A seismogram is a record of ground motion over time, and reading one comes down to recognizing a few key features: the arrival of different wave types, the size of their swings, and the timing gaps between them. The tall, jagged wiggles you see are not random noise but a layered story about where an earthquake happened, how big it was, and how deep beneath the surface it originated. Once you know what each part of the trace represents, even a basic seismogram becomes surprisingly readable.
What a Seismogram Actually Shows
A seismogram plots ground motion on the vertical axis against time on the horizontal axis. The baseline, when nothing is happening, sits as a more or less flat line. When seismic waves arrive, the line begins to deflect up and down. The height of those deflections (the amplitude) tells you how strong the ground shaking is at that moment. The spacing between peaks and troughs tells you the frequency of the waves. Short, tight wiggles are high-frequency; long, rolling undulations are low-frequency.
Modern seismometers typically record motion in three directions: vertical (up-down), north-south, and east-west. A single earthquake therefore produces three traces, each capturing a different component of the ground’s movement. Some analytical techniques combine all three components to reconstruct the full three-dimensional motion of the incoming waves, which helps determine both the direction the waves came from and what type of wave is arriving.1GeoScienceWorld (Bulletin of the Seismological Society of America). Three-component analysis of regional seismograms
P-Waves and S-Waves on the Trace
The first signal to appear on a seismogram after an earthquake is the P-wave, short for primary wave. P-waves compress and expand the rock they travel through, like a slinky being pushed and pulled. They move fast through both solid rock and liquids, so they always arrive first. On the trace, you identify the P-wave arrival as the moment the line first breaks away from the quiet baseline. The initial deflection is usually modest in amplitude compared to what follows.
The second big arrival is the S-wave, or secondary wave. S-waves shear the rock side to side (or up and down) rather than compressing it, and they travel slower than P-waves. On the seismogram, the S-wave arrival shows up as a sudden jump in amplitude along with a noticeable drop in frequency. There is often a clear change in the character of the wiggles: the trace goes from the tighter oscillations of the P-wave coda to broader, higher-amplitude swings when the S-wave hits.2Computers & Geosciences. A novel algorithm for identifying arrival times of P and S Waves in seismic borehole surveys
After the P and S arrivals, the seismogram continues to show motion for a long time. Some of this comes from surface waves, which travel along Earth’s outer shell and tend to have even larger amplitudes and lower frequencies than body waves. The rest is coda, a drawn-out tail of scattered energy bouncing around inside the crust. In a large earthquake, the coda can last minutes or even hours.
How Seismometers Capture Ground Motion
Almost all seismometers work on the same basic principle: a heavy mass is suspended inside a frame that is bolted to the ground. When the ground shakes, the frame moves with it, but the mass lags behind due to its own inertia. The instrument measures the difference between the ground’s motion and the mass’s motion. Most modern instruments are tuned to be sensitive to ground velocity, meaning the output voltage at any instant is proportional to how fast the ground is moving at that moment.3Elsevier / ScienceDirect. Chapter 5 – Seismometry (Modern Global Seismology)
This matters for reading seismograms because a velocity trace looks different from a displacement trace. On a velocity seismogram, the wiggles emphasize higher-frequency content. When researchers want to see how far the ground actually moved, they mathematically integrate the velocity record to get displacement. This is often done for early-warning calculations and for comparing events of different sizes.
How the Time Gap Between P and S Tells You Distance
Because P-waves travel faster than S-waves, the time gap between their arrivals grows with distance from the earthquake. If you are close to the epicenter, the P and S arrivals are almost on top of each other. At greater distances, the gap stretches. This relationship is well-calibrated: for a given region’s geology, there are standard travel-time tables that convert the S-minus-P time into distance in kilometers.
A single station can only tell you how far away the earthquake was, not in which direction. To pin down the actual location, seismologists use readings from multiple stations. With three or more stations, each contributing a distance estimate, the earthquake’s epicenter falls where those distance circles overlap. In practice, modern networks use dozens or hundreds of stations and sophisticated algorithms that simultaneously solve for the earthquake’s location in three dimensions.
How Magnitude Gets Extracted
The size of the wiggles on a seismogram is directly related to the energy released by an earthquake, but the relationship is not as simple as “bigger wiggles equal bigger earthquake.” Amplitude drops with distance, so a magnitude-3 earthquake recorded nearby can produce taller wiggles than a magnitude-6 event recorded on the other side of the planet. Magnitude scales correct for this distance effect.
