The Carl Heiland Lecture Series takes place every Wednesday at 4:00 PM during the fall and spring semesters. Each week, we are joined by a distinguished speaker from academia, industry, or government on a topic pertinent to the geosciences. The lecture series is a public event open to all members of the Mines community and beyond.
We are excited to announce that the Heiland Lecture will be available remotely this fall, both for the safety of our speakers and attendees, and to enable you to see each lecture from wherever you are in the world. The schedule is posted, below; watch for more information, including abstracts and links to tune in.
Fall 2020 Schedule
August 26, 2020
No Lecture Scheduled
September 2, 2020
Unconventional reservoir characterization using seismic-band DAS records
Dr. Ariel Lellouch
Postdoctoral Research Fellow, Geophysics
Stanford Exploration Project, Stanford University
Distributed Acoustic Sensing (DAS) can record seismic wavefields with an unprecedented spatial resolution. We focus on many of its possible uses for reservoir characterization, concentrating on the seismic frequency band and the array processing approaches taken to utilize it. We first illustrate how records from a vertical DAS array deployed in an Enhanced Geothermal System project can be simply and effectively processed to construct velocity models, detect microseismic events and earthquakes, locate them, and estimate their magnitude. Then, we show how a horizontal DAS array deployed in an unconventional shale reservoir records dispersive guided waves propagating for hundreds of meters with frequencies as high as 700 Hz. Thanks to the high spatial resolution of DAS, guided waves are recorded unaliased despite their very short (<10 m) wavelengths. We compare the field observations with semi-analytical and wave-equation modeling approaches. Guided waves are also strongly affected by the interaction with open fractures induced by hydraulic stimulation. We use a horizontal cross-well acquisition of perforation shots in a simple, geometrical analysis of horizontal fracture growth. We also observe secondary diffractive signals, some of which are still unexplained. We conclude by discussing several future applications of guided waves, including preliminary EFWI results.
September 9, 2020
Bayesian inversion and model selection with poorly informed geological spatial priors
Dr. Niklas Linde
Professor, Environmental Geophysics
University of Lausanne, Switzerland
Bayesian theory allows combining conceptual geological understanding with indirect hydrogeological and geophysical data. Conceptual geological knowledge represented in the form of a training image (e.g., a geological map, photograph of an outcrop) is traditionally used to sample prior model realizations using multiple-point statistics (MPS) algorithms. In the context of Markov chain Monte Carlo (MCMC) inversion, I highlight the advantages offered by replacing MPS-simulations with draws from deep generative models (e.g., variational autoencoders, generative adversarial networks). Since our prior knowledge about spatial statistics is typically weak, I emphasize approaches to Bayesian model selection enabling comparison of competing conceptual models. Finally, I discuss deep neural networks that are trained to map measurement data (vector) into a subsurface model (image), thereby, bypassing formal inversion.
September 16, 2020
Seismic investigations over and within hypervelocity impact structures, at three sites
Dr. Douglas Schmitt
Professor and Stephen and Karen Brand Endowed Chair of Unconventional Energy
Department of Earth, Atmospheric, and Planetary Sciences
Hypervelocity impacts are a fundamental geological process within the solar system and on the earth. On the earth, many of such structures ore heavily eroded and buried; and often they are only first revealed (often inadvertently) in geophysical surveys. The geophysical anomalies detected result from extreme changes to the physical properties of the earth materials during shock deformation due to high pressure metamorphism, melting, and damage. The seismic geophysical changes introduced to these materials are studied through borehole seismic investigations at the Bosumtwi Structure in Ghana and the Chicxulub Structure off the Yucatán more recently, and surface seismic imaging at the Bow City Structure, Alberta. These all show anomalously low seismic velocities that is primarily due to the high degree of damage the materials have suffered. These zones, too, are attenuating to seismic waves due to both extensive wave absorption and scattering. Work is progressing on attempting to determine the ratio between these from the Chicxulub seismic data set as this may provide some insight into the levels of damage and the scale of heterogeneity. Understanding this better may, too, assist in the interpretations of future upper crustal geophysical investigations from the Martian and Lunar surface, information that may be key to the development of habitations.
September 23, 2020
Performance issues and scalability of ML tasks for Geoscience
Dr. Mauricio Araya
Senior Computer Scientist and Senior R&D Manager
Total E&P Research and Technology USA
Machine Learning (ML)-based applications in geosciences are covering most fundamental problems and maturing, the natural next steps for these applications include training and testing on larger datasets for more complicated tasks. Unfortunately, scaling ML models is not straightforward, and just taping on existing High Performance Computing (HPC) techniques is not enough. In order to reach accuracy targets and increase parallel performance, scaling ML models requires deep understanding of the ML task. In this talk we will discuss the effect of carefully selected batch size intertwined with scaling strategies for training of a segmentation task is analyzed.
