Kandyce, SEES Earth System Explorer, 2024 - Mission Mosquito

SEES 2024 Experience

Getting the opportunity to participate in the NASA SEES Earth System Explorers program has positively impacted my educational and career journey in more ways than I could've imagined. Throughout this internship, I gained an immense amount of knowledge in Earth science, utilizing the GLOBE Observer app, ArcGIS, Collect Earth Online, Python, and many other tools to better understand the environment around us. During the last few weeks of the program, a team of seven other interns and I completed our own research project called "Harnessing Citizen Science to Enhance Land Surface Temperature Prediction." In this article, I will highlight the analysis and research done during my time as an intern, as well as my team's project.

GLOBE Observer App

To start off, our first task was to collect field data around our communities by applying the Adopt-a-Pixel research framework and creating a 3km x 3km Area of Interest (AOI) map. Once I made my AOI grid consisting of 37 points within a 9 km² square radius, it was time to take six directional photos at each spot - North, East, South, West, Up, and Down. In the end, I was able to visit and take pictures at 30 locations due to the other seven being inaccessible, either being on private property or in deep forest areas. An example of one of my GLOBE Observer directional photos is shown below!

I learned that taking these directional photos utilizing the GLOBE Observer app can help track land cover changes that occur overtime in my area and that information can be used to compare with satellite data.

ArcGIS Online

To further analyze our AOI locations, we developed maps with feature layers and dashboards on the ArcGIS Online platform. Each location had an associated smaller 100 m x 100 m box to define the boundaries of where the samples should be taken. The image shown below is my AOI sample grid map consisting of 37 points illustrated on ArcGIS. 

The image shown right is a zoomed in version of one of my AOI points, displaying the aforementioned 100 m x 100 m sampling square. Ideally, the directional photos that I take on the GLOBE Observer app would be at the center point indicated by the orange dot. However, human error is inevitable in science so we declared that anything within the smaller sample square boundary is still considered acceptable.

The image shown left shows my attempt to be as accurate as possible and get as close as possible to the orange center point. Although I wasn't able to get exactly on it, my location was still within the 100 m x 100 m sampling square boundary.

ArcGIS was a helpful tool to perform analysis on our maps, specifically creating buffer feature layers that provided further analyzation.

Collect Earth Online (CEO)

After performing analysis on ArcGIS, we began to label specific land cover elements (building, impervious surface, tree, grass, etc.) for each of our 37 location within our AOI on the Collect Earth Online (CEO) platform. Within each location, 100 points were generated for us to manually label (total of 3700 labels) with the help of Sentinel-2 satellite sourced imagery. The image shown right is an example of what one completed CEO labeled area would look like.

The CEO tool was especially useful in quantifying land cover and observing how each element may contribute to bigger environmental issues. For example, more impervious surface cover and less vegetation could correlate with higher land surface temperatures.

Team Research Project: Harnessing Citizen Science to Enhance Land Surface Prediction

In the second half of our internship, we formed smaller working group teams to began the process of brainstorming and carrying out our research project. My team's project focused on creating machine learning models utilizing citizen science sourced data to predict land surface temperature in the context of the rising issue of surface urban heat islands around metropolitan areas. Surface urban heat islands are classified as developed areas indicated by higher land surface temperatures compared to surrounding areas. While exploring different online resources about our topic, our group started to notice the pattern of how more impervious surfaces (such as concrete, asphalt, and more) in an area resulted in higher land surface temperatures because of less vegetation being present. Later on, our team used this observation as one way to explain why some areas may potentially experience higher land surface temperatures than others. The image shown below displays the temperature distribution along various community types.

Source: World Meteorological Organization

After our preliminary research was completed, our team gained additional knowledge of the chosen topic and why it is important. The next step now was to extract, retrieve, and prepare our data from various sources to use as training data for our machine learning models. Training datasets used for models included downward GLOBE Observer land cover photos collected by the Earth System Explorers program (further labeled for land cover through Zooniverse with the help of citizen scientists), mean land surface temperature from the Landsat-8 satellite, and Collect Earth Online. Our team then developed Random Forest and XGBoost models trained with those numerous datasets and compared their results. I helped develop visualizations through Python for our Random Forest models to illustrate the process it goes through to develop a prediction. Creating these gave me a better understanding of how the model worked and revealed how one decision made in a tree based on feature importance could affect the whole prediction value in the end. Once we successfully achieved results from our model, our group started to develop our deliverables which included a collaborative research paper, poster, presentation, and archived codes. The products created for our project were then presented at the NASA SEES Symposium and submitted to the International Virtual Science Symposium (IVSS) and American Geophysical Union (AGU).

Reflection

In summary, going through this internship has taught me more about the environment around me and how remote sensing plays a huge part in unveiling climate issues that occur. I was exposed to a wide range of tools such as NASA Worldview, Landsat Time Series, ArcGIS, Collect Earth Online, and so much more that helped contribute to my team's overall research project. I particularly enjoyed getting the opportunity to combine both my interests of computer science and the environment in this program to then view how they can benefit each other. This experience has provided me with an immense amount of knowledge that I will cherish and apply throughout my educational and professional career, not only the data tools but also the mentors that helped me along the way. 

About the author, Kandyce 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.