Different magnitude scales use different parts of the seismogram. The body-wave magnitude uses the amplitude of P-waves measured at roughly one-second period. The surface-wave magnitude uses the amplitude of surface waves at around 20-second period. These two scales do not always agree, and the relationship between them has been a source of confusion in seismology for decades. An influential study showed that body-wave versus surface-wave magnitude comparisons were thrown off because earlier researchers assumed that the original body-wave formula applied to one-second waves, when in fact it had been derived for waves with periods closer to five seconds.4Bulletin of the Seismological Society of America. Scaling relations for earthquake source parameters and magnitudes The moment magnitude scale, which is the standard for large earthquakes today, avoids these frequency-dependent problems by estimating the total energy release from the entire waveform rather than a single peak measurement.
Reading Earthquake Depth from Waveforms
Depth is one of the trickiest parameters to extract from a seismogram, and it is also one of the most important. A magnitude-6 earthquake at 10 kilometers depth causes far more surface damage than the same magnitude at 200 kilometers. Fortunately, depth leaves a clear fingerprint on body-wave recordings.
When an earthquake occurs below the surface, some of its energy travels upward, bounces off the surface, and then heads back down into the Earth. These reflected waves, called depth phases, arrive at distant stations shortly after the direct P-wave. The time gap between the direct P arrival and the first depth phase is controlled almost entirely by how deep the earthquake was: deeper earthquakes produce a larger gap because the surface reflection has to travel a longer round-trip path.5Reviews of Geophysics. Depth determination for shallow teleseismic earthquakes: Methods and results For shallow earthquakes, these depth phases (labeled pP and sP) arrive so close behind the direct P-wave that they can overlap with it, creating a distinctive interference pattern in the frequency content that analysts learn to recognize.6Bulletin of the Seismological Society of America. Using the Effects of Depth Phases on P-wave Spectra to Determine Earthquake Depths
Identifying depth phases on a busy seismogram is not always straightforward. Some automated methods now use array-processing techniques, where signals from many closely spaced sensors are combined to pick out the faint depth-phase arrivals from background noise and estimate how deep the earthquake was.7Bulletin of the Seismological Society of America. Earthquake Depth Estimation Using the F Trace and Associated Probability
Signals That Are Not Earthquakes
Not every squiggle on a seismogram is an earthquake. Seismometers are sensitive enough to record a wide variety of other signals, and learning to tell them apart is a major part of reading seismograms in practice.
Volcanic tremor is a continuous or semi-continuous signal that looks very different from an earthquake’s sharp onset. It can appear as a sustained hum with a peaked spectrum (energy concentrated at specific frequencies) or as a more chaotic, irregular vibration. These oscillations are excited by magma and gas flowing through underground channels in a process analogous to how air excites vibrations in a wind instrument. Changes in the tremor’s amplitude or frequency can indicate that the plumbing inside a volcano is shifting, which makes tremor monitoring a key tool in eruption forecasting.8Journal of Geophysical Research: Solid Earth. Volcanic tremor: Nonlinear excitation by fluid flow
Ocean waves produce a persistent background hum called microseismic noise. This noise shows up on seismograms worldwide, even at stations thousands of kilometers from any coast. It is strongest at periods around 6 to 14 seconds and tends to track storm activity in the ocean. Research using seismic arrays in northern Europe has shown that the relationship between ocean wave height and the amplitude of microseismic noise at a given station is roughly linear, with the most efficient noise generation happening at coastlines near Norway and the northern British Isles.9Oxford Academic. Linking source region and ocean wave parameters with the observed primary microseismic noise You will sometimes hear seismologists refer to this as “the Earth’s heartbeat.” On a quiet day with no earthquakes, microseismic noise may be the dominant signal on the trace.
Other non-earthquake signals include quarry blasts, sonic booms, trains, construction equipment, and even footsteps near a sensitive instrument. Experienced analysts learn to recognize these by their waveform shape, frequency content, and time of day. Quarry blasts, for instance, tend to happen during working hours and produce a sharp onset followed by a distinctive surface-wave pattern.
