension, making the lava flow more easily. When the flow wasn’t strong enough, the bubbles retained their shape and acted more like solid particles, increasing resistance. As the lava cooled and large bubbles escaped, the lava’s overall resistance to flow increased.
Crystal content also rose along the flow, from about 6% at the vent to 18% at 12.5 kilometers. Farther downstream, cooling and crystal growth became the main controls on the lava’s movement, marking a shift from the bubble-dominated dynamics near the vent.
Flow velocity also changed over time due to processes occurring 40 kilometers away at Kīlauea’s summit. Near-daily collapse events—when parts of the summit crater fell downward—sent pressure pulses through the volcano’s plumbing system, triggering surges in the lava channel. Maximum flow velocities within 12 hours of a collapse were about 80% higher than during quieter periods. The amount of lava moving through the channel peaked about 3.5 hours after each event and then declined over the next 40 hours.
Improving Lava Flow Predictions
The study tested a lava flow simulation model called PyFLOWGO and found that how it represents bubbles strongly affects its predictions. When PyFLOWGO was run with incorrect bubble assumptions for the fissure 8 flow, the model predicted the flow would stop short of 4 kilometers, when the actual flow traveled 13 kilometers. When configured with decreasing bubble content and rigid bubble behavior, PyFLOWGO more accurately reproduced the flow’s observed velocities and length.
In addition to model accuracy, model accessibility and usability are also critical. Lev and her colleagues are developing a cloud-based platform to bring together modeling tools and datasets. Their goal is to make these types of simulations more accessible to researchers and hazard managers during volcanic eruptions.
The study was coauthored by Jasper Baur and Janine Birnbaum of the Lamont-Doherty Earth Observatory; Brenna A. Halverson and Alan Whittington of the University of Texas at San Antonio; Hannah R. Dietterich of the U.S. Geological Survey Alaska Volcano Observatory; and Julia Hammer of the University of Hawaii at Mānoa.
