Skip to main content

Analysis of Second Earthquake Reported in a century with Epicenter Beneath Lake Michigan

August 12, 2026

On Wednesday July 8, 2026, at 2:38 pm (19:38:44 UTC) a magnitude 2.9 (M2.9) earthquake occurred nearly 20 km offshore Evanston beneath Lake Michigan and at a depth of 5 ± 2 km. With a 16-km diameter circle of uncertainty around its epicenter, this “Evanston earthquake” is the second earthquake ever to be instrumentally recorded beneath Lake Michigan. The first earthquake had the same magnitude and occurred approximately 300 km further north, offshore and east of Door Peninsula, Wisconsin, just 2 years ago.

Given the Evanston earthquake’s vicinity near the densely populated Chicago area and the high competency for propagating seismic waves of the tectonically stable crust in this part of North America, hundreds of people felt the earthquake’s weak to light “shaking” and reported it to the United States Geological Survey (USGS) within the first 24 hours of its occurrence. News media reported at least as many people who felt it as who did not feel it. One person reported the water becoming “choppy” for a moment, which is a plausible result of the passing seismic waves.

 

Earthquake history around Lake Michigan

Whether or not similar earthquakes have occurred before these two in 2024 and 2026, especially during the previous century, is hard to say. The country was not densely instrumented with seismometers and the seismometers that were operating did not always have the needed sensitivity to pick up all M2.9 earthquakes in the broader region around them. This is expressed in the fact that eight of the eleven earthquakes for the region around Lake Michigan in the USGS catalog from 1900 to 2011 are M3.8 or higher, nearly one magnitude point higher than the Evanston Earthquake (Figure 1). Such higher-magnitude earthquakes are about ten times less frequent, which might imply that smaller magnitude earthquakes, such as the off-shore Door Peninsula and Evanston earthquakes, must also have occurred, but were not detected, recognized, recorded, and cataloged. There are three notable exceptions among the pre-2011 earthquakes that have magnitudes below 3.8:

map

Figure 1. Map of earthquake epicenters from the 1900-2026 (end date: July 9, 2026) catalog of the USGS (grey circles). Circle sizes are proportional to earthquake magnitude. The yellow epicenter is for the recent Evanston earthquake. The northern Illinois portion of this map is provided with more detail in Figure 2.

It is notable that most of these earthquakes are found on the northern Illinois side of Lake Michigan. The largest earthquake in Northern Illinois was the 1909 M5.1 earthquake (known as the “Aurora earthquake”) near Sandwich, Illinois.  Huysken and Fujita (2013) infer an earthquake epicenter for the Aurora earthquake on the Sandwich Fault Zone: an area of parallel faults that extends northwest for about 135 km from Will County to Ogle County, IL (Kolata et al., 1978). Closer to the Evanston earthquake but unrecorded by the USGS catalog, a few earthquakes or earthquake-like events have been reported in the southwest suburbs of Chicago in 1977, 1997, and 2010 (Carpenter et al., 2011; NBC Chicago, 2010).

In 2011, EarthScope’s USArray project (Meltzer, 2003), which our team at that time participated in, installed sensitive broadband seismometers roughly 70 km apart across the Lake Michigan region and the USA. This significantly decreased the minimum magnitude of earthquakes that could be detected and located. The USGS catalog counts 12 earthquakes between 2011 and 2026, only one 2015 M4.2 earthquake in Michigan has its magnitude above 3.8; the other 11 have smaller magnitudes. This roughly tenfold increase in total earthquake count, compared to the previous century, is commensurate with the higher occurrence frequency of lower magnitude earthquakes, down to M2.5, that are being cataloged. In other words, it is plausible that M2.9 or smaller earthquakes have occurred in the Lake Michigan area all throughout the previous century without being detected, recognized, recorded, or cataloged. Given that intraplate earthquake epicenters tend to be scattered across broad areas rather than concentrated along a major fault line, as interplate earthquakes tend to do, it is unlikely, though possible, that the subsurface of Lake Michigan would have been spared till now.

Finding tiny earthquakes through machine learning.

In addition to operating more sensitive seismic instrumentation since 2011 to facilitate the detection and location of epicenters of smaller magnitude events, new machine learning techniques, developed in recent years, are providing additional detections of earthquake signals in seismic waveform data recorded by these instruments. These signals are so obscured by noise that they went undetected by conventional algorithms and human eyes. In this manner and with the available data, our team already detected 4 recent small-magnitude earthquakes that are not in the USGS catalog, and we located their epicenters on the southern portion of the Sandwich Fault Zone (Table 1).

