Cameron believes the democratization of data can improve people’s lives and help protect the planet. At SkyTruth, he’s advancing projects to do just that.
My name is Cameron Reaves. I’m a student at Harvard College and an intern at SkyTruth this spring semester pursuing two of my passions: tech and the environment. My interest in these fields started in high school, when I spent weekends coding web applications for computer competitions and I discovered my passion for environmental issues after taking an environmental science class my sophomore year. I was stunned that human activity causes so many complex and intersecting environmental issues. What inspired me was that there is always hope. We can find solutions to the thorniest problems impacting living and non-living systems: we just need people to address them. I wanted to be one of those people! When I arrived at Harvard, I decided to concentrate in Environmental Science and Public Policy. Over the last four years, I’ve taken coursework in the physical sciences, data science, policy and economics. As I began looking for post-graduate positions, I started seeking opportunities to merge technology and the environment. When I discovered the SkyTruth (remote) internship on a “tech for good” job board, I realized that SkyTruth was a perfect fit.
After reading a few SkyTruth blog posts, I reached out to intern coordinator Brendan Jarrell over email to ask about spring internship opportunities. When we finally met over video chat, I felt surprisingly comfortable. We talked about our past experiences and bonded over Ultimate Frisbee, a pastime we both enjoy. I would discover later on that everybody at SkyTruth had the same welcoming and laidback energy. My first staff meetings told me everything I need to know about working here: I felt set at ease by my colleagues’ jokes about their haircuts, Jar Jar Binks Zoom backgrounds, and pet updates.
The projects that I am working on this spring include monitoring the Southeast Asian region for oceanic bilge dumping (the illegal dumping of oily waste at sea), validating geographic points for Project Inambari (which detects unregulated gold mining in the Amazon rainforest), building up datasets for a road detection algorithm, and designing a dashboard that visualizes changes in vegetation in the Colorado River watershed to identify trends in habitat health. Although I had previous experience with the programming language Python and ArcGis, an online mapping and analysis tool, I was unfamiliar with Google Earth Engine and QGIS, the tools used in my projects. However, the onboarding process with Brendan illustrated how all of these ideas, tools, and processes fit together. The SkyTruth internship program does a great job of bringing new interns up to speed and feeling ready to contribute to ongoing work.
On my first day monitoring satellite images of the ocean for our bilge dumping project, Cerulean, I found an oil slick off the coast of Malaysia. This was really exciting, more so because it initially seemed that this particular spill went unseen by Cerulean’s machine learning algorithm. By finding missed spills by hand, we are able to improve the algorithm so it can continue to get better and better. It turns out, however, that there was a small bug in the tool that interns use that mislabeled the satellite image I found. In fact, Cerulean had actually detected this image! After talking with Brendan about the issue, he immediately knew exactly what to fix. The next day, the Ocean Monitoring tool was updated, and we are now using the corrected edition. I was excited to be able to help improve our processes, even on day one.
For another project, I’ve created novel data by digitizing roads in forested areas using satellite imaging. This will add variance and diversity to the inputs in our model that detects roads linked to illegal mining sites, helping it learn to identify the different ways roads appear in imagery. Roads often provide the first opening for harmful environmental activities to follow. I selected a forested location on the border of West Virginia and Kentucky that had limited human infrastructure. I classified roads into wide (greater than three pixels wide) and narrow (less than three pixels wide). This will help the algorithm recognize that roads can appear as different sizes. Hopefully, we can repeat this process and see significant improvements in our model’s performance.
So far, I’ve been able to meet (remotely) with some of my colleagues and get to know them better. I’ve had several long chats with Jona Raphael, a Machine Learning Engineer in charge of Cerulean. I was familiar with neural network concepts because of previous coursework, but after discussing the design of Project Cerulean and fast.ai (a Python based deep learning library of existing computer code), I am on much better footing. I plan on learning more about pixel segmentation, data augmentation, and deep learning over the course of my internship, but I’ve got a good start thanks to Jona. (Thanks Jona!)
I also met with Ry Covington, SkyTruth’s Technical Program Director. We connected about the obscure but crucial field of Science and Technology Studies (STS). His graduate dissertation considered the spatial origins and distribution of technical knowledge creation. In essence, how does technical knowledge spread and how does that affect important environmental decision-making? Fascinating stuff! I am familiar with this field of research because of a required course taught by Professor Sheila Jasonoff at Harvard, one of the most brilliant people I’ve ever met. I look forward to more thought-provoking conversations with Ry, and helping with his current research on mapping vegetation trends in the Colorado River Basin!
SkyTruth’s work is meaningful to me for several reasons. For one, I am able to learn awesome techniques to decipher millions of pixels captured by satellites orbiting Earth at thousands of miles per hour. That alone is rewarding enough. But through these data science techniques, I am also contributing to conserving the natural environment and preventing climate change. The mission of civic technology is incredibly compelling to me. I believe that providing services, analysis and data to the public results in impactful improvements to people’s lives. The democratization of data, when it is accessible and utilized for anyone interested, is a worthwhile goal. These ideas are what get me excited and motivated to work in this space. I’m so grateful to SkyTruth for providing the opportunities to engage in this work to young folks like myself. I look forward to learning a whole lot and helping out wherever I can.