Mapping Ancient Oceans

Mapping Ancient Oceans

This guest post is by Dr Grace Shephard, a postdoctoral researcher in tectonics and geodynamics at the Centre of Earth Evolution and Dynamics (CEED) at the University of Oslo, Norway. This blog entry describes the latest findings of a study that maps deep remnants of past oceans. Her open access study, in collaboration with colleagues at CEED and the University of Oxford, was published this week in the Nature Journal: Scientific Reports. This post is modified from a version that first appeared on the CEED Blog.

Quick summary:

There are several ways of imaging the insides of the Earth by using information from earthquake data. When these different images are viewed at the same time, a new type of map allows geoscientists to identify the most robust features. These deep structures are likely the remains of extinct oceans, known as slabs, that were destroyed hundreds of millions of years ago. The maps are computed at different depths inside the Earth and the resulting slabs can be resurrected back to the surface. Along with a freely available paper and website, the analysis yields new insights into the structure and evolution of our planet in deep time and space.

Earth in constant motion

The surface of the Earth is in constant motion and this is particularly true of the rocks found under the oceans. The crust – the outermost layer of the planet – is continually being formed in the middle of oceans, such as the Mid-Atlantic Ridge. In other places, older crust is being destroyed, such as where the Pacific Ocean is moving under Japan. A third type of locality sees the crust shifted along laterally, such as the San Andreas Fault in San Francisco. These three types of locations are often referred to as plate boundaries, and they connect up to divide the Earth’s surface into tectonic plates of different sizes and motions.

Where plates plunge into the mantle are termed subduction zones (red lines Figure 1, below). The configuration of these subduction zones has changed throughout geological time. Indeed, much of the ocean seafloor (blue area in Figure 1) that existed when the dinosaurs roamed the Earth has long since been lost into the Earth’s mantle and are now known as slabs. The mantle is the domain beneath the outer shell of our planet and extends to around 2800 km depth, to the boundary with the core.

The age and fabric of the seafloor contains some of the most important constraints in understanding the past configuration of Earth. However, the constant recycling of oceans means that the Earth’s surface as it is today can only tell us so much about the deep geological past – the innards of our planet hold much of this information, and we need to access, visualize, and disseminate it.

Figure 1. A reconstruction of the Earth’s surface from 200 Million years ago to present day in jumps of 10 Million years. Red lines show the location of subduction zones, other plate boundaries in black, plate velocities are also shown. Continents are reconstructed with the present-day topography for reference. Based on the model of Matthews et al. (2016; Global and Planetary Change). Credit: G Shephard (CEED/UiO) using GPlates and GMT software.

Imaging the insides

Using information from earthquake data, seismologists can produce images of the Earth’s interior via computer models – this technique is called seismic tomography. Similar to a medical X-ray scan that looks for features within the human body, these models image the internal structure of the Earth. Thus, a given seismic tomography model is a snapshot into the present-day structure, which has been shaped by hundreds of millions to billions of years of Earth’s history.

However, there are different types of data that can be used to generate these models and different ways they can be created, each with varying degrees of resolution and sensitivity to the real Earth structure. This variability has led to dozens of tomographic models available in the scientific arena, which all have slightly different snapshots of the Earth. For example, deep under Canada and the USA is a well-known chunk of subducted ocean seafloor (see ‘slab’ label in Figure 2). A vertical slice through the mantle for three different tomography models shows that while overall the models are similar, there are some slight shifts in its location and shape.

Importantly, seismic waves pass through subducted, old, cold oceanic plates more quickly than they do through the surrounding mantle (in the same way that sound travels faster through solids than air). It follows that these subducted slabs can be ‘imaged’ seismically (usually these slab regions show up as blue in tomography models such as in Figure 2 and as shown in this video by co-author Kasra Hosseini. The red regions might represent thermally hot features like mantle plumes).

Figure 2. Vertical slices through three different seismic tomography models under North America and the Atlantic Ocean (profile running from A to B). The blue region outlined by black dashed line is related to the so-called Farallon slab. While it is imaged in all three models the finer details of the slab geometry and depth are different. Model 1 is S40RTS (Ritsema et al., 2011), 2 is UU-P07 (Amaru, 2007) and 3 is GyPSum-S (Simmons et al., 2010).

For other geoscientists to utilize this critical information, for example to work out how continents and oceans moved through time, requires a spectrum of seismic tomography models to be considered. But several limiting questions arise:

Which tomography model(s) should be used?

