Jordan Smith and Chase Lamborn, from Utah State University, recently completed a study of fire impacts on fishing in the Kenai River from the 2019 Swan Lake fire. Their study–funded by the Joint Fire Science Program— combined a literature review with interviews of local experts to identify impacts. The Kenai River is important: not only is it the most popular sportfishing destination in Alaska, averaging 275,000 angler days per year, it also produces 1/3 of the commercial salmon harvest in the Cook Inlet basin. Smith interviewed a small but diverse group of stakeholders who had extensive experience with the KR watershed, including agency resource managers, fishing advocates, people from non-profits, tribal members, and business owners. In addition to fire impacts, the study established a model of the Kenai River as a social-ecological system, which could be used to determine impacts from other kinds of disturbance.
Interviewees pointed out that a certain amount of luck, such as the lack of heavy rains post-fire to add big sediment loads as well as the fire’s location missing key Chinook spawning watersheds—limited any direct reported impacts of fire on fish. There were, however, impacts to resource users and businesses—primarily due visitors avoiding the area and road/river closures which restricted access during a brief, but critical, period of the summer. Nevertheless, a terrific early 2019 sockeye run (3-5x above preceding few years) helped to offset impacts on the sport fishery by encouraging anglers with high bag limits and success rates.
The literature review part of the study highlighted potential impacts to rearing and spawning habitats, water quality and fish passage. Most sobering are examples where populations failed to recover after fire, but these are the exception, not the rule. Adverse impacts are most likely from high-severity fires becasue they can lead to erosion and flooding. These events can induce loss of stream fishes, and generally require 3-10 years to recover when spawning habitat is affected. For the Kenai River, early-run chinook salmon were identified as the most vulnerable to this type of event. Although Smith et al. did not directly measure water temperature, stream flow, sediments, or mercury levels following fire on the Kenai, they provide a useful literature review of examples from elsewhere. They point out that with stream temperatures increasing and flows decreasing in the western continental US (a combination which can be deadly for fish), the threat of fire-related warming may become more serious in the future than it has been in the past.
To thwart runaway climate warming, the global community is struggling to find strategies to limit carbon dioxide (CO2) emissions that are steeply climbing. Increasing boreal wildfires in Alaska and Canada also threaten to increase CO2 emissions and could contribute potentially 12 gigatons to the world’s carbon headache by mid-century.
Fire Management strategy could make a difference: A research team from The Woodwell Climate Research Center and Union of Concerned Scientists wondered whether fire management offered a realistic way to slow down the release of legacy carbon in boreal forests, giving Nature and humans time to adapt and implement other mitigation strategies. How much would it cost to keep Alaskan wildfires at their historic level, avoiding climate-induced predicted increases? And was it even possible to make a difference? In short, the study found that—yes—more fire suppression could keep nearly 1/3 (4 Gt ) of that carbon in the ground in Alaska and Canada. The study tries to estimate costs associated with carbon savings and compares them to other carbon-sparing strategies being used or planned. Project goals are below are from a presentation given to Alaskan fire managers last fall.
Download our short Research Brief above (and/or you can access the full scientific article, open access, HERE:
Phillips, et al. 2022.
Escalating carbon emissions from North American boreal forest wildfires and the climate mitigation potential of fire management.
The face of a scientist: does that conjure an image of a certain gender, race, and age? Albert Einstein perhaps? Those stereotypes are changing: meet Dr. Yaping Chen–a rising star of science with a spectacular track record. The last 3 years she has come up with one mind-boggling revelation after another about how fire works in the Alaska tundra. After a MS degree in environmental engineering in China, Dr. Chen completed her PhD in the lab of the venerable Dr. Feng Sheng Hu at the University of Illinois. I first met her presenting a poster on the Nimrod Hill fire (Imuruk Lake, on the Seward Peninsula) at an American Geophysical Union meeting in 2019. The work was novel, ingenious, and suggestive of new ways to study fires with new computational and remote sensing tools. That was just the tip of the iceberg–or the thermokarst, if you will! Since then Dr. Chen has published numerous diverse research studies improving our understanding of dueling post-fire successional trajectories in tundra, improved burn severity mapping of legacy tundra fires, and fire regime effects on carbon balance. Her most recent paper outlines the role of tundra fire vs. climate warming in thawing permafrost in Alaska tundra statewide! If you’ve missed any of these important papers for your collection, links are included below. Now Dr. Chen is a post-doctoral researcher at the Virginia Institute of Marine Science, continuing her work on unraveling impacts of climate change. Thank you, Dr. Chen for all you’ve revealed to us in Alaska!
