Out of sight

Seismologists and volcanologists are unusual amongst scientists in that for the most part, what they're studying are things that are permanently unavailable for direct observation.

Oh, sure, they can access the results on the Earth's surface; fault lines, lava flows, uplift or subsidence from magma movement, and so on.  But the actual processes -- the stuff down there that is causing it all -- is inaccessible.

The deepest hole ever dug is the Kola Superdeep Borehole, on the Kola Peninsula near the Russian border with Norway, which is an impressive twelve kilometers deep; but when you realize that's only one-thousandth of the diameter of the Earth, it puts things into perspective.  Even so, it was deep enough that the bottom had a measured temperature of 180 C -- hot enough to boil water, but far from hot enough to melt rock.  (It bears mention that a claim circulating last year that they'd gone down fourteen kilometers, hit temperatures of 1000 C, and could hear the screams of the damned -- because, apparently, they'd punctured a hole into hell -- was unfounded.)

So the fact remains that much of geological science is based upon inference -- not only using surface processes to infer what's happening in Earth's interior, but using data such as earthquake wave traveling speed to figure out what the mantle and core are made of, whether they're liquid or solid or somewhere in between, and how all that stuff in there is moving around.  And being inferential, our understanding of deep geologic processes is constantly subject to revision.

Which brings us to a study out of Utrecht University that appeared in the journal Nature last week, about a discovery showing that deep in the Earth's mantle there are two continent-sized subterranean "islands" at least a half a billion years old -- showing that the stuff down there isn't mixing around quite the way we thought it was.

The upper mantle has been thought of as basically a big recycler.  As pieces of the Earth's crust get forced down into subduction zones (marked by the oceanic trenches that neighbor some of the most tectonically-active regions on Earth), it melts and gets mixed into what's already down there.  Being colder than the surrounding rock, everyone thought the process was slow; other than the bits that get hot enough to melt and then rise to the surface, causing volcanoes like the ones in the North American Cascades, Andes, Caribbean, Italy, Japan, and Indonesia, the rest just has to sit down there till it blends into the material surrounding it.

Apparently some of this will need to be rethought, because these "islands" in the mantle are still holding together despite being so old that they "should have" completely melted away by now.

One of the chunks is under Africa and the other under the Pacific Ocean, and they were located by using the paths and speeds of seismic waves, giving them the moniker of LLSVPs (Large Low Seismic Velocity Provinces).  "Nobody knew what they are, and whether they are only a temporary phenomenon, or if they have been sitting there for millions or perhaps even billions of years," said Arwen Deuss, who co-authored the study.  "These two large islands are surrounded by a graveyard of tectonic plates which have been transported there by subduction, where one tectonic plate dives below another plate and sinks all the way from the Earth’s surface down to a depth of almost three thousand kilometers."

You might be wondering how they figured out that they are a half a billion years old, given that they're way out of reach of direct study.  That, in fact, is the most fascinating part of the study, and has to do with the fact that rocks which cool quickly (such as obsidian and basalt) have much smaller crystals than ones that cool more slowly (like granite and gabbro).  The molecular reassembly that results in crystal formation takes time, especially in thick, viscous liquids like magma, so if lava is rapidly cooled on the surface it doesn't have time to form crystals.

"Grain size is much more important," Deuss said.  "Subducting tectonic plates that end up in the slab graveyard consist of small grains because they recrystallize on their journey deep into the Earth.  A small grain size means a larger number of grains and therefore also a larger number of boundaries between the grains.  Due to the large number of grain boundaries between the grains in the slab graveyard, we find more damping, because waves lose energy at each boundary they cross.  The fact that the LLSVPs show very little damping, means that they must consist of much larger grains."

Large grain size = a long time spent underground.  Mineralogist Laura Cobden, who specializes in mineral crystallization rates in igneous rock, estimated that based on the inferred crystal size in the two "islands," they've been down there, relatively undisturbed, for around five hundred million years.

[Image from Deuss et al.]

So that's our cool science news from the geologists for today.  Two islands in the mantle that are stubbornly resisting melting away.  Why these structures have been so persistent is beyond the scope of this study; but as with all science, finding out something's there is the first step.  After that, the theorists can figure out how to explain it all.

Even if they never have a chance to see it.

****************************************

Earthquakes and sharpshooters

A guy is driving through Texas, and passes a barn.  It's got a bullseye painted on the side -- with a bullet hole in the dead center.

He sees two old-timers leaning on a fence nearby, and pulls over to talk to them.

"Did one of you guys make that bullseye shot?" he says.

One of them says, a proud smile on his face, "Yeah.  That was me."

"That's some amazing shooting!"

The man says, "Yeah, I guess it was a pretty good shot."