Are models based certain data types more likely to pick up a feature?

How many models are sufficient to say that a deep slab can be imaged robustly?

Voting maps of the deep

To facilitate solutions to these questions, a novel yet simple approach was undertaken in the study. Different tomography models were combined to generate counts, or votes, of the agreement between models – a sort of navigational guidebook to the Earth’s interior (Figure 3).

Figure 3. An interactive 360° style image for the vote map at 1000 km depth. The black and red regions highlight the most robust features (high vote count = likely to be a subducted slab of ocean) and the blue regions are the least robust areas (low vote count). Coastlines in black for reference. Image: G Shephard (CEED/UiO) using 360player ( and GMT software. More depth slices and options can be also imaged at our website.

A high vote count (black-red features in Figure 3) means that an increased number of tomography models agree that there could be a slab at that location. For the study in Scientific Reports the focus was on the oldest and deepest slabs, but the process can be undertaken for shallower and younger slabs, and for other features such as mantle plumes. The maps show the distribution of the most robust slabs at different depths – the challenge is to now try and verify the features and potentially link them to subduction zones at the surface back in time.

One way to achieve this is to assume that a subducted portion of ocean will sink vertically in the mantle, and then to apply a sinking rate to connect depth and time. This enables pictures that link the surface and deep Earth, like the cover image, to be made. A sinking rate of say, 1.2 centimeters per year, means that a feature that existed at the surface around 100 Million years ago might be found at 1200 km depth.

Many studies have started to undertake a similar exercise on both regional and global scales. However, because these vote maps are free to access, showcase a lot of different models and can be remade with a sub-selection of them, they serve as an easy resource for the community to continue this task.

Secrets in depth

A bit like dessert-time discussions about the best way to cut a cake, so too are the ways of imaging and analyzing the Earth (Figure 4). Do you slice it horizontally and see things that might correspond to the same age all over the globe? Or slice vertically from the surface to see a spectrum of ages (depths) at a given location? Or perhaps a 3-D imaging would be most insightful? Whichever choice is made for the vote maps, many interesting features are displayed.

Figure 4. Vote maps visualized using alternative imaging options on a sphere. Credit: G Shephard (CEED/UiO) using GPlates software

By comparing the changes in vote counts with depth, some intriguing results were found. An apparent increase in the amount of the slabs was found around 1000-1400 km depth. This could mean that about 130 Million years ago more oceanic basins were lost into the mantle. Or perhaps there is a specific region in the mantle that has “blocked” the slabs from sinking deeper for some period of time (for example, an increase in viscosity).

The vote maps and their associated depth-dependent changes hold implications on an interdisciplinary stage including through linking plate tectonics, mantle dynamics, and mineral physics.

Of course, the vote maps are only as good as the tomography models that they are comprised of – and by very definition, a model is just one way of representing the true Earth.

A resource for the community

Having accessed a variety of tomography models provided by different research groups or data repositories, this study was facilitated using open-source software (Generic Mapping Tools and GPlates).

An important component of reproducible science and advancing our understanding of Earth is to make datasets and workflows publicly available for further investigations.

An online toolkit to visualize seismic tomography data is being developed by the co-authors and a preliminary vote maps page is already online. Here, vote maps for a sub-selection of tomography models can be generated, including with a choice in colour scales and with overlays of plate reconstruction models. More functionality will soon be available – so watch this space!

By Grace Shephard, a postdoctoral researcher in tectonics and geodynamics at the Centre of Earth Evolution and Dynamics (CEED)

Contact information for more details: Grace Shephard –

Full reference to the article, freely available to the public:

Amaru, M. L. Global travel time tomography with 3-D reference models,. Geol. Ultraiectina 274, 174 (2007).

Matthews, K.J. K.T. Maloney, S. Zahirovic, S.E. Williams, M. Seton, R.D. Müller. 2016. Global plate boundary evolution and kinematics since the late Paleozoic. Global and Planetary Change. v146. doi: 10.1016/j.gloplacha.2016.10.002

Ritsema, J., Deuss, A., van Heijst, H. J. & Woodhouse, J. H. S40RTS: a degree-40 shear-velocity model for the mantle from new Rayleigh wave dispersion, teleseismic traveltime and normal-mode splitting function measurements. Geophysical Journal International 184, 1223-1236, doi:10.1111/j.1365-246X.2010.04884.x (2011).

