Research Internships

In the summer entering junior year, I was fortunate enough to combine my passions of environmental science and computer science in the Stanford Earth’s internship. I was able to conduct a formal research report with two other interns on creating complex analyses based on paleontological data:

​

Extinction within the Paleozoic era has been studied in the past, but there still lacks a comprehensive understanding of how extinction risk changed throughout it.

​

Our research project aims to bridge this gap by exploring extinction risk in relation to major Paleozoic phyla and ecological characteristics. Using R, we analyzed the Stanford Earth Body Size dataset, which includes extensive data (n=8816) on Paleozoic marine animals. In Step 1, regression coefficients were formed, indicating whether being in one of the 6 phyla in each period of the Paleozoic era conferred greater or less extinction risk. In Step 2, the examined ecological characteristics included ocean acidification resilience, feeding patterns, body volume, length, surface area, motility, tiering, circulatory systems, and respiratory organ type. In Step 3, 6 binomial machine learning models were created using the traits from Step 2 to determine whether an individual genus went extinct in a particular period. Our Step 1 results confirm that within these timeframes, while certain phyla have greater extinction risk, extinction risk was not uniform across these groups. Our Step 2 results show certain traits provided advantages and disadvantages for an organism's extinction risk.

One interesting pattern was that the only consistently non-significant traits were body length, area, and volume. Likewise with Step 1, extinction risk for each ecological characteristic varied across the Paleozoic. Finally, in Step 3, the results were largely successful. Most of the six models had an accuracy above 80% with the highest being 92% in the Silurian. The areas under the Precision-Recall and the Receiver Operating Characteristic Curves were all in the acceptable (>0.6) range, demonstrating that the model has low false positive/ negative rates and is able to distinguish between what trait indicates extinction or survival for each period.

Our research project identified phyla at risk of extinction in each period of the Paleozoic, determined which natural traits incited greater extinction risk, and demonstrated machine learning models trained on fossil descriptors can predict when an individual genus became extinct. Our results confirmed that extinction risk is not consistently dependent on a singular factor nor is it constant across every period of the Paleozoic era

​

Ultimately, I have grasped so much in this brief period like how body size research can shed light on underlying processes that shape biodiversity patterns. It was a privilege to have the mentors of Dr. Monarrez and Mr. Pimentel-Galva, who instructed us on programming in R and conducting a research project. They were always willing to support us whenever we needed help for the abstract, data collection, and anything else. I am also grateful to Dr. Saltzman and Dr. Payne, who organized this entire program with guest speakers and all the opportunities provided. This amazing opportunity allowed us to present our findings to the Stanford Faculty and at the Winter 2021 American Geophysical Union conference.