Why are some faults so creepy?
We dive into a new Nature paper that makes a surprising claim about why some faults slide all the time, while others get stuck.
Citation: Bradley, K., Hubbard, J., 2024. Why are some faults so creepy?. Earthquake Insights, https://doi.org/10.62481/9fadae3b
Although most faults remain locked between earthquakes, some slip at a steady pace, releasing tectonic stress without causing shaking. A new paper published in Nature argues that this behavior is tied to the geometry of fault systems, “contradict[ing] the previous understanding that creep occurs along faults with high roughness, heterogeneous structure and certain compositions.”
We replicate their analysis, and find that the apparent correlation between fault network geometry and creep rate can better be explained by differences in the geology surrounding the fault. When geology is taken into account, the proposed correlation disappears.
In our opinion, the far-reaching claims of this Nature paper do not seem to be supported by the evidence.
A quick note: Like in many of our posts, we are not experts in this subfield. We are learning as we go. We may get things wrong occasionally. If anyone identifies an error, we will happily update this post! All of our scientific paper analyses always remain free to read and open to comment.
Earthquake Insights is an ad-free newsletter written by two earthquake scientists. Our posts are written for a general audience, with some advanced science thrown in! To get these posts delivered by email, become a free subscriber. If you would like to support our work here, please also consider a paid subscription.
A paper published this week in Nature by Jaeseok Lee and coauthors proposes a new reason why some faults have it easy - they can slip without sticking, gliding along in quiet peace - while other faults have to go through the more dramatic earthquake slick-slip cycle.
This paper touches on one of the hot topics of earthquake science. Slow and steady slip on a fault is called ‘creep’. Fault creep in the shallow crust is not very common, geologically speaking, but over the last few decades we have started to measure this slow and continuous slip on some faults. Understanding the conditions for fault creep can tell us a lot about how faults actually work.
A lot of effort has already gone into understanding the factors that could lead to fault creep. The usual geological challenge applies: we don’t actually get to observe the actively slipping fault surface (except in very rare cases like SAFOD, where the San Andreas Fault was drilled right through). Nevertheless, based on field, laboratory, and theoretical studies, there are a few factors that are widely accepted as important.
Physical theories of fault friction all agree that if enough stress is applied, any fault will eventually begin to creep. Basically, the points of frictional contact that lock the two sides of the fault together are not infinitely strong, and will begin to smear out or fail when stress is high enough, allowing the fault to slide. This creep caused by unusually high stresses is often observed after large earthquakes, when sections of the ruptured fault can continue to slip slowly for some years to decades (a process called postseismic creep). As the earthquake-induced stresses are relieved by the slip and decay away, the creep slows down and ultimately disappears.
However, some faults seem to creep even at low stresses (a process called interseismic creep, since it occurs between earthquakes, not tied to pulses of earthquake-related stress). This creep prevents or reduces the accumulation of seismic potential energy: if the creep rate is the same as the long-term fault slip rate, then no seismic energy is accumulated and the fault does not become more dangerous over time.
The usual interpretation of interseismic fault creep is that it happens when the sides of the fault are just too slippery to catch together, preventing or at least reducing stick-slip behavior. (For the cognoscenti, we have deliberately decided not to discuss more advanced concepts like rate-and-state friction. Also, while rock strength and the coefficient of sliding friction are actually different things, we will use the terms interchangeably here. This is just to keep things simple, but we recognize that the topic is more complex than we are making it out to be.)
There are a few reasons why a fault might be slippery. First, the rocks that the fault cuts through could be made out of low-friction materials, allowing the fault to slide at low stress.
What kinds of materials might allow this behavior? Well, one example is serpentinite, a green metamorphic rock that is created when mantle-like rocks react with water.
Serpentinite itself is quite weak, but it is probably still strong enough to make a fault lock up and become stick-slip. However, serpentinite isn’t the end game of slipperiness. In a fault zone, serpentinite can convert to talc through even more reaction with water, and talc is far slipperier. (Once a main component of baby powder, talc is now mineral-non-grata because it can contain asbestos, inherited from the original serpentinite.)