Table 1 Source parameters of four uncatalogued earthquakes near the Sandwich Fault Zone of Northern Illinois. Source depths were fixed at 10 km. The rotation angle of the error ellipses are with respect to the horizontal/x axis (pointing east) and positive counterclockwise.

epicenters

Origin time (UTC)

Latitude (°)

Longitude (°)

Error Ellipse

Semi-major axis (°)

Semi-minor axis (°)

Rotation Angle (°)

2016-03-26T08:43:49.140

41.462

-88.398

0.0510

0.0462

-33.294

2016-03-26T09:24:38.359

41.458

-88.418

0.0695

0.0413

-19.649

2016-06-20T08:41:49.710

41.314

-88.516

0.104

0.0473

-34.157

2024-07-21T17:17:09.160

41.600

-88.749

0.0967

0.0511

-125.532

 

map

Figure 2.  Map of an amalgam of earthquake epicenters in northern Illinois from multiple sources: the 1900-2026 catalog of the USGS (black stars) also used in Figure 1, the 1795-1977 catalog of Stover et al. (1979) of historical earthquakes without magnitudes (blue stars), the 2011-2014 catalog of the Array Network Facility that monitored data from EarthScope’s USArray’s Transportable Array (Astiz, personal communication) (red stars), and our team’s epicenter determinations for earthquakes detected by machine-learning (Thomas, 2026) (cyan stars). The black and red epicenters are accompanied by their year of occurrence and inferred magnitudes. The blue epicenter estimates are accompanied by their year of occurrence and a roman numeral that refers to the shaking intensity on the Modified Mercalli Intensity Scale of 1931 (Wood and Neumann, 1931). The cyan epicenters have magnitudes below M2.5, which have not yet been quantified. The green lines are inferred fault zones referenced by Huysken and Fujita (2013). It is likely that many other fault zones exist in the subsurface, most of which are dormant or healing through low seismic activity within our tectonically stable region of North America (Dieterich, 1994; Magnani et al., 2017; Heimisson and Segall, 2018).

 

Seismic waveform data

For understanding the context of the Evanston Earthquake, we analyze its seismic waves, which have been recorded by multiple seismic stations: USArray’s legacy network of seismometers (N4, LM, OH) and a few stations from the US National Seismic Network (US); Northwestern University’s professional seismometer, L44A, adopted from USArray in 2014; and a series of Raspberry Shake geophones (higher-frequency seismometers) installed on Northwestern’s Evanston and Chicago campuses. Figure 3 shows a record section of seismograms from these broadband seismometers, whose data are shared through a 42-year old data management facility, formerly associated with the Incorporated Research Institutions for Seismology (IRIS), renamed to EarthScope in 2023. Figure 5 shows the records of on-campus geophones. L44A’s data are also easily accessible through a range of tools provided by EarthScope and through Earthtunes, a smartphone application that sonifies seismic data by increasing its infrasonic pitches by a number of user-specified octaves. An audible example of the Evanston Earthquake is at this URL. Figure 4 shows photos of our team maintaining station L44A on July 13, 2026.record section

Figure 3. Record section of vertical component seismograms of the Evanston earthquake recorded by several stations of the N4, LM, OH networks as well as station L44A, supplemented with data from two stations of the US network (see the Data References section). The seismograms are band-pass filtered between 0.5 and 12 Hz. Geophone records are in Figure 5. The maximum vertical ground acceleration 6 mm/s2, less than 1% of g (the acceleration of gravity that keeps us on solid footing) and registered at L44A, and the maximum horizontal acceleration there was 30 mm/s2 in the north-south direction.

 

 field photos

Figure 4. Photos of our team maintaining and repairing station L44A on July 13, 2026. Anticlockwise from top right: The team before maintenance, clearing overgrowth by the solar panels, damage to the clock cable as a result of a dislodged PVC pipe, the team after maintenance. Listen to the crackle of our work at this URL; we recommend using the Earthtunes application directly to listen to this entire day of data at 9, 13, and 7 octaves of pitch elevation (at which of these three did you hear the earthquake signals that arrived during our work?).

 

CRACS data

Figure 5. Vertical component seismograms of the (A) Evanston Earthquake on July 8 at 2:38 PM local time and (B) its possible aftershock on July 9 at 10:42 AM local time (15:42 UTC), recorded at six Raspberry Shake seismometers installed within the Evanston (top five records) and Chicago (bottom record) campuses of Northwestern University.

 

Data Analysis

First, we identified direct P waves that propagated from the earthquake’s hypocenter to seismic stations on land and recorded whether they first moved the ground up or down, these binary data are called P polarities (Figure 6).

