Tag Archives: Plate reconstructions

Plate tectonics drive tropical reef biodiversity dynamics

A new paper is out in Nature Communications – a study by Fabien Leprieur and co-authors (including me) on how plate tectonics influences the biodiversity dynamics of tropical reefs. Previously published paleo-shoreline estimates (see data on Github and Heine et al paper) have were used as base to model paleo-bathymetry and time-dependent spatial diversification patterns of tropical marine reefs – here’s the abstract:

The Cretaceous breakup of Gondwana strongly modified the global distribution of shallow tropical seas reshaping the geographic configuration of marine basins. However, the links between tropical reef availability, plate tectonic processes and marine biodiversity distribution patterns are still unknown. Here, we show that a spatial diversification model constrained by absolute plate motions for the past 140 million years predicts the emergence and movement of diversity hotspots on tropical reefs. The spatial dynamics of tropical reefs explains marine fauna diversification in the Tethyan Ocean during the Cretaceous and early Cenozoic, and identifies an eastward movement of ancestral marine lineages towards the Indo-Australian Archipelago in the Miocene. A mechanistic model based only on habitat-driven diversification and dispersal yields realistic predictions of current biodiversity patterns for both corals and fishes. As in terrestrial systems, we demonstrate that plate tectonics played a major role in driving tropical marine shallow reef biodiversity dynamics.

The paper is available as open-access on the Nature Communications website. Associated data can be downloaded from Figshare.

The Saharan Atlantic ocean makes ripples (edited)

Our  “Saharan Atlantic ocean” paper has just been featured in GEOLOGY’s “Research Focus” article in the March issue. The focus article is entitled “Roadmap to continental rupture: Is obliquity the route to success?” is written by Cythia Ebinger and Jolante van Wijk and is available as open access. This is fantastic news!

Update #1,2,3,4,5: So by now, there’s been quite some science (and popular) media buzz with both Sydney Uni and GFZ Potsdam having released press info on our article – it’s been overwhelming. Spiegel Online (one of the largest German online news outlets) & Sueddeutsche Zeitung (also one of the largest papers) reported in Germany even going a bit beyond the standard press release texts. As Sascha and I hoped, our hypothetical “Saharan Ocean” image (see here) facilitated the take up of the story quite a bit.

According to Altmetric, the article is currently ranking as #7 in terms of impact which makes it one of the highest ever scores for articles published in GEOLOGY.

Links here (I’m updating those from time to time):

Why there is no Saharan Atlantic Ocean

Our paper “Oblique rifting along the Equatorial Atlantic ocean: Why there is no Saharan Atlantic Ocean” (Doi: 10.1130/G35082.1) is now available online, I believe a pre-issue publication listing will follow this week. We use plate kinematic and 3D numerical modelling to explain why the Equatorial Atlantic ocean formed in the Early Cretaceous time (around 120-100 Million years ago). Here’s a summary of the paper in simple terms:

Every schoolchild can recognise continents or parts of them based on their shape. But why does Italy look like a boot, why is Australia an island-like continent and what sculpted Africa’s margins? In this study we address the underlying processes that shaped Earth’s continental plates when the last era of supercontinents came to an end, between about 150 and 100 My years ago.

At the time when dinosaur evolution peaked, the southern continents were still united in the supercontinent Gondwana. However, vast continental rift systems comparable to the present East African rift, extended between present-day South America and Africa as well as within the African continent. These rifts are preserved as deep sedimentary basins in the subsurface of the African continent and along continental margins and document processes where continental crust is stretched like chewing gum. The so-called South Atlantic and West African rift systems were about to split the African-South American part of Gondwana North-South into nearly equal halfs, generating a South Atlantic and a Saharan Atlantic Ocean (see Image). In a dramatic plate tectonic twist, however, a competing rift along the present-day South American and African Equatorial Atlantic margins, won over the West African rift, causing it to become extinct, avoiding the break up of the African continent and the formation of a Saharan Atlantic ocean.

Our work elucidates the reasons behind the success and failure of these rift systems by coupling plate tectonic and advanced 3D numerical models of continental lithosphere deformation. We find that rift obliquity acts as a selector between successful and aborted rift systems, explaining why the South and Equatorial Atlantic Ocean basins formed and other rifts became aborted. Our modelling also sheds lights on the dynamics of rifting, suggesting that feedback loops caused a ten-fold acceleration in the velocities of the South American plate once the Equatorial Rift System had sufficiently weakened the last remaining continental bridge between both plates.