The old-timer's friend gives a derisive snort.  "Don't let him fool you, mister," he says.  "He got drunk, shot a hole in the side of his own barn, and the next day painted the bullseye around the bullet hole."

This is the origin of the Texas sharpshooter fallacy, the practice of analyzing an outcome out of context and after the fact, and overemphasizing its accuracy.  Kind of the bastard child of cherry-picking and confirmation bias.  And I ran into a great example of the Texas sharpshooter fallacy just yesterday -- a Dutch geologist who has gone viral for allegedly predicting the devastating earthquake that hit southeastern Turkey and northwestern Syria on February 6.

The facts of the story are that on February 3, a man named Frank Hoogerbeets posted on Twitter, "Sooner or later there will be a ~M 7.5 earthquake in this region (South-Central Turkey, Jordan, Syria, Lebanon)."  This, coupled with the fact that the day before, the SSGEOS (the agency for which Hoogerbeets works) had posted on its website, "Larger seismic activity may occur from 4 to 6 February, most likely up to mid or high 6 magnitude. There is a slight possibility of a larger seismic event around 4 February," has led many to conclude that they were either prescient or else have figured out a way to predict earthquakes accurately -- something that has eluded seismologists for years.  The result is that Hoogerbeets's tweet has gone viral, and has had over thirty-three million views and almost forty thousand retweets.

Okay, let's look at this claim carefully.

First, if you'll look at Hoogerbeets's twitter account and the SSGEOS website, you'll see a couple of things right away.  First, they specialize in linking earthquake frequency to the weather and to the positions of bodies in the Solar System, both of which are correlations most scientists find dubious at best.  Second, though, is that Hoogerbeets and the SSGEOS have made tons of predictions of earthquakes that didn't pan out; in fact, the misses far outnumber the hits.

Lastly, the East Anatolian Fault, where the earthquake occurred, is one of the most active fault zones in the world; saying an earthquake would happen there "sooner or later" doesn't take a professional geologist.

[Image licensed under the Creative Commons Roxy, Anatolian Plate Vectoral, CC BY-SA 3.0]

What seems to have happened here is that the people who are astonished at Hoogerbeets's prediction have basically taken that one tweet and painted a bullseye around it.  The problem, of course, is that this isn't how science works.  You can't just take this guy's one spot-on prediction and say it's proof; in order to support a claim, you need a mass of evidence that all points to a strong correlation.

Put a different way: the plural of anecdote is not data.

No less an authority than the United States Geological Service has stated outright that despite improvements in fault monitoring and our general knowledge about how earthquakes work, quakes are still unpredictable.  "Neither the USGS nor any other scientists have ever predicted a major earthquake," their website states.  "We do not know how, and we do not expect to know how any time in the foreseeable future.  USGS scientists can only calculate the probability that a significant earthquake will occur (shown on our hazard mapping) in a specific area within a certain number of years."

So what Hoogerbeets and the SSGEOS did was basically nothing more than an unusually shrewd guess, and I'd be willing to bet that the next "sooner or later" prediction from that source will turn out to be inaccurate at best.  Unfortunate, really; having an accurate way to forecast earthquakes could save lives.

But realistically speaking, we are nowhere near able to do that -- viral tweets and spurious bullseyes notwithstanding.

****************************************

A whole lot of shakin'

I didn't realize how complicated it is to calculate the magnitude of an earthquake.

Most of us are probably familiar with the Richter Scale, the one most commonly used in the media.  It was developed in 1935 by seismologist Charles Francis Richter to give a standard scale to measure the power of earthquakes.  The scale is logarithmic; each increase in one on the scale represents a ten-fold increase in intensity.  The scale is based upon the displacement amplitude on a seismograph at a distance of one hundred kilometers from the epicenter, starting with a magnitude 0 earthquake causing the needle to move with an amplitude of one micron.  The scale extends up to an unspecified "greater than 9" -- because at that point, pretty much everything in the vicinity, including the seismograph, gets completely pulverized.

When you start looking more closely, though, the problems with the scale start to become obvious.  First of all, if the measurement is being made one hundred kilometers from the epicenter, the terrain in between is a significant factor.  Tremors passing through material with a high amount of shear (such as sand or mud) will lose intensity fast, as compared to ones going through a material that is rigid (such as solid rock).  Second, the origin of the earthquake usually isn't at the epicenter, which is the point on the surface nearest the source; the origin is the hypocenter, directly underneath -- but which can be at any depth from right near the surface down to hundreds of kilometers down.  (The deepest earthquake ever recorded was a minor tremor off the island of Vanuatu in 2004, which had a hypofocus 736 kilometers deep.)  Then there's the fact that earthquakes can be of different durations -- a less powerful earthquake that lasts longer can do as much damage as a more powerful, but shorter, tremor.

