Eyes on Earth Episode 113 – EROS Science Leader’s Vision
Detailed Description
Terry Sohl helped develop the National Land Cover Database, NLCD, when he first arrived, and now he’s overseeing significant improvements to the widely used product. But that’s just one part of his new role. In this episode of Eyes on Earth, Sohl provides an overview of the science efforts at EROS and how artificial intelligence and machine learning help scientists focus more on the work that helps society. He also shares his vision for the future.
Details
Sources/Usage
Public Domain.
Transcript
JANE LAWSON:
Hello, everyone, and welcome to another episode of Eyes on Earth, a podcast produced at the USGS EROS Center. Our podcast focuses on our ever-changing planet and on the people at EROS and around the globe who use remote sensing to monitor the health of Earth. My name is Jane Lawson, and I'll be hosting today's episode where we're talking about the future of EROS science under the newly announced leadership of Terry Sohl. Terry came to EROS three decades ago from a background in Landsat use. His first task at EROS was significant and enduring, helping develop the National Land Cover Database, or NLCD, based on Landsat satellite imagery. The NLCD stands today as the definitive land cover database for the United States. Terry went on to help with the Land Cover Trends project and then developed the FOREcasting SCEnarios of Land Use (Change), or FORE-SCE, framework to model historical and future land use.
Currently, Terry is overseeing an effort to incorporate improvements into the NLCD with a new release planned for later this year. Terry joins us today to talk about his new role as chief of the EROS Integrated Science and Applications Branch, and the scientific aspects he considers important to EROS and the remote sensing world going forward. Welcome, Terry, to Eyes on Earth.
TERRY SOHL:
Well, thank you, Jane. Glad to be here.
LAWSON:
First, let's talk a little about the science branch at EROS. Give us an idea of the breadth of science going on here and the people dedicated to this work.
SOHL:
OK. I think I actually want to start with the people because they're the ones that make it all work. I started here in 1993, and I've had the privilege of working with some of the biggest names that have ever gone through the field of remote sensing, and that really is what has made EROS special over the years is the continuing pipeline of talent and big names in the remote sensing field and the partnerships that we develop with similarly big names. So from a perspective of what makes science work at EROS over the years, it really has been the cast of characters that we've had. In terms of the breadth of work, we do have six major focus areas for the Science and Applications Branch. Each one of them provide just a little bit of a different flavor to what's happening on the landscape. Overall, we're trying to look at the landscape, look at how it changes, look at the causes and the drivers of that change and then try to identify what the consequences are for society. But those six general focus areas include land change monitoring, which is generally using something like the long-term Landsat archive to look at things and how they've changed over time. Terrain mapping and monitoring, so looking at topography and elevation change, including coastal change. We do wildland fire science. A lot of this work is with projects like the LANDFIRE project that's looking at fire fuels and fire danger and fire modeling across the United States to try to inform hazard and risk and vulnerability. We have human health and food security, including some projects that are looking at famine on a global scale. We have vegetation, water and climate dynamics. This really is a broad set of projects that include things like looking at the agricultural change and evapotranspiration and crop condition. And then finally, we have Landsat science, which is looking at the basics of the Landsat mission and how we can better support baseline products for the future.
LAWSON:
Can you give us some examples of how EROS science is helping to better the lives of people in the U.S. and around the world?
SOHL:
Well, I'll start here in South Dakota. From a perspective of locally, how the science that we develop at EROS is supporting a number of important societal issues, in South Dakota is relatively unknown by even the local population. But think of agriculture, and if you've ever been to EROS, we're out in the middle of the country, we're surrounded by corn and soybeans and, you know, just from the perspective of supporting the mapping of what's on the landscape and where agriculture is, where natural habitat is, looking at things like cropland condition. We have a product such as evapotranspiration that is a proxy for looking at water use over time so we can evaluate how crops are performing in drought conditions or other conditions. Speaking of drought, we also have products that look at drought at a large scale across the United States that feed into things like the National Drought Monitor that are done by the National Drought Mitigation Center down in Lincoln. Or we have things like fire, and so in the Black Hills, the uniquest ecosystem compared to the rest of South Dakota, but we have a very active fire program that's mapping and monitoring fires, the severity, post-fire regeneration and what the impacts are on the landscape. And even things like hunting. We know that hunting is a big driver of economic development in the state, particularly in the fall during pheasant season. But we've done work looking at things like climate change and land use change and how that's impacting bird populations, for example. And so for a species like sharp tailed grouse, we can look at how that particular species is likely to react in the future. And that relates to future availability for hunting. Just in the state of South Dakota, there's many, many issues. From a global perspective, too, the example I love to give is the FEWS project, Famine Early Warning System. This is a project using remote sensing data to look at drought conditions, agriculture, and it feeds directly into some of the work that's done by the UN and other entities to try to provide famine relief for populations that are at risk.
