If you had asked me what “citizen science” was before this internship with the NASA SEES (STEM Enhancement in Earth Science) Earth System Explorers, I most likely would have conjured a description that involved studying the human body. Now, however, I feel confident in saying that citizen science is a dynamic, extraordinary meaningful process by which everyday citizens contribute to real-world scientific research. This internship allowed me to take on the roles of both a citizen scientist and a researcher, deeply engaging with my local community and environment while simultaneously considering issues related to climate change, social justice, and sustainable development on a large scale. I’ll take you through a few highlights of my incredible July and July, broken into chronological phases, although rest assured that some phases overlapped or are still ongoing!
Phase 1: Data collection through the GLOBE Observer App
After getting to know my fellow interns (shoutout to Working Group 2!), our amazing mentors, and our equally amazing peer mentors, Peder sent us out on a mission. I followed his enthusiastic instructions as I downloaded the GLOBE Observer app. Then, armed with a map of 37 distinct points in a 3 km x 3 km grid around my neighborhood, I set out walking and taking pictures. Yes, really! I went to each spot and stealthily snapped pictures in all four directions, plus up and down, before filling out the tedious land cover descriptions on GLOBE Observer. Of course, many locations fell on private property, so I had to strategically analyze Google Maps to get as close as possible without breaking any laws or seeming too suspicious.
I divided my research into several expeditions, each lasting an hour or two, as I had to do quite a bit of walking. Although data collection led to a few tired and busy afternoons, it allowed me to explore parts of my surrounding area I had never seen before or notice parts of the environment I had previously walked or driven past. For example, I never realized just how many evergreen trees there are surrounding every single house in my neighborhood. Although I still had no idea how to use this data, it was gratifying to know I was contributing to real-world land cover data as my photo count ticked up on the ArcGis website.
A few beautiful places I got to explore while hunting down points on my ArcGis map
(Oh yeah, and after realizing I chose the wrong setting on the app while taking the majority of my photos, I had to go back and redo them to get more accurate location data. That was fun.)
Everyone's amazingly ambitious ideas on Padlet
Role reversal: now I was the researcher! After a few weeks of data collection, plus learning about remote sensing and hearing from real scientists like Dr. Bri Lind and Dr. Di Yang, we were ready to explore our data and develop a research topic idea. Numerous past SEES interns also spoke to us about their projects, including Nikita Agrawal, who later furthered the skills she developed in SEES to develop a machine-learning model for wildfire prediction. My Working Group was all passionate about machine learning as well. After a tsunami of ideas dumped onto our Padlet, we settled on developing a machine learning model that would incorporate land cover data to predict surface temperatures in areas around the world.
Of course, I was stressed out! It was July 1st, and I had to complete an entire research project by July 19th along with a poster, video, and presentation. I had no prior experience with machine learning, and neither did the majority of my group mates. Nonetheless, we felt reassured that each of us had individual strengths we could bring to the team to make our project a reality. Having capable, experienced mentors was also a huge plus.
Phase 3: Land Surface Temperature Research
We had selected our idea of creating a machine-learning model to predict land surface temperatures, but how would we do it? When? Why? We didn’t know, but we quickly delved into research about urban heat islands–a phenomenon where urban areas become hotter than surrounding areas due to an abundance of impervious surfaces (concrete, etc) and a lack of cooling vegetation. Learning about the social implications of heat islands was intriguing. They were a topic I had never previously studied, but the more I investigated, the more it became clear that urban heat islands not only exacerbate public health risks but also contribute significantly to energy consumption and greenhouse gas emissions.
Researching heat islands for our project provided further motivation to create a predictive model for land surface temperature, as modeling temperatures is crucial for protecting the health of people and our environment.
Urban heat islands in a nutshell. Credit: NASA/JPL-Caltech
Examples of directional photos taken via GLOBE Observer. (Our group focused on the "down" photos, although we later realized adding in other directions would most likely have improved our model's accuracy.)
Of course, experts have been modeling surface temperatures for years, so I was initially unsure how our group of eight high school interns would contribute anything to the field. However, with Peder’s guidance, I realized that the dataset of GLOBE photos collected by SEES interns had yet to be explored in any scientific context. We set out to study how (and if!) citizen-collected data could increase the accuracies of predictive models for surface temperature. Specifically, we wanted to see if land cover data from GLOBE Observer “down” photos correlated to surface temperatures within a region. (Concrete surfaces would most likely be hotter, for instance, while areas with grass or water would be cooler.)