Simmons, N. A., Forte, A. M., Boschi, L. & Grand, S. P. GyPSuM: A joint tomographic model of mantle density and seismic wave speeds. Journal of Geophysical Research: Solid Earth 115, doi:10.1029/2010JB007631 (2010).




Educators: apply now to take part in the 2018 GIFT workshop!

Educators: apply now to take part in the 2018 GIFT workshop!

The General Assembly is not only for researchers but for teachers and educators with an interest in the geosciences also. Every year the Geosciences Information For Teachers (GIFT) is organised by the EGU Committee on Education to bring first class science closer to primary and high school teachers.

The topic of the 2018 edition of GIFT is ‘Major events that shaped the Earth’. This year’s workshop will be taking place on 9–11 April 2018 at the EGU General Assembly in Vienna, Austria.

Teachers from Europe and around the world can apply to participate in the 2018 edition of GIFT, and to receive a travel and accommodation stipend to attend the workshop, by November 15. Application information is available for download in PDF format, a document which also includes the preliminary programme of the workshop.

Not sure what to expect? More information about GIFT workshops can be found in the GIFT section of the EGU website. You can also take a look at a blog post about the 2015 workshop and also learn what the workshop is like from a teacher’s perspective here. You might also find videos of the 2017 workshop useful too.


Imaggeo on Mondays: A prehistoric forest

Imaggeo on Mondays: A prehistoric forest

This stunning vista encompasses the south-western wilderness of Tasmania as seen from the Tahune air walk 60 m above the Huon river valley. In front lies the beginning of a huge UNESCO World Heritage Site, covering almost a fourth of the area of Tasmania. The site mostly consists of a pristine, temperate rainforest of Gondwanan origin that is home to the tallest flowering trees in the world; Eucalyptus spp. reach up to 100 m height in this region.

“I have never tasted the sense of a more remote place than this one. Give me more,” says Vytas Huth, who captured this stunning shot.

Gondwana was a supercontinent, consisting of present day Africa, South America, India, Madagascar, Australia and New Zealand. It formed when the even larger supercontinent of Pangaea broke up 250 million years ago.

Slowly, Gondwana started to break apart too. India tore away first, followed by Africa and then New Zealand. By the end of the Cretaceous, 65 million years ago, only South America, Australia and Antarctica remained joined.  It took a further 20 million years before Australia and Antarctica separated.

By the time Australia started being pulled northwards, the first glaciers were forming on Antarctica, as it began freezing over. Atop the old rocks which made up its bulk, animals and plants of ancient origin, travel northwards with the Land Down Under.

Because India and Africa broke away from the supercontinent so early on, few hallmarks of ancient Gondwana wildlife are left in their present biodiversity. In contrast, Australia and Tasmania remained connected to Antarctica and South America much longer and there are clear similarities in species across these continents.

“Fossil evidence suggests that temperate rainforest once extended across Australia, Antarctica, South America and New Zealand around 45 million years ago. Such fossils and the surviving species in Tasmania provide evidence of the ancient link to Gondwana”, reports the Tasmania Parks & Wildlife Service.

Imaggeo is the EGU’s online open access geosciences image repository. All geoscientists (and others) can submit their photographs and videos to this repository and, since it is open access, these images can be used for free by scientists for their presentations or publications, by educators and the general public, and some images can even be used freely for commercial purposes. Photographers also retain full rights of use, as Imaggeo images are licensed and distributed by the EGU under a Creative Commons licence. Submit your photos at


Is it an earthquake, a nuclear test or a hurricane? How seismometers help us understand the world we live in

Is it an earthquake, a nuclear test or a hurricane? How seismometers help us understand the world we live in

Although traditionally used to study earthquakes, like today’s M 8.1 in Mexico,  seismometers have now become so sophisticated they are able to detect the slightest ground movements; whether they come from deep within the bowels of the planet or are triggered by events at the surface. But how, exactly, do earthquake scientists decipher the signals picked up by seismometers across the world? And more importantly, how do they know whether they are caused by an earthquake, nuclear test or a hurricane?  

To find out we asked Neil Wilkins (a PhD student at the University of Bristol) and Stephen Hicks (a seismologist at the University of Southampton) to share some insights with our readers.