Historically, wildfire detection in Alaska relied on keeping an eye on neighborhoods and on lightning storms. Much of the detection around the state was done by aerial patrols beginning in 1973. In fact, in 1978 the BLM had 12 aircraft dedicated solely to fire detection and another six smokejumper aircraft which often did loaded patrols after widespread thunderstorm activity!
Alaska Department of Forestry T-28 Trojan on the ramp at Ft. Wainwright, 1988. (Photos by Linn Clawson). Below, pilot and observer were required to wear parachutes on missions in the WW2 vintage fighter-turned-detection aircraft
Then, at the turn of this century, it became apparent that the weather satellites MODIS Terra and Aqua could detect heat signatures of fire—fortunate timing because the price of contracted aircraft had skyrocketed and those surplus WW2 airplanes were mostly out of service.
Alaskan fire managers excitedly tested the use of spaceborne images to make wildfire detection faster and less expensive. Although these satellites are now nearing the end of their useful life (after 20 years in orbit), the VIIRS instruments aboard two newer satellites are starting to provide data.
Alaskan fire managers again are eager to make use of the new capacity, and are receiving help from key partners, inside and outside of Alaska. Several talks and posters at the 2020 American Geophysical Union (AGU) meeting held virtually in December highlighted important facets of ongoing efforts to harness the latest science and technology for use in Alaska.
The above figure (R. E. Wolfe, et al. FIRMS US/Canada – An Extension of NASA Near Real-Time FIRMS for the Forest Service and Inter-agency Wildfire Management Community—Fig. 1) diagrams the extensive network of agency and institution partnerships that have been established to gather and serve fire detection data to meet fire agencies needs around the country. With respect to the polar-orbiting satellites carrying some of these sensors, including VIIRS, Alaska find itself advantaged by twice as many daily satellite passes. Even more exciting, the Alaska Satellite Facility (ASF)—in the Geophysical Institute on the UAF campus– downloads the raw data directly, without waiting for processing and server functions in the lower 48. In the last couple years, ASF and GINA (Geographic Information Network of Alaska) have teamed up to feed VIIRS satellite detection hotspots directly to a digital map layer that can be accessed by fire managers less than one hour after the satellite passes. Compare the simplicity of the Alaska model with the above path for MODIS “rapid-response” data!
This kind of state-of-the-art service has taken a lot of logistical planning, hard work, and scrambles for funding both at the University of Alaska as well as the Alaska Interagency Coordination Center. The hard works pays dividends by giving fire management in Alaska a much needed boost against the background of longer, hotter fire seasons with flat suppression funding.
Not only has the satellite “hotspot” data proved useful for finding fires, it also is boosting our ability to monitor, model, and predict fire spread. A talk by Dr. Chris Waigl outlined how maximizing the use of three key remotely-sensed data streams: snow-off date to help start fire danger index models, improved wall-to-wall rasterized visuals of fuel drying by region, and improved regional algorithms to maximize the accuracy and sensitivity of satellite hotspot identification, were all used to good effect by Alaska fire management in 2020. Having the remotely sensed data to help with prioritization and decision-making was especially strategic during the COVID pandemic when managers were striving to spare staffing and reduce travel to rural villages.
Figure 3 Example of VIIRS fire heat points use in fire spread monitoring, P. Hickman.
The potential utility of remotely-sensed data for wildfire management has been recognized by scientists and agencies for some time, but it’s not always easy to bring a product from the laboratory to the operations room. There have been many discussions at various national levels and even a grassroots workshop in Fairbanks “Opportunities to Apply Remote Sensing in Boreal/Arctic Wildfire Management & Science”sponsored by NASA and facilitated by the Alaska Fire Science Consortium in 2017 addressing the potential application of remotely-sensed data. And although the story seemed very successful, given the above, it’s not over! Canada plans to launch the world’s first dedicated wildfire monitoring satellite constellation, WildFireSat, in 2025. And now we are thinking of ways to harness Artificial Intelligence for fire detection and spread monitoring! Stay tuned.
