Blogs

SEES 2021 Ria J - Combining Machine Learning and Remote Sensing

SEES2021 Earth System Explorer

Have you ever wondered how NASA develops accurate models of the earth's surface from space? How do we know the topography of the planet, the behavior of the water cycle, or the changes to sea level?  My experience with the NASA SEES Earth System Explorers virtual internship this summer gave me a glimpse into the process of how such data is collected, analyzed, and ultimately presented: remote sensing. The internship allowed me to empirically learn about the world around me through a combination of coursework, literature review, hands-on science, and collaborative research.

 

In preparation for the internship, I completed an Earth and Space Science course that introduced me to the functionality of many of NASA's satellites. In the SEES Earth Explorers and Mosquito Mappers cohort, I learned about remote sensing to collect land cover data and environmental data in different regions. I read papers on citizen science and remote sensing techniques to characterize the presence of mosquito populations and mosquito vector-borne illnesses in various communities. I was also able to see the relationship between an environment's properties and the mosquito presence first hand through hands-on science. I set up traps, placed them in different areas of my backyard, and eagerly awaited results. I compared the successes and failures of this process with my peers, and it was exciting to share stories about traps getting knocked over or discovering larvae in them. This initial exposure to remote sensing data, combined with my programming background, sparked my interest in researching the applications of computer science techniques to environmental data and mosquito ecology.

 

For the final project, my research team and I took a machine learning approach to epidemiological modeling. We focused on predicting Dengue virus outbreaks in Rio de Janeiro due to the disease's recent prominence in the area, the lack of a cure, and the abundance of available data. We obtained outbreak data using Brazil's Notifiable Diseases Information System (SINAN) and used Giovanni, a web interface to access satellite and surface remote sensing data. Because the Dengue virus is a high-risk disease that significantly impacts Rio de Janeiro, my team and I wanted to develop a method to predict outbreaks to help with preparedness and improve response efforts' efficacy. We tested and compared various supervised machine learning models using factors like humidity, temperature, population density, and precipitation as input and outbreak data as a corresponding output. Our best models, the long short-term memory (LSTM) neural network and random forest model, were able to achieve high accuracies, with coefficients of determination that exceeded 0.99!

 

This project revealed the power of machine learning and data analysis. We were able to use remote sensing data to fuel an algorithm that can significantly improve conditions in Rio de Janeiro. In the end, this internship was an incredible opportunity that allowed me to gain more experience with scientific research, meet like-minded individuals from all over the nation, connect with scientists, and do some hands-on science! More importantly, it allowed me to find the answers to some of the questions I have always asked myself, including how NASA has discovered so much about the world around us.


Ria is a  high school student at Dublin High School in California. Her virtual internship is part of a collaboration between the Institute for Global Environmental Strategies (IGES) and the NASA  Texas Space Grant Consortium (TSGC) to extend the TSGC Summer Enhancement in Earth Science (SEES) internship for US high school (http://www.tsgc.utexas.edu/sees-internship/). She shares her experience this summer in this guest blog post.

 

More Blog Entries