A brief historical digression
In 2010, a sordid legislative affair took place in California, which lists serpentine as the official state rock. Before the geologists begin to collectively rise from their seats and shake their fists, yes, serpentine is a mineral - the rock formed mainly of serpentine is called serpentinite. The difference is important, but the deep state does not care. Anyway, the State Senate originally passed a bill about compost with unanimous support, but then (!!!) entirely swapped out the language of the bill to instead dethrone serpentine as the state rock. We imagine this was done in the darkest moment of night, at a big round table with all participants cast in shadow.
The reason for this dark dealing was that serpentinite can contain chrysotile asbestos, the less carcinogenic version of asbestos.
When this revised bill was revealed to the public, a loose alliance of geologists, educators, and concerned citizens collectively rose from their seats, fists shaking, and mounted a vibrant defense of serpentinite. They pointed out the many geological and ecological virtues of the rock, and derided the guilt-by-association argument. Ultimately, they won: the mineral serpentine remains California’s state rock:
We look forward to a ten-episode Netflix adaptation of this entire affair.
OK, back to fault creep.
Another way to get a fault to creep is to fill the fault plane with pressurized water. High pore pressure effectively pushes the two sides of the fault apart, reducing the frictional contact area and allowing slip at lower stress. Some minerals also become much weaker in the presence of water, effectively lubricating the fault. Interestingly, the slipperiest minerals already have a lot of water caught up in their crystal structure, which can be released along the fault surface by temperature increases, thus increasing the pore pressure - so weak minerals and fluid pressure can work together to help a fault creep in several ways.
Thus, there are a lot of reasons why rock type might directly affect fault creep, and many studies have clearly demonstrated the close association between fault creep and geology. The explanation works so well that Kyle once suggested to a seismology grad student visiting a newly identified creeping segment of the Sumatran Fault that she should keep a lookout for serpentinite along the fault trace - and she subsequently brought a nice chunk back!
Coming back to the issue at hand, the Nature paper by Lee et al. takes a very different approach: it claims that the overall shape of the network of nearby faults determines where creep can or cannot happen. Where faults all agree - i.e., they all line up in the same direction on a map - creep happens. Where they don’t line up, the system locks up and stick-slip behavior happens. This result would indicate that fault geology is really not as important as we thought.
To get to this conclusion, the paper basically documents a strong correlation between two observables: creep rates and fault geometries.
Why is the paper printed in Nature? Well, the results are presented as being pretty revolutionary:
Our findings indicate that the accommodation of regional stress loading in large-scale geometrical incompatibilities may be a primary controlling factor on the seismogenic nature of faults.
…
Our findings contradict the previous understanding that creep occurs along faults with high roughness, heterogeneous structure and certain compositions.
…
Our study introduces a new perspective on earthquake behaviour, highlighting the importance of slip resistance caused by geometric incompatibilities rather than variations in laboratory-measured frictional properties.
So, we can see that there is a lot at play here.
Let’s take a look at what the new study does, and then dig into some of the details. Once we get a handle on the data and methods, we can ask ourselves whether the results really do add to our understanding of why some faults creep.
As with our other critical discussions of recent prominent papers, it is important to remember that this analysis simply represents our own reading of the paper and our own look at the data. Those who are particularly interested should certainly read the original paper and public peer review document, and make up their own minds about any disagreements. Because Nature papers are never published as open access, we suggests that those without a university affiliation but wanting to read the paper should contact the authors directly.
What does the paper say?
We will first try to summarize the methods of the paper as simply we can, and then we will look at the analysis and conclusions.
The Method
Lee et al. take advantage of a new compilation of fault creep rates for California, published by Kaj M. Johnson and others in 2022. That study brought together many different measurements of fault creep in order to support official calculations of seismic hazard. Lee et al. saw potential in this dataset: it would allow them to compare creep rates on faults to other observations, to see if there were interesting correlations. So, the authors reprocessed the measured creep rate data for California, resampling to a 10 km along-fault spacing to avoid over-weighting data from closely spaced measurements. This is a simple calculation that is not prone to major errors.
The meat of the paper is the comparison of creep rates with the local misalignment of mapped faults. Misalignment (technically, the misalignment ratio) measures the degree to which faults in a given neighborhood all line up with one another. This is a new metric that was developed in 2021 in another paper with several of the same coauthors.
