The goal of this project is to use Ps receiver functions, which are a passive seismological technique, to date detect upper mantle discontinuities in the African subsurface. Ps receiver functions are often limited in resolution due to an overprinting of the seismic signal due to the multiple reverberations that follow the direct P wave and converted Ps wave. To avoid, the overprinting of these reverberations over the converted wave signal, this project uses a new methodology for analyzing receiver functions called CRISP-RF to improve the precision of the receiver function technique. CRISP-RF stands for Clean Receiver function Imaged using SParse Radon Filter.
The most relevant paper to what I am doing now is called "Moho depth variation in southern California from teleseismic receiver functions" (Zhu and Kanamori, 2000). This study was the first use of a method called H-k stacking of receiver functions, which is what I am doing but with a different study area. The main point of this study was to show that you can constrain the depth to the Moho and the Vp/Vs ratios through a stacking algorithm if you have a receiver function with your Ps wave and the multiples.
This is an example of the final step in a process called H-k stacking. The depth to the Moho, and the body wave velocities are the main two factors that determine when a sesimometer records the P-s singal. What H-k stacking does is that it compares what we observe (receiver function) to a forward model. In this forward model, h-k stacking uses the two parameters (H and Vp/vs) to generate a "fake" receiver function. After that, we match the fake singal to the real singal by multiplying the amplitudes. The best forward model has the biggest product between the fake signal amplitude and the real signal amplitude. By tweaking the paramters around, you can compute the best pair of numbers that fits your station. In this case the best pair is H = 36.9 km and Vp/Vs = 1.78. This is with an assumed Vp = 6 km/s, based on literature.
Now that we are getting close to the end of the internship, my main goal is to get reasonable H-k stacks for all of the available stations.
I created this map using m_map. The black dots are all of the unusable stations and the blue triangles are the 52 stations that were used. The shading of the continent represents elevation. I am using this study area because I want to have values of the depth to the Moho of as much of Africa as possible.
One problem I'm currently facing is that one of my figures is missing data and I'm not sure why. One of my successes is efficiently generating the receiver functions for all 89 stations and describing the quality of each of them.
Something that I have so far learned in my time here is that researchers spend a lot of time thinking of ways to make a figure better or more intuitive. They have already figured out what they want to show, and the technicalities of how to do it. But they spend more time than I do thinking about ways to make their figures "sell" better, to the editors and to the scientific community. So I think to present information graphically at an advanced level, you probably need to do a lot more double-checking and critique than usual. This is an important skill to develop because it actively causes you to improve your data and presentation both at the same time. And something I'm doing to do more of this is just to start double-checking my figures because a lot of the time I just create them and move on as long as they make sense.
For receiver functions, in this project, the location of the earthquakes themselves isn't the study area, but rather the station that the signal is recorded by. In this project, all of the stations are on the African continent. Most stations are in South Africa and East Africa. There were almost 1,100 stations of interest, but then after preprocessing the data to determine if it is usable, only remain. This brings up the main strength and weakness of this project in general. Africa has a geographically sparse distribution of seismic stations, and ontop of that, the filtering that takes place in the preprocessing step renders even more stations unusable. However the workflow for this project (CRISPR) is good at dealing with this sparsity.
I'm not sure if the data set that I'm using is a widely used data set. Because my data set is basically all the seismic stations on Africa in the iris database, my guess is that it is widely used. I also started with downloading the 1000 stations from iris but a previous graduate student had already had the list of stations (they used it for their paper).
For URISE summer 2023, here in Rochester, NY, I want to get better at articulating complex ideas succintly. I also want to build my computer skills and interpretive skills. I also want to do as much as I possibly could during my time here. So far, talking to friends and family members has helped me practice on explaining what I'm doing here. I've also been participating in the skills building workshop when I have free time. Recently though, there's more to do here and it's getting busier.