 

map

Figure 6 Map of the Evanston earthquake’s P polarities (black=upwards, red=downwards), interactively observed in the waveform data (Figures 3 and 5) for each of the stations on this map. Triangles represent so-called impulsive P polarities and circles represent harder to identify, so-called emergent P polarities.

 

Next, we identified the Central US (CUS) model of Herrmann (1979) as an appropriate seismic velocity model for the southern Lake Michigan region, based on qualitative agreement with receiver functions we constructed from teleseismic earthquakes recorded at L44A. A seismic velocity model is a model of subsurface properties that determine the propagation speed of seismic waves. Receiver functions are processed seismograms that highlight seismic waves generated boundaries between subsurface layers.

Then, by re-tracing the paths taken by these P waves back to the hypocenter, we inferred a faulting mechanism for the earthquake from the pattern of these up and down P wave polarities. Faulting mechanisms are typically represented on a map as an imprint of the bottom half (a stereonet projection) of a four-quadrant beachball, with the quadrants separated by two perpendicular planes: the fault plane and the plane normal to the slip vector on the fault plane. The orientation of the faulting mechanism reflects the strain tensor and the regional or local stress tensor field influencing the strain. Typically, the two opposing shaded quadrants represent extension and tensile stress, whereas the other two opposing quadrants represent contraction and compressive stress; both as deviations from the overall compressive pressure from the weight of 5 km of overlying rocks (overburden).

As suggested by the red-black pattern in Figure 6, it was challenging to find the best way to rotate and orient a four-quadrant beachball so that it produces this pattern of P polarities. Nevertheless, two different team members independently read polarities from the seismogram records and used two different methods to match a faulting mechanism to the polarities. Figure 7 shows both results. While these are valid preliminary results and encouraging in their coarse agreement with one another and strike-slip faulting, they can in theory be further constrained with amplitude values of Pand S waves.

Our preliminary strike-slip mechanism (Figure 7) is similar to but slightly rotated clockwise from faulting mechanisms for several of the better recorded, larger earthquakes in Figure 2, which generally have NE-SW oriented axes of relative contraction (Herrmann et al., 2011; Huysken and Fujita, 2013).

 

 faulting mechanisms

Figure 7. Map view of stereonet projections of faulting mechanism solutions for the Evanston earthquake. Left: Mechanism obtained from the impulsive P polarities in Figure 6, using an application called focmec (Snoke, 1984; Snoke et al., 2003) and allowing for the beachball to disagree with up to two polarities. Circles and triangles represent up and down polarities, respectively. The lines are possible fault planes. The T axis (axis of extension) points just W and down from N, whereas the P axis (axis of contraction) points westwards, roughly towards Kenilworth, IL, and also slightly down. Symbols on the edge of the circle correspond to the stations that are closest to the epicenter, including L44A, L46A, and the on-campus Raspberry Shakes. Right: Mechanism obtained from the P polarities indicated on the beachball alongside the corresponding station name using an application called SKHASH (Skoumal et al., 2024). Circles and plusses represent down and up polarities, respectively. The lines are possible fault planes. The grey quadrants represent extension and tensile stress and the white quadrants represent contraction and compressive stress. Note that in both mechanisms the NE-SW trending fault plane causes P waves of opposite polarities to reach Evanston and Chicago, respectively, where they were recorded by the on-campus geophones.

 

Furthermore, we searched for additional earthquake occurrences in the weeks before and after the Evanston earthquake.  We applied the machine learning algorithm EQTransformer (Mousavi et al., 2020) to our on-campus seismic data to detect potential fore- and aftershocks of the Evanston earthquake. The method detected one possible aftershock in the morning of July 9 (see Figure 5), which is partially hidden within the noise, but was still spotted visually by a team member who walked by the live data display of Northwestern University’s Raspberry Shake seismometers after the quake. We estimate its magnitude as about 2.2.  The waveforms of this potential aftershock are not overwhelmingly similar to those of the Evanston earthquake, which explains both why the more traditional method of matched filtering did not produce this or other aftershock detections, and that the EQT-detected aftershock might have occurred in a slightly different place, potentially closer to the on-campus geophones and farther from the broadband seismic stations, which did not record this aftershock in a detectable way.

