135 years ago, an unexpected ~M7 earthquake shook the Charleston, SC area. This project aims to sort through seismic data collected from 2011 - 2012 and apply deep machine learning and template matching techniques to help determine the orientation of the fault that produced the earthquake. By building a new catalog with deep learning and using the template matching system to take another look at the new catalog, we hope we can add several more data points to the existing data sets to get a clearer picture of the fault geometry.
It's been a crazy, chaotic couple of weeks, but I've been loving it! Even when it's stressful, it's amazing knowing that the hard work I'm putting in is meaningful not only for my research but for my personal development as a future scientist. I've encountered several frustrations and setbacks (lost data, memory issues, broken scripts, no space left on my drive etc) that I've had to troubleshoot, and it's really built up my resillience and my understanding of realistic expectations. I've been "fighting" a lot with EQTransformer, but now I know exactly how to use it to work with my dataset.
I presented where I was with my research project to the intern group over Zoom, and I enjoyed getting their feedback while shaking off some of my nerves. I still get nervous before presenting, even in low-key, firnedly settings, so I appreciated another oppertunity to work to refine my skills in this area. We've all put in so much hard work and it's really cool to check in on our progresses.
I'm moving on to the 2nd part of my project which is matched filtering. Wish me luck!
This week, the interns put together mini-presentations detailing how our projects are coming along since we're halfway through our internships at his point. I spent most of the first half playing a maching learning program called EQTransformer (EQT). EQT can take a continous waveform and pick the P and S arrival times instead of being picked manually by a person. We were curious how it would perform on my dataset since it consists of very small, intraplate earthquakes.
The event catalog I'm working with has 134 events, and to test EQT, I downloaded waveforms from IRIS using Obspy 1 hour before and 1 hour after each event. The bad news first...EQT could only identify just over half of the events. The good news is I used the generic model for phase picking, and it's possible to train a custom model to an individual dataset. I don't think I'll have time to do that while I'm here, but I'm interested in building and testing a model customized to my dataset in the future to test its performance. It would be awesome if a model could accurately pick phase arrivals for small, intraplate quakes. More good news: of the events it did detect, it was very accurate in picking the P and S phase arrivals. Poster teaser: the official results will be presented at AGU! 😊
On a personal note, I've been living with my cousin and her American Bulldog, Roxie, while I'm here in Atlanta. It has been amazing catching up with my cousin because I haven't seen her much since she moved down here, and I feel very lucky I have them as a support system while I'm here. I've never lived with a dog before, and Roxie and I have been enjoying our daily walks and play sessions. This weekend, we'll be driving our to my aunt and uncle's new house about 45 minutes from here to help them paint and lounge at their pool. The internship has been fun, but obviously stressful at times, and I'm looking forward to a nice, relaxing weekend with family.
Hope everyone is well!
I created a super rough draft of my map area using ArcMap. There's a ton of cropping and refining that needs to be done (definately don't need the entire east coast!), but it's getting there. I'm going to switch it so my inset is smaller and contains the US map so that the current inset picture is the larger and more promient of the two. Then I need to label the roads, add some rivers, the detected quakes and their focal mechanisms (indicated with beach balls) and create a legend indicating which stations are temporary and which are pemanent. It's a work in progress, but I should complete sometime next week.
I encountered some frustration working with gmt, which is why I switched to ArcMap which is much more familiar and enjoyable for me to work with. My greatest frustration turned into a success I'm proud of though. I learned it's ok to be ok with not knowing everything. I struggled with getting gmt to produce images I liked until I said "ya know what, this might not be for me, and I can do this in a different way" and I did that other way instead. I will still get to the same place at the end, so who cares if I took a slightly different route!
Anyway, back to the map: it is meant to show a large scale and close up view of my study region where small intraplate EQs are happening. I think it's important to have a picture of the east coast of the US so people can understand where the area is and see that it differs from several other seismic areas because it isn't at an active plat boundary. Eventually, I'll put a reference box around the Summer area in South Carolina and label the station map as Summerville so people can connect to where the earthquakes are being detected. Once I broke through my gmt struggle, I've really enjoyed revisiting ArcMap to create a GIS project outside of the classroom.
Thanks for reading!
It’s hard to believe I’m at the end of my 5th week, but here I am! The days have been flying by and my internship is already halfway over already. Time certainly flies…
For the past 2 weeks, I have been busy downloading data from IRIS using a program called Obspy and converting the data into miniseed format. I run the mseed files through a machine learning program called EQTransformer. EQTransformer takes my data, looks for earthquake events, and picks the P and S wave arrivals for me depending on the probability threshold I’ve set in the parameters. I had 2 years and 8 stations worth of data to work with, so it took nearly 2 weeks to get through it all and figure out exactly what I was dealing with. I’m happy to say I wrapped up last Friday and am ready to start some early analysis of the effectiveness of EQTransformer compared to 2 catalogs I was working with using event times detected by HypoDD and Hypoellipse. Next, on to learning eqcorrscan!
During week three, we were tasked to create an “elevator speech” that conveys the relevance of our work in 30-60 seconds. As you can probably tell from my blog posts, I am a pretty long-winded, roundabout talker, so this was a challenge for me! In addition, I am still wrapping my head around the work I’ve been doing here and trying to put into my own words what my project is about in the simplest way possible.