Another problem is that earthquakes can result in differences in the oscillation of the waves relative to the direction they're moving.  This is largely due to the fact that there are three basic sorts of faults.  There are thrust faults or convergent faults, where two tectonic plates are moving toward each other; what happens then can be one plate being pushed underneath the other (subduction), which is what causes the quakes (and the volcanoes) in Indonesia and Japan, or the two plates kind of smashing together into a jumble, which is the process that created the Himalayas.  There are extension faults or divergent faults, where the two plates are moving apart; this usually creates smaller but more frequent quakes, and lots of volcanism as magma bubbles up from the underlying mantle.  This is happening in Iceland, and is also the cause of the Great Rift Valley in Africa, which will eventually peel off the Horn of Africa (Somalia and parts of Ethiopia, Kenya, and Tanzania) and open up a new ocean.  Last, there are strike-slip faults or transform faults, where the plates are moving in opposite directions on each side of the fault, such as the famous San Andreas Fault in California.

Map of the (known) tectonic plates [Image is in the Public Domain courtesy of NASA/JPL]

The problems with the Richter Scale have led to the development of several other scales of intensity, such as the Surface-wave Magnitude Scale (which is pretty much just what it sounds like, and doesn't take into account source depth), the Moment Magnitude Scale (which is based on the amount of energy released as measured by the amount and distance of rock moved), the Duration Magnitude Scale (which figures in how long the tremor lasts), and so on.  But these all use different numerical benchmarks, and given that the Richter Scale is more widely known, a lot of people have continued to use that one despite its downsides.

The reason all this comes up is a new study from the University of Southampton that has identified evidence of what appears to be the biggest earthquake known; an almost unimaginable 9.5 on the Richter Scale quake that happened in Chile 3,800 years ago.  Trying to find the epicenter brings up yet another problem with measuring quake intensity, because the evidence is that this particular quake originated from the rupture of a part of the thrust fault between the Nazca Plate and the South American Plate off the coast of the Atacama Desert -- a rupture that was one thousand kilometers long.

The result was a tsunami that deposited marine sediments and fossils of oceanic animals several kilometers inland, and then traveled across the Pacific Ocean and slammed into New Zealand, tossing boulders the size of cars over distances of hundreds of meters.  That region of the Atacama Desert had been inhabited prior to the quake -- astonishing considering how dry and inhospitable the place is -- but it was (understandably) abandoned by the survivors for a long while afterward.

"The local population there were left with nothing," said geologist James Goff, who co-authored the study.  "Our archaeological work found that a huge social upheaval followed as communities moved inland beyond the reach of tsunamis.  It was over a thousand years before people returned to live at the coast again, which is an amazing length of time given that they relied on the sea for food.  It is likely that traditions handed down from generation to generation bolstered this resilient behavior, although we will never know for sure.  This is the oldest example we have found in the Southern Hemisphere where an earthquake and tsunami had such a catastrophic impact on people’s lives.  There is much to learn from this."

The obvious next question is, "Could this happen again?"  The answer is not just that it could, but it will.  Probably not in the same spot, but somewhere along the many tectonic boundaries in the world.  Nor do we know when.  Earthquake prediction is very far from an exact science.  We have instruments like strain gauges to estimate the tension rock is experiencing, but that doesn't tell you what's going on deeper in the ground, nor when the rock will fracture and release that energy as an earthquake.  Predicting volcanic eruptions is much easier; vulcanologists have gotten pretty good at detecting magma movement underground, and recognizing when a volcano is likely to blow.  (This is why the ongoing hoopla about the Yellowstone Supervolcano is all hype; sure, it'll probably erupt again, but some time in the next hundred thousand years or so, and it's showing no signs of an imminent eruption.)

The Earth is a dynamic planet, and the plates on the surface are in constant motion, jostling, coming together, moving apart, a bit like ice sheets on a river when they begin to break up in the spring.  You can't help but be fascinated by the amount of power it's capable of -- a catastrophic release of energy so large that the scales we've developed to measure such things are all but incapable of expressing.

**************************************

Induction, inference, and the fate of six Italian seismologists

There are two basic kinds of reasoning; deduction and induction.  Deduction consists of putting together statements about generalizations, or categories, and then drawing a specific conclusion from those statements.  Induction is, in a way, the opposite; analyzing specific instances of phenomena, and then inferring a general, overarching pattern from them.

Neither is infallible.  Deductive logic, for example, is only as good as the premises.  The argument, "All dogs have tails; boxers do not have tails; therefore boxers are not dogs" is a perfectly valid piece of deduction (it follows the pattern called modus tollens), but it leads to a wrong conclusion because it started out with a false premise (that all dogs have tails).  In induction, it is the process of inference that can lead you astray; there might be instances you haven't considered, causalities about which you were unaware.  No one was more aware of this than Einstein -- when congratulated on experimental data having proven his Special Theory of Relativity, Einstein soberly replied, "A thousand experiments could not prove me right, but one could prove me wrong."

