Projects
Contents
Projects¶
This section keeps track of some of my unpublished projects that may or may not be related to my primary research. These projects showcase various applications of coding/data science to earth science topics.
2023¶
The past Martian climate may support present-day liquid water under the SPLD¶
Project done with peers at the Karthaus School for Ice and Climate
Group Project found here
This group project looked at forcing a time-dependent heat equation within ice with past Martian climates to see if the time dependence could result in remnant liquid water existing at the Martian south pole.
I primarily worked on running a suite of sensitivity tests to see instances where liquid water could potentially exist in present day
2022¶
Using Machine Learning Techniques to Emulate Ice Sheet Velocities from a Parallel, High Resolution Ice Sheet Model¶
Project for a Machine Learning in Earth and Environmental Engineering course
Project found here
I develop and compare traditional NN, CNN, LSTM, CNN+LSTM models and their efficacy in emulating outputs from the Parallel Ice Sheet Model.
Hurricane tracks and their economic impacts in a changing climate:¶
Project for a Climate Prediction Challenges course
Group project found here
This project used hierarchical clustering to identify two different “classes” of hurricanes that pose a financially damaging risk. It then investigates the evolution of these hurricanes with respect to a changing climate.
I specifically conducted all the data cleaning/processing and initial k-means clustering for analyzing “financially damaging” hurricanes and how they may evolve with climate change. I also created most of the data visualizations.
A physics guided machine learning model for lake temperature prediction:¶
Project for a Climate Prediction Challenges course
Project found here
This project had us fine-tuning an existing process-guided LSTM model that predicts lake temperature profiles
Our team attuned this model to use a dynamic learning rate with an SGD optimizer
I specifically further attuned this model by recognizing that basal depth temperatures of the lake do not vary much, so I trained it on only the top 60% of depths, which still produced similarly accurate predictions for the entire depth profile.
Using Shapely values to determine key features in ML-based pCO2 reconstructions:¶
Project for a Climate Prediction Challenges course
Project found here
As the title suggests, we conducted a shapely analysis to identify what features a ML-based pCO2 recontruction relied on the most.
Through this analysis, we identified a possible correlation between a misrepresentation of biological activity in the Southern Ocean and a failure to accurately capture seasonlality in the same area.
Using Machine Learning Techniques to Emulate Ice Sheet Velocities from a Parallel, High Resolution Ice Sheet Model¶
Project for a Machine Learning for Earth and Environmental Engineering course
Project found here
Tested different types of neural networks for their ability to emulate a high resolution ice sheet model
2021¶
A correlation analysis of meltwater runoff and glacier velocities on the Greenland Ice Sheet¶
My final project for a Research Computing in Earth Science course
Project found here
This project involved displaying spatial averages of runoff and velocity over time to compare trends across the entire GIS, while also having 2d heatmaps of the cross correlation to see if only specific regions are strongly influenced by runoff