Tina is a remote sensing scientist with over 10 years of experience working at the crossroads of spatial analysis and machine learning. She is an active member of the FOSS4G community and an OSGeo charter member. At TellusLabs, a Boston startup, she is responsible for turning raw images into agricultural and environmental insights that help answer critical questions facing our society. In her previous position at the Woods Hole Research Center (WHRC), she linked field measurements with remotely sensed optical, LiDAR, and radar products to model ecosystem responses to changes in the environment. Tina earned an M.S. in Natural Resources from the University of New Hampshire and an Honors B.A. in Environmental Science from Saint Anselm College.
Tina was interviewed for GeoHipster by Randal Hale.
Q: Tina Cormier, where are you located and what do you do?
A: I live in Brunswick, Maine. As far as what I do, my answer is “way too many things”! But at work, I am a remote sensing scientist on the data science team at Telluslabs, a Boston startup. Like most startups, it’s a very fast-paced environment. We are a small (but growing) team, and I’m constantly amazed at how much we accomplish in a short period of time.
We use machine learning to combine decades of remote sensing images with in situ reference data. Every single day we incorporate new images and ground data into our system. Why? We want to leverage the information locked inside of this unprecedented historic record of the earth to answer critical questions that we care about — questions about the environment and questions that affect our economy. Right now, we are primarily focused on agriculture and building a living map of the world’s food supply, but our tech stack is structured to allow us to quickly branch into other important sectors as the team grows and as we hire the resources to do so.
My specific role involves converting raw satellite imagery into “insights”, or features that are meaningful for our modeling team. For example, millions of raw satellite reflectance values may not be very meaningful, but when we can turn them into a Crop Health Index, now we’re talking. Even more valuable insights begin to coalesce when we can compare today’s crop health index value to a long term average, or when we can turn on a rapid detection and alerting system for extreme anomalies in the growing regions that we monitor. Then we can start to answer questions about the status of the world’s food supply on any given day or season.
I also work on creating raster visualizations (typically developed in QGIS) for our web app, Kernel. Day to day, I spend a lot of time writing code, primarily in R — though I’m determined to get a better handle on Python and become fluent in PostgreSQL/PostGIS — right now, I’d say I’m conversational at best!
I’m something of a compulsive FOSS4G user and evangelist, which is why I recently became a charter member of OSGeo (that just means I get to vote on things — mostly new board and charter members). In the last couple of years, I’ve worked hard to bring R into the light for geospatial data science folks via social media and presenting at conferences. Last year, I presented three talks and a workshop at FOSS4G, which was tons of fun!
Q: How do you get to be a charter member of OSGeo?
A: To become a charter member of OSGeo, you must be nominated by an existing charter member (thanks, Alex Mandel!) and demonstrate a number of positive attributes with regard to the open source community. My nomination was based largely on my role as a geospatial R evangelist — a role I didn’t necessarily want, but the mix of a large gap in representation with oppressive guilt made me do it! In particular, my almost excessive participation at last year’s FOSS4G conference in Boston was largely responsible for my eventual nomination/election to the group.
Q: How did you get into the Geospatial Field? Was it an end goal of college?
A: So, no. Geospatial was not an end goal or even on my radar. I got my undergraduate degree in Environmental Science from Saint Anselm College. There, I worked on a project where we radio-tagged turtles and tracked them over the course of a year — turns out, they travel surprising distances through all sorts of habitats — I mean, they are turtles, so who knew? We used triangulation to figure out where we were each day and put our turtle sightings on a map — no GPS! I’m laughing to myself about this now. But really, I loved that project so much. That said, I had no idea that I was doing GIS or even what it was.
Fast forward through a few confused and frustrating years post-graduation where I did everything from coaching soccer and teaching high school (something I hope to go back to at some point later in my career) to working for a pharmaceutical company as a clinical data manager, checking to make sure drug trial protocols were followed. One day I woke up and said to myself, “Ok, I need to make a move here, I’m going to grad school.” And so it was. I knew I wanted to stay in the environmental field and sort of assumed I’d go the PhD/professor route.
The following fall, I started a Natural Resources/Remote Sensing Masters degree at the University of New Hampshire. I still had no idea what GIS or remote sensing was, but I got a teaching assistantship that paid my way and it all sounded quite interesting, so I decided to dive in. I had to get up to speed pretty quickly, though, as I was charged with teaching GIS and remote sensing classes that I had never taken before! My Master’s thesis was an exciting (to me) combination of remote sensing, GIS, and machine learning — I built a model that predicts vernal pool locations based on image and GIS-based predictors. My journey can pretty accurately be described as a fortunate series of chances, risks, and leaps of faith that somehow worked in my favor and landed me in a career that I enjoy. And the rest, as they say, is history!
Q: You mentioned R and you did a workshop at FOSS4G in Boston on R which was pretty well received (I tried to sneak in and couldn’t). What is R and why do you like — possibly love — R? I don’t know enough about it but I’m trying.