One hundred years after the German scientist Alfred Wegener developed first ideas of continental drift, this study provides a new keystone in understanding the rules which govern continental extension and tectonic plate motion ultimately sculpting Earth’s continents into the shapes as we recognise them today.

The  abstract of the paper:

Rifting between large continental plates results in either continental breakup and the formation of conjugate passive margins, or rift abandonment and a set of aborted rift basins. The nonlinear interaction between key parameters such as plate boundary configuration, lithospheric architecture, and extension geometry determines the dynamics of rift evolution and ultimately selects between successful or failed rifts. In an attempt to evaluate and quantify the contribution of the rift geometry, we analyze the Early Cretaceous extension between Africa and South America that was preceded by ∼20–30 m.y. of extensive intracontinental rifting prior to the final separation between the two plates. While the South Atlantic and Equatorial Atlantic conjugate passive margins continued into seafloor-spreading mode, forming the South Atlantic Ocean basin, Cretaceous African intraplate rifts eventually failed soon after South America broke away from Africa. We investigate the spatiotemporal dynamics of rifting in these domains through a joint plate kinematic and three-dimensional forward numerical modeling approach, addressing (1) the dynamic competition of Atlantic and African extensional systems, (2) two-stage kinematics of the South Atlantic Rift System, and (3) the acceleration of the South America plate prior to final breakup. Oblique rifts are mechanically favored because they require both less strain and less force in order to reach the plastic yield limit. This implies that rift obliquity can act as selector between successful ocean basin formation and failed rifts, explaining the success of the highly oblique Equatorial Atlantic rift and ultimately inhibiting the formation of a Saharan Atlantic Ocean. We suggest that thinning of the last continental connection between Africa and South America produced a severe strength-velocity feedback responsible for the observed increase in South America plate velocity.

The associated data for the plate kinematic model is available full and for free as open data from the Datahub.org pages (http://datahub.io/dataset/southatlanticrift) of my earlier South Atlantic paper in Solid-Earth.net. More detailed explanations and animations will follow later.

Lastly, here’s our guess for how the world might look liked like if a Saharan Atlantic ocean had formed:

The world as it might have looked like if the West African Rift system had been "successful" in forming a "Saharan Atlantic Ocean basin". We explain in our paper why this did not happen.

The world as it might have looked like if the West African Rift system had been “successful” in forming a “Saharan Atlantic Ocean basin”. We explain in our paper why this did not happen. Made with GPlates and image manipulation.

Working with plate velocities in GPlates

GPlates v1.3 can display and extract plate velocities. Depending on your work, you might have different requirements for these domains, plus there are a few pitfalls on the way. Currently, there are two three  ways to create velocity domains in GPlates:

  1. Create a global set of points at regular spacing which stay fixed absolute and the plates move across them. This method will report velocities of whatever plate will be on top of these points. This method is generally used when working with global models and when one wants to export boundary conditions for global geodynamic modelling (the CitcomS and Terra mesh generation options). Use  Features → Generate Velocity Point Domains to create a velocity domain setup according to your needs. Note that this also allows to create a regular spaced lon/lat grid of distributed points. The features are gpml:MeshNode features in the GPlates Geological Information Model.

    Adding a global layer of velocity domain points using GPlates' built-in function

    Adding a global layer of velocity domain points using GPlates’ built-in function

  2. Alternatively, you can just go and create a set of points wherever you require them – you might have a few small plates which fall through the cracks when creating the global velocity domains. So here, you just go and create a few point features, assigning them the right properties in the ‘Create Feature’ dialogue. Create and save the layer after you are done. You can automatically assign PlateIDs using GPlates’ cookie-cutting tool afterwards which saves you typing in PlateIDs manually.

    Manually creating a point feature collection where velocities are to be extraced

    Manually creating a point feature collection where velocities are to be extracted

  3. Use QGIS and the Vector toolbox to create a regular spaced point grid in OGR Vector format (ESRI Shapefile, OGR GMT etc) and load that into GPlates, cookie-cut – or rather – assign plate ids and there is a regional, regular-spaced point set which you can use as velocity domains.

    Creating a regular spaced point grid to be used as velocity domain, using QGIS.

    Creating a regular spaced point grid to be used as velocity domain, using QGIS.

