Landsat Archive Fills in Tundra Fire Gaps

An example of a burn scar on Landsat imagery

Tussock growing on a burned area of the Alaska tundra

The Arctic foothills and Brooks Range mountains in Alaska

By Earth Resources Observation and Science (EROS) Center July 11, 2024

Curiosity can be a strong motivator. It certainly was for Eric Miller of the Bureau of Land Management (BLM) Alaska Fire Service.

A man wearing protective gear stands in front of a low prescribed fire in the background — Fire ecologist Eric Miller with the Bureau of Land Management Alaska Fire Service is shown at a prescribed fire.

The fire ecologist didn’t know much about remote sensing when he and some colleagues started noticing burned areas on the Arctic Slope that they knew were not listed in the Alaska Large Fire Database, which includes fires dating to 1939.

If the fire record was missing the tundra fires they stumbled across, had it missed others in the past, too? And did the increasing amount of fires in the record in recent years really indicate a trend, or just improved detection methods?

Their quest to find answers stemmed from more than their personal curiosity. For lack of other sources, researchers use the Alaska Large Fire Database for studying the tundra—in some cases, the effects of a changing climate. So the more complete the record, the better their results can be.

Miller had tried looking for burned areas in Google Earth with no luck. But eventually, he and his colleagues, including a couple of U.S. Geological Survey (USGS) scientists familiar with remote sensing, decided the best approach would be to turn to the longest Earth observation satellite record available: Landsat, whose 50-year-plus archive is kept and added to continually at the USGS Earth Resources Observation and Science (EROS) Center.

Miller’s thorough search of the Landsat archive yielded some intriguing results for the Brooks Range Foothills area, a stretch of Alaska tundra that’s north of the Brooks Mountain Range and south of the Beaufort Coastal Plain on the northern coast of Alaska.

Miller found 57 previously unknown fires, representing 41% of the area’s new total of 138.
Fire frequency has increased significantly in recent years.
Most unknown fires were recent rather than historical, despite coinciding with better detection technology.
Some known fires in the fire database were listed with inaccurate locations and sizes.

Analysis ‘Just Kind of Snowballed’

Wildfires in remote tundra areas, with colder conditions and few witnesses, have long been thought rare. Fires aren’t often fought there because they may not be known, firefighting resources are in use elsewhere, or no structures need protecting. Many past fires have become known simply because of planes passing nearby or spotters chasing storms, noting the approximate location and size of fires.

Imagery from Landsat satellites represented the best option for a more thorough record. So Miller acquainted himself with EarthExplorer, the Landsat data access website hosted by USGS EROS, and started looking for the 8 or 10 burned areas he and his colleagues had noticed in the landscape.

A gray map of the northern part of Alaska showing mountains in relief and dotted with open circles and cross symbols — The locations of recorded (pluses) and previously unrecorded (circles) tundra burns in the Alaska Large Fire Database. The Brooks Range Foothills ecoregion is outlined in white. Figure from: Unrecorded Tundra Fires of the Arctic Slope, Alaska USA

“I found them all,” Miller said. “But the thing is, once you start looking, you start finding all these other fires. It just kind of snowballed. It was never my intent to go out and write this paper or do this analysis or anything. It just kind of happened—scanning around in Landsat, finding all these different fires.”

In late 2022, Miller started a systematic search of the Brooks Range Foothills in the entire Landsat archive, which dates to 1972. He became quite familiar with the EarthExplorer site as he combed individually through a total of 790 scenes—an average of 17 per year with up to 32 for one year. Not included in the search were seven years in the 1990s, when there were gaps in data for the northern high latitudes. The project also used some historical aerial photographs available through EarthExplorer.

Burned areas had to meet certain criteria to be counted as an unrecorded fire, including being identified in at least two scenes.

The unrecorded fires Miller found ranged in size from a tenth of an acre to 689 acres, for a total of about 4,163 acres, or 1% of the total previously known fires on the Arctic Slope. Still, they represent about two-fifths of the total number of fires now known.

Landsat helped improve the accuracy of fires already recorded, as well. A passenger next to a pilot might have estimated the coordinates and size, or called in a location based on vague landmarks. Miller found a truer picture from Landsat. Some burned areas that he found, he assumed were unrecorded until he realized they were just inaccurate.

