Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article


Doctor of Philosophy




Atkinson, Gail M.

2nd Supervisor

Ghofrani, Hadi


Source parameters of earthquakes play a key role in the understanding of earthquake behavior and modelling of seismic hazard. They describe the size of earthquakes, including how much energy is generated during the rupture, and how the ground motion is distributed over different frequency bands and azimuths. The last decade’s increase in induced seismicity caused by oil and gas production has led to an interest in understanding the underlying earthquake processes and how they can be modelled. This thesis is divided into three studies, each examining source parameters of induced earthquakes in North America.

In the first study, I show that for earthquakes in Central US the variability of ground-motion prediction equations (GMPEs), known as sigma, can be reduced by adjusting the basic input source parameters of location and magnitude. Sigma is an important seismic hazard parameter because it exerts significant control over the expected ground motions at return periods used in seismic design. Refinements in magnitude were shown to reduce sigma more than refinements in location. This reflects that between-event variability is not completely accounted for by magnitude in the GMPE, as it is also influenced by other source parameters such as stress drop.

In the second study, I examine stress drop and corner frequency in the Western Canada Sedimentary Basin (WCSB) using the Empirical Green’s Function (EGF) method. Large azimuthal variations are found in the corner frequencies for earthquakes, which indicates rupture directivity, a phenomenon which can have implications for observed high-frequency ground motion. By modelling the directivity using a Haskell (1964) model, earthquake corner frequencies are retrieved despite the region’s sparse seismic network.

Finally, in the third study, I show that the stress drops obtained from the previous WCSB EGF study can be used as proxies for the GMPE “stress parameter”. I also test whether they provide equivalent measures of the high-frequency content of the earthquake source. I find that GMPE stress parameters tend to yield lower corner frequency values in the forward rupture directivity direction when comparing individual earthquake records. This can be partly attributed to the trade-off between source and site effects in GMPE modeling.

Summary for Lay Audience

The last decade has seen an increase in man-made earthquakes, referred to as induced seismicity, due to developments within the oil and gas industries (e.g., Rubinstein and Mahani, 2015; Atkinson et al., 2016). This has led to an increase in earthquake hazard in regions such as the Western Canada Sedimentary Basin (WCSB) in Canada and the Central US (Oklahoma, Texas). A key question for assessing the hazard is whether these induced earthquakes behave in the same way as natural earthquakes, and if they can be modelled using the same assumptions. This thesis takes a closer look at earthquake source parameters of induced seismicity in order to increase the knowledge of the underlying processes. The thesis is divided into three parts, where the first part examines how the variability of ground-motion prediction equations (GMPEs) are affected by small changes in the source parameters magnitude and location. GMPEs are used by engineering seismologists to describe the expected ground motion levels in a region, and the variability of GMPEs will tell the user how much uncertainty there is around the expected level. The smaller the variability, the more precise are the estimations. The second part of the thesis focuses on one source parameter which determines the high-frequency content of ground motion. High-frequency ground motion is important to most types of ordinary structures, so understanding how the high-frequency content from induced earthquakes differs from that of natural earthquakes is key to assessing the hazard. Finally, the last part looks at the high-frequency ground motions using two different methods that are commonly used within the field of seismology, to see whether alternative methods give us the same information.

Holmgren_2020_Thesis_esupp.zip (122 kB)
Electronic Supplement