One important thing to know is that GPlates utilises layers for different types of data, so here’s a little digression and some background info. This layer business is much like GIS software has vector and raster layers, and other layers which are the result of some computations/combination of other layers – or like image manipulation applications like GIMP or Photoshop. The following layer types exist in GPlates (n.b. ideally a detailed overview of the layers with examples should follow here but currently there is no time, I suppose this might be found in the GPlates documentation somewhere):

  • ReconstructionTree: Rotation files/models are automagically assigned to the yellow layer.
  • ReconstructedGeometries: Essentially all feature data one works with gets stuffed into the green layer type. Shapefiles, standard GPML files (not topologies), OGR GMT files all belong to this layer type.
  • Co-registration: Features related to data mining and association checks are combined in this layer type. This is one example where a layer does not correspond directly to a feature collection (ie a single file on disk). Instead, the user selects a set of seed points (a feature collection) and target geometries/rasters (another feature collection/raster) to generate a new data type.
  • Calculated velocity fields: Another one of of those layers where the layer=feature collection equation breaks down. Here we load a point feature collection and combine it with other data such as topological polygons or static polygons to compute velocities.
  • Resolved topologial networks:   This (brown) layer type is similar to the topological geometries but creates triangulated networks used for deformation.
  • Resolved topological geometries: Topologically closed polygons are created from the combination of a rotation file (ReconstructionTree) and a feature layer (ReconstructedGeometries).
  • Reconstructed Raster: Raster data gets loaded into the red layer type.
  • 3D scalar field: Volume-rendering of scalar fields as, for example, from seismic tomography

Now that this is sorted, let’s generate some velocity fields.

After completing either of those steps above you should have a point layer from which you can now extract velocities. Depending on the way you created these points, the steps to display velocities might differ a bit.

Generating plate velocity vectors:

  1. Create a global velocity domain layer using (1) from above or create your velocity points using method 2 or 3 from above.
  2. Once you have created the velocity domain layer using option (1), GPlates will automatically load a new feature layer with the name of velocity point layer as you specified. If you have  used the “Latitude Longitude” option in GPlates, your default layer name will be something like lat_lon_velocity_domain_SomeIntegervalue and have a green color in the layer window (the ReconstructedGeometry layer). In addition, GPlates will create a aqua turquoise colored layer (the “Calculated Velocity Fields” layer) with the same name. That is where ze aktion happens.
  3. Creating a global velocity domain layer will automagically create a "ReconstructableFeature" layer (green) and a "Calculated Velocity Fields" layer (turquoise) - see arrows. Above, the different layer types are displayed from the drop-down menu.

    Creating a global velocity domain layer will automagically create a “ReconstructableFeature” layer (green) and a “Calculated Velocity Fields” layer (turquoise) – see arrows. Above, the different layer types are displayed from the drop-down menu.

  4. The next thing you need is either a static polygon feature collection set or a topologically closed polygon feature collection and of course a rotation model loaded. The next steps will illustrate why.
  5. Only if you have created your point velocity domain layer manually): Add a new blank “Calculated Velocity Fields” layer throug clicking the big at the top of the layer window.
  6. The key now is to connect the Velocity layer with other feature data. This can be a global polygon data set or a set of static polygons. Open the aqua-colored layer by clicking the little triangle on the left and  under the “Inputs” category you see the headings “Velocity domains” (and your generated velocity domain reconstructable geometry layer is already connected to this. If it is not connected, you need to add it here. This is  the case where you created it manually. Under the “Velocity surfaces” heading, click on the “Add new connection”  field and select the polygon data. If one is set to export velocity boundary conditions for geodynamic modelling this of course requires continuously closed polygons, covering the surface of the Earth at any reconstruction time. Such topologically closed polygon data is provided on the EarthByte website. If you are working regionally, just select whatever polygon feature collection you are working with.
    The "Calculated Velocity Fields" layer unfolded.

    The “Calculated Velocity Fields” layer unfolded.

    Adding regional polygons as surfaces for the velocity calculations

    Adding regional polygons as surfaces for the velocity calculations

  7. In the case that you are using global velocity domain points and polygons, make sure that in the “Velocity and Interpolation options” section,  Calculate velocities is set to “of surfaces” . If you have a set of points that are assigned plate ids, one would want to set this to “of domain points”.
  8. Once you have connected your surface domains with your velocity domains, you should see velocity vectors displayed in the main window. The settings for the vector display can be altered in the layer dialogue.

    Screenshot showing calculated velocities at a set of global domain points. Note that velocities are only displayed where surface polygons overlap with the points.

    Screenshot showing calculated velocities at a set of global domain points. Note that velocities are only displayed where surface polygons overlap with the points.

  9. If needed these vectors can be exported using the “Export…” function under the “Reconstruction” menu in a wide flavour of formats.

