EOA Member Spotlight – Caitlin Adams Senior Data Scientist at FrontierSI

This article was originally featured in the Earth Observation Australia newsletter on Thursday 31 August, 2023. 

My career in Earth observation began after a fortunate series of events in early 2019. I had just completed my PhD in Astrophysics and was excited to continue working at a start-up where I had interned as a data scientist the previous year. There was just one problem: the start-up founders had decided to move on to other opportunities, meaning I was out of a job. After a frantic search, I stumbled upon what seemed like the perfect role: a data science position focused on showing people how to use free and open satellite data at a not-for-profit called FrontierSI. I had experience working with spectral data (albeit looking at galaxies rather than our planet) and a deep passion for teaching and enabling others. Plus, as a newbie in Earth observation, I figured I would be able to build examples that catered to those new to the field, without the bias of assumed knowledge. I was hired at FrontierSI in March 2019 and have been with the company ever since.

My four-and-a-half years at FrontierSI have been highly rewarding. I have had the opportunity to work on incredible state, national, and international projects, collaborating with all sorts of brilliant people. For this member spotlight, I want to reflect on the three pieces of work that have shaped my career to date.

My first major piece of work at FrontierSI was collaborating with the Digital Earth Australia team in Geoscience Australia (DEA) to develop resources that demonstrated how free and open satellite data could be used for different applications, such as agriculture and forestry. I was able to learn directly from the incredible remote sensing scientists on the team and, in turn, I developed accessible and intuitive demonstrations that could be accessed by people across Australia, helping to get the message out about DEA.

In particular, I built a highly productive collaboration with Robbi Bishop-Taylor, Claire Krauss, Bex Dunn, Claire Phillips, Chad Burton, and numerous others, where we focused on developing new examples and analyses. Over time, we built a shared culture that enabled us to work effectively together, building useful tools, reviewing each other’s work, and striving to make our work accessible to a wider audience. This culture has evolved over the years, with new additions to the team bringing fresh new ideas. It is a great joy to me to see this team continually working to improve the usability of Earth observation data across Australia.

In early 2020, I began working on Digital Earth Africa, a program started by Geoscience Australia but now run by the Program Management Office in Pretoria, South Africa. Early in the project, I teamed up with Ee-Faye Chong to develop a short online course on how to use Digital Earth Africa, which grew into an online learning platform with multiple courses and an Africa-wide community sharing their knowledge about Earth observation analyses.

Around the time I was working with Digital Earth Africa, I embarked on an exciting project with Victoria’s Department for Environment, Land, Water, and Planning (now the Department of Transport and Planning), led by Cat Gilbert and John White. The Department produces Vicmap—the state’s authoritative catalogue of spatial data—and had previously completed a proof-of-concept project with FrontierSI demonstrating that machine learning could be used to detect trees in aerial photography. My task was to take the proof-of-concept and produce a state-wide dataset, along with the process that could be used to update the data in future years. This was my first major foray into machine learning, but over the course of 18 months, I learned a lot and worked hard to build state-wide vegetation extent and density products. I had a lot of great support from Lachlan Hurst (Full-Stack Developer at FrontierSI), as well as getting to work with Orbica (experts in geospatial machine learning) to create and refine our training data.

In the last two years, I have worked on many great and diverse projects with the team at FrontierSI and our collaborators, including looking at crop type prediction with Digital Earth Africa, helping make derivative satellite products more useful for aerial firefighting with Digital Earth Australia and Natural Hazards Research Australia, and understanding requirements for an open standard for machine learning training data as part of an Open Geospatial Consortium Testbed. During this time, our Earth observation team has grown significantly, and I feel very privileged to work alongside Fang Yuan, Lavender Liu, Claire Fisk, and Madeleine Seehaber on a wide variety of interesting Earth observation projects.

After a whirlwind start to my career, I felt incredibly honoured to be recognized with the inaugural award for Earth Observation Australia for early career contributions to the community and its capabilities. This was made more special by the fact that Robbi Bishop-Taylor received the highly commended award, and that we were able to receive these awards side by side after many years of productive collaboration and friendship. Robbi is an incredible researcher and science communicator, and I hope we’ll continue to have opportunities to work together over our careers.

Looking forward, I am taking on more responsibilities as a technical leader in FrontierSI’s Earth observation projects, with my speciality being machine learning. I am excited to continue building derivative products from satellite imagery that help people better understand and manage the complex world and ecosystems in which we live, work, and play.