Sunday, January 18, 2009

Predictions, Forecasting and Eruptions

A half-joking query that I'm sure every volcanologist has encountered in their career is: "When will the volcano erupt? When's the 'big one' coming?" A major misconception of the public is that volcanologists can predict eruptions. Volcanologists, on the other hand, prefer to say that they can sometimes forecast eruptions - but not all the time, and often only in rough terms. This is a situation where semantics can cause a rift between scientists and the public, and it's the responsibility of both the scientist and the layperson to help fix it - the scientist by defining their word choice, and the layperson by realizing that in volcanology, forecast and predict don't mean the same thing.

(Photo of the eruption of Mount Pinatubo, Phillipines, in 1991; courtesy US Geological Survey Photographic Library.)

In volcanology, a prediction would mean saying that a volcano will erupt with a specific type and magnitude of phenomena, at a very specific time. If I were to say, "Mount St. Helens will produce a Plinian eruption in the next month," I would be predicting the volcano's behavior. There's also a good chance I'd be wrong. Volcanologists do not make predictions.

A forecast, however, involves probabilities, much like a weather report. "There is a high probability that Mount St. Helens' activity will increase in intensity in the next few weeks," would be an example of a forecast. (NOTE: THIS IS NOT AN ACTUAL FORECAST - real ones for Mt. St. Helens can be found here.) A forecast is a statement of what volcanologists think may happen based on current monitoring and evidence of past eruptions. But even when they're based on the best data available, forecasts can be wrong; they are, however, a more responsible way of answering the "when will it blow" question.

Activity forecasts are usually associated with currently active volcanoes, or volcanoes which are showing signs of becoming active. If, for example, an earthquake swarm occurs under a volcano, scientists might take this to mean that magma is moving through the volcano's plumbing system, if they are the right kind of earthquakes. Or, an inflation of a volcano's flanks might indicate an influx of new melt into the chamber of an already active volcano. Volcanologists have learned what the general signs are that might indicate volcanic activity is about to start or change, but these signs vary significantly depending on the volcano.

In addition, most volcanoes have not been studied long enough to understand the full range of their behavior. 30 or 40 years might represent a volcanologist's entire career, but in geologic terms, it's an incredibly short period, especially considering that volcanoes can erupt over timescales of hundreds of thousands of years. (In addition, eruption styles can change, so even if scientists understand a volcano's current behavior, they can still be surprised by sudden shifts in eruption style.)

What this all boils down to is that NO ONE, at this point in the evolution of volcanology, can say exactly what a volcano will do. Volcanoes are constantly surprising even the people who have spent their lives studying them. Because volcanoes are natural systems, there is always a chaotic element to their behavior (scientists call this aleatoric uncertainty). Aleatoric uncertainty can be reduced by acquiring knowledge about a system, but can never be completely eliminated. Volcanologists do their best with the knowledge that they have, and try to take into account the full range of phenomena that a volcano can produce, but hazard analysis and mitigation are still difficult.

This is why websites like this one, which advertises a computer program that supposedly forecasts volcanic eruptions, are extremely misleading and even dangerous*. It is no great stretch to say that a volcano that has been active in the last decade will continue to be active, as the website's "forecast" does in many cases, but those volcanoes could just as easily cease their activity. Computer simulations of the plumbing systems of volcanoes are still being refined, and we'll probably never really be able to "see" inside a volcano with the accuracy that will allow us to predict its behavior. There are simply too many variables to take into account: different eruptive materials, structures, weaknesses, melt compositions and rheologies, gas contents, etc etc.

The Volcano Listserv recently distributed this email about the website I mentioned:
Please be advised that Volcano Listserv does not endorse the forecasts
made by R.B. Trombley or others that have not been vetted in the
peer-reviewed scientific literature. To our knowledge, Dr. Trombley does
not have training as a volcanologist, and his previous reports have
raised concerns among a number of volcano practitioners and
organizations (including WOVO, the World Organization of Volcano
Observatories) about the possibility of misinterpretation.

Dr. Jonathan Fink
Founder and Editor, Volcano Listserv

Dr. Warner Marzocchi
Co-chairman WOVO (World Organization of Volcano Observatories)

-----Original Message-----
From: VOLCANO [VOLCANO@asu.edu] On Behalf Of Kimberly Genareau
Sent: Sunday, January 11, 2009 7:54 PM
To: VOLCANO@ASU.EDU
Subject: Year 2009 Volcano Eruption Forecast

****************************************************************
From: R. B. Trombley,Ph.D. <swvrc@usa.net>
****************************************************************

The global volcano eruption forecast for the year 2009 as forecasted
by ERUPTION Pro. 10.7
can be found at the following URL:

http://www.intlvrc.org/forecast.htm
I'm going to repeat the warning: DO NOT use this website, or anything like it, as a means of deciding whether you are in danger of a volcanic eruption. Listen to the volcanologists who study a particular volcano, your local Civil Defense or other disaster prevention agency, the National Guard, etc. They are the ones who have the most experience in dealing with eruptions, and that experience is more valuable and reliable than anything a computer program can spit out.

*I did notice that the website for the program contains a disclaimer stating that the software is not the be-all and end-all of eruption forecasting, but I take issue with the fact that the software is in any way touted as a legitimate tool for hazard mitigation. Especially by a person who has absolutely no qualification as a volcanologist, and who is trying to sell an MS-DOS-based program for almost $500. I certainly wouldn't buy it - the program or the claims.

3 comments:

BrianR said...

Good post.

My current job involves stratigraphic 'prediction' and, like any prediction or forecasting, an awareness of the uncertainties is critical. Sometimes you have a lot of data and uncertainty is reduced (but not eliminated completely, as you point out) and other times you only have a bit of data and the uncertainty is higher.

I think a lot of geoscience is 'uncertainty management' ... collecting data, learning about every nook and cranny of a problem ... all to help reduce uncertainty.

What isn't communicated too well to the general public is that there are commonly broad range of uncertainties ... it's a matter of degree. But, your post just helped communicate this better!

GabbroGuy said...

A reporter finally did an expose on this guy (http://www.azcentral.com/arizonarepublic/news/articles/2010/06/05/20100605arizona-volcano-expert-questioned.html#ixzz0qBORUy7k). GVP has a short discussion of the danger of the related problem of inappropriate creation of alert levels (http://www.volcano.si.edu/faq/index.cfm?faq=07), which Trombley also does.

Jessica Ball said...

I've been following the news reports with interest - the bit in this post about Trombley kind of got buried in the bottom. (Glad you noticed it, though!) It really just kills me that journalists are so unaware of resources like the USGS and academic researchers that they'll call up this guy for an interview. Then again, it's a good chance for us to speak up.