However, there is a vast public misperception that both of these methods of reasoning are (or should be) infallible.  Logic, of either flavor, should always lead you to correct answers.  More to the point, if scientists know what they are doing, they should be able to get it right every single time.  If they don't get it right, something serious is amiss -- perhaps they have a political agenda that they are trying to foist.  Maybe they fudged their data to get grant money.

Or maybe they're just criminally malfeasant.

That last one is the chilling conclusion reached in Italy Monday regarding six geologists, who the courts declared guilty of manslaughter because of their failure to predict the earthquake in L'Aquila in 2009 that killed over 300 people.  The judge sentenced each of them to six years in prison, and the government agency for which they worked (the National Institute of Geophysics and Volcanology) to pay 7.8 million euros ($10 million) in damages.  [Source]

This is such a bizarre miscarriage of justice that I barely know where to begin.

Saying "scientists are human, and therefore fallible" is only the shallowest layer of why this verdict is absurd.  It's more than just six fallible individuals who made a regrettable mistake; it's a complete misunderstanding of what science and inductive reasoning does, and in fact what it is capable of doing.  Scientific inference is never going to give you certainty; even in fields about which a great deal is known, and about which the mechanisms are generally understood (let's say, biological evolution), there will always be pieces of the puzzle that don't seem to fit.  There are still species (plenty of them) whose position on the Grand Tree of Life is poorly understood, and therefore subject to revision; there will always be features, adaptations, and structures still to explain.  If they weren't, well, we biologists would be out of a job, wouldn't we?  There'd be nothing left to research.

The necessity of maintaining an awareness of uncertainty, of living on the edge of what is known, is even more pronounced when you are in a realm of science about which the mechanisms are only partly understood.  Climatology falls into this category -- and so does seismology.  While we understand a great deal more about these subjects than we did fifty, or even twenty, years ago, they are not yet at the point of being models that can predict with anything near 100% accuracy.  (And as I pointed out above, 100% isn't reachable no matter what.)

The fact that scientists in general, and especially ones in fields that are perched on the edge of what is explainable, get it wrong sometimes is inevitable.  More importantly, these missteps aren't indicators of some hidden agenda, or of outright malfeasance; they are indicators that we don't fully understand the system being studied.  Which we already knew, right?  The fact that climatologists are nearly unanimous in attributing the climate changes we've seen in the past century to anthropogenic carbon dioxide doesn't mean that they can yet tell you how those changes will manifest in weather events day after tomorrow, or how far those trends will continue, or what the ultimate result will be for the Earth's climate.  The fact that seismologists understand a great deal about plate tectonics doesn't mean that they can tell you when and where the next major earthquake will strike.  We simply don't have enough information yet to make those kinds of pinpoint-accuracy predictions.

I trust science.  I trust the majority of scientists.  At the same time, I am always aware of its limitations and boundaries, and the fact that by nature, inductive reasoning gives you a tentative, incomplete picture of the world.  "Scientists are always at the drawing board," astronomer Neil DeGrasse Tyson said.  "If they're not at the drawing board, they're not doing science.  They're doing something else."  The six scientists who are facing prison terms in Italy are where they are because inevitably, the scientific process generates partial solutions and uncertain predictions.  That's simply how science works.  Progress is made in science not by someone having a flash of insight and figuring out the "right answer," but by slow, painstaking motion toward a model that seems to be consistent with everything that is observed, the majority of the time.

To put it bluntly, the judge who ruled them guilty of manslaughter apparently has no understanding whatsoever of science as a process.  My fear is that this verdict will place a further chill on scientific research -- a scary thought, in a time when accusations of political bias, and claims that pure research is a waste of money, are already chipping away at the public's perception of science as a worthy endeavor.  Who will want to publish a supported, but controversial, result if now you not only can be accused of having a secret agenda, or wasting money, but be found criminally responsible if your model turns out to be wrong, if your predicted results don't come to pass?  In one way, saying "scientists are only human" is relevant -- scientists have lives, and families, and value their freedom, and if they think that some idiot judge is going to imprison them for manslaughter because they failed to predict an earthquake, they are likely to leave the field altogether.  Which, incidentally, two other Italian seismologists, Mauro Dolce and Luciano Maiani, did on Tuesday after hearing about the verdicts.

The whole thing is a travesty of justice.  I don't know enough about the Italian judicial system to know with any certainty how likely it is that an appeal will be successful, but it is to be hoped that these men will ultimately be freed and their names cleared of these charges.  If not, I fear for the future of science, which remains our best and most reliable method for finding out about the world we live in.