How Local Geology Warps the Signal
Two seismometers the same distance from an earthquake can record dramatically different seismograms depending on what the ground beneath each one is made of. Soft sedimentary basins act like bowls of jelly: seismic waves slow down when they enter the basin, their amplitude grows, and the shaking lasts longer because energy bounces back and forth between the basin edges and the hard rock underneath. The bottom of such a basin acts as a strong reflector that traps seismic energy inside.10Earthquake Science. Multi-parameter modeling and analysis of ground motion amplification in the Quaternary sedimentary basin of the Beijing-Tianjin-Hebei region
This amplification can be substantial. Three-dimensional simulations of the Nenana Basin in central Alaska found that average amplification ratios reached about four on the horizontal components and seven on the vertical component, driven primarily by the basin’s three-dimensional shape rather than just the shallow soil properties.11Journal of Geophysical Research: Solid Earth. Analysis of Seismic Wave Amplification in Sedimentary Basins Using 3D Wavefield Simulations: Nenana Basin, Central Alaska This means a seismogram recorded inside such a basin will show amplitudes several times larger and a longer duration of shaking than a nearby station sitting on bedrock. If you are comparing seismograms from different stations, you always need to account for the local geology before concluding anything about the earthquake itself.
Earthquake Early Warning from the First Seconds
Earthquake early warning systems exploit the fact that P-waves arrive before S-waves and before the destructive surface waves. The idea is simple: detect the P-wave, rapidly estimate the earthquake’s magnitude, and broadcast an alert before the stronger shaking arrives. Every second of warning counts, giving people time to take cover and automated systems time to shut down trains, close gas valves, or pause surgeries.
One widely studied approach measures the peak displacement in the first three seconds after the P-wave arrives, a parameter called Pd. Research in southern California established a clear relationship between Pd, the earthquake’s magnitude, and the distance from the source. For earthquakes below about magnitude 6.5, the magnitude estimated from Pd agreed with catalog magnitudes to within roughly 0.2 magnitude units, which is accurate enough for practical warning purposes.12Geophysical Research Letters. Magnitude estimation using the first three seconds P‐wave amplitude in earthquake early warning
More recent work has explored using changes in the signal-to-noise ratio of P-wave arrivals as a simultaneous trigger and magnitude estimator, aiming to reduce the delay between detecting a P-wave and issuing a magnitude estimate.13PubMed Central. A Synchronous Magnitude Estimation with P-Wave Phases’ Detection Used in Earthquake Early Warning System For very large earthquakes, the three-second window may not capture enough of the source, and newer parameters that accumulate information over slightly longer windows are showing promise in providing more stable estimates.14Scientific Reports. Magnitude determination for earthquake early warning using P-alert low-cost sensors during 2024 Mw7.4 Hualien, Taiwan earthquake
Machine Learning Is Changing How Seismograms Are Read
For most of seismology’s history, identifying wave arrivals on seismograms was done by hand. A trained analyst would examine the trace, mark where P and S waves arrived, note the polarity of the first motion (up or down), and catalog the event. This is painstaking work, and global networks now record far more data than humans can manually process.
Deep learning has stepped in to fill the gap. Convolutional neural networks trained on millions of manually picked seismograms from Southern California can now pick P-wave arrival times with a standard deviation of just 0.023 seconds compared to human analysts and determine the direction of first motion with about 95% precision. Remarkably, these automated systems pick more usable first-motion polarities than the human analysts did, roughly doubling the number of focal mechanisms that can be computed from the same dataset, without sacrificing quality.15Journal of Geophysical Research: Solid Earth. P Wave Arrival Picking and First‐Motion Polarity Determination With Deep Learning
Another approach, PhaseNet, takes an entire window of continuous seismic data as input and outputs a probability curve showing where P and S arrivals most likely occur. Unlike traditional methods that use a sliding window to compare short-term and long-term signal power, PhaseNet processes the whole segment at once and can differentiate between P and S phases, something that simpler detection algorithms cannot do on their own.16Geophysical Journal International. PhaseNet: a deep-neural-network-based seismic arrival-time picking method Time-frequency representations, where the seismogram is converted into a visual map of how energy is distributed across frequencies over time, have also proven effective as inputs for neural networks, achieving detection accuracies near 99% in tests on Canadian earthquake data.17Seismological Research Letters. Seismic Event and Phase Detection Using Time–Frequency Representation and Convolutional Neural Networks
What Seismograms Reveal About Earth’s Interior
Reading a seismogram is not just about the earthquake that produced it. Every seismic wave that arrives at a station has traveled through some portion of Earth’s interior, and the path it took is encoded in its arrival time, amplitude, and waveform shape. When a wave hits a boundary between different rock layers, it splits: part reflects back, part transmits through, and both P and S waves can convert into each other at the boundary. This mode conversion means that a single earthquake can produce dozens of distinct arrivals, each of which took a different route through the planet.18Elsevier. The Structure and Interpretation of Seismograms
By cataloging these arrivals from thousands of earthquakes recorded at stations around the world, seismologists have built up detailed models of Earth’s layered structure: the crust, the mantle, the liquid outer core, and the solid inner core. Shadow zones, where certain wave types fail to appear at expected distances, were the clues that revealed the liquid core over a century ago. Modern seismic tomography works like a medical CT scan, using travel-time variations from millions of earthquake-station pairs to map slow and fast regions inside the mantle, which correspond to hotter and cooler rock.
