Re-evaluating Remote Dynamic Earthquake Triggering by First Establishing Background Seismicity Rates

Kristine L. Pankow and Debi Kilb

AGU Meeting, 2018, abstract #S51B-05

The phenomenon of remote dynamic triggering has been well-established since the 1992 Landers, California earthquake triggered earthquakes from California to Yellowstone. Since then, there have been many other studies that have clearly shown a direct link between remote mainshocks and small magnitude local seismicity. With the development of matched filter and other detection algorithms, there is an increasing number of instances documenting remote triggering. However, in many studies, the triggering criteria is based on examining the number of earthquakes in small time windows surrounding the arrival of the teleseismic waves and rarely is the null hypothesis, that the rate change is merely a random occurrence, fully tested. In this study, we examine 33-years (1984-2017) of catalog data from four catalogs (Anza, CA, Montana, Utah, and Yellowstone). We bin these data into three-year increments, determine a magnitude of completeness, and then determine the average number of events within a given time window (5- 12- and 24-hours). We count the number of events in every time window using a 1-hour sliding window over the three-year catalogs. This nets over 26,000 measurements for each three-year window, providing a statistically meaningful dataset to compute mean and standard deviation. To identify times when the counts are significantly high (probabilities ≥ 96%), we use five times the standard deviation of the mean following Chebyshev’s Inequality Theorem (CIT). We compare the times of significant rate increases to the P-arrival times of 524 mainshocks (M6.5 to 8). This new method identifies previously identified sequences following the Landers and Denali Earthquakes and identifies up to 20 potential cases of remote triggering including triggering in Yellowstone following the 1985 Mexico earthquake. In comparison, the previous method that looks at rates using events within a chosen time window pre- and post- the arrival of the mainshock identifies 40 to over 100 instances (depending on the time window) of remote triggering, many of which fail the null hypothesis test when using the newly calculated background rates. To get instances of remote triggering on par with the previous method, one would need to use two or three times the standard deviation as a threshold, well below the criteria established by the CIT.

 

Exploring subtle temporal changes in earthquake catalogs to guide identification of dynamically triggered events

Debi Kilb and Kristine L. Pankow

SSA Meeting, 2018

In studies of remote dynamic triggering, the common practice is to be reactionary, in that the impetus is to select a large earthquake and then systematically search for an increase in small earthquakes at remote distances. This approach requires specifying a duration (e.g., 5, 12, 24 hrs) used to search for seismicity rate changes. Rarely is the null hypothesis, that the rate change is merely a random occurrence, fully tested. Here, we take a different approach and simply count the number of events (above the magnitude of completeness) in a moving window of a given duration (e.g., 5, 12, 24 hrs), creating a catalog of count values for each sub-window over the full duration of a catalog.  As expected the count increases substantially when there are local mainshock/aftershock sequences; however other rate increases can be identified as well. To identify times when the counts are significantly high (probabilities >= 96%), we use five times the standard deviation of the mean following Chebyshev’s Inequality Theorem.  Applying this method to regional earthquake catalogs, we successfully find statistically high counts following the 2002 M7.9 Denali, Alaska, earthquake in Yellowstone and Utah indicative of the already known remote triggering. In addition, we find evidence of potential remote triggering in Montana following the 2002 M7.6 Papua Indonesia earthquake, and in Yellowstone following the 1985  M8.0 Michoacan, Mexico, earthquake. There can be a strong trade-off between the length of the moving window and the robustness of the results. If the window becomes too small, an unrealistically large number of times may be flagged.  These results indicate that we can use a simple earthquake count within temporal windows to identify potential remote triggering. 

 

Identifying New Earthquake Templates Adds Valuable Information to Induced Seismicity Sequences

Lisa Linville, Kris Pankow, Debi Kilb, and Justin Rubinstein

SSA Meeting, 2017

Comprehensive reprocessing of seismic data from sedimentary basins in the Central United States during the occupation of EarthScope’s Transportable Array adds a significant number of new sources to existing earthquake catalogs. Subsequently, using subspace detection to increase the resolution of time-histories for sources near active injection wells reveals prolific seismic sequences in areas where other catalogs show no seismicity, such as sections of the Denver Basin. In other cases, under the same scrutiny, seismic gaps persist despite decades of injection (e.g., Williston Basin, North Dakota). In areas where robust levels of seismicity are well documented in existing catalogs, new events can either echo event families from relatively stable and consistent sources (e.g. Dagger Draw Field, New Mexico) or greatly diversify known families (e.g. more than double in Cogdell, Permian Basin). For both cases, catalog enhancement through subspace detection alters event time-histories when new templates contain non-redundant information. We observed that cataloged events are not necessarily the largest events in a sequence based on waveform amplitude and duration. We further observed that the character and complexity in a sequence is not necessarily tracked through the largest magnitude events. The implication of these observations is that additional templates can significantly change interpretations based on correlations between event locations or time-histories and pumping records or fluid migration models.