LAWSON:
Sounds like an exciting set of science work to start overseeing. What does your leadership role entail, and how does it feel to take the lead in an area you've worked in for 30 years?
SOHL:
It's kind of funny how a career sneaks up on you, but I do think it's a natural progression because you start as a young scientist, and you're supporting other scientists at the start of your career. You eventually move to principal investigator, and you're leading projects. And then a lot of times you come to a decision point, you know, are you going to continue with research, or are you going to go into management? And it's really been in the last five years that I've moved more towards the management side. I think my experiences set me up pretty well with - you know, I've been a contractor at EROS for a number of years. I've been USGS, supported a team, I've led a team, and especially in the last five years, I've been involved with a lot of higher-level initiatives within USGS and the like. But even so, it's definitely a change from a remote sensing scientist research type of role. But one of the quotes that I just absolutely adore that I've tried to guide my career on is Tom Loveland, one of the biggest names in remote sensing, who was an icon here for years until he tragically passed away a couple of years ago. But one of the quotes that he gave was that if you're too comfortable in your job, you're not doing it right. For me, five years ago, I was too comfortable in my job, and moving into the management side of things has been a challenge, but I really love it, and I particularly love the part of trying to set a vision, so moving beyond the view of one individual project or one specific application and instead trying to look at the vision across the whole organization, how all the pieces can fit together, how to manage the resources to best serve society. And that challenge is really a lot of fun.
LAWSON:
So let's focus on your vision for the future. What goals do you have for the EROS science branch, and how do you aim to accomplish them?
SOHL:
Well, you know, I started, again, in 1993, and up until the early 2000s, and even to this day, for many people, we've always been known as the EROS Data Center. I think it was 2004 that we officially had the change to the Earth Resources Observation and Science Center, you know, focusing on the science side. But to many people we're still a data center, and that's the one impression that I am always going to fight because, you know, even in the science branch, we're known for producing data. We produce big national-scale data sets like the National Land Cover Database or LANDFIRE or other datasets that we put out on the web, and people access and use. And that's not going to change. We're always going to be a creator of these large national-scale data sets, the gold standard of these data, if you will. But my goal is to move more towards integrated and interdisciplinary science, and that means getting closer to the actual stakeholders and being more than a set of individual projects. And right now, just with the structure of EROS over the years, we have funding from a lot of different sources, we have a number of different projects, and my primary goal is to try to tie that more tightly together where we can break down some of those stovepipes between the projects, where we can look at larger-scale questions of what's happening on the landscape, how it's impacting society, and answer them in a way that no individual project can, but collectively as a group, we can. I think it's that bigger picture of seeing how all the pieces fit together to answering those questions that really sum up the goal that I have for the branch overall.
LAWSON:
Let's dive a little bit into artificial intelligence, machine learning or AI/ML. How exactly can that benefit the EROS science projects and the people who rely on all the information that those projects provide to the world as it changes?
SOHL:
Machine learning and AI is something we've actually been using for quite some time. And if you think of a regression tree, it's a form of machine learning that we've been using for almost 20 years. And so AI, in terms of some of the more advanced techniques like convolutional neural nets, are tools that have also been out there for a while. It's just now we finally have the technical capability, the hardware, to support it. So I'm going to go back to when we started national-scale land change monitoring. So that first National Land Cover Database, we started it in 1993. I would come in, and I would take a big tape, and you would load up that big tape, and you'd wait half an hour, 45 minutes for that one Landsat scene to load onto a computer system. Back then, those data weren't georegistered, so we had to pick ground control points. We had to georegister those images ourselves. That took time. To actually process the image, it would take three or 4 hours to georegister one image. Then you had to run some algorithms to classify it. There was a lot of manual interpretation in reading the outputs of those early classification algorithms. AI is a tool that allows us to move past a lot of that early work that we did and really facilitate time series analysis, really facilitate long-term land change monitoring.
LAWSON:
Can you give us an example of how AI has helped to bring data faster or products to the public more efficiently?
SOHL:
The biggest example we have right now is, as you mentioned at the start, we are in the stage of merging LCMAP and NLCD projects, and those projects have their own form of machine learning. And so on the NLCD side, we've used regression tree and other approaches for quite some time. The algorithm that's used for LCMAP is also a form of machine learning. What we're doing right now is the next step. We're taking it to a more complex form of machine learning called convolutional neural nets. And that particular AI approach is something that for us in the science branch is very promising. It's something that allows us to look at characteristics of the data that traditional algorithms don't. So instead of looking just at the spectral information, it allows us to also look at some of that temporal information, or it allows us to potentially bring in ancillary data to the analysis in one sitting.
LAWSON:
So it feels like even though you have been using aspects of AI/ML for a while, that it still has a lot of unrealized potential ahead.
SOHL:
There's no doubt. I look at AI right now as something that's that's helping us modernize and completely uproot our methodologies in terms of how we produce products. In the future, I look at it, again, kind of as a preprocessing step that really allows us to move more of our resources out of pure data production into true analysis. And when I look at AI and ML going forward, I see a continued automation of those forms of tasks that were once very manual and very time consuming. And so where I would love to go in the science branch as we move forward is - in even our bread and butter products, something like land cover or drought products, invasive species, what have you, I'd love to have AI and some of the new approaches that we're developing automate the production of those keystone products to the point that much less of our effort goes into the development of the data itself, and it frees up a lot of our time to do the applications. Because the one thing that has been a challenge for us at EROS, I think, over the years is that direct link to the stakeholders on the ground that are using remote sensing data and the data we provide for applications such as biodiversity, hydrology, weather, climate. AI, to me, is a tool that's going to free up a lot of our time on the data preparation side and allow us to spend more of our time on that data analysis side where it really matters and has an impact on society. The time has come that there's enough competition, there's enough collaborative opportunity with folks in private industry and academia and the public world that that really needs to be part of our future and that we're using these tools from an AI perspective. But we're doing it in tandem, benefiting from some of the financial and long-term commitment priorities that the government can bring to bear versus some of the more agile approaches that private industry can provide.
LAWSON:
So some of our listeners may know that in your off time, you are an avid birdwatcher and photographer. What do you find fulfilling about that? It seems like there's a vast difference between working with imagery that's collected, you know, 438 miles above Earth and then taking these incredibly detailed close-up photos.
SOHL:
There are definitely differences. It's been kind of weird over the years, and particularly when photography has gone into the digital world, because having decades of experience in processing digital imagery actually is very beneficial when you're talking about bird photography right now, because you're using a sensor now with a digital SLR camera that's got a lot of comparability to something like a remote sensing sensor, and the processing of the data, too, in terms of how you process that and and create a final image that's useful, there actually are some corollaries. But overall, you know, for me, I am an avid nature lover, and to me that's what birdwatching and photography is about. When I went out previously, it used to be that I would focus primarily on the photography, and if I saw some good birds, but I didn't get some photos, I tended to think of that trip as not a success. Now there are times I go out, I don't even bring my camera. I just love being in nature. I love hiking, I like getting away from it all and just connecting with what's on the ground.
LAWSON:
Do you have any closing thoughts for us, Terry, about your new role at EROS, or EROS science going forward?
SOHL:
The one thing I do want to focus on a lot over the next year is communication and that. You know, I've mentioned that we're no longer the only game in town. We're at a stage where we do have a lot of people in the private industry and elsewhere that are potential partners, that are producing products that are similar to what we're producing. And as such, there's, I think, an underappreciation for what EROS science brings to bear. Again, we have 30 years of expertise in land change monitoring, and that's something that isn't going to be matched by the private industry right now. We have a resource in Landsat with a long term record of the landscape that provides us a really unique opportunity to look at things like the impacts of climate change over the long term. And again, that's a niche that private industry really isn't going into. But overall, you know, people are often familiar with projects like NLCD because it is the most widely used land cover product in the world. It has just a massive impact both within the federal community and within society overall. But I think now that anybody can go out on the computer and create an image, land cover image, even, on their own pretty quickly, I think people take something like what we do for granted. We are still the gold standard. One quote that I really like is from Matt Hansen, University of Maryland at the time, and his quote was, you know, with a plethora of all this new technology - and I'll paraphrase to protect innocent ears - but the quote was, you know, it's very easy to produce a bad map. And that's true. You know, with the technology we have right now, it's very easy to go online and to create a product that has a remote sensing base that you can create very quickly, but it's not very high quality. My goal over the next year is communication, to try to make people realize the value of this history that we have at EROS, the gold standard product that we have produced over time, the value of EROS science overall. I do think partnerships is a major part of that moving forward, just making folks aware of the work that we do in our science branch and EROS overall is the key part of our future.
LAWSON:
Thank you, Terry, for joining us for this episode of Eyes on Earth, where we have explored the future of science work at EROS. And thank you to the listeners. Check out our EROS Facebook and Twitter pages to watch our newest episodes. And you can also subscribe to us on Apple and YouTube podcasts.
VARIOUS VOICES:
This podcast. This podcast. This podcast. This podcast. This podcast is a product of the US Geological Survey Department of Interior.