September 30, 2020
Hunting the magnetic field through ocean drilling
Dr. Lisa Tauxe
Distinguished Professor of Geophysics
Scripps Institution of Oceanography
University of California, San Diego
Earth’s magnetic field has been the target of scientific investigation for over four centuries yet the basic fact that the field switches polarities, though suspected for over a century was not proven until the early sixties. This fact was key to the plate tectonic revolution and part of the rationale to begin drilling the ocean floor with the Deep Sea Drilling Project over fitty years ago. And the study of the Earth’s magnetic field hos remained on integral port of ocean drilling throughout the history of endeavor. In addition to flipping polarity, the Earth’s magnetic field changes both direction and strength on time scales from decades to millennia. Human observations of field directions provide constraints for field behavior since the age of maritime exploration starting in the fifteenth century but field strength measurements only started in the 19th century, so understanding of the geomagnetic field requires the use of “accidental” records such as sediments and igneous rocks. Because 70% of the surface of the Earth is covered by ocean, marine records are essential to get a global view of history of the Earth’s magnetic field and records beyond a few million years require ocean drilling. The International Ocean Discovery Program maintains cores from fifty years of drilling. Magnetic measurements on these cores continue to provide clues as to the timing and nature of magnetic reversals, attempted reversals (excursions), and the rise and fall in field strength since the Jurassic.
October 7, 2020
Turning fiber optic cables into the next-generation seismic networks
Dr. Zhongwen Zhan
Assistant Professor of Geophysics
California Institute of Technology
Seismology is one of the main approaches to study quakes and image Earth and planetary interiors. However, deploying large-scale dense seismic networks has been challenging. Distributed acoustic sensing (DAS) is an emerging technology that converts every few meters of a long (currently 10’s of km) optical fiber into a seismic strainmeter. At its most basic level, DAS works by shining a laser pulse into the fiber from one end and interrogating the “echo” of Rayleigh scattering from intrinsic fiber defects. DAS provides a scalable and affordable way to deploy a dense seismic network, by installing dedicated fiber cables or leveraging existing telecommunication fiber networks. In the last three years, we have been exploring the potential of DAS in the next generation seismic networks on different scales. More specifically, we test DAS in earthquake detection, structure inversion, and hazard assessment. In this talk I will give an overview of these efforts and our vision for the future.
October 14, 2020
No Lecture Scheduled
October 21, 2020
The long tail of induced seismicity
Dr. Jake Walter
Oklahoma Geological Survey
University of Oklahoma
The rate of earthquakes across the United States mid-continent has dramatically increased since 2009, concurrent with a surge in activity to extract hydrocarbons from unconventional plays. Oklahoma, over the last decade, experienced a year in 2015 where there were ~900 M3.0+ earthquakes against a prior tectonic background rate of just 2-3 M3.0+ earthquakes per year. During that surge in small earthquakes, 4 of the 5 largest earthquakes occurred, all greater than or equal to M5.0. These earthquakes caused moderate but not widespread damage to the rural communities in which they occurred. The number of earthquakes M3.0+ has subsequently declined to 194 and 64 in 2018 and 2019, respectively. The scientific community has since identified wastewater disposal into the Ordovician-age Arbuckle Group as the most plausible cause of the surge in earthquake activity. In the talk, I will present recent research related to the ongoing decline in seismicity since 2015, aftershock properties of Oklahoma earthquakes, and the elevated potential for inducing seismicity in areas where active injection and/or hydraulic fracturing are already ongoing. Finally, I will discuss a hopeful roadmap to the future where scientists, industry, and regulators can work quickly to identify induced seismicity and mitigate damage to lives, property, and public opinion. The lessons learned in Oklahoma are relevant for a future where carbon dioxide removal and storage are practiced on a widely-distributed scale.
October 28, 2020
Tsunami assimilation for rapid forecast: Application to Cascadia and New Zealand
Dr. Anne Sheehan
Cooperative Institute for Research in Environmental Sciences (CIRES)
University of Colorado – Boulder
In the data assimilation method, long used in weather forecasting, real-time observations are continuously assimilated to produce a forecast. Here we apply the data assimilation method to seafloor pressure data from temporary ocean bottom seismic experiments offshore Cascadia and New Zealand. The tsunami observations from these dense arrays are used to update the wavefield at each time step to forecast the coastal tsunami in the vicinity of the array, and knowledge of the tsunami source is not required. For the Cascadia example, we use the tsunami from the magnitude 7.8 2012 Haida Gwaii earthquake as recorded on 45 seafloor pressure gauges of the Cascadia Initiative ocean bottom seismic experiment. In the New Zealand example, we use the tsunami from the 2009 magnitude 7.6 Dusky Sound earthquake as recorded on 15 seafloor pressure gauges from the MOANA ocean bottom seismology experiment off the west coast of the South Island. Our results illustrate that with an appropriate array of sensors, models continuously updated with real-time tsunami wave data could be used for rapid tsunami forecast. In our case, the data were not available in real-time, thus only a retrospective simulated forecast was produced. Given the high cost of cabled ocean bottom instrumentation and low latency tsunami buoys used for real-time transmission of tsunami data, we investigate the use of non-traditional observation strategies such as ship position data in the tsunami data assimilation framework.
November 4, 2020
November 11, 2020
November 18, 2020
November 25, 2020
Thanksgiving Break – No Lecture Scheduled
December 2, 2020
No Lecture Scheduled
Subscribe to future Heiland Lecture Announcements