“A lot of those old burns were off by miles,” Miller said.

Stumbling Across Past Tundra Burns

Aerial photo of an expanse of green tundra interrupted by a large irregularly oval shaped brownish burn scar and small ponds — In 2017, a small burn scar from a tundra fire was found during field work related to the 2007 Anaktuvuk River Fire. In the background photo, Eric Miller and a colleague are seen surveying the burned area, where tufts of tussock grow green with white flowers.

The Bureau of Land Management Alaska Fire Service, based in Fort Wainwright, manages wildland fire operations and suppression on all Department of the Interior and Alaska Native lands. That totals more than 230 million acres, among them BLM, National Park Service and U.S. Fish and Wildlife Service lands.

In the Alaska Fire Service, Eric Miller has worked in science-related roles like monitoring and research, including some field work. The Landsat analysis of tundra fires stemmed from observations made during field work in the aftermath of the 2007 Anaktuvuk River Fire. That lightning-caused fire had burned 256,000 acres of tundra, making it the largest recorded fire in Alaska that year.

Miller’s predecessor and then supervisor, Randi Jandt, put together a team of scientists including Miller to set up plots on the Anaktuvuk burn scar to monitor the site’s recovery over time. Visits took place in 2008-2011 and 2017, with a follow-up planned soon. Among others joining Miller for the 2017 visit were Ben Jones and Carson Baughman, both then geographers at the USGS Alaska Science Center who have experience in remote sensing.

In 2017, they observed new burn scars in that area from fires that were not included in the fire database.

That added to Miller’s past experiences. In 2011, he had been intrigued by white spots on distant hilly slopes. “We got back in the helicopter and had a look, and sure enough, it was a burn from the previous year,” Miller said. The year 2010 had been especially active for tundra fires.

“The fire had burned up all the debris and released a whole bunch of nutrients and stimulated these tussocks to flower white,” he said of the tufts of sedges that grow on the tundra. “So that was super interesting. We left that burn and headed back to Toolik and found another white patch on the way. That was another instance of fires that were not going to be in the official fire record.”

Sources/Usage: Public Domain. View Media Details

A burn scar from a fire on the Alaska tundra

Increasing Fire Frequency

One advantage of Landsat’s long historical record is the ability to see trends over time. With confidence in a more complete and accurate record in the Alaska Large Fire Database, Miller and his colleagues wanted to more definitively say whether fires were increasing in the tundra over time.

They looked at all previously recorded and newfound wildfires in the Brooks Range Foothills in the 1972-2022 timeframe that were 24.7 or more acres (about the size of the smallest fire that could be detected by the first Landsat satellite, to eliminate a bias toward newer satellites with higher resolutions). They found that half of those fires had occurred in the first 37 years, and the other half in the last 13 years. In other words, fire frequency had increased nearly threefold.

“Landsat tells us the trend is actually real. There are more fires up there now. I think that probably was one of the most gratifying results.”
- Fire ecologist Eric Miller, BLM Alaska Fire Service

Looking at unrecorded fires, they expected to find more from before 2002, when the MODIS sensor from another satellite routinely began capturing the Arctic Slope. Data from MODIS and a more recent sensor, VIIRS, are relied on to detect active fires, so this technology was expected to make the later record more complete.

However, they found the opposite: Only 7 of the 57 unrecorded fires had occurred prior to 2002, with MODIS and VIIRS both missing most of the later fires. The fires may have been too small for the sensors’ coarser resolution to detect, or too short-lived to be captured by a satellite pass.

Big Project, Important Results

Miller admits that sitting and staring at scene after Landsat scene for two or three weeks was a challenge.

“It was a bigger project than I thought it was going to be,” he said. “I powered through it. It’s almost like playing slots or something—every time you’d find a new fire, it’s a dopamine hit. It wasn’t as bad as you might think. It was relaxing in a lot of ways just to be able to scan.”

And the results are worth it. Miller is especially happy to provide scientists with a more robust and accurate record to use. Researchers have found that tundra wildfires can accelerate permafrost thawing and change the makeup of plant communities.

