Computer-Oriented Geoscience Lab

Our research

We focus our efforts on creating and applying new methods for 3D modeling and data processing, mostly for gravimetry and magnetometry. Below, you’ll find the main research themes that we pursue.

Magnetic microscopy

The magnetization that gets locked in some minerals (known as ferromagnetic minerals) at the time of their formation is one of a few gateways we have into the Earth’s distant past. So far, researchers have mostly only been able to make bulk measurements of magnetization from collected rock samples. Emerging magnetic microscopy technology is now allowing us to image thin-sections and distinguish the magnetic fields of the individual minerals that make up the rock sample.

Our group is collaborating with experts in paleomagnetism to develop new methods and software for data processing and interpretation, unlocking the huge potential of these exciting new datasets.

Red and blue map showing dipolar anomalies at a scale of 20 micrometers.
Example magnetic microscopy dataset from a thin-section of a speleothem showing tiny magnetic anomalies (on the order of 20 µm in size) that are associated with the ferromagnetic minerals found in the rock.

Antarctic magnetic data and geothermal heat flow

Antarctica is a vast and poorly explored continent that is suffering greatly from the impacts of human-induced climate change. One parameter that is crucial to understanding how Antarctica’s ice sheets will respond to climate change is the flow of heat coming from the Earth’s crust. This so-called geothermal heat flow is needed to constrain models of how ice sheets flow and how the Earth’s crust rebounds upwards once ice mass is displaced, influencing sea-level rise. One of the few ways we have to determine heat flow over the entire continent is through indirect geophysical observations, like magnetic anomaly data.

We are working hard on improving the way that airborne and satellite magnetic data are processed and merged, aiming to reach the highest resolution possible from the data. Our group is also tackling the challenging inverse problem of producing heat flow estimates from the magnetic data.

Map of Antarctica overlaid by red, white, and blue points representing magnetic measurements. The data coverage has a lot of gaps.
The ADMAP2 compilation of open-access airborne magnetic anomaly data for Antarctica. We are improving the compilation by adding more recent surveys and standardizing the data, as well as developing a new process for merging these data with satellite observations.

Geophysical data processing and interpolation

It’s undeniable that a machine learning frenzy has taken over the world. Geoscientists have been doing similar things for decades, for example the equivalent sources technique in gravity and magnetics and spline interpolation are basically supervised-learning methods. Given the many similarities, it makes sense to apply and adapt machine learning techniques and best-practices to these geophysical problems.

At the lab, we are employing machine-learning methods like ensemble techniques and cross-validation to overcome key computational challenges in equivalent-source methods and data interpolation (AKA gridding).

Two maps side by side of the area, with coastlines visible on the upper part, the left map showing red-white-blue dots scattered throughout the continental part, the right map showing a continuous red-white-blue distribution of colors.
Results from our interpolation of millions of ground gravity data from Australia to a uniform height using gradient-boosted equivalent sources. Left: Section of the ground measurements of gravity disturbance (red means positive and blue means negative) from several surveys with different spacings and distributions. Right: Same section of the interpolated grid at a uniform height, which is able to retain the high resolution of the denser surveys while filling in the gaps in the data.

3D geophysical inverse problems

Our ultimate goal as geophysicists is to understand the inner structure and dynamics of the Earth from surface observations. This is a tough mathematical and computational problem: it’s an ill-posed inverse problem, to which a solution might not exist, not be unique, and not be stable.

We develop methods to overcome these challenges and solve different kinds of inverse problems that arise in geophysics.

3D rendering of white blocks with topography and some red blocks spread in a thin vain representing the estimated iron ore body.
Modeled density anomalies associated with iron-ore formations derived from observed gravity disturbances.

Forward modeling gravity and magnetic data

A key component for solving an inverse problem is first solve the forward problem (predicting observed data from a known model of the subsurface). This is often a challenging problem to solve, both mathematically and computationally. Having a forward model that is both accurate and performant is the key to a reliable solution to an inverse problem.

One of our earliest research themes is the development of methods and software for modeling gravitational fields caused by a tesseroid (a segment of a sphere). This is a surprisingly difficult task but is crucial to model geology at continental and global scales. We also produce software that can model different geometries, like points, spheres, prisms, polygons, etc.

Drawing of a curved prism sitting on top of a curved Earth. The prism is cut by white lines that represent the discretization.
A tesseroid (spherical prism) discretized using our adaptive algorithm.