A flowchart showing our simplified process of data collection
Our research involved numerous tedious stages of data preparation: writing Python scripts to query the GLOBE API, learning to use Zooniverse so that citizen scientists could help us label photos based on land cover, and extracting labeled satellite data from Collect Earth Online to go along with our images. Then, of course, we had to learn about machine learning. We were fortunate to discover NASA ARSET training for machine learning models, including Random Forest and XGBoost, in the context of analyzing satellite data. I took on the role of training the Random Forest model, and there were many painful hours of dealing with error messages in Google Colab and waiting many minutes for small sections of code to run. However, nothing beat the gratification when I finally trained a model and saw it making realistic predictions for the first time. It took weeks of fine-tuning our models and datasets before we got to a place where we could begin working on our data, but I loved the process. I was delving into my passion for coding while seeing how our data could make a true difference in a real-world environment context.
The moment of glory: a few actual vs predicted values for land surface temperature generated by the Random Forest model
Smooth collaboration was also key to the success of our project. There were many hours spent over Zoom and Google Meets with each other brainstorming ideas, asking questions, going to office hours, and sharing what we had learned from various tutorials or conversations. We quickly learned to manage our time well–difficult in a project divided between eight different people living in four different time zones! To help with this, we separated into two groups and had each one focus on a different ML model. We also communicated as much as possible when we were struggling and tried to set clear goals before each weekly meeting. Although none of us had ever met before the internship, we quickly learned each other’s strengths and weaknesses, and I could not have completed any part of our project without the support of my amazing teammates and mentors.
Our model wasn’t perfect, but it provided us with confidence that more high-quality land cover data could be used to enhance future models for surface temperature prediction. We did our best to communicate our findings and their significance through our research paper. When we finally watched our recorded presentation at the SEES Symposium on July 23, I felt a ton of pride in what we had accomplished. I was also amazed by the in-depth, meaningful research projects that other working groups accomplished, which spanned from agrovoltaics to the socioeconomic correlations of heat islands. I was equally impressed with projects from the entire NASA SEES intern groups, which delved into many intriguing fields relating to engineering, space science, and sustainability.
If you're curious, you can check out our presentation at minute 43:15 of the SEES Symposium.
Phase 4: Reflection, Community Climate Chronicles, & More!
We were almost finished. One poster. One abstract. One presentation. A forty-page written research paper. However, even after the symposium, we had a lot to learn and finalize. One aspect of science we emphasized, with the help of Andrew Clark, was the importance of making our research accessible and reproducible through the principles of “open science.” To do so, we learned to use GitHub to store all our code products and datasets, as well as Zenodo to link our research paper to our unique ORCiD identifiers and data. Additionally, with Rusty and Cassie’s help, we worked on a plain language abstract for our research report. This is essentially a version of the abstract that eliminates scientific jargon to be easily understandable by everyone, not just specialized scientists.
Besides polishing our writing for submission to IVSS (the International Virtual Science Symposium) and AGU (the American Geophysical Union), we also finished up our Climate Chronicles blogs of our local areas! We used these blogs to document our communities’ histories while incorporating the scientific data we collected to tell a meaningful story about change over time. Delving into my local neighborhood, a small city famous for its wineries and nature outside a much larger urban metropolis, I learned so much about urbanization occurring in the region over time. This was even reflected in the data I collected for GLOBE, as well as satellite data for the AOI (Area of Interest) I studied over the last 40 years. It was empowering and surprising to see how my research reflected meaningful change in the place I have lived my whole life, and I feel empowered now that I have created a platform to share this story with others. You can check out my Climate Chronicles here:
Community Climate Chronicles: Woodinville, WA
Overall, I found this internship incredibly valuable, from the individual learning modules I did in May to the final paper I helped craft in July. I met so many amazing people and learned so much about machine learning, land cover research, open science, citizen science–and myself! Moving forward, I will take the principles of citizen science to heart as I attempt to be an active, engaged member of all my communities. I had always planned to pursue a career in a STEM field, mostly likely computer science; however, I now feel empowered to apply computer science skills to create real-world change in a field like environmental research or biomedical engineering. I feel so fortunate to have been a part of the 2024 Earth System Explorers, and I strongly encourage any eligible readers to pursue an internship with the NASA SEES program!
You can also check out this same blog via ArcGIS StoryMaps.
About the author, Anna is a rising senior from Seattle, Washington. This 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 U.S. high school (http://www.tsgc.utexas.edu/sees-internship/). This guest blog shares the NASA SEES Earth System Explorers virtual internship in 2024