List of the AGU talks/posters referenced with links:
This AFSC research brief takes a look at early Alaska fire history from the 1940s. The “Zombie” Fires of 1942 is a historical narrative of an exceptional fire event related to the Alaska Railroad, including an early description of a holdover fire burning over winter.
Can remote sensing products help mitigate the loss of on-the-ground resources due to the COVID-19 pandemic?
Chris Waigl and the EPSCoR Boreal Fires Science team are rapidly developing new tools to aid with the fire season in Alaska. Products include, enhanced access to daily snow cover extent and fire danger maps, and highly focused fire-detection algorithms. The tools aim to provide data that can be integrated into existing systems facilitating direct applications for users, including fire operation managers.
Remote sensing of daily snow cover extent
Spring 2020 saw the introduction of a new daily snow cover extent mapping product for the state of Alaska. The source data from the NOAA National Ice Center is based on near-real time readings from the Interactive Multi-sensor Snow and Ice Mapping System (IMS). This satellite multi-sensor can differentiate between snow, ice, water, and snow-free ground with high levels of accuracy. The snow cover product is available seasonally for download as a vector file and as web-browser map with near-real time updates through the Alaska Interagency Coordination Center (AICC) mapping service, and year-round (with limited updates) from the Boreal Fires Team. Inter-annual comparisons of snow cover (Figure 1) can be made by geographic zone or throughout the state. This snow classification data could potentially be improved by validation through Moderate Resolution Imaging Spectroradiometer (MODIS) and Visible Infared Imaging Radiometer Suite (VIIRS) products, National Weather Service snow depth data, and citizen science projects that measure snow depth.
Improving access to spatial representations of Alaska Fire Danger Ratings
The Canadian Forest Fire Danger Rating System (CFFDRS) combines fire occurrence prediction systems, fire weather indices, and fire behavior systems to establish a fire danger rating. MesoWest produces fire danger ratings from CFFDRS for Alaska. The Boreal Fires team helps make this data more accessible by processing the MesoWest GeoTIFFs into a format that can be more easily used for webmapping by AICC. These fire danger ratings are available on the AICC web-mapping service and also hosted by the Boreal Fires Team. The fire danger ratings, known as Spruce Adjective Ratings, are grouped into low, moderate, high, very high, and extreme classes. These discrete groupings along with provincially specific parameters can create harsh differences in adjacent areas at province and international borders. CFFDRS has proven to have great application to Alaska. The Boreal Fires Team hopes that making the danger ratings more accessible will open the door for fine tuning the data to seamlessly fit Alaska, and lead to improved integration into fire behavior and analysis tools for the state.
VIFDAHL (VIIRS I-band Fire Detection Algorithm for High Latitudes)
VIIRS fire detection has shown to be invaluable for remote fire detection at high latitudes. VIFDAHL compliments VIIRS by subsetting high fire-danger areas and known fire locations. This information is particularly important for fire operations managers. Low-intensity detections have direct application to spotting residual fire hazards, which can help with resource prioritization Having additional inputs for where fire is now, particularly low-intensity detections, is helpful to identify ignition sources for fire behavior models.
The Environmental Protection Agency (EPA) has just ~15 official air quality monitoring sites around the immense area of Alaska to monitor air pollutants that can affect human health. Wildfire smoke, for example, produced about 60,000 tons of PM2.5 in 2018 (400,000 acres were burned –just a moderate fire season for Alaska!) If data from lower quality private and academic air sensors (called “Purple Air”) could also be used, we could add an additional 100 monitoring sites to better understand and forecast air quality. NASA ABoVE scientists Allison Baer and Tatiana Loboda from the University of Maryland compared EPA and Purple Air sensor data and came up with calibrations that correlate extremely well (coded T&RH—see example graphic below). You can view their Interactive Poster at the 6th ABoVE Science Team meeting—this week (Jun 1-4): https://astm6-agu.ipostersessions.com/default.aspx?s=09-98-87-A0-E6-1A-FA-E4-79-58-CF-F8-B6-54-4B-79
Example correlation from one private air quality monitoring station in Fairbanks.
Spatiotemporal patterns of overwintering fire in Alaska
Rebecca Scholten and Sander Veraverbeke – Vrije Universiteit Amsterdam
What are holdover and overwintering fires?
Fires can appear to be out, but retain smoldering combustion deep in the fuelbed and flare up again when the weather favors flaming behavior and fire spread. This phenomenon occurs not unfrequently in boreal forests of North America, and presents a well-known challenge to firefighters. Over the last two decades, fire managers noted increasing occurrences where fires survive the cold and wet boreal winter months by smoldering, and re-emerged in the subsequent spring.
Scientists and managers seek better understanding of how these fires sustain during such unfavorable conditions. Fire managers have already started targeting locations where they expect fires to flare up again. However, they are missing detailed information on the environmental and climatic factors that facilitate these fires. This information is crucial to detect fires at an early stage and keep firefighting costs low. A research group at Vrije Universiteit Amsterdam is studying when and where these holdover fires emerge and how their occurrence is tied to specific geographic locations.
Mapping overwintering fires from satellite data
Since 2005, fire managers reported data on 39 holdover fires that survived winter in Alaska. However, the location and emergence date of these fires were used in conjunction with satellite data to develop an algorithm for overwintering holdover detection. From satellite imagery, we can only observe fires that are large enough to generate a considerable amount of heat and burn a large enough area. Consequently, 32 out of 39 reported overwintering fires were too small (all smaller than 11 ha, 25 out of 32 smaller than 1 ha) to be detected from space. The location and emergence date of these small overwintering fires were used for the calibration of an algorithm focused on large overwintering fires. From the remaining seven large reported overwintering fires, our algorithm classified 6 out of 7 as overwintering fire. In addition, our approach revealed 9 large overwintering fires that were not reported by agencies between 2002 and 2018 in Alaska. A results paper is currently in preparation.
The spread rate of smoldering fires is known to be very low, and a smoldering fire would spread only between 100 and 250 m in an entire year (Rein, 2013). So, overwintered fires usually emerge within or close to the previous year fire (Fig.1) and can re-emerge with flaming behaviour as soon as favourable burning conditions appear in spring develop in to flaming forest fires before the major lightning-induced fire season. The onset of warm and dry conditions varies from year to year depending on the winter and spring temperatures and precipitation. These variables also shape the regional snowmelt day, which can be inferred from satellite observations. Indeed, our research indicates that holdover fires usually re-emerge within 50 days after the regional snowmelt. Overwintering fires are more likely to occur the year after a large fire
year (Fig. 2).
Can we predict where overwintering may re-emerge?
It is not only important to know when these fires emerge, but also where. We therefore analyzed spatial drivers of the overwintering fires we detected. Our research indicates that holdover fires are facilitated in those regions of a fire perimeter that had burned deeper into the organic soil the year before. Deep burning is a characteristic of a high severity fire. We also observed that overwintering fires were more likely to emerge in lowland areas with black spruce-dominated forest. Overwintering fires thus have some temporal and spatial predictability. Monitoring the edges of fire perimeters from the preceding year in lowland forested peatlands early in the fire season, and especially after a year with large burned area, may prove beneficial to extinguish flare-ups from overwintering fires before they develop into a large flaming forest fire. This could be a cost-efficient strategy for fire management agencies. In addition, this would preserve terrestrial carbon by safeguarding it from combustion.
Figure 2: Years with a large burned area (grey bars) are more likely to generate
overwintering flare-ups (orange bars) than years with less burned area
Rein, G. (2013). Smouldering Fires and Natural Fuels. In C. M. Belcher (Ed.), Fire Phenomena and the Earth System: An Interdisciplinary Guide to Fire Science (pp. 15–34). https://doi.org/10.1002/9781118529539
Turetsky, M. R., Benscoter, B., Page, S., Rein, G., Van Der Werf, G. R., & Watts, A. (2015). Global vulnerability of peatlands to fire and carbon loss. Nature Geoscience, 8(1), 11–14. https://doi.org/10.1038/ngeo2325
As discussed in the Feb. 7 Fire Science Highlight, burn severity in Alaska is best related to the amount of consumption of the forest floor—not the degree of tree canopy mortality as is in temperate pine and fir forest. Yet the most commonly applied metric to map burn severity using satellite remote sensing does not correlate well with substrate burn severity. The change in Normalized Burn Ratio (dNBR; Key and Benson 2003) is based on comparing a pre- and a post-fire image. However, NBR thresholds for severity differ from one fire to another and among different years: similar numbers don’t indicate the same severity levels (D. Chen et al. 2020). And with tundra fires, sometimes it works, other times not. This problem has dogged fire effects and ecology studies in Alaska for some time (see list of papers in Sean Parks November 2019 presentation) leading French et al. (2008) to conclude: “Satellite remote sensing of post-fire effects alone without proper field calibration should be avoided.”
2008 Transect photo from Anaktuvuk River tundra fire (R. Jandt)
Recently, we’ve seen some promising new methods used to improve satellite remote sensing of burn severity in boreal forest. Whitman et al. compared several indices including a relativized index that facilitated comparisons between different fires in Canada. She told us about it at the Opportunities to Apply Remote Sensing in Boreal/Arctic Wildfire Management and Science Workshop in 2017—here’s her presentation if you missed it: Improving Remotely Sensed Multispectral Estimations of Burn Severity in Western Boreal Forests. Loboda et al. ( 2020) found single images using just NIR (near-infrared) bands of Landsat did better than NBR in discriminating tundra fire severity. Sean Parks is attempting to harness the power of Google Earth Engines and cloud-based computing to use multiple images to further define the ecological burn severity (Parks et al. 2019)—this work is kicking off at the University of Montana. He also found that unusual aspects of some fires in Alaska (pre-existing beetle kill, short fire return interval) contribute to poor performance of the standard index (see his recorded November, 2019, Association of Fire Ecology meeting presentation HERE). And Yaping Chen, from the University of Illinois, explored using indices based on Visible and NIR bands (which have a large archive of available imagery going back to the early 1970’s) to evaluate tundra fire severity. Her paper (Y. Chen et al. 2020) points to a VNIR index called GEMI as a “robust surrogate to NBR in Arctic tundra ecosystems, capable of accurately estimating fire severity across fire seasons, tundra fires, ecoregions, and vegetation types.” The fact that GEMI is not as influenced by different vegetation types as dNBR gives it a distinct advantage mapping tundra burn severity.
Being able to more accurately map burn severity levels from space would give ecologists a boost for understanding why fires sometimes induce radical changes in ecosystems while other times the system self-replaces in a very short span. For example, Yaping Chen used GEMI to reconstruct burn severity on older tundra fires like the 1977 example below and tie it to thermokarst effects (like catastrophic lake drainage or ponding) resulting from the fires (poster presented at AGU meeting December 2019). We look forward to more exciting products and tools coming from these research teams!
Reconstructed fire severity map of the 1977 OTZNNW 38 tundra fire computed with dGEMI using Landsat MSS imagery.
In Alaska, we know that post-fire recovery of spruce forest generally takes one of two major pathways. Fires in spruce forest burn with high intensity and it is typical for 90-100% of the standing trees to be killed in the fire, so trees primarily regenerate by seeds released from the cones—often preserved and dried in the dead snags with the heat of the fire. What happens next is largely dependent on the amount of forest floor moss layers consumed in the blaze: if much is consumed, maybe even leaving patches of mineral soil, we consider this a high severity fire, whereas if only a few centimeters of moss duff have been removed, we consider the fire severity to be low. Black spruce
Photo by USFS, PNW (2004).
readily re-establishes itself after low-to-moderate severity fires because its abundant and relatively large seeds can germinate and survive a few dry spells in the peat-like substrate of the remaining organic forest floor. This is called “self-replacement”. Other tree species, including white spruce and the deciduous trees prefer a more exposed mineral seed bed, which may offer more consistent moisture and nutrient availability. After higher severity fires, grass and fireweed are notable early on and followed by a period of shrub and deciduous tree seedlings and re-sprouts. The dominance of the forest