Graticules for plate tectonic reconstructions

Plate tectonic reconstructions require to have some present-day markers so that any person reading or looking at the results can correlate the paleo plate positions and continents with present day. Things did indeed look quite a bit different back then… Usually the present-day coastlines are used a such a marker, but as sealevel has varied extensively over the geological history, displaying an Early Cretaceous reconstruction at, say 110 Million years, with present-day shorelines might be a bit misleading. In fact one could probably say that it is plainly wrong.

So what’s the big deal about this you might ask. One key aspect of graticules is usually that they are not “features” in the sense of tangible geospatial data, but rather a “decorative” overlay. GPlates also displays a fixed graticule (the thin gray lines spaced at 30 degrees) on the globe. However, in plate tectonics, if we go back in time, we require that such lines and decorations become ‘reconstructable’ back through geological history. So we need data, not decorations. During the 1980’s in the famous PLATES project at UTIG, a special data type called ‘Gridmarks’ was invented which was pretty much a set of crosshairs, centered at equally spaced increments, mimicking a graticule which could be reconstructed. Take these individual crosshairs, assign them a lifespan and a PlateID and one could simply reconstruct them as continental outlines and other geospatial features were. All worked quite happily with this concept and this old file.

When I started to work on South Atlantic plate kinematics, I realised that, albeit being quite useful, some smaller plates would simply be missed by the grid marks. Also, one would also have to reassign plate ids for any new set of polygons one is working with and sometimes a bit of an update to the way things are being done is also refreshing. The routines which were used to generate these gridmark files (in the old PLATES *.dat format) were written in Fortran and I don’t have a compiled version for my OS at hand and knowing the pain associated with this exercise I opted for a writing a new routine from scratch in Python tailored for use with GPlates.

The result of this is  a short script called “CreateGraticuleLines.py” in my gptools repository on Atlassian’s BitBucket which borrows command line options from GMT. I’ll give a brief overview about the usage. First,  either clone the toolbox using

git clone https://bitbucket.org/chhei/gptools.git

or simply download a zipped archive (either from the overview or from the Downloads page under ‘branches’). Open the terminal, cd to the place where you have downloaded the archive, unzip and type:

cd gptools
chmod u+x CreateGraticuleLines.py
./CreateGraticuleLines.py --help

So now you can create regular spaced graticule lines either for a global coverage at 5 degree line spacing (the default settings, equivalent to -Rd -I5) or for any other bounding box and line increment you require. For example

./CreateGraticuleLine.py -R-10/40/40/70 -I1

will create the following file for GPlates (by default the output name for the file is “Graticule.gpml”):

Graticule lines feature collection showing a 1-degree graticule covering most of Europe. Note that all lines are individual features.

Graticule lines feature collection showing a 1-degree graticule covering most of Europe. Note that all lines are individual features.

Once you load plate polygons into GPlates along with the graticule lines, you can use the cookie-cutting functionality in GPlates to cut the lines and assign individual plate IDs along with an “age of appearance” (set to 0 Ma by the script). It should then somehow look like this:

A global graticule (using the default settings) cookie-cut and age-assigned along with plate polygons.

A global graticule (using the default settings) cookie-cut and age-assigned along with plate polygons.

Once you have cookie-cut and age-assigned the graticule lines, you can reconstruct them like any other feature in GPlates. Note that areas without plate ID and changed ages of appearance will not display once you step back beyond 0 Ma (present day). Here’s another screenshot:

Graticule lines as reconstructable features - cookie cut to plate polygons and rotated back to 100 Ma.

Graticule lines as reconstructable features – cookie cut to plate polygons and rotated back to 100 Ma.

If you now export your reconstruction as set of GMT files, you can now use GMT’s psxy to plot a graticule mesh on top of your reconstruction maps using various line styles. A last example from my South Atlantic maps:

An example of the application of the reconstructed graticules here as thin gray dashed lines (Heine et al., 2013)

An example of the application of the reconstructed graticules here as thin gray dashed lines (Heine et al., 2013)

Similarly, this will also work when using GPlates’ SVG export. You can of course also export your graticule file in GPlates to different formats – such as ESRI Shapefile or OGR GMT plain text format (“Save as” functionality in GPlates’ feature manager).

Happy map making. For bug reports and improvement suggestions please use the BitBucket issue tracker or the commenting functionality here.

