Geophysics From Small to Big Data

Abstract

In this thesis, I focus on the induced seismicity and glaciology topics. The amount of data available to support my models varies from non-existent (chapter 2) to massive (chapter 5). Central United States, a tectonically quiet region, has recently experienced a surge in seismic activity. Studies correlate the increase in seismicity with the rates of saltwater disposal. The saltwater is co-produced with oil during its extraction from shales and then injected into deeper geological formations. Because of the hydraulic communication between the saltwater reservoir and faults in the crystalline basement, the on-fault pore fluid pressure increases, which may nucleate earthquakes. The off-fault permeability changes dramatically during induced earthquakes. In chapter 2, I develop a model for this permeability evolution based on micromechanics. This model can simulate pore fluid pressure history during induced earthquakes. This will allow to infer the distribution of seismicity on these activated faults. Induced earthquakes are often small or moderate but they contribute to seismic hazard. Since most of the earthquake detection methods from seismograms are designed for large events, they miss the low-magnitude seismic events. In chapter 5, I present the first convolutional neural network for earthquake detection and location from a single waveform. I apply it to detect earthquakes in Oklahoma (USA). It detects 20 times more earthquakes than previously cataloged by the state survey and runs orders of magnitude faster than established methods. Ice streams in West Antarctica are separated from the nearly stagnant ice by zones of highly localized deformation known as shear margins. In chapter 3 and 4, I present results supporting the idea that internal melting at the shear margins controls the ice streams' width and stabilizes their migration. First, I show that a significant portion of the ice column at the margins of most of these ice streams is temperate (chapter 3). The shear margins support less lateral stress and, hence, most of the resistance to the downstream ice flow comes from basal resistance. Then, I investigate the hypothesis that the high basal resistance at the shear margins is a consequence of channelize drainage (chapter 4), in agreement with limited observations. I highlight the intimate relationship between internal melting at the margins and subglacial hydrology and show how this stabilizes the margins' migration.