Seismometers are highly sensitive and they are able to detect a magnitude 5 earthquake occurring on the other side of the planet. Also, most seismic monitoring stations have sensors located within a couple of meters of the ground surface, so they can be fairly susceptible to vibrations at the surface. Seismologists can “spy” on any noise source, from cows moving in a nearby field to passing trucks and trains.

A nuclear test

On Sunday the 3rd of September, North Korea issued a statement announcing it had successfully tested an underground hydrogen bomb. The blast was confirmed by seismometers across the globe. The U.S.  Geological Survey registered a 6.3 magnitude tremor, located at the Punggye-ri underground test site, in the northwest of the country. South Korea’s Meteorological Administration’s earthquake and volcano center also detected what is thought to be North Korea’s strongest test to date.

However they occur, explosions produce ground vibrations capable of being detected by seismic sensors. Mining and quarry blasts appear frequently at nearby seismic monitoring stations. In the case of nuclear explosions, the vibrations can be so large that the seismic waves they produce can be picked up all over the world, as in the case of this latest test.

It was realised quite early in the development of nuclear weapons that seismology could be used to detect such tests. In fact, the need to have reliable seismic data for monitoring underground nuclear explosions led in part to the development of the Worldwide Standardized Seismograph Network in the 1960s, the first of its kind.

Today, more than 150 seismic stations are operating as part of the International Monitoring System (IMS) to detect nuclear tests in breach of the Comprehensive Test-Ban Treaty (CTBT), which opened for signatures in 1996. The IMS also incorporates other technologies, including infrasound, hydroacoustics and radionuclide monitoring.

The key to determining whether a seismic signal is from an explosion or an earthquake lies in the nature of the waves that are present. There are three kinds of seismic wave seismologists can detect. The fastest, called Primary (P) waves, cause ground vibrations in the same direction that they travel, similar to sound waves in the air. Secondary (S) waves cause shaking in a perpendicular direction. Both P and S waves travel deep through the Earth and are known collectively as body waves. In contrast, the third type of seismic waves are known as surface waves, because they are trapped close to the surface of the Earth. In an earthquake, it is normally surface waves that cause the most ground shaking.

In an explosion, most of the seismic energy is released outwards as the explosive material rapidly expands. This means that the largest signal in the seismogram comes as P waves. Explosions therefore have a distinctive shape in the seismic data when compared with an earthquake, where we expect S and surface waves to have higher amplitude.

Forensic seismologists can therefore make measurements of the seismic data to determine whether there was an explosion. An extra indication that a nuclear test occurred can also be revealed by measuring the depth of the source of the waves, as it would not be possible to place a nuclear device deeper than around 10 km below the surface.

Yet while seismic data can tell us that there has been an explosion, there is nothing that can directly identify that explosion as being nuclear. Instead, the IMS relies on the detection of radioactive gases that can leak from the test site for final confirmation of what kind of bomb was used.

The figure shows (at the bottom) the seismic recording of the latest test in North Korea made at NORSAR’s station in Hedmark, Norway. The five upper traces show recordings at the same station for the five preceding tests, conducted by North Korea in 2006, 2009, 2013 and 2016 (two explosions in 2016). The 2017 test, is as can be seen from this figure, clearly the strongest so far. Credit: NORSAR.

When North Korea conducted a nuclear test in 2013, radioactive xenon was detected 55 days later, but this is not always possible. Any detection of such gases depends on whether or not a leak occurs in the first place, and how the gases are transported in the atmosphere.

Additionally, the seismic data cannot indicate the size of the nuclear device or whether it could be attached to a ballistic missile, as the North Korean government claims.

What seismology can give us is an idea of the size of the explosion by measuring the seismic magnitude. This is not straightforward, and depends on knowledge of exactly how deep the bomb was buried and the nature of the rock lying over the test site. However, by comparing the magnitude of this latest test with those from the previous five tests conducted in North Korea, we can see that this is a much larger explosion.

The Norwegian seismic observatory NORSAR has estimated a blast equivalent to 120 kilotons of TNT, six times larger than the atomic bomb dropped on Nagasaki in 1945, and consistent with the expected yield range of a hydrogen bomb.


Nuclear tests are not the only hazard keeping our minds busy in the past few weeks. In the Atlantic, Hurricanes Harvey, Irma and Katia have wreaked havoc in the southern U.S.A, Mexico and the Caribbean.