The “GROT” – A new rotation file format for GPlates

I had written earlier about the new file format (“GROT” – pronounced g-rot) for GPlates rotation files. After probably 3 decades of plain text rotation files with freeform comments, things like direct links to DOIs and bibliographies, structured document metadata and the option to include Hellinger style rotation statistics in the rotation files finally arrive in the plate tectonic reconstruction world 2.0. The technical details of the new format are described in a supplementary PDF to the GPGIM Paper by Michael Chin. Now, the next release of GPlates will be able to read and write those files natively, below are a few screenshots (the release of GPlates v1.3 is pending). But first – what is metadata and why could it potentially be useful in the context of plate tectonic rotation models? Here’s a little description from the Dublin Core website:

The word “metadata” means “data about data”. Metadata articulates a context for objects of interest — “resources” such as MP3 files, library books, or satellite images — in the form of “resource descriptions”. As a tradition, resource description dates back to the earliest archives and library catalogs. The modern “metadata” field that gave rise to Dublin Core and other recent standards emerged with the Web revolution of the mid-1990s.

Now, such information is important when it comes to the ‘semantic web’ and having machines to process such files – something Matt Hall from Agile recently commented on “Machines can read, too“). This means there is information about the author, access rights, description of the content and coverage (spatial and temporal) and contributors (where there is no publication for a rotation sequence yet), plus some info on how to programmatically process the data, encoded. The GROT format is naturally not as engineered as GPlates’ GPML files or GeoSciML, but I think it strikes a good balance between easy enough to edit by hand in a text editor while maintaining some structure in order to automatically process the file content.

Here’s a little talk through the new features of the GROT files – nothing seems to have changed in the Total Reconstruction Sequence View in GPlates apart from an additional ‘Show metadata’ button at the right hand side of the window (the column names you see in the window are from an earlier development version and will not show up in the pending release version):

So far nothing seems to have changed in the "View Total Reconstruction Sequences" window

So far nothing seems to have changed in the “View Total Reconstruction Sequences” window (the column names shown were only in a development version and will not be in the release candidate of GPlates)

Now, if the individual sequences are expanded, the metadata is revealed and the individual columns populated:

However, when expanding the individual sequences you are now exposing chunky parts of metadata for each moving plate rotation sequence or the individual rotation.

However, when expanding the individual sequences you are now exposing chunky parts of metadata for each moving plate rotation sequence or the individual rotation – again this was only in an earlier development version and will not be in release 1.3.

That little “Show metadata” can be used at various locations, depending where you have focused the view. Here we click on a single pole:

Individual moving plate rotation sequence metadata

Individual moving plate rotation sequence metadata

Now, note the structured fields in the right window – Here the user can specify geological timescales used for the rotation sequences, magnetic anomaly chron IDs, the reference(links to a local BibTeX database/ and DOI of the publication. In addition author names (AU), modification time (T) and plate pair (PP) can be added as valuable comments. In future incarnations there will hopefully be a hyperlink so that one can just click on the DOI in GPlates and  get automagically taken to the publication.

There is one other interesting part to the GROT format, namely the file metadata – this is a whole chunk of structured text at the top of the file (similar to what one uses in LaTeX or in a well-written script header). This metadata encapsules global information for the GROT file which GPlates can understand, namely:

  • The GROT version (for future-proofing updates to the format)
  • a Dublin Core metadata section as discussed above
  • Data relevant for the revision history of a rotation file
  • Data to connect the rotation file with a bibliographic database
  • Information about the geological time scales used to convert magnetic anomaly chrons and stratigraphic ranges to absolute ages

Here’s a screenshot of the metadata from the GROT file header:

GROT File header metadata

GROT File header metadata

Now a closer look at the contributors:

Details about contributors to a rotation file as part of the Dublin core metadata set

Details about contributors to a rotation file as part of the Dublin core metadata set

Now that all sounds probably quite complicated, but here’s a raw vision of how this looks like in a text editor with syntax coloring (using TextMate and my GROT bundle, see also this post here):

A raw vision of a part of the DublinCore file description

A raw vision of a part of the DublinCore file description

Raw vision of the moving plate rotation sequence

Raw vision of the moving plate rotation sequence

With the new GROT file format, the possibility also now exists to link certain features into the rotation file. In the following screenshot, the rotation of South America (201) relative to Africa (701) is specified based on picks of magnetic anomaly chrons (@CHRONID”…”). This structured format now allows to (theoretically) extrat all magnetic picks of e.g. CHRONID=”CAn18″ from a geospatial pick database with a simple where the plate id = 201 and the conjugate plate ID = 701.

Detailed annotations of stage poles allow linking to geospatial features, such as magnetic anomaly picks expressed through a CHRONID tag.

Detailed annotations of stage poles allow linking to geospatial features, such as magnetic anomaly picks expressed through a CHRONID tag.

So with the upcoming release of GPlates v1.3, finally structured, meta-data rich rotation files have arrived. More examples and some updated on the final release features should follow later.