“We all had the idea that there is more fire up there than we can document, and the harder you look, the more you’ll find it. Eric’s work and that paper confirm that,” said geographer Carson Baughman of the USGS Alaska Science Center, who contributed to the study. Baughman uses remote sensing to help answer a variety of historical scientific questions.

Portrait of a man wearing outdoor gear with a lake and distant mountains in the background — Carson Baughman is a geographer who works with Landsat satellite data at the USGS Alaska Science Center.

Sources/Usage: Public Domain. View Media Details

A burn scar from a tundra fire on the Arctic Slope in Alaska

Landsat ‘Only Game in Town’ for Historical Data

A portrait of a man in front of a background of trees — Matthew Hethcoat

The United States isn’t the only country interested in learning more about the true number of fires in the tundra.

Matthew Hethcoat and his colleagues of the Canadian Forest Service used a more automated process than Miller’s to explore unrecorded tundra fires over the larger Canadian Arctic and sub-Arctic tundra region.

But Hethcoat had to admire Miller’s ambition. “He basically went systematically through almost every single record. That is some dedication,” said Hethcoat, who also cited Miller’s 2023 publication in his own 2024 publication.

Hethcoat works in remote sensing for the Canadian Forest Service’s fire group based in Edmonton, Alberta. Like the Alaska Large Fire Database, the Canadian National Fire Database had a gap in the record of tundra fires because of a lack of monitoring and mapping. Hethcoat and his colleagues set about to create a baseline of known fires so further research could explore fire trends.

Hethcoat relied on machine learning along with manual validation to identify fires in Landsat imagery from the years 1986-2022. His group also used Landsat to map fire perimeters for inclusion in another Canadian database, the National Burned Area Composite.

“Having that historical period is what really makes Landsat stand out. I think it’s what has made it so valuable to the global scientific community,” Hethcoat said. “We can really start to look over 40 years at really important trends and drivers and questions that society is facing. Clearly, Landsat is the only game in town if you want to start looking at those historical data.”

Striking Increase in Canadian Fires Added

Four different maps with spots of various colors in assorted places, with a color and size key on the right — Panel (a) of this figure shows the spatial distribution of newly detected tundra fires in northern Canada that occurred between 1986 and 2022 (orange circles), with individual fires north of 66°N shaded lighter orange. Fires previously mapped in the National Burned Area Composite are black stars. Panels (b–d) are centered over each sub-region of fire clusters, showing the fire area as graduated point sizes and the year of burning as colors. Figure from: Unrecorded Tundra Fires in Canada, 1986–2022

Landsat enabled Hethcoat to find 206 out of what are now 275 known fires in the Canadian Arctic and sub-Arctic region from 1986-2022. Like in Alaska, most of the fires were rather small in size; the new discoveries represent a 32.8% increase to the previously known total area burned. Hethcoat is sure some small fires still were missed, but the record is now far more complete than before.

Hethcoat’s group also looked at MODIS and VIIRS data. While three additional fires were found via that data, Hethcoat had come up with a similar result as Miller: Most unrecorded fires did not show up in that data.

The goal of Hethcoat’s study was to establish a baseline, so it did not consider causes or trends of fires. Further studies are expected to examine those questions, along with how a changing climate is driving trends and what effects are being produced.

“We wanted to at least begin to build that dataset of known fires to then start to understand: Are we seeing changes? What are the climatic or weather events or phenomenon surrounding a particularly bad fire year or clusters of fire that we’re seeing in the landscape?” Hethcoat said.

He looks forward to the improvements coming from the next generation of satellites in the joint USGS-NASA Landsat program. Landsat Next, scheduled to launch in late 2030/early 2031, will provide more frequent and more detailed information about the Earth.

“Given the commitments that the Landsat program has made, and there is a plan for the Landsat Next mission, there’s a clear continuity and a well-thought-out plan going forward. I think that is a huge global service because we’re facing a lot of pressing global challenges,” Hethcoat said.

“As someone who uses remote sensing, I’m optimistic that having that kind of information, and access to the kind of scale and timeliness of information like that, will enable us to address some of the challenges that we’re facing as a society.”