Citizen Seismometers and Low-Cost Networks
You no longer need institutional funding to contribute useful seismograms. Low-cost instruments like the Raspberry Shake, a small seismometer that plugs into a home internet connection, have created dense citizen-science networks in cities around the world. A study in Wellington, New Zealand, tested whether these citizen-owned devices could improve earthquake locations when combined with the professional GeoNet seismic network. The results showed that integrating Raspberry Shake data meaningfully improved location precision, and even the fact that station coordinates are deliberately blurred (to protect users’ home addresses) did not undermine the results much, provided the overall network geometry was good.19Seismica. Using citizen science Raspberry Shake seismometers to enhance earthquake location and characterization: a case study from Wellington, New Zealand
If you have a Raspberry Shake or similar device at home, you can pull up your own seismograms in near real-time. The basics of reading them are the same: look for the P-wave onset, the S-wave onset, note the time gap, and compare with nearby stations. You will also see plenty of non-earthquake signals, from traffic to door slams to your washing machine’s spin cycle. Learning to distinguish these from genuine quakes is half the fun and is exactly the same interpretive skill professional seismologists use, just at a smaller scale.
Seismograms from Mars
The reading principles that work on Earth transfer directly to other planets. NASA’s InSight lander carried a seismometer to Mars and recorded thousands of marsquakes between 2018 and 2022. The largest, cataloged as S1222a, produced a record that lasted more than eight hours and included clearly identifiable body waves and surface waves. Analysis of the long, scattered surface-wave tail of that event allowed researchers to estimate how strongly Mars scatters and absorbs seismic energy, finding that the planet’s scattering quality factor is much lower than Earth’s in the same frequency range, meaning seismic energy bounces around inside Mars more readily before being absorbed.20Geophysical Research Letters. Seismic Scattering and Absorption Properties of Mars Estimated Through Coda Analysis on a Long‐Period Surface Wave of S1222a Marsquake
Martian seismograms look recognizably like Earth seismograms in some ways and strikingly different in others. The lack of oceans means there is no microseismic noise from waves crashing on shorelines. Instead, the dominant background signals come from atmospheric pressure fluctuations and wind vibrating the lander. The absence of plate tectonics means marsquakes are driven by different processes, primarily cooling and contraction of the planet’s interior. But the core skill of reading a seismogram, identifying arrivals, measuring their timing and amplitude, and working backward to figure out what happened, is identical.
Digitizing the Paper Trail
Before the digital era, seismograms were recorded as ink traces on smoked paper wrapped around slowly rotating drums. Archives around the world hold millions of these paper records, spanning much of the 20th century. The data locked inside them is invaluable for studying long-term earthquake patterns and understanding historical events, but it cannot be used for modern analysis until it is converted from an image into a digital time series.
Software tools like DigitSeis have been developed to automate this conversion. The process involves scanning the paper record, using pattern recognition to identify and separate the trace from the background grid, and then converting the trace’s position into a time-amplitude series that researchers can analyze with the same tools they use for modern digital recordings.21Progress in Earth and Planetary Science. DigitSeis: software to extract time series from analogue seismograms A project at the Royal Observatory of Belgium applied this workflow to a large archive, converting scanned TIFF images of historical seismograms into digital vector-format files suitable for geophysical modeling.22PubMed Central. Computer Vision Algorithms of DigitSeis for Building a Vectorised Dataset of Historical Seismograms from the Archive of Royal Observatory of Belgium These recovered records extend the instrumental earthquake catalog back by decades, improving our understanding of seismic hazard in regions where historical earthquakes are the only evidence of what the faults there are capable of producing.