Unzen and the inclined spine
Another week, another cheeky self-promoting post. This time, I have published a new paper in Solid Earth on my research on the eruption of Unzen volcano in Japan. The project involves a volcanic spine, thousands of volcanic earthquakes, and a whole medley of analytical tools. Interested? Read on...
Let's start from the beginning. Lava spines are a curious and spectacular formation found at volcanoes around the world. These plugs of lava are squeezed out of lava domes during eruptions, grinding as they go, and eventually reaching hundreds of metres in size. One such spine grew at Unzen volcano in Japan in 1994, and is the focus of the new paper. Extensive field and experimental investigations have already looked this event (e.g. Adrian already wrote about his research in a previous post), but my research was looking at whether seismic data recorded could help shed more light on the processes occurring as the spine grew.
So what exactly happened during the eruption at Unzen volcano in the early 90's? It began with a series of phreatic explosions as magma neared the surface, and was shortly followed by the growth of a new lava dome. Unfortunately, this lava dome was very unstable and was often collapsing, producing thousands of destructive pyroclastic density currents, forcing the evacuation of over 10,000 local residents. One of these currents was responsible for the death of 43 scientists and journalists who had gathered to watch the eruption. As the eruption was coming to a close in late 1994, the aforementioned spine was pushed out of the top of the dome, synchronous with some curious patterns in earthquakes and deformation.
If we plotted the number of earthquakes detected per hour, a curious pattern emerges. As spine begins forming, a regular 40 hour cycle can be seen. By using statistical tools to analyse the earthquake count, we confirmed that the pattern is real. The same pattern was seen in tilt records from the same time period at Unzen, and is also very similar to cycles observed at Soufrière Hills volcano in 1996-97.
All pretty cool so far, but how much more could we learn from the seismic data? Another interesting phenomenon observed at a number of volcanoes are groups of repeating earthquakes, or clusters (also known as multiplets or families). These clusters very likely come from the same source and their characteristics can give us more information about processes inside the volcano. At Unzen volcano, from 1st October to 15th November, we found 29 clusters in the earthquake dataset. The two largest clusters, cluster 1 and 2, were much bigger than the others and almost synchronous with the appearance of the spine.
Using two separate techniques, called singular value decomposition and coda wave interferometry, we can use these clusters to detect any subtle but detectable changes inside the volcano. To keep a long explanation short, these techniques look at relative changes in the repeating earthquake waveforms. What we find are small changes in arrival times for the earthquakes in each cluster. We interpreted this as the result of movement of the cluster source, combined with very small changes in the properties of the rock through which the earthquakes travelled.
To help explain our results, we came up with a concept for what may be happening just beneath the surface of the volcano during the spine formation. A key fact about this spine is that it was pushed out of the dome at a steep angle, rather than vertically (like at Mt. Pelée). This meant that, as the spine was growing, cooling down, and getting denser, this was affecting the pressure on its margins. This resulted in the sources for clusters 1 and 2 moving up and down (respectively) from their original locations on opposite sides of spine.
Now you're probably wondering what these findings mean in the greater scheme of things. We now have a better idea of what was happening inside Unzen volcano in late 1994. We've also demonstrated that by using evidence from field, experimental, and geophysical investigations we can come up with some really interesting and better constrained models. Additionally, by describing and analysing the seismic data from this episode on spine extrusion, it can contribute to a better understanding of seismic data from monitored volcanoes around the world.
Thank you for reading about my research. You can download the full open access Solid Earth paper here. If you have any questions or comments, feel free to get in touch!
 |
| Incandescent glow seen on the newly forming lava dome at Unzen volcano in 1991. Pyroclastic density current deposits can be seen in the foreground. These currents, also known as pyroclastic flows, were derived from collapses of the highly unstable dome. Photo credit: Fumiaki Kobayashi |
Let's start from the beginning. Lava spines are a curious and spectacular formation found at volcanoes around the world. These plugs of lava are squeezed out of lava domes during eruptions, grinding as they go, and eventually reaching hundreds of metres in size. One such spine grew at Unzen volcano in Japan in 1994, and is the focus of the new paper. Extensive field and experimental investigations have already looked this event (e.g. Adrian already wrote about his research in a previous post), but my research was looking at whether seismic data recorded could help shed more light on the processes occurring as the spine grew.
