No, we did not ‘predict’ an eruption

The last seven days have been a bit of a whirlwind for me. Not only did I start my new job as a lecturer at the University of Exeter (Cornwall campus), but I had a paper published in Scientific Reports. The paper generated quite a media buzz with all sorts of flashy headlines attached. Unfortunately, not all these headlines were strictly correct – there was definitely some sensationalised exaggeration going on, a point correctly picked up on in Erik Klemetti’s blog about how to be a savvy science-news reader. While my co-authors and I are obviously grateful for our research to get a bit of press, I want to take a second to explain the research and what it means in my own, un-edited words.

A selection of the news headlines…

First and foremost, we did not ‘predict’ an eruption. We never used that word, and we never will. A more common word to use in volcanology, and natural hazards research in general, is ‘forecast’. A forecast differs from a prediction as it includes a certain level of uncertainty, i.e., something MAY happen in some period of time, whereas a prediction might say something WILL happen on some given date (e.g., Friday 13th January 2017). You can think of it in similar terms to the weather, where you get a forecast of what the weather MIGHT do over the next 24 hours (and everyone knows that the forecast becomes less and less reliable as time period goes up – never trust a 5 day forecast, especially in Cornwall!). But enough about prediction vs. forecast – I’ll save that for its own blog post some other time*.

Our paper was about the Sakurajima volcano / Aira caldera magmatic system in Japan. Sakurajima is Japan’s most active volcano, producing regular small and localised explosions most days. However, the eruptive history of this volcano is also littered with more violent and deadly events. A large eruption in 1914 killed 58 people, produced huge lava flows and caused the ground to subside a metre leading to widespread flooding. Of concern to the local authorities now is whether this volcano is likely to see an increase in its current level of activity and produce an eruption similar in scale to the 1914 event.

Sakurajima volcano sits on the southern rim of Aira caldera

To answer this question our research aimed to improve the understanding of the volcanic system. We used data from highly accurate GPS sensors to measure the ground movement around the volcano resulting from the build-up of magma below, and then included data from seismometers and heat-flow measurements. Using advanced, 3D numerical models (that don’t rely on numerous over-simplistic assumptions like the previous models in use) we were able to reconstruct the magma reservoir that feeds the volcanic system. With our approach we could more accurately estimate the magma reservoir location, size and shape, as well as the rate, timing and mechanism of magma supply.

The results matched with other observations and presented a consistent view of the developing magma system – something of a rarity in Earth sciences. They showed that the rate of magma supply was greater than the rate of magma currently being erupted – causing the magma reservoir underground to swell.

Then, to apply our results to some basic eruption forecasting, we took the assumption that for another 1914-sized eruption, the volcano needs to store the same amount of magma that the 1914 eruption threw out – about 1.5 cubic kilometres. Using the magma supply and eruption rates we estimated, and assuming these rates are constant over time (not necessarily true), this would take roughly 130 years from 1914.

So we did not predict when a larger eruption would take place, we estimated the amount of time it would take to store enough magma for a larger, 1914-sized, eruption. An eruption that may or may not happen, or that may happen earlier or later as magma eruption and supply rates change. Our results could even be used to estimate a sliding scale of eruption size with time if all rates do remain constant.

A rough estimate of eruption size with time for a large (paroxysmal) eruption at Sakurajima, assuming all the rates we calculated in our research remain constant, and the entire volume that is stored is subsequently erupted.

Another cool point comes up when we apply our results to a past time period. There was also a large eruption in 1892, and if we use our 130 year timeframe again to accumulate 1.5 cubic kilometres of magma we would estimate a large eruption around 1912 – just two years different to the 1914 event. Not bad, eh?

And that was it **. There were no harbingers of doom in the paper, ‘just’ some fancy new models and application of the results to simple eruption forecasting.

Enjoying the sun and eruptions at Sakurajima in 2013.

* I predict that won’t happen in the next 7 days ;)

** Well, there was actually a lot more that went into the paper, of course. What was equally interesting for us was the fact that our new models provided much better insights (in our opinion) than the models that had been used previously, and these old models have been applied to dozens of volcanoes around the world since the late 1950’s. This leaves plenty of work to be done in the future! Plus more improvements can still be made to the techniques we developed, such is the world of science. Obviously this is not as sexy as apparently ‘predicting’ an eruption, and was not as widely reported by the media.

Published by James Hickey

Volcanologist | Sportsman | Blogger @Between_a_Rock | Sci Comm & Outreach | PostDoc @UOBEarthScience | Outdoor Enthusiast | Geophysicist | @VUELCO_Project

2 thoughts on “No, we did not ‘predict’ an eruption

  1. Any micro gravity work being done? Would be useful to help further constrain the processes and help distinguish the type of magma being intruded.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s