The procedure for calculating misalignment is pretty simple. For a given point in California (or elsewhere), select all fault segments from the USGS active fault database that fall within a radius of 15 kilometers of the point. Project each fault segment (each line defined by two points on the fault) onto a north-south line, and add together all the projected lengths. If all of the faults segments in the area run straight north-south, then the result will simply be the sum of the segment lengths. In contrast, if they all run exactly east-west, then the result will be zero because there is no north-south component to any segment. Repeat this calculation for every compass direction, keeping track of the shortest and longest projected lengths. Finally, calculate the misalignment ratio by dividing the shortest length by the longest length.
If that explanation isn’t clear, you can ignore it and simply use this description:
If the faults in a neighborhood are all parallel to each other, there will always be one direction with zero length, and the misalignment ratio will be 0. This is the lowest possible value.
If instead there are many faults equally distributed in orientation, all with the same length, then the projected length in each compass direction will be the same, and the misalignment will be 1. This is the highest possible value.
The misalignment ratio is a very simple metric!
So, that’s it - the data here are really quite straightforward. The rest of the paper is simply comparing these two datasets and discussing the results. The authors provided their complete source data and code, allowing us to easily replicate the results and play around with the methods ourselves. We would like to note that all figures are regenerated from the provided code and are not copied from the paper itself. Many thanks to the authors for that!
The Results
So, what did they find? Here’s the map that shows the two datasets together - their figure 1, which can be perfectly reproduced using the provided code. The colored circles show the creep rate data (resampled to 10 km spacing), and the colored shadow behind the points shows the fault misalignment, which was calculated at each circle location and then interpolated to make a continuous raster. The faults themselves are shown in white.
There are a few technical issues we would like to highlight at this point.
First, the dataset does not include offshore fault traces, which are contained in a different database file in the ZIP file distributed by the USGS. For some western coastal faults, it is very likely that inclusion of these traces would change the fault misalignment value.
Second, the coloring of the creep rate points in the above figure is incorrect, compared to the legend. The San Andreas Fault at 121°W (left of the center of the figure) creeps at a rate of almost 30 mm/yr and should plot as almost white, but the plotted color indicates a creep rate of only just above 1 mm/yr. Looking at the code, we see that the value plotted is actually the log10(abs(creep rate)), not creep rate.
Third, because the absolute value of the creep rate is being plotted, some fault segments with a negative creep rate plot as if they are positive. Negative creep rates suggest that the fault is slipping backward - opposite the expected slip direction. In Johnson et al. (2022), the negative rates are proposed to be measurement problems and not actual backwards slip. We think these segments should probably be treated as having a creep rate of 0.
We modified the code to use a different color stretch for the creep rate data, with a linear rather than logarithmic stretch. In order to make it possible to see variations at low creep rates, this color bar saturates at ±10 mm/yr. Note that we use different colors for positive versus negative creep rates. The blue dots are where the creep rate is negative in the supplied dataset.
OK, those are some important problems to take note of, but it doesn’t seem like they would strongly affect the analysis or the results. Let’s move on to the statistics.
The statistical analysis is presented as violin diagrams (although the left-most violin looks more like a mountain dulcimer to us…) that compare the misalignment of creeping versus locked faults (figure 2 of Lee et al.). For those who haven’t seen violin diagrams, they are basically probability histograms turned sideways. Where the violin is wider, there are more values represented at that y-value. 50% of the data points fall within the vertical range of the black box, and the median value is shown by the light yellow dot (along with its uncertainty). Basically, if two violin plots are similar-looking and have a lot of overlap between the black boxes and medians, then the data going into the two plots are similar. Where the boxes/medians don’t overlap vertically, there is a significant difference.
Panel a at left separates the creep rate data into two classes: creep rate ≤ 3mm/yr (the violin labeled ‘Locked’) and creep rate > 3mm/yr (the violin labeled ‘Creeping’). This is the most important panel because it compares creeping vs. non-creeping faults.
The locked fault areas have a much wider range of fault misalignment (a taller violin) compared with creeping fault areas, and the median values are statistically different. The black boxes overlap a bit. (The left-most violin is wider because more data points count as locked than as creeping.)
Panel b at right shows that faults with the lowest misalignment values also have the largest range of creep rates. However, the tall and thin chimney of the left-most violin illustrates that only a few samples actually contribute to that extended range. In fact, these all come from the Parkfield area of the San Andreas Fault, where the creep rate is closest to the plate motion rate.
The conclusion of the study comes straight from this figure. Basically, a population of creeping vs. not creeping faults can be readily distinguished using only the local fault misalignment. This, the title of the paper: Fault network geometry (misalignment ratio) influences earthquake frictional behavior (creep rate).
