we gave a talk
this week in SimPEG: week of May 15
This week, myself, SEOGI KANG, Dom Fournier, and Craig Miller gave a talk for the BC Geophysical Society. The talk was held downtown, and ~30 people attended — mainly practicing geophysicists who work here in BC. This was our first SimPEG talk for a local Vancouver crowd. Our slides are available here.
I started off the talk with an overview of SimPEG. I borrowed my first slide from the talk that Doug gave in Banff this past summer: as the field of geophysics is progressing, the gap between people who are working on applied, multidisciplinary problems and those working on algorithmic problems is ever increasing. Yet progress is only really made when we can bridge that gap — apply the state of the art tools to the challenging multidisciplinary problems we are facing. This is the space we are working to play in with SimPEG.
So what can you actually do with SimPEG?? Dom takes over to get started on answering this question. He introduces the Kevitsa deposit — a porphyry Ni-Cu-PGE deposit in Finland. From seismic surfaces interpreted by Emilia Koivisto, we defined a synthetic geologic model. From there, Dom introduced the potential fields data, starting with Gravity. He constructed a density model based on the field data; from there he demonstrated that the using compact norms in the regularization, we can hope to image the lower limit (in depth) of the intrusion (code). The magnetic data are particularly interesting over Kevitsa as there are large negatives in the data over the centre of the intrusion, indicating the presence of remanence (notebook). How best to invert data in such a setting is ongoing research in Dom’s thesis.
The potential fields data gave us an indication of the camp-scale geology, but isn’t directly sensitive to the mineralization. Next up, Seogi introduces electrical and electromagnetic methods over Kevitsa. DC resistivity and Airborne time-domain (VTEM) electromagnetic surveys were conducted over Kevitsa. From the VTEM data, we see that the sedimentary units are conductive and there are indications that the mineralized zone is conductive. Similarly, the DC resistivity data indicate both the sedimentary units and mineralized zones are conductive. Using these data, Seogi constructed a synthetic electrical conductivity model of Kevitsa. Using the TDEM code in SimPEG, he demonstrated the type of signal we would expect to see from an isolated, compact body (notebook). Using the DC code, he showed not only how we can simulate data, but also visualize how electrical currents flow in the earth; this is a powerful way to ask survey design questions (notebook). He finished the discussion on electrical / EM methods by showing a DC resistivity inversion — these results indicate that we should be able to not only detect, but also resolve the mineralization.
Craig Miller wrapped up the show with a discussion of the work he has been doing with Gravity data over Laguna del Maule, a volcano in Chile. This is a different style of scientific challenge than Kevitsa as the data coverage and supporting geologic data (eg. from drilling) is limited. Instead, there are conceptual models of how a magma chamber is structured; can inversion be used to ask questions and build up a conceptual model of the magma chamber at Laguna del Maule? As the data coverage over the site is sparse, particularly over the main magma chamber which is beneath a lake, Craig used a checkerboard test to examine how station distribution impacts our ability to recover a model from gravity data. This test demonstrated that there should be sensitivity to structures beneath the lake, so he then conducted an L2 “smooth” inversion. This inversion showed a low-density structure where we expect to see the magma chamber, however, being a smooth inversion, it did not much edge definition. As a next step, he conducted a mixed norm inversion, this sharpened up the edges and enabled interpretation of the density model in terms of a thermodynamic model including: melt, crystal and volatile proportions. You can check out his paper and the code used to invert the gravity data.
Following the presentation, we took questions and had a number of conversations. A few of the highlights there were “Is it really free for use” — yes, and on the modelling side, the ability to interrogate the fields and fluxes, for example by looking at the currents in a DC resistivity survey, differentiates SimPEG from many other currently used simulation codes.
In other news:
- We launched http://computation.geosci.xyz, a web resource aimed at providing examples and tutorials for computational geophysics, specifically simulations and inversions. It is still rough around the edges!, but we are always game to hear feedback and welcome ideas for new contributions
- A new beta-release of SimPEG, SimPEG 0.6.0b1, was released. You can see the most recent release notes here. You can upgrade to the most recent SimPEG version using
pip install SimPEG --pre
See you next week.
