A new study is raising doubts about the accuracy of earthquake records in the Cascadia region. Researchers looked at various turbidite layers from the Cascadia subduction zone, which date back about 12,000 years. By using a specialized algorithm, they evaluated how well these turbidite layers corresponded with one another. The findings revealed that the links among the turbidite samples were mostly no better than random chance. This suggests that the connection between turbidite layers and historical earthquakes is more ambiguous than previously believed.
The Cascadia subduction zone, located in the Pacific Northwest, has generated significant and damaging earthquakes that have devastated forests and triggered tsunamis reaching as far as Japan.
The last major earthquake occurred in 1700, but it is unlikely to be the final one. This region is now filled with bustling cities that host millions of residents.
Determining how often earthquakes occur and predicting the timing of the next significant quake is an ongoing scientific endeavor that requires examining geological signs of past earthquakes, such as disturbed rocks, sediments, and altered landscapes.
However, a study conducted by scientists at The University of Texas at Austin and their partners is challenging the dependability of a lengthy earthquake record derived from geological deposits known as turbidites found on the ocean floor.
The researchers analyzed various turbidite layers from the Cascadia subduction zone, which date back roughly 12,000 years, employing an algorithm to judge how well the turbidite layers correlated with each other.
The results indicated that in many instances, the correlation among the turbidite samples was no more reliable than random chance. Because turbidites can form from multiple causes, not just earthquakes, this finding implies that the relationship between turbidites and historical earthquakes is less certain than once thought.
“We want those referencing the timeframes of Cascadia subduction earthquakes to realize that these timelines are being scrutinized by our research,” stated Joan Gomberg, a research geophysicist with the U.S. Geological Survey and co-author of the study. “It’s crucial to conduct more research to fine-tune these timelines. What we do know is that Cascadia has been seismically active in the past and will continue to be so in the future, so it is essential for people to be prepared.”
The study’s outcomes do not necessarily alter the estimated earthquake occurrence rate in Cascadia, which is thought to be about once every 500 years, according to the researchers. This current frequency estimate relies on various data sources and interpretations, not solely on the turbidites examined in this research. However, the findings underscore the need for further investigation into turbidite layers, in particular, and their relation to large earthquakes.
Co-author Jacob Covault, a research professor at the UT Jackson School of Geosciences, noted that the algorithm presents a quantitative method that offers a consistent approach to interpreting ancient earthquake records, which are typically based on more descriptive geological analysis.
“This tool yields reproducible results, allowing everyone to observe the same findings,” Covault explained, co-principal investigator for the Quantitative Clastics lab at the Jackson School’s Bureau of Economic Geology. “While there may be room for debate regarding the results, at the very least, it establishes a baseline with a reproducible method.”
The study has been published in the GSA Bulletin and includes researchers from USGS, Stanford University, and the Alaska Division of Geological & Geophysical Surveys.
Turbidites are formed from underwater landslides and consist of sediments that have settled back to the seafloor after being disturbed by turbulent sediment flows. These layers display a distinct gradient, with coarser particles at the base and finer ones on top.
However, multiple situations can create turbidite layers. Earthquakes can trigger landslides that produce turbidites when the seafloor is disturbed, but storms, floods, and various other natural events can also cause similar effects, albeit typically over smaller geographical areas.
Currently, linking turbidites to past earthquakes usually involves locating them in geological core samples taken from the ocean floor. When a turbidite appears in roughly the same location across various samples over a wide area, it may be considered evidence of a past earthquake, according to the researchers.
While carbon dating can help pinpoint timing, uncertainties remain regarding whether samples that emerge around the same time and location are indeed connected by the same event.
To better understand the relationships between different turbidite samples, the researchers decided to apply a more quantitative approach—specifically, an algorithm known as “dynamic time warping”—to their turbidite data. This algorithmic technique has been in use since the 1970s and can be found in various applications, from voice recognition to enhancing graphics in virtual reality experiences.
This marks the first application of the algorithm to the analysis of turbidites, according to co-author Zoltán Sylvester, a research professor at the Jackson School who adapted the algorithm for this study.
“This algorithm has been a vital component of many projects I have undertaken,” Sylvester commented, “but it remains underutilized in the geosciences.”
The algorithm detects similarities between two samples that may change over time and assesses how closely the two sets of data match.
In voice recognition technology, it identifies key phrases, even if spoken at varying speeds or tones. For turbidites, it recognizes shared magnetic characteristics among different samples that may appear different depending on their location yet originate from the same event.
“Linking turbidites is quite challenging,” co-author Nora Nieminski, manager of the coastal hazards program for the Alaska Division of Geological & Geophysical Surveys, remarked. “Turbidites typically exhibit significant lateral variations that reflect their differing flow dynamics. Thus, it is not expected for turbidites to retain identical deposition characteristics over large or even short distances, especially along active margins such as Cascadia or across diverse depositional environments.”
The researchers also subjected the correlations derived from the algorithm to additional scrutiny by comparing them with correlation data generated using synthetic data from 10,000 pairs of random turbidite layers. This synthetic comparison served as a control for assessing any coincidental matches within the actual samples.
The researchers applied their technique to magnetic susceptibility data for turbidite layers in nine geological cores collected during a scientific expedition in 1999. They discovered that in most cases, the relationships among previously correlated turbidite layers were no better than random. The only exception to this pattern was seen in turbidite layers located relatively close to one another—no more than approximately 15 miles apart.
The researchers stress that the algorithm represents just one analytical approach for turbidites, and incorporating additional data could influence the correlation levels between the cores. However, based on these findings, simply finding turbidites at the same time and location in the geological record is insufficient for definitively linking them.
Moreover, while algorithms and machine learning techniques can assist with this analysis, the responsibility for interpreting the results and directing future research lies with geoscientists.
“We aim to answer questions rather than merely apply the tool,” Sylvester stated. “However, when you engage in this type of work, it compels you to think critically.”
