Scientific fraud announced in two studies of the 2016 M=7.0 Kumamoto, Japan, earthquake

Posted on by

2016年M7熊本地震にまつわる二件の研究不正について
By Ross S. Stein, Ph.D., Temblor CEO

 

In one case, the fault rupture and its associated seismicity had been shifted toward a volcano in support of the author’s hypothesis; in the other, seismograms from temporary near-fault stations had been manufactured. In both cases, evaluation committees and universities called for retractions.

 

Citation: Ross S. Stein (2019), Scientific fraud announced in two studies of the 2016 M=7.0 Kumamoto, Japan, earthquake, Temblor, http://doi.org/10.32858/temblor.018

 

On Tuesday, March 26, Kyoto University held a press conference to announce that a panel of outside experts had concluded that Prof. Aiming Lin of Kyoto University had fabricated, mislocated, distorted, and misrepresented data in a paper published in the journal, Science, in October/November 2016. Endorsing the conclusion of the panel, the University called for the paper to be retracted by Science. Retraction has, so far, has not occurred, nor has Science published an ‘Editorial Expression of Concern’ which can appear when the integrity of a paper is questioned.

Prof. Lin told the investigation committee that he made careless mistakes because he was not accustomed to using drawing software, but he argued that the paper’s conclusion—that the 2016 rupture was arrested by the Aso volcano—was nevertheless correct. The Evaluation Committee and University administration respectfully disagreed. The Kyoto University administration stated in the press conference that while Prof. Lin would be subject to unspecified disciplinary actions, the investigation committee had exonerated his five coauthors on the study, four of whom also are associated with Kyoto University.

 

What is Prof. Lin accused of doing?

Lin et al. based their argument on the 2016 right-lateral fault rupture (whichever side you are on, the other side moves to the right), and its associated seismicity extending into the Aso caldera, where they contended the rupture ran into the magma body feeding the volcano, and stopped. The Lin et al. fault rupture (Sites 1 to 9, below) are seen to extend well into the caldera. But the independently mapped rupture (sites 1 to 5) terminates well outside of the caldera, contradicting Lin et al.’s argument that the magma stopped the rupture. Further, the Committee found that Site 5, which Lin et al identify as the site of peak right-lateral slip, in fact records only normal slip. Here are three key figures from the Evaluation Committee report, slightly modified for readability, and with the Japanese annotations removed.

 

A comparison of Lin et al. fig. 2A and the Evaluation Committee’s rectified figure on a shaded relief map. The Lin et al. rupture for the Kumamoto earthquake (black lines, above) extend into the caldera; the independently mapped rupture (red and green lines, below) do not. The caldera itself is mislocated in the Lin et al. map.
A comparison of Lin et al. fig. 2A and the Evaluation Committee’s rectified figure on a shaded relief map. The Lin et al. rupture for the Kumamoto earthquake (black lines, above) extend into the caldera; the independently mapped rupture (red and green lines, below) do not. The caldera itself is mislocated in the Lin et al. map.

 

The Evaluation Committee further found that the maps in Lin et al. were distorted (stretched) N-S vs. E-W, and that the caldera rim (white line above) was mis-registered by several kilometers in the Lin et al. (2016) maps. Many of the field points are mislocated in the map. For example, Lin et al. placed Site 5 on the caldera rim, but it is actually 2.8 km outside the rim. Sites 8 and 9, which Lin et al. place 1.5 km apart, are in fact 50-80 m apart.

 

A comparison of Lin et al. figs. 2A and 2B (middle and bottom panels) with the Evaluation Committee’s rectified and re-registered figures (top and middle panels).
A comparison of Lin et al. figs. 2A and 2B (middle and bottom panels) with the Evaluation Committee’s rectified and re-registered figures (top and middle panels).