Several community members have proposed the following hypotheses for the occurrence of this unusual earthquake:

  1. Hypothesis: A local heat wave in Chicago and over most the eastern USA may have caused or facilitated the earthquake. Our take: This is unlikely, given that the temperature of the Earth’s crust 5 km underground is already at least 50 oC, well above the atmospheric temperatures of the heat wave, and that atmospheric temperature fluctuations do not transfer deeply into the subsurface and nowhere near as deep as where earthquakes occur (though seasonal atmospheric temperature fluctuations do affect your planted tulip bulbs).
  2. Hypothesis: Glacial isostatic adjustment may have caused or facilitated the earthquake. Our take: Based on the spatial scale of this phenomenon, this would be equally likely or unlikely as it being the origin for or contributor to most other mid-continent earthquakes between the Rocky Mountains and the Appalachians. There is still not enough known about why, how, or when intraplate earthquakes occur, but if this is a or the main reason, it would apply to many intraplate earthquakes in the Midwest.
  3. Hypothesis: Climate change may have caused or facilitated the earthquake. Our take: The effects of climate change are not limited to the atmosphere. Climate change affects the cryosphere, hydrosphere, biosphere, anthroposphere, and –indirectly– the lithosphere (crust). That said, we have an insufficiently abundant catalog of earthquakes under Lake Michigan to investigate if and how climate change would or could have influenced the crust beneath Lake Michigan.
  4. Hypothesis: Changes in Lake Michigan’s water levels, which were at a multi-decadal high in 2020 and have dropped by about one meter since, caused or facilitated the earthquake. Our take: Indeed, the associated drop in water weight would slightly change stresses in the shallow crust beneath Lake Michigan, by about 10 kPa. It is conceivable that such a relatively rapid change increased the odds of the M2.9 quake to occur, for example by lowering the normal stress on the fault that slipped. However, similar drops in water level have occurred about five times over the past century (Figure 8). As outlined above, it is possible that these prior drops also led to low-magnitude earthquakes like the Evanston earthquake in the past. However, we do not know whether such earthquakes occurred, owing to not having had the ability to unambiguously detect, recognize, record, catalog, and locate the epicenters of these low-magnitude earthquakes before EarthScope’s USArray project in 2011. Moreover, in many regions sea tides change water levels by more than 1 m within 6 hours, that is, much more rapidly, without systematically changing earthquake statistics. Back to the Great Lakes, Yao et al. (2022) did not find a convincing correlation between Lake Erie’s water level changes and somewhat clustered off-shore and on-shore seismicity near the lake between 2013 and 2020. Nonetheless, to reject this hypothesis we’d need much more data. 
  5. Hypothesis: Plate-wide stresses within the North-American plate’s lithosphere caused or facilitated the earthquake. Our take: The North American tectonic plate is sandwiched between subduction zones and mid-oceanic ridges around most of its periphery, rendering its oldest, stiffest, central portion in a general state of compression, with minor shear components resulting from stress heterogeneity (Zoback, 1992; Levandowski et al., 2018). However, this stress field’s roughly NE-SW oriented compressional axis does not entirely agree with that of our preliminary faulting mechanisms (Figure 7). There is still not enough known about why, how, or when intraplate earthquakes occur, but if this is a or the main reason, it would apply to many intraplate earthquakes in the Midwest.

 

water levels

Figure 8. Water levels in Lake Michigan and other Great Lakes since the year 1918 (from NOAA). The x-axis labels are years. The curve in the top frame shows that Lake Michigan’s water level has dropped about a meter over the past half decade.

 

News programs of Chicago TV stations requested DEEPS’ commentary and data on the earthquake, which they reported on here:

 

Authors, Acknowledgements, and Disclaimer

This seismic analysis was prepared by the DEEPS seismology group, specifically by Victor Agaba, Albert Kabanda, Yoweri Nseko, Lucas Schirbel, Ann Thomas, and Suzan van der Lee, during the month that followed the earthquake. We thank Amanda Morris for connecting us to the news media, Emile Okal for numerous earthquake discussions including on this one, and Steve Matz for installing the Raspberry Shake sensors on campus. This article has not been peer-reviewed. We reserve the right to submit all or portions of this work for publication in a professional journal.

 

Data References

US: Albuquerque Seismological Laboratory (ASL)/USGS. (1990). United States National Seismic Network [Data set]. International Federation of Digital Seismograph Networks. https://doi.org/10.7914/SN/US

N4: Albuquerque Seismological Laboratory/USGS. (2013). Central and Eastern US Network [Data set]. International Federation of Digital Seismograph Networks. https://doi.org/10.7914/SN/N4

LM: Athena Geophysics LLC. (2016). Michigan Seismic Network [Data set]. International Federation of Digital Seismograph Networks. https://doi.org/10.7914/SN/LM

NW: Northwestern University (2014): Northwestern University Distributed Seismometers [Data set]. International Federation of Digital Seismograph Networks. https://doi.org/10.7914/bbsy-k258

OH: Ohio Geological Survey. (1999). Ohio Seismic Network [Data set]. International Federation of Digital Seismograph Networks. https://doi.org/10.7914/SN/OH

NOAA lake water levels (Fig. 7): National Oceanic and Atmospheric Administration,   https://tidesandcurrents.noaa.gov/products.html, accessed 7 Aug 2026.