So what exactly happened during the eruption at Unzen volcano in the early 90's? It began with a series of phreatic explosions as magma neared the surface, and was shortly followed by the growth of a new lava dome. Unfortunately, this lava dome was very unstable and was often collapsing, producing thousands of destructive pyroclastic density currents, forcing the evacuation of over 10,000 local residents. One of these currents was responsible for the death of 43 scientists and journalists who had gathered to watch the eruption. As the eruption was coming to a close in late 1994, the aforementioned spine was pushed out of the top of the dome, synchronous with some curious patterns in earthquakes and deformation.
 |
| A side-on view of the spine as seen at the top of the Unzen lava dome in 2007. The final dimensions of the spine were 150 long, 30 m wide, and nearly 60 m high. Photo credit: taedhk/Panoramio |
If we plotted the number of earthquakes detected per hour, a curious pattern emerges. As spine begins forming, a regular 40 hour cycle can be seen. By using statistical tools to analyse the earthquake count, we confirmed that the pattern is real. The same pattern was seen in tilt records from the same time period at Unzen, and is also very similar to cycles observed at Soufrière Hills volcano in 1996-97.
All pretty cool so far, but how much more could we learn from the seismic data? Another interesting phenomenon observed at a number of volcanoes are groups of repeating earthquakes, or clusters (also known as multiplets or families). These clusters very likely come from the same source and their characteristics can give us more information about processes inside the volcano. At Unzen volcano, from 1st October to 15th November, we found 29 clusters in the earthquake dataset. The two largest clusters, cluster 1 and 2, were much bigger than the others and almost synchronous with the appearance of the spine.
 |
| Figure 3 from the paper. In (A) you can see all the clusters found from 1st Oct to 15th Nov 1994. Each line is a cluster, with each dot representing an individual earthquake in that cluster. The largest clusters, cluster 1 and 2, are highlighted. In (B) and (C), you can see the general waveform and frequency content of cluster 1 and 2, respectively. (C) details the hourly earthquake counts, and showing how much of the dataset is made by clsuters 1 and 2. You can also see the 40 hour cyclicity from 23rd Oct onwards. |
Using two separate techniques, called singular value decomposition and coda wave interferometry, we can use these clusters to detect any subtle but detectable changes inside the volcano. To keep a long explanation short, these techniques look at relative changes in the repeating earthquake waveforms. What we find are small changes in arrival times for the earthquakes in each cluster. We interpreted this as the result of movement of the cluster source, combined with very small changes in the properties of the rock through which the earthquakes travelled.
To help explain our results, we came up with a concept for what may be happening just beneath the surface of the volcano during the spine formation. A key fact about this spine is that it was pushed out of the dome at a steep angle, rather than vertically (like at Mt. Pelée). This meant that, as the spine was growing, cooling down, and getting denser, this was affecting the pressure on its margins. This resulted in the sources for clusters 1 and 2 moving up and down (respectively) from their original locations on opposite sides of spine.
 |
| The conceptual model derived from the observations made from seismic data recorded during the spine extrusion at Unzen volcano. This illustrates how the tilted spine affected the location of the sources for clusters 1 and 2, making them move up and down the spine margin, respectively. |
Now you're probably wondering what these findings mean in the greater scheme of things. We now have a better idea of what was happening inside Unzen volcano in late 1994. We've also demonstrated that by using evidence from field, experimental, and geophysical investigations we can come up with some really interesting and better constrained models. Additionally, by describing and analysing the seismic data from this episode on spine extrusion, it can contribute to a better understanding of seismic data from monitored volcanoes around the world.
Thank you for reading about my research. You can download the full open access Solid Earth paper here. If you have any questions or comments, feel free to get in touch!
Comments
Post a Comment