A: R is an open source software project and programming language. It is held in pretty high regard by academics and data scientists, and is becoming more mainstream among spatial analysts as well. For those who want to automate their work through coding, R is essentially a fully functional command line GIS. The most important reason that I use R (or any programming language) is because it offers repeatability, automation, and documentation of my work — YUP, I just did that…RAD!
I will admit that I didn’t always love it. I had a hard time learning it, and that process involved a lot of foul language. Fortunately, I had some great mentors to pull me through, including an amazing group of former colleagues from the Woods Hole Research Center.
What I like (love?) about R is that I can script my entire workflow — from data cleanup/wrangling (for which R is exceptional), to spatial and statistical analysis, to publication of beautiful figures to tell my story — all in one environment. And once I’ve coded my workstream, I have a complete record of what I did, including which files I used and how I processed them. Working with terabytes (maybe petabytes?) of data — many thousands of images and files — there is no option about programming; it is a necessity to automate my work. R does have some drawbacks though — the biggest of which is that it does everything in memory. Advances in technology have provided a lot of ways to work around the memory limitation though, including better hardware as well as easier ways to chunk up the data and distribute processing. As a spatial data scientist, R is the complete package, with possibly the exception of cartography. While I’ve seen folks do some neat things with maps in R, my go-to for a single, really nice map is still QGIS.
Q: So when you’re at college you meet this dashing young man whom you eventually married — correct?
A: Yes, sir! We met in graduate school. I had been there for about 6 months and kept hearing about this other student, Jesse, who was doing field work in New Zealand. One day around Christmas he showed up in the lab (to my delight). He ended up being my TA for one of the remote sensing classes. Hope there is a statute of limitations for that sort of impropriety, but we are married now, if that helps his case.
Q: Two married people in the GIS field — do you both sit around and talk about spatial things?
A: Sometimes, but over the years we have developed a code of conduct regarding work talk. Jesse and I have worked together for a long time, including 2.5 years at the Southern Nevada Water Authority in Las Vegas and 6 years at the Woods Hole Research Center. Early on, we agreed that we would only talk about work during our commute. At home, work is off-limits. And today, even though we don’t work together anymore, the same pretty much holds true. We’ll talk about each other’s days over dinner, and we may discuss a programming puzzle now and then, but for the most part, we keep work at work. It’s a nice separation that keeps us mostly sane.
Q: OK — You’re working at Tellus. You’ve worked at Woods Hole Research Center. What is the most exciting thing you’ve done up till now in the industry?
A: That’s a tough question, as I’ve had the good fortune of doing a lot of fun things during various jobs, including field work in all sorts of environments, from the tropics to the desert (not always fun, but always interesting). Some of my most wonderful work memories come from teaching technical workshops in various parts of the world. During my time at WHRC I taught workshops here in the US, but I also frequently traveled to South America — Colombia, Peru, Bolivia — thank you, undergraduate minor in Spanish! My farthest trip was to Nepal, which was just amazing. The workshops were designed to build capacity in remote sensing, programming, and forestry within indigenous groups, government agencies, and non-profit organizations in developing countries. I made some lifelong friends who were gracious enough to share their culture with me, teach me to salsa, and even introduce me to their families and friends. I’m forever grateful for the opportunity to share my experience while seeing some incredible places and meeting equally incredible people.
Q: Which is better: A horse or bicycle? Why?
A: I suppose there are pros and cons to each. To my knowledge, bicycles are not spooked by plastic bags, motorcycles, mailboxes, or any other brand of vicious predator you may happen upon in the course of a ride. A bicycle is not typically going to buck you off, though I feel like I may have been bucked off by a bike in the past. It won’t walk away when you try to get on, demand carrots and apples (and frisk your pockets looking for them if you don’t deliver), ask for a scratch in just the right spot (the belly on my guy), knicker when you arrive, look longingly after you when you leave, play with you, choose to be your partner, and it won’t love you back. A bicycle does not have a mind of its own, with memories of positive and negative experiences. It doesn’t have good days and bad days, and it cannot learn, grow, bond, and communicate with you. So, while horses can be dangerous for many of the above reasons, those are also the reasons why horses will always be better than bicycles, in my opinion.
Q: Anything you want to tell the world?
A: I guess having worked in tech/geospatial for “several” years now, I could offer some advice.
- Don’t be afraid to try something new, and don’t be afraid to fail and break things. If you never fail, you aren’t pushing yourself hard enough!
- Learn to code (see #1).
- Impostor syndrome will get you nowhere. Focus on your strengths and what you bring to a situation, and don’t lament not knowing what everyone else seems to know. They don’t.
- Help people. You didn’t get wherever you are on your own — pay it forward when you can.
- Take time off. It never feels like a good time, but you need to do it, so just do it.
- Adopt an animal (unrelated, but still important!).
I’m still working on all of these… except #6… I’ve probably done enough of #6 for the moment.