A peer-reviewed paper that found a correlation between solar activity and earthquakes is sending tremors through the scientific community. Salon reached out to scientists to inquire after the research paper's seismic conclusions, and discovered a surprising amount of rancor among skeptical geophysicists and seismologists.
The paper, published in Nature's Scientific Reports, theorizes that clusters of protons from the sun may correlate to large earthquakes on Earth. The sun, like all active stars, constantly vents matter in the form of solar wind, which includes particles like electrons, protons, and ionized helium atoms. The distribution and density of these particles vary, as the sun has variable "weather" just as the Earth does.
Researchers Vito Marchitelli, Paolo Harabaglia, Claudia Troise and Giuseppe De Natale looked at data on earthquakes and solar activity to ascertain if there was a correlation. Their data consisted of worldwide earthquake data as well as measurements of solar protons from the Solar and Heliospheric Observatory, a joint satellite of NASA and the European Space Agency that was launched in 1995 and is still active.
Understanding the researchers' thesis requires a brief foray into statistics. If earthquakes happened at completely random times of day, one could map them on a chart depicting time, and the dots would be randomly distributed, with no apparent pattern. Moreover, the rate at which they occur would be statistically constant, albeit random — meaning they would favor no one time over any other one.
But evidently, earthquakes don't fit this pattern.
"Large earthquakes occurring worldwide have long been recognized to be non Poisson distributed," the researchers write. The Poisson distribution is a statistics term that formally describes a situation described above, wherein events occur at a constant rate and independently of previous events. Radioactive decay is a good example of something else that observes the Poisson distribution: individual events of atomic decay are randomly distributed, but over time they average out to something predictably constant, which is why physicists are able to predict precise half-lives for any isotope.
Because earthquake timing seems to fit some unseen pattern, they posit that big earthquakes must then "involv[e] some large scale correlation mechanism, which could be internal or external to the Earth." They continue:
"Till now, no statistically significant correlation of the global seismicity with one of the possible mechanisms has been demonstrated yet. In this paper, we analyze 20 years of proton density and velocity data, as recorded by the SOHO satellite, and the worldwide seismicity in the corresponding period, as reported by the ISC-GEM catalogue."
They conclude, "We found clear correlation between proton density and the occurrence of large earthquakes [of magnitute greater than 5.6], with a time shift of one day. The significance of such correlation is very high, with probability to be wrong lower than 10^-5" [meaning one in 10000].
In statistical terms, that would imply a very confident correlation between the two events, meaning proton density and the occurrence of these large quakes. Indeed, one in 10,000 is close to the gold standard of statistical certainty.
Despite this, multiple geophysicists and seismologists were very skeptical of the paper's conclusion for other reasons. Salon spoke with several who were critical of the findings.
"My first thought is that what keeps the global time distribution of earthquakes from being Poissonian is the occurrence of aftershocks," Tom Parsons, a research geophysicist at the United States Geological Survey, told Salon by email. "Once a catalog is declustered and aftershocks are removed, then it is difficult to show non-random inter-event times. As aftershocks are primarily a local phenomenon, it strikes me as unlikely that they are caused by global-scale solar activity."
In other words, Parsons notes that earthquakes aren't randomly distributed, and thus wouldn't adhere to the Poisson distribution, because often earthquakes aren't independent events: one earthquake can trigger a follow-up quake, or vice-versa. "Declustering" refers to separating related "clusters" of earthquake events, such as a quake and its accompanying aftershocks.
Co-author Giuseppe De Natale, when approached about this criticism, contended to Salon by email that "the large scale correlation among worldwide earthquakes is not only due to simple 'clustering,' meant as main shock-aftershocks sequences."
De Natale clarified that the researchers "absolutely don't say that the aftershocks are caused by solar activity." (Indeed, correlation does not equate to causation, as high school math teachers everywhere are fond of reminding us.) "In fact in this case we would not get a so high statistical significance for the inferred correlation," he continued.
De Natale also shook off the criticism about "clustering" of earthquakes and aftershocks undermining his thesis. He told Salon that the correlation they found pertains to "a very peculiar behavior of worldwide seismicity, which we are presently noting and studying, evidencing a tendency of global seismicity [. . .] to occur clustered in times, but at mutual distances of several thousands of [kilometers]; so, they cannot be 'aftershocks' in the common sense." He argued that this "seems to be strictly related to the triggering effect of proton density (and hence of solar activity)."
De Natale pointed to other earthquake incidents, including ones that were very deep underground, as "isolated events" that prove that the aftershock criticism is unwarranted. Even large, shallow earthquakes, De Natale says, did not exhibit Poisson distributions even after isolating them from aftershocks, meaning that they appeared to be correlated to something.
Greg Beroza, Director of the Southern California Earthquake Center and a professor of geophysics at Stanford University, had a similar reaction to the paper as Parsons.
"The claim seems extraordinary and the physical mechanism is obscure. This is an example of statistical seismology, and there are many potential pitfalls associated with that kind of work," Beroza told Salon by email.
De Natale responded to Beroza's criticism, too. "We agree that the physical mechanism we propose is still very qualitative and needs more experimental evidence," he said. "However, the evidence of correlation is extremely strong with probability to be wrong (i.e. just by chance) less than 1/100,000. Furthermore, we used several different, very advanced statistical techniques to confirm the correlation."
He continued: "So, it is possible that the physical mechanism is not the appropriate one, but, in our opinion, it is almost impossible to deny the existence of a clear correlation between proton density and worldwide earthquakes."
Another top seismologist called into question the rigorousness of the study and expressed deep skepticism about its conclusions.
"I'm not an enthusiast," John Emilio Vidale, a seismologist at the University of Southern California, told Salon. "There are a lot of red flags in that paper. I'm frankly surprised it made it through review. I have to admit I didn't go through the numbers and try to reproduce their statistical analysis, which would really be the key, but they kind of misunderstand a number of things and they approach the problem in a pretty roundabout way, so I'm fairly skeptical."
He added, "They make a whole set of arbitrary rules with who knows how many pre-parameters and then make these complicated plots whose meaning really isn't clear. So it's not a direct test of their ideas."
De Natale responded to this criticism: "We use just one parameter to infer the correlation: the level of proton density. Moreover, we use different and very advanced statistical procedures to statistically test the correlation." De Natale cited one of these statistical procedures, a "renowned" method to "test seismological forecast algorithms" known as the Molchan method.
De Natale added, "I understand these methods, for people not used to handle them, can be difficult to understand; but they are very sound — and the statistical results, showing such a high significance, are almost impossible to deny."
Whether or not solar activity is correlated with earthquakes on Earth, earthquake prediction has long been considered a holy grail of sorts for geologists and actuaries alike. Indeed, a robust means of predicting earthquakes could save lives, and aid disaster relief and preparedness. Yet probing deep within the Earth, to the point where faults might be observable, is an impossible task for any sort of scientific imaging due to the density of rock; rather, modern seismology relies on indirect observation and observational seismic data instead. Indeed, recent advances in earthquake prediction technology — such as those that predicted the increasing likelihood of a large Southern California quake — derive from computer modeling using seismic records and geologic models.