 

The Evaluation Committee found that the fault slip model, from Koketsu et al. (2016), had been shifted by 5 km and rescaled, and with the epicenter repositioned 3.5 km. When rectified to its proper position, it is evident that the principal slip at depth, and not just at the surface, stopped short of the caldera, contradicting the Lin et al. thesis. Lin et al. Sites 4 and 5 are seen to lie in slip saddles (thin red lines connecting upwards to single arrowheads), not slip peaks as Lin et al. had argued Lin et al. (dashed red lines connecting downward to the double arrowheads).

 

A comparison of Lin et al. fig 1B with the Evaluation Committee’s superimposed boxes used by the NIED to plot seismicity cross-sections. They found that the boxes shown in Lin et al. designated by Roman numerals do not correspond to the numbered boxes used by the NIED. The base map has only one lat/lon mark, making its registration difficult and its distortion hard to detect.
A comparison of Lin et al. fig 1B with the Evaluation Committee’s superimposed boxes used by the NIED to plot seismicity cross-sections. They found that the boxes shown in Lin et al. designated by Roman numerals do not correspond to the numbered boxes used by the NIED. The base map has only one lat/lon mark, making its registration difficult and its distortion hard to detect.

 

Lin et al. used seismicity maps and cross-sections to argue that aftershocks accompanied the rupture into the caldera. But the Committee found that the seismicity cross-sections shown in Lin et al. figure 1B (above) were rotated clockwise by 30° and mis-located by up to 10 km eastward, giving the false impression that the active fault trace and aftershocks extended into the caldera (compare red numbered polygons with white roman-numeraled boxes above). As a result of the distortion and relocation of these cross-sections, the faults drawn in fig. 1C cross-sections do not connect to their surface traces.

 

Sloppiness or manipulation?

Immediately after a large damaging earthquake, field observations are difficult to conduct. Even using handheld GPS units, some mislocation of field points is possible. But the Evaluation Committee found that the mislocations and distortions altered the data in a manner to support the hypothesis of the authors, rather than as random errors associated with field reconnaissance. Deferring judgement on Prof. Lin’s intentions, they nevertheless called this fraud and falsification.

 

Fabricated Kumamoto seismograms by another author

On 18 March 2019, Osaka University announced that Dr. Yoshiya Hata, Associate Professor of the Graduate School of Engineering of Osaka University, faked some of the data used in at least five papers about the Kumamoto earthquake. Dr. Hata resigned his Osaka post and subsequently died. 

Although Dr. Hata claimed that he had installed a seismograph and collected the observation data shown in these papers, the Evaluation Committee empaneled by Osaka University found that he had instead fabricated the seismograms by manipulating data, essentially stretching and rescaling the wiggles of the seismogram, observed by seismographs installed by other institutes. The Investigation Committee has also determined that he falsified the theoretical calculation values to make the fabricated data appear more convincing. The Investigation Committee did not find that the co-authors were involved in the research misconduct.

 

The seismograms for the permanent KIK and MTO seismic stations are real; the TMP1-3 (temporary installations) seismograms are faked, and have been retracted by the journals in which they were published. The peak accelerations for the TMP stations are higher than for the others (1000-1500 Gal, or 1.0-1.5 g), which would have made them important for building design criteria.
The seismograms for the permanent KIK and MTO seismic stations are real; the TMP1-3 (temporary installations) seismograms are faked, and have been retracted by the journals in which they were published. The peak accelerations for the TMP stations are higher than for the others (1000-1500 Gal, or 1.0-1.5 g), which would have made them important for building design criteria.

 

In September 2017, one of Dr. Hata’s coauthors, Dr. Hiroyuki Goto from the Disaster Prevention Research Institute of Kyoto University, placed a statement on his website in Japanese and English warning that the Kumamoto seismograms contained “wide reaching errors,” errors he explained that another seismologist had brought to his attention nine months earlier, but that he had not fully investigated, for which he apologized.  

 

Why weren’t these errors discovered before publication?