Another aside. Why don’t violin plots ever actually look like violins? Other instruments clearly deserve more recognition as good substitutes for kernel density plots.
Our initial thoughts about the results
A common refrain in science is that correlation is not causation. When a correlation is demonstrated, this is typically only the first step toward establishing an actual causal relationship, if one exists. The strength of the correlation has no bearing on whether there is a cause-effect relationship.
In geology, we often have difficulty proving cause-effect relationships. This is because there are usually many factors at play, which arise together from the same physical system, and therefore strongly correlate with each other. To make things worse, we don’t even get to run experiments that control for confounding factors.
So, when a cause-effect relationship is proposed, many geologists will instinctively start to think about whether there are other correlated factors that are more compelling.
What about in this case? Well, we personally think there are strong correlations that explain the observations equally well, or better, and make more sense.
Here’s a simple demonstration: what happens if we simply separate the data by longitude, and make violin plots for western faults vs eastern faults? Let’s choose longitude -119.4° as our separation line - you’ll see why in a minute. Here’s the new version of Panel a (Panel b does not change because the data are all the same).
Looking at the new Panel a, we see that there is an even stronger separation between the two datasets (east vs west), compared with the published figure (creeping vs. locked). Now the vertical black bars do not even overlap, so the 50% of central data in each violin are mutually exclusive. This means that east vs. west is a stronger predictor of misalignment ratio than locked vs. creeping.
How that could happen? First, let’s look at a map of fault misalignment ratio for the whole region. To make this map, we used the provided code to calculate the values on a 0.1° grid, and interpolated the resulting data; this is pretty similar to the 10 km along-fault spacing used in the study. We also plot the fault map, with faster-moving faults colored orange/red and slower-moving faults colored white. (This is not creep rate, but the long-term slip rate provided in the fault database.) The offshore faults (shown in brown) were not included in the original study, and are not included in our misalignment calculations either.
There is a clear geographic pattern, with low misalignments west of -119.4° (give or take - now our previous selection becomes clear!), and higher misalignments to the east; the boundary is indicated with a dashed white line.
Does fault network geometry depend on longitude? No; longitude is an artificial construct of our north-centric worldview. However, by separating the creep rate data by longitude, we have also effectively divided them by geological setting: the Pacific - North American plate boundary plates changes fundamentally moving from east to west.
East of -119.4°, the plate boundary is actually partly collisional, with the Pacific Plate sideswiping North America as it moves northwest (much like a Ford F650 shouldering its way out of a packed Chili’s parking lot). Among other effects, this sideways collision has raised the Transverse Ranges and caused the Big Bend of the San Andreas Fault.
West of -119.4°, the plate boundary faults are perfectly aligned with plate motions.
We can see the difference easily by looking at GPS velocity vectors:
In the east, the vectors commonly are at an angle to the major faults. This is a simple result of the partially collisional tectonics.
In the west, the arrows are parallel to the major faults; the faults in this area are almost perfects transforms.
Collision zones are extremely complex from a geological perspective, because the fault systems are necessarily unstable and are forced to shift and evolve as strain accumulates in the crustal crumple zone. Many of the faults are thrusts and not strike-slip faults, so their surface traces are unlikely to be linear even in the best case. On the other hand, transform faults like in the west can retain their simple geometry no matter how far they slide, and the surrounding crust isn’t really forced to deform very much. This is a simple explanation for the higher fault misalignments in southeastern California compared with western California.
So, geography clearly correlates with fault misalignment, but what about creep rate?
You may have noticed that the areas of high creep rates mainly occur near the coast in western California. Below, we plot fault misalignment against creep rate, coloring the points by longitude. The blue-green points (west of -119.4°) tend to have higher creep rates than the yellow-orange-red ones (east of -119.4°). The highest creep rates generally are sourced from longitudes between -122° and -120° - light blue-greens.
At this point, we need to break out a geological map. (Finally!)
California is a great mosaic of crust that was assembled by multiple tectonic tear-ups and collisions over many millions of years. This history has left behind some distinct geological domains, with different rock types of different ages cropping out at the surface. When we look at a generalized geological map, one possible relationship between fault creep and geology jumps out at us. (Below, we have plotted the creep rate data in a black (not creepy) vs white (creepy) color scheme.)