This is an incredibly useful task for me to master because I am developing a skill I will continue to use my entire life. Outside of AGU, I’m going to have to explain to classmates at UW – Seattle, my family and friends, my coworkers and pretty much everyone I’ll ever meet what I do as a scientist and why I am doing it, so learning how to concisely detail my work will serve me well. I’m looking forward to sharing my speech with Jack and getting his feedback. I’ll continue to ask for feedback from my peers and mentors, because this is something that should always be practiced and refined.
Last week, Michael said that our accomplishments during our internship is logarithmic, and I totally understand what he means. Up until the middle of last week, I didn’t feel like I had accomplished anything. In hindsight, I’ve accomplished so much since I’ve been here and am finally wrapping my head around all the work I’ve done. Now I have a better understanding of what I’m doing and what the path is moving forward. I have a much clearer picture of the work I’ve been doing and what I’ll be presenting at AGU, and I’m getting really excited that my project is starting to take form.
Until next week!
Now that I'm halfway through my second week, I'm starting to feel a lot more comfortable with the work I will be doing. I love taking the bus to the Georgia Tech campus every day and getting settled into my workstation in the geophysics lab. All the graduate students I've met in Dr. Peng's group have been incredibly friendly and helpful, and I'm really enjoying getting to know Jack as we take this journey together.
For my project, I'm utilizing a dataset of earthquakes in the Summerville, SC region from 2011-2012 deployed by M. C. Chapman, Jacob N. Beale, Anna C. Hardy, and Qimin Wu. These 8 seismic stations recorded events for one year specifically in hopes of obtaining aftershock signals from the 1886 Charleston M ~7 earthquake. The instruments used were 3 component, 2 Hz geophones sampled at 100 samples per second. I've also collected additional events for my catalog using the IRIS Wilbur3 site, which supplied 2 more events recorded in the area in 2011-2012. I am still looking for more data to add to my catalog.
The strength of the dataset is that the 8 stations deployed by Chapman et. al were placed in a location specific to the hypothesized event epicenter from 1886. The weakness is only 8 stations were deployed, so there is limited data available to analyze. That's where I come in: hopefully, by utilizing deep machine learning to pick my P and S arrival times to create a template and matched filtering techniques to scan the available data with said template, I can detect smaller events within the dataset to help image the substructure of the area and add to the investigation of the focal mechanisms and geometry of the earthquake fault.
As for the skills from the mentoring rubric, I've been reflecting on using software and technology to obtain or collect data. One of my biggest areas for opportunity is getting more familiar with writing command line codes and analyzing data through my terminal. Part of the reason I was drawn to this internship was to spruce up my computer savvy, and I'm enjoying the challenge of learning Linux commands and utilizing programs like SAC and Obspy to view and analyze my data. To demonstrate it proficiently would mean being able to replicate what I've learned without having to rely on tutorial notes. I'm getting there! It's important to develop these skills to quickly and accurately be able to look at and describe datasets and use visual aids to help others understand too. I'll be continuing to develop these skills every day and I use them over and over again as I work on my project.
This weekend, my cousin is taking me to hike in Amicalola Falls State Park via the East Ridge Loop Trail. I've been loving Atlanta and am super excited to get out of the city for a bit to explore the surrounding area. I'm really bad at remembering to take pictures, but I'll try and upload some for my next blog post.
Until next time,
Yesterday, after a whirlwind Finals Week at UW - Seattle, I flew arcoss the country and arrived in Atlanta to start my internship! I was so excited last night to get settled into my place and kick off the next 10 weeks working at Georgia Tech in the geophysics lab of the Earth and Atmospheric Sciences department. I was able to explore the campus this morning and get started with the intro research for my project in the lab today. Dr. Peng and I sat down in person for the first time, and it feels great to be back on a campus and interacting with faculty and students again. It is such a gorgeous campus, everyone is super friendly, and I'm very grateful and excited to be here. Plus, Jack will be here working on a different project with Dr. Peng, so I can't wait for us to hang out together!
There are several things I'd like to accomplish while I'm here, and here is my early list, in no particular order:
1) Brush up on my python knowlege. I took an intro course over 2 years ago and my memory is definately a bit fuzzy. Looking forward to jumping in an learning the ins and outs of some python codes we'll be using.
2) Gain deeper knowlege of MATLAB. I feel more comfortable with MATLAB than Python, so my focus will be python so I can catch up. However, I don't want to lose my MATLAB skills and there are a few MATLAB packages I can explore here as well.
3) Explore Atlanta! I chose Georgia Tech because I wanted something completely different than Seattle to push me outside my comfort zone a little. I grew up in the Midwest and have been in the Pacific Northwest for the last four years, so it's time to discover the south. My cousin lives here, and I'm looking forward to connecting with her and hopefully getting a trip down to Savannah with her while I'm here.
4) Master the local public transportation system. The MARTA rail is a 6 minute ride from my apt to the GT stop, but I want to explore several routes, particularly ones that take me to nearly thrift stores 😊
5) Build my network. I'll be applying for grad schools in the fall, and the more people I talk to, the more informed decision I can make. I also want to build a huge network of people I can collaborate with in the future.
6) Make a great research poster. I'm so excited to be presenting at AGU in the fall and want my exicitment to be reflected in quality work.
7) Find some good hiking trails. I've heard there are several beautiful trails near Atlanta, and my hiking boots will be itching to go find them soon.
8) Learn how to cook. Yeah, I don't do that 😊. Might as well try while I'm getting a fresh start with several other things in my life.
That's all for now, but I can't wait to keep you all posted on what's happening down here in Atlanta. Hope you're enjoying your day!