Science depends on researchers who, when asked by journal editors, volunteer to review papers for their quality, originality, persuasiveness, and importance. Reviewers can choose to be named or anonymous when their reviews are transmitted to the authors, and they can also make comments that only the editor will see. While reviewing is a time-consuming effort that does nothing to advance one’s career, it is a collective community endeavor that scientists engage in to advance, distill, and diffuse knowledge, for which we take pride. Unlike reviewers, editors are not anonymous, but they are generally also volunteers, except at Science and Nature, where they are trained science professionals. But here’s what’s key: Most reviewers assume that the data in the manuscript are authentic; they view their role as deciding if the data support the author’s conclusions, rather than conducting a forensic analysis of its veracity. And so, reviewers, editors—and was the case here, even coauthors—can miss clues that something is not right. [Full disclosure: I am a former editor of the Journal of Geophysical Research, and former Chair of the Board of Journal Editors of the American Geophysical Union, and so dealt with fraud accusations, editor-author disputes, and co-author disputes, forwarding those deemed credible for greater scrutiny. So, many of these reviewing and editing mistakes and lapses apply to myself as well].

 

Self-corrective but slow

Fortunately, science is self-corrective, in the sense that a large number of subsequent papers on the Kumamoto earthquake reported results at odds with Lin et al., and took issue with its findings. But even though publication can be fast (4 months from the earthquake to publication for Hata et al, and 6 months for Lin et al), self-correction can be slow, as practiced by the journals, and by institutions of the authors that empanel evaluation committees, as is evident from the 28-month period from publication these papers to public exposure.

 

What next?

Papers that appear in peer-reviewed journals are not necessarily right; rather, they are not obviously wrong. While that may seem like a low bar, in fact it is a demanding and critical filter. Publications that are subsequently identified as fraudulent mislead both scientists and the public, and so must be retracted. The Hata-authored papers with the faked seismograms were retracted this week by their journals. If Lin et al. (2016) is not also retracted, the damage—both to our understanding of the Kumamoto earthquake, and to the integrity of science—will continue.

 

Citation: Ross S. Stein (2019), Scientific fraud announced in two studies of the 2016 M=7.0 Kumamoto, Japan, earthquake, Temblor, http://doi.org/10.32858/temblor.018

References

Asahi Shimbun (2019), ‘Kyoto academic used tampered quake charts in Science article,’ 27 March 2019, http://www.asahi.com/ajw/articles/AJ201903270040.html

Hata, Y., Goto, H. and Yoshimi, M. (2016), Preliminary analysis of strong ground motions in the heavily damaged zone in Mashiki Town, Kumamoto, Japan, during the main shock of the 2016 Kumamoto Earthquake (Mw7.0) observed by a dense seismic array, Seismological Research Letters, 87, 1044-1049.

Koketsu, K., H. Kobayashi, H. Miyake (2016), “Generation process of the 14th and 16th April 2016 Kumamoto earthquakes,” http://taro.eri.u-tokyo.ac.jp/saigai/2016kumamoto/index.html#C.

Kyoto University Press Release and Linked Evaluation Committee Reports (26 March 2019) http://www.kyoto-u.ac.jp/ja/about/events_news/office/kenkyu-suishin/kenkyu-suishin/news/2018/190326_1.html

Notice of Retraction by Seismological Research Letters (2019)
https://doi.org/10.1785/0220190066

Lin, A., T. Satsukawa, M. Wang, Z. Mohammadi Asl, R. Fueta, and F. Nakajima (2016), Coseismic rupturing stopped by Aso volcano during the 2016 Mw 7.1 Kumamoto earthquake, Japan, Science, 354, doi:10.1126/science.aah4629.

NIED Hi-Net seismicity and cross-sections associated with the 2016 Kumamoto earthquake, http://www.hinet.bosai.go.jp/topics/nw-kumamoto160416/?LANG=ja (2016)

Outline of the results of the investigation into allegations of specific research misconduct that occurred at Osaka University, 15 March 2019

RetractionWatch.com (2019), https://retractionwatch.com/2019/03/18/late-researcher-faked-kumamoto-earthquake-data-university-finds/

Latest posts by Ross Stein (see all)