West of -119.4°, the plate boundary faults cut through the rocks making up the Coastal Ranges (brown and green colors). These are mainly accreted Mesozoic oceanic rocks and sediments, which were scraped off of a subducting slab long ago, along with large slivers of oceanic crust and even slivers of mantle rocks. These rocks are the perfect geological hosts for serpentinites. For example, the picture at top left (by Andrew Alden) shows an exposure of the Calaveras Fault - the striated green rocks are serpentinites.
East of -119.4°, the rocks are very different. While the geology is really complex, most of the exposed rocks are crystalline - granites, granulites, gneisses, and many other metamorphic rocks - mostly colored a shade of red on the map. These represent a very different type of crust, one that is less likely to host slippery rocks like serpentinites.
Where the plate boundary fault crosses longitude -119.4°, the geology of the wall rocks fundamentally changes.
This geology was kneaded, proofed, and baked many millions of years before the San Andreas Fault system was even a shared twinkle in the eyes of the Pacific and North American plates. The route that the San Andreas Fault takes from the Gulf of California to the Mendocino Triple Junction certainly takes advantage of the pre-existing geology.
So, we argue that this geology is a confounding variable of profound importance. To explore the relationship between fault geometry and creep rate, we really should be looking at sets of faults that have roughly the same lithology - removing this confounding factor.
So, let’s try that. Let’s run the study again - comparing the misalignment of faults for creeping vs locked faults - but only in western California (longitude < -119.4°). This is a test of the variability of fault misalignment versus creep rate, that avoids mixing the different geological domains. The map below shows the awkwardly chopped-out data:
The plots above show that there is almost complete overlap of the creeping vs. locked data. In other words, fault misalignment does not differentiate fault creep, at least in western California.
What about eastern California? There are a lot fewer creeping segments there.
Well, we can’t make a violin (world instrument) plot for the creeping sites because there are too few. So, the comparison can’t really be made. However, the quantile boxes and medians still overlap, and we can’t conclude that there is a measurable difference.
So, in our opinion, the geology provides a simple and sufficient explanation for the observed correlation between fault misalignment and creep rates.
What if we look at only one creeping fault?
Maybe a better way to test this idea is to look at a single fault that shows a wide variation in creep rate and also a variation in misalignment. By severely limiting the variation in geological setting, this should give us a chance test how well one factor explains the other, at least for one fault.
So, we selected points along the Bartlett Springs Fault (sensu lato), shown by the yellow dots on the subsetted map at left. Geologists who know the Barlett Springs Fault personally might have something to say about the actual variation in geology along the fault length, but we don’t personally know much. In our opinion, the scatter plot at right shows no clear relationship between misalignment ratio and creep rate.
What about the fact that the one section of the San Andreas Fault with very high creep rate (20-30 mm/yr) has particularly low fault misalignment? That doesn’t seem disputable. Is this enough of an argument on its own?
Well, no. Consider the Elsinore Fault in southeastern California. This active fault system is not included in the creep rate dataset, because it is not known to creep. However, as an isolated, linear strike-slip fault, it has particularly low misalignment ratios - just as low as the fastest creeping fault in California - and yet there is no creep.
(We note that the long-term slip rate on the Elsinore Fault is much lower than that on the San Andreas - just a few mm/yr. So, the maximum creep rate on the Elsinore Fault could never reach the heights of the creeping San Andreas. Still, if the Elsinore Fault behaved like the creeping San Andreas, we would presumably see a measurable amount of creep on this fault.)
By the way, the Elsinore Fault cuts through batholithic rocks - granites.
A quick note about creep rates
Lee et al. base their analysis on a comparison of creep rate with fault misalignment ratio. We think this is problematic. Ultimately, it is not the actual magnitude of the creep rate that matters - it is the fraction of the long-term fault slip rate taken up by creep. A fault slipping 1 mm/yr on average, and creeping at 1 mm/yr is creepier than a fault slipping 30 mm/yr on average and creeping at 20 mm/yr (the other 10 mm/yr taken up by coupling).
Only where the entire plate motion is taken up on one main fault can creep ever become very fast. Where multiple faults exist in parallel, they are capped by the need to share the slip between them. For instance, the Superstition Hills Fault - near the southern tip of the fault system - has a long-term slip rate of 1.7-5.5 mm/yr. The reported creep rate on this fault is widely variable - ranging from 0.6-7.2 mm/yr. These numbers leave room for the possibility that all of the long-term slip on this fault is accommodated as creep. That would mean that in the plot below (Extended Data Figure 3 of Lee et al.), the magenta data point on the far right should look a lot more like the tall purple one on the left - if the data were normalized by long-term slip rate. In other words, the data allow for high creep fraction even for faults that have high misalignment.
It seems like this study should have examined creep fraction rather than creep rate. How much that would affect the results, given the other issues we have talked about, is unclear. Also, calculating the creep fraction requires making two separate measurements quite accurately (long-term slip rate and creep rate), which is is hard to do. (Although similar datasets of fault slip rate do exist.)
Testing the hypothesis outside of California
While the analysis presented in the main paper focuses on California, one of the few places with a lot of fault creep data, the Extended Data section also includes two test cases from other fault systems.
Based on the results of those tests, Lee et al. state:
The consistent negative correlations observed in other creeping faults globally indicate that the influence of fault complexity on fault creep may be a universal behaviour.
That is a pretty strong statement. So, let’s look at these two results in some detail - keeping track of whether these correlations could also be explained by geology.
Panels a) and b) show the data from the North Anatolian Fault in Turkey:
The North Anatolian Fault data do seem to show that high creep rates occur only on the isolated central section of the fault, where misalignment is low (we note that the fast creep at the western most point is known to be postseismic creep and should technically not be considered here).
These data might support the hypothesis - as long as geological explanations don’t already account for the correlation.
Kaduri et al. (2017) performed detailed field analyses of the composition of fault gouges in outcrops all along the same section of the North Anatolian Fault. They showed a clear relationship between the composition of fault gouges and the creep rate of the fault - where micas and clays exist in the gouge, the fault zone is narrow and the fault creeps. We notice that Kaduri et al. (2017) is actually cited in the Lee et al. (2024) manuscript, but only to indicate the section of the fault that creeps - there is no mention of their actual field observations or conclusions.
The figure below shows the conclusions of Kaduri et al. (2017).
The main point is seen by comparing panels a and b. Panel a shows the creep rate along the fault. Panel b shows the clay content - the slippery mineral content - of the fault gouges. Where abundant clays were found, the fault was more likely to be creeping. Where clays were not found, the fault definitely does not creep. Thus, the pattern of creep can be directly explained by local material properties of the fault zone.
Kaduri et al. conclude: “A clear correlation is shown between shallow creep and near-surface fault gouge composition: locked segments of the NAF are mostly composed of massive limestones without clay gouges, whereas creeping segments comprise clay gouges that contain low-friction minerals.”
Thus, it seems like the North Anatolian Fault does not escape the problem of confounding geological factors.
What about the Chaman Fault? Panels c) and d) below show the data.
Well, let’s say that - at least to us - the visual correlation between fault misalignment and creep rate in the scatter plot (panel d) is less than compelling in this case. But even if we did accept that there is some kind of statistically meaningful correlation, could it also be due to geology?
According to Barnhardt (2016): “Geologic field surveys report that the eastern side of the Chaman fault consists of flysch-type sedimentary assemblages that likely originate from the Makran accretionary prism, while the western side of the fault variously comprises of flysch, marine limestones, plutonic granites, and ultramafic intrusions [Jones, 1960; Lawrence and Yeats, 1979]. Ophiolites and serpentinite bodies have been identified along the active trace of the Chaman fault and other faults throughout the Chaman transform boundary [e.g., Asrarullah and Abbas, 1979; Gansser, 1979; Lawrence and Yeats, 1979; Tapponnier et al., 1981].”
A cross section from Tapponier et al. (1981) shows ophiolites (in black) along the Chaman Fault, where it aligns with the Kabul Suture. Ophiolites - remnants of ancient oceanic crust - are primo serpentinite habitat.
To better see this for ourselves, we spliced together a recent geological map of Afghanistan and a much older map of Pakistan (the map on the right in the figure below). We then roughly aligned the geology with a map of creeping vs. locked segments along the Chaman Fault produced by Barnhart (2016) (the map on the left). Using the geological map, we then roughly drew in the location of “green” geological units in the area surrounding the Chaman Fault. These units are mapped as mafic volcanics - likely sources of ‘slippery’ fault gouges.
Looking at the map, it seems like there is some correlation between the green blobs and the areas of observed fault creep. This suggests that slippery rocks are a potential explanation for the variation of creep rate along this fault, as well.
Is this a robust analysis? No - it’s ten minutes of cut-and-paste geology, using very generalized geological maps. However, it does highlight that the spatial patterns of creep along the Chaman Fault could also have (likely also has?) a geological explanation. Unfortunately, it does not seem likely that a major field campaign across the Pakistan-Afghanistan border is likely to solidify the observations along the Chaman Fault any time soon. If any reader is familiar with better maps/data for the Chaman Fault, do please let us know!
While it is fun to look at geology around the world, it takes a lot of time. Luckily, a review paper has recently been published that compiles the identified or proposed causes of fault creep globally. In most cases, the creep has been associated with variations in fault strength specific to the geology of the fault. However, we don’t know to what degree the variation along each fault has been related to the variation in these properties.
Reading the Peer Review
Nature is one of the few paid journals that offers up the complete peer review history for papers, although not the originally submitted manuscript or the correspondence with the editors. This is a tremendous advance in open science because it allows readers to see which parts of the paper received scrutiny, and see how the authors dealt with objections or onboarded suggestions. For open reviews to meet their great potential, people need to read them and discuss them, knowing that papers are always a work in progress and that it is only the final paper that should be subjected to criticism.
The peer review file contains a lot of information on sections of the original paper that were dropped, concerns of the reviewers that were dealt with by reducing the strength of the language of the claims, concerns about causation vs correlation, plus the lack of a physical model or calculation to support the newly proposed mechanism. We highly recommend that interested readers take a look for themselves.
Final thoughts, we promise!
Lee et al. (2024) propose that diverse near-field fault orientations somehow impede the ability of faults to creep.
We like the fault misalignment metric, because it is easy to calculate and clearly correlates with geological setting. It is easy to believe that this metric could be used to tease out more information about fault systems, although in practice there might be some problems. The fault misalignment ratio distills complex geological processes down to a bare number that may not have an easy interpretation. As an example, thrust faults often have a highly sinuous surface traces, which will produce higher misalignment even if only a single, perfectly planar fault exists in the neighborhood. In addition, it seems like the metric must depend on the approach to mapping the fault system; most fault maps are incomplete, choosing to include only a subset of fractures based on connectivity and slip rate. Before this metric is applied in statistical studies like this, it should be evaluated more carefully to understand its sensitivity to a variety of factors.
The proposed statistical correlation between fault misalignment and creep rate is probably just the shadow of a much more obvious correlation between creep rate and fault geology. This is most apparent for California, where the geology is well known, but also appears very plausible for the two global examples.
In the end, we think that the confounding factor of geology, which is so strongly supported by decades of research, contaminates the statistical analysis done in this paper and invalidating its conclusions.
In our opinion, this paper stands on very “shaky ground.”
References:
Aslan, G., Lasserre, C., Cakir, Z., Ergintav, S., Özarpaci, S., Dogan, U., Bilham, R. and Renard, F., 2019. Shallow creep along the 1999 Izmit Earthquake rupture (Turkey) from GPS and high temporal resolution interferometric synthetic aperture radar data (2011–2017). Journal of Geophysical Research: Solid Earth, 124(2), pp.2218-2236. https://doi.org/10.1029/2018JB017022
Barnhart, W.D., 2016. Fault creep rates of the Chaman fault (Afghanistan and Pakistan) inferred from InSAR. Journal of Geophysical Research: Solid Earth, 122(1), pp.372-386. https://doi.org/10.1002/2016JB013656
Cappa, F., Scuderi, M.M., Collettini, C., Guglielmi, Y. and Avouac, J.P., 2019. Stabilization of fault slip by fluid injection in the laboratory and in situ. Science advances, 5(3), p.eaau4065. https://doi.org/10.1126/sciadv.aau4065
Chu, S.X., Tsai, V.C., Trugman, D.T. and Hirth, G., 2021. Fault interactions enhance high‐frequency earthquake radiation. Geophysical Research Letters, 48(20), p.e2021GL095271. https://doi.org/10.1029/2021GL095271
Gratier, J.P., Richard, J., Renard, F., Mittempergher, S., Doan, M.L., Di Toro, G., Hadizadeh, J. and Boullier, A.M., 2011. Aseismic sliding of active faults by pressure solution creep: Evidence from the San Andreas Fault Observatory at Depth. Geology, 39(12), pp.1131-1134. https://doi.org/10.1130/G32073.1
Guillot, S., Schwartz, S., Reynard, B., Agard, P. and Prigent, C., 2015. Tectonic significance of serpentinites. Tectonophysics, 646, pp.1-19. https://doi.org/10.1016/j.tecto.2015.01.020
Harris, R.A., 2017. Large earthquakes and creeping faults. Reviews of Geophysics, 55(1), pp.169-198. https://doi.org/10.1002/2016RG000539
Hetland, E.A., Simons, M. and Dunham, E.M., 2010. Post-seismic and interseismic fault creep I: model description. Geophysical Journal International, 181(1), pp.81-98. https://doi.org/10.1111/j.1365-246X.2010.04522.x
Johnson, K.M., Murray, J.R. and Wespestad, C., 2022. Creep rate models for the 2023 US National Seismic Hazard Model: Physically constrained inversions for the distribution of creep on California faults. Seismological Society of America, 93(6), pp.3151-3169. https://doi.org/10.1785/0220220186
Lee, J., 2024. Data and code for ‘Fault network geometry influences earthquake frictional behavior’. Zenodo https://doi.org/10.5281/zenodo.10982013
Lee, J., Tsai, V.C., Hirth, G., Chatterjee, A. and Trugman, D.T., 2024. Fault-network geometry influences earthquake frictional behaviour. Nature, pp.1-5. https://doi.org/10.1038/s41586-024-07518-6
Lockner, D.A., Morrow, C., Moore, D. and Hickman, S., 2011. Low strength of deep San Andreas fault gouge from SAFOD core. Nature, 472(7341), pp.82-85. https://doi.org/10.1038/nature09927
Moore, D.E. and Lockner, D.A., 2011. Frictional strengths of talc‐serpentine and talc‐quartz mixtures. Journal of Geophysical Research: Solid Earth, 116(B1). https://doi.org/10.1029/2010JB007881
Muksin, U., Bauer, K., Muzli, M., Ryberg, T., Nurdin, I., Masturiyono, M. and Weber, M., 2019. AcehSeis project provides insights into the detailed seismicity distribution and relation to fault structures in Central Aceh, Northern Sumatra. Journal of Asian Earth Sciences, 171, pp.20-27. https://doi.org/10.1016/j.jseaes.2018.11.002
Tapponnier, P., Mattauer, M., Proust, F. and Cassaigneau, C., 1981. Mesozoic ophiolites, sutures, and large-scale tectonic movements in Afghanistan. Earth and Planetary Science Letters, 52(2), pp.355-371. https://doi.org/10.1016/0012-821X(81)90189-8
Very impressive analysis and breakdown of this paper. I haven’t gotten to read this yet so I really appreciated the summary.
A few of my thoughts:
- The arguments the authors put is compelling for orientation based controls on creeping vs. locking on fault systems and I don’t believe there is any fundamental flaws, and thus it is a strong support to their claim, excluding the argument of wanting more data. Which is why it was published in Nature :)
- Broadly I agree with you that there could probably be more data to support their claim, but this is a ubiquitous problem in geology where there is almost always going to be a lack of data and analysis in pretty much any study. Eventually, you run out of space and time to do a completely comprehensive analysis.
- As far as your argument that geologic controls are *probably* more important than alignment, I completely agree. California is a really complex region to study fault geometry and pointing primarily to alignments is possibly missing a lot of factors. On the coast, you have historic Farallon plate subuction uplift and possible fluid interactions in the crust, stress interactions with the Sierra Nevada-Great Valley block acting as a backstop and transferring stresses to the east, and rotation/extension in the Basin and Range from the northward movement of the Mendocino Triple junction.
- Just one example I’m very familiar with in terms of fault locking is the Alaska subduction zone off the Alaska peninsula. Two neighboring segments have vastly different fault behavior, where one is nearly completely locked, and the other almost entirely creeping. The trench orientation is uniform and plate age is consistent so fault alignment is a non-factor. A similar case exists in Mexico. Obviously subduction zones are a different animal and much larger singluar faults than a complex continental fault system like California, but just one example excluding fault orientations as any kind of factor in slip behavior of a fault.
Thanks again for your analysis, really enjoyed it!
I learned a lot!