 

Other References

Carpenter, P., McMakin, D., and Albrecht, M (2011). Seismicity of Northern Illinois: Insights from Recent Earthquakes and a New Review of Historical Resources (Abstract). Geological Society of America Annual Meeting.

Dieterich, J. (1994). A constitutive law for rate of earthquake production and its application to earthquake clustering. Journal of Geophysical Research, 99, 2601-2618, doi:10.1029/93JB02581.

Heimisson, E. R., and Segall, P. (2018). Constitutive law for earthquake production based on rate-and-state friction: Dieterich 1994 revisited. Journal of Geophysical Research: Solid Earth, 123, 4141–4156, doi:10.1029/2018JB015656.

Herrmann, R. B. (1979). Surface wave focal mechanisms for eastern North American earthquakes with tectonic implications, J. Geophys. Res. 84, 3543–3552.

Herrmann, R.B., Benz, H. and Ammon, C.J., (2011). Monitoring the earthquake source process in North America. Bulletin of the Seismological Society of America, 101(6), pp.2609-2625.

Huysken, K. T. and Fujita, K. (2013). A Reevaluation of the Intensities of the Northern Illinois Earthquake of 1909. Bulletin of the Seismological Society of America, 103 (5):2810–2823.

Levandowski, W., Herrmann, R.B., Briggs, R., Boyd, O. and Gold, R. (2018). An updated stress map of the continental United States reveals heterogeneous intraplate stress. Nature Geoscience, 11(6), pp.433-437.

Larson, T. H. (2002). The Earthquake of 2 September 1999 in Northern Illinois: Intensities and Possible Neotectonism. Seismological Research Letters, 73(5), 732-738.

Magnani, M.B., Blanpied, M.L.,  DeShon, H.R. and Hornbach, M.J. (2017). Discriminating between natural versus induced seismicity from long-term deformation history of intraplate faults. Sci. Adv. 3, e1701593, doi:10.1126/sciadv.1701593.

Meltzer, A., (2003). EarthScope: Opportunities and challenges for earth-science research and education. The Leading Edge, 22(3), pp.268-271.

Mousavi, S. M., Ellsworth, W. L., Zhu, W., Chuang, L. Y., and Beroza, G. C. (2020). Earthquake transformer—an attentive deep-learning model for simultaneous earthquake detection and phase picking. Nature communications, 11(1), 3952.

NBC Chicago (2010). ”Miner” Earthquake. NBC Chicago. URL https://www.nbcchicago.com/local/earthquake-hits-lagrange/1884555/.

Skoumal, R.J., Hardebeck, J.L. and Shearer, P.M. (2024). SKHASH: A Python package for computing earthquake focal mechanisms. Seismological Research Letters, 95(4), pp.2519-2526.

Snoke, J.A (1984). A program for focal mechanism determination by combined use of polarity and SV-P amplitude ratio data. Earthquake notes 55: 15.

Snoke, J. A., Lee, W. H. K., Kanamori, H., Jennings, P. C., & Kisslinger, C. (2003). FOCMEC: Focal mechanism determinations. International handbook of earthquake and engineering seismology, 85, 1629-1630.

Stover, C.W., Reagor, B.G., and Algermissen, S.T (1979), U.S. Geological Survey, Miscellaneous Field Studies Map MF-1143, Seismicity Map, Illinois, Reston, Virginia.

Stover, C.W., Reagor, B.G., and Algermissen, S.T. (1985), U.S. Geological Survey Open-File Report 84-225, Denver.

Thomas, A.M. (2026). Urban Intraplate Seismology and Geoscience Engagement in the Chicago Area. PhD Dissertation.

Wood, H.O., Neumann, F. (1931). Modified Mercalli intensity scale of 1931. Bulletin of the Seismological Society of America, 21 (4): 277–283, doi:10.1785/BSSA0210040277.

Yao, D., Huang, Y., Xue, L., Fu, Y., Gronewold, A., and Fox, J. L. (2022). Seismicity around Southern Lake Erie during 2013–2020 in relation to lake water level. Seismological Society of America, 93(4), 2268-2280.

 Zoback, M.L. (1992). Stress field constraints on intraplate seismicity in eastern North America. Journal of Geophysical Research: Solid Earth, 97(B8), pp.11761-11782.