 

Tilt Trivia: A Multiplayer App Teaching Induced Seismicity Concepts

Debi Kilb, Alan Yang, Nathan Garrett, Victoria Hilke, Kris Pankow, Justin Rubinstein, and Lisa Linville

SSA Meeting, 2017

Today’s technology is opening up new ways to learn. Here, we introduce TILT TRIVIA, a suite of Jeopardy! quiz-style multiple player games for use on mobile devices and tablets (Android or Apple) to help students learn simple definitions and facts. The game was built using the Unity engine and runs on a server, allowing the multiplayer functionality to run seamlessly, all day, every day. Although the game is configurable to any topic, here we focus on our game that teaches players about induced earthquakes. From a suite of 15-20 questions, 6 questions are chosen at random for gameplay. A single TILT TRIVIA game takes 3-5 minutes to complete and allows up to 5 players to play simultaneously in the same gamespace. To begin, each player selects a zany avatar to represent them in the game. While in the competitive playing field, players are presented with a question and then they simply “tilt” their tablets until their avatar rests on the correct answer marker (i.e., true or false). Because the playing field is constantly tilting, keeping your avatar on the correct answer requires a continual counter-tilting motion of the tablet to maintain your position within the game. A countdown timer is incorporated, requiring players to answer the questions as quickly as possible. At the end of a game, a leaderboard displays the players’ scores and rankings, a metric that motivates repeat play. Players have the option of jostling other players off of the correct answer in the hopes to net the highest score. Join us and play today (http://siogames.ucsd.edu/TiltTrivia/index.html)!

 

 

Template hunting for small events using Frequency Domain Array processing in regions of induced seismicity

Lisa Linville, Kristine Pankow, Debi Kilb, Nathan Garrett

SSA Meeting, 2016

It is common practice to use waveform templates from seismicity catalogs in waveform matching techniques to increase the size and completeness of seismic catalogs. These improved catalogs are key in characterizing responses to changes in local conditions in areas subject to induced seismicity (e.g. where extraction and/or injection occurs). However, regional catalogs rarely contain a complete set of templates, especially for small events (< M2.5), resulting in consistently omitted yet important event families from the analysis. Here, we introduce an automated frequency domain coherence filter method developed for seismic arrays that successfully recognizes small magnitude events.  The method is based on identifying low-level increases in band-limited energy sums in the frequency domain over multi-station distances, allowing us to assemble a more complete set of event templates. In this study, we tune the algorithm for data recorded by the EarthScope/USArray Transportable Array stations deployed in sedimentary basins in the Central United States.  We find false positive rates remain low (<10%) and, with very few exceptions, ComCat cataloged events are fully recovered by our method. Within the Forest City Basin, centered between Iowa and Missouri, we have identified over 250 auto-detected and analyst-reviewed events in the year 2011 that were not previously cataloged. Applying waveform matching to continuous data using these new event templates allows us to evaluate if these new templates belong to already known event families or if they identify new source zones. This frequency domain detection method helps to clarify the diversity of templates and their characteristics within a region that ultimately can help establish priorities in induced seismicity monitoring.

 

Enhancing Seismicity Catalogs for Basins in the Central United States

Lisa Linville, Kristine Pankow, Debi Kilb, and Justin Rubinstein

EarthScope Meeting, 2015

Higher resolution seismic catalogs for active basins in the Central United States (CUS) can help answer fundamental science questions related to fluid-injection and induced earthquake potential that are of increasing concern to the public. Data from Earthscope’s Transportable Array (TA) coupled with composite methods of earthquake detection will enable us to probe CUS basins for more complete background seismicity rates and spatial distributions, in addition to potentially illuminating subtle rate changes related to fluid injection and transient stressing. Here, we outline our approach to obtaining a more comprehensive understanding of basin seismicity using TA data through a joint method of frequency domain earthquake detection, clustering and visualization, combined with subspace (Harris and Paik, 2006) event detection. These methods capitalize on discriminatory spectral characteristics to decrease non-earthquake false detections and rely on clustering and array visualization to validate and associate related earthquake detections. These enhanced catalogs will help reduce the current catalog completeness levels to help determine if transient stresses from the passage of relatively large surface waves dynamically trigger events within these basins or otherwise change the distribution of induced seismicity.