Hurricanes in the Atlantic can occur at any time between June and November. According to hurricane experts, we are at the peak of the season. It is not uncommon for storms to form in rapid succession between August, September and October.

The National Hurricane Centre (NHC) is the de facto regional authority for producing hurricane forecasts and issuing alerts in the Atlantic and eastern Pacific. For their forecasts, meteorologists use a combination of on the ground weather sensors (e.g. wind, pressure, Doppler radar) and satellite data.

As hurricane Irma tore its way across the Atlantic, gaining strength and approaching the Caribbean island of Guadeloupe, local seismometers detected its signature, sending the global press into a frenzy. It may come as a slight surprise to some people that storms and hurricanes also show on seismometers.

However, a seismometer detecting an approaching hurricane is not actually that astonishing. There is no evidence to suggest that hurricanes directly cause earthquakes, so what signals can we detect from a hurricane? Rather than “signals”, seismologists tend to refer to this kind of seismic energy as “noise” as it thwarts our ability to see what we’re normally looking out for – earthquakes.

The seismic noise from a storm doesn’t look like distinct “pings” that we would see with an earthquake. What we see are fairly low-pitched “hums” that gradually get louder in the days and hours preceding the arrival of a storm. As the storm gets closer to the sensor, these hums turn into slightly higher-pitched “rustling”. This seismic energy then wanes as the hurricane drifts away. We saw this effect clearly for Hurricane Irma with recordings from a seismometer on the island of Guadeloupe.

What causes these hums and rustles? If you look at the frequency content of seismic data from any monitoring station around the globe, noise levels light up at frequencies of ~0.2 Hz (5 s period). We call these hums “microseism”. Microseism is caused by persistent seismic waves unrelated to earthquakes, and it occurs over huge areas of the planet.  One of the strongest sources of microseism is caused by ocean waves and swell. During a hurricane, swell increases and ocean waves become more energetic, eventually crashing into coastlines, transferring seismic energy into the ground. This effect is more obvious on islands as they are surrounded by water.

As the hurricane gets closer to the island, wind speeds dramatically increase and may dwarf the noise level of the longer period microseism. Wind rattles trees, telegraph poles, and the surface itself, transferring seismic energy into the ground and moving the sensitive mass inside the seismometer. This effect causes higher-pitched “rustles” as the centre of the storm approaches. Gusts of wind can also generate pressure changes inside the seismometer installation and within the seismometer itself, generating longer period fluctuations.

During Hurricane Irma, a seismic monitoring station located in the Dutch territory of St. Maarten clearly recorded the approach of the storm, leading to an intense crescendo as the eyewall crossed the area. As the centre of the eye passed over, the seismometer seems to have recorded a slightly lower noise level. This observation could be due to the calmer conditions and lower pressure within the eye. The station went down shortly after, probably from a power outage or loss in telemetry which provides the data in real-time.

Seismometers measuring storms is not a new observation. Recently, Hurricane Harvey shook up seismometers located in southern Texas. Even in the UK, the approach of winter storms across the Atlantic causes much higher levels of microseism.

It would be difficult to use seismometer recordings to help forecast a hurricane – the recordings really depend on how close the sensor is to the coast and how exposed the site is to wind. In the event of outside surface wind and pressure sensors being damaged by the storm, protected seismometers below the ground could possibly prove useful in delineating the rough location of the hurricane eye, assuming they maintain power and keep sending real-time data.

At least several seismic monitoring stations in the northern Antilles region were put out of action by the effects of the Hurricane. Given the total devastation on some islands, it is likely that it will take at least several months to bring these stations back online. The Lesser Antilles are a very tectonically active and complex part of Earth; bringing these sensors back into operation will be crucial to earthquake and volcano hazard monitoring in the region.

By Neil Wilkins (PhD student at the University of Bristol) and Steven Hicks (a seismologist at the University of Southampton)

References and further reading

GeoSciences Column: Can seismic signals help understand landslides and rockfalls?

NORSAR Press Release: Large nuclear test in North Korea on 3 September 2017

The Comprehensive Nuclear-Test-Ban Organization Press Release: CTBTO Executive Secretary Lassina Zerbo on the unusual seismic event detected in the Democratic People’s Republic of Korea

First Harvey, Then Irma and Jose. Why? It’s the Season (The New York Times)

NOAA  National Hurricane Center

IRIS education and outreach series: How does a seismometer work?


Get every new post on this blog delivered to your Inbox.

Join other followers: