More than a decade ago, Tom Loveland sat down to sketch out a few thoughts on land change and the Landsat archive.
The archive was and remains rich with history, adding new observations of the entire planet every eight days. But at that point, Landsat-based land change research was about comparing points in time – this year versus five years ago, to 10 years ago, and so on.
What if Loveland and his fellow scientists at the USGS Earth Resources Observation and Science (EROS) Center looked at all the data, as far back in time as possible?
What would it take to make that happen? What could be learned if they pulled it off?
Thanks to the support of EROS leadership and countless hours of hard work and innovation from current and former EROS scientists, partners and collaborators, the world is about to find out.
The USGS Land Change Monitoring, Assessment, and Projection (LCMAP) initiative has released its first suite of land cover change and landscape change products, an unmatched record of land cover and change for the United States, one that characterizes the dynamics of the landscape across the conterminous United States through 33 years of its history.
“I’m relieved. It’s been a long time,” said Loveland, the former EROS Chief Scientist who oversaw the LCMAP initiative until his retirement in 2018. “It’s pretty exciting to see it finally coming out after so many years of advocacy for the concept, the technical formulation, and ultimately, the implementation. It’s a real testimony to the crew that’s stuck with it all the way through.”
Collection 1 Release: ‘Really the Beginning’
For the last two years, EROS Research Geographer Jesslyn Brown has led the crew charged with the task of moving LCMAP from concept to reality.
LCMAP’s five land cover change and five land surface change science products were made possible through a sophisticated change detection algorithm that mines every high-quality Landsat pixel for information, as well as by advances in processing power in recent years, high performance computing resources similar to the EROS-based Denali system, and a host of technological advances over the past decade.
The technology made it possible, Brown said, but it didn’t make the work easy. The team dealt with multiple setbacks, algorithm tweaks, and powered through test runs as they worked to fine-tune the products for accuracy and reliability. The reference dataset used to validate the product and guide users in best practices – itself a product with wide-ranging scientific value – took five years of collaborative human labor with the U.S. Forest Service.
“The activity of monitoring the Earth’s surface is fraught with challenges, and we’ve tried to deal with those challenges,” Brown said. “There were times it was really frustrating, but we didn’t stop. I’m incredibly proud of this team.”
The release of LCMAP Collection 1 Science Products, Brown said, represents a starting point for an initiative that promises both annual updates – a faster clip than users of land cover data are accustomed to – and the addition of more detail to meet user needs as time passes and products are refined.
In short, the LCMAP team has no intention of resting on its laurels.
“Yes, this time series approach is amazing and allows us to see more (change) than we’ve ever been able to see before,” Brown said. “These products are stunning, but they’re not perfect.”
Pete Doucette, head of the Integrated Science and Applications Branch at EROS, told the team in a meeting the week before Collection 1 went live through the EarthExplorer data portal that the achievement is the beginning of a new chapter.
"Although the release marks an endpoint of several years of work, it also can be seen as a beginning,” Doucette said.
From Comparison to a Deeper Understanding
As Brown and Doucette look to the future, so do their colleagues in the science community. Zhuoting Wu, a physical scientist with the Land Resources Mission Area and key collaborator in the LCMAP initiative, said information stretching three decades into the past is momentous.
Even in a rapidly advancing field whose progress is driven by cloud computing resources, ever-expanding pools of data from sources such as the European Space Agency’s Sentinel-2 satellites, LCMAP’s meticulous approach, temporal depth, quality control and nationwide coverage stand out.
“To be going back to the 1980s, that’s a first,” Wu said. “This is the longest we’ve been able to go back in time looking at land change and land cover. That’s the game-changer.”
The value lies not only in the availability of three decades of annual change, but in what LCMAP reveals about each year. LCMAP’s land change mapping approach sorts each 30-meter plot of land into a land cover class—such as developed, wetlands, or grassland—for each year, along with probabilities, and highlights class changes. Its land surface change products offer even greater insight, providing information such as the time of year that change was detected and the magnitude of change—even ephemeral or impermanent changes from incidents such as tornadoes, hurricanes, floods, or fires.
Moving from land cover mapping to change detection has broad implications. Tapping into what Wu calls the “gold mine” of the Landsat archive creates an opportunity for the remote sensing community, land managers and even the general public to move a step beyond simple before-and-after comparisons to a deeper understanding of the nature of landscape change.
“People have traditionally looked at land cover change – forest to agriculture, or agriculture to urban, Wu said. “But most of the change is more gradual and conditional. We’ll see drought or an insect outbreak in a forest, but it’s still a forest. It’s just a stressed forest. If we just look at a land cover map, but we don’t know what happened to that forest throughout the year. With LCMAP and mapping spectral (land surface) change, we get a convergence. It tells us the condition of the land in a way we can tie to other things. It has implications for fire behavior, for climate feedback or for water resources.”
The data have already sparked some interest from the research community. Heidi Sieverding of the South Dakota School of Mines and Technology attended an LCMAP workshop in fall 2018, and is currently working on a paper that aims to assess LCMAP data as a tool to study of the impact of mountain Pine Beetles in the Black Hills. Sieverding and her colleagues have already worked with provisional data, provided through EROS, on various watershed assessment and modeling projects.
“It is good to hear that the product has been fully released,” Sieverding said in early June. “It will be useful in our research and that of several of our collaborators.”
LCMAP Collection 1 Science Products are available through EarthExplorer, as well as through an interactive web viewer that allows users to explore the datasets for their area of interest.
Loveland: LCMAP adds to EROS Legacy in Land Change Science
Loveland is excited to see where research teams take the data. More data depth, produced more quickly, has the potential to refine our understanding of the landscape, he said, moving from viewing change as a steady march to a more nuanced view of change as dynamic and fluid, inclusive of linkages and feedback loops that tie landscape change to changes through climate, public policy, human behavior, natural disasters and other factors.
“I think they have something that is unprecedented. I’m excited about the potential for this to start changing our view of land change,” Loveland said. “The ability to start looking at all the processes that are going on is sitting in front of us now with the datasets.”
It’s fitting, Loveland said, that those datasets come from EROS. Advancing the world’s understanding of land change is the EROS mission. Loveland oversaw LCMAP, but has been involved with a wide range of projects in the run-up to it. So have other now-retired scientists who guided LCMAP, including John Dwyer, Alisa Gallant, and Jim Vogelmann.
Loveland also credits former EROS Director Frank Kelly, who stepped up to sell the LCMAP concept to USGS headquarters long before the tools to make it happen were available.
“Understanding the role of humans on the landscape was a pretty bold step by the USGS (through the launch of the first Landsat),” he said. “EROS really stepped up when we created things like Global Land Cover, the National Land Cover Database, LANDFIRE, Land Cover Trends, and now LCMAP. We at the Center have been constantly pushing to innovate as we move into the future. That can only be done through reading the tea leaves from the (user) community, advocating for those needs, and having the technical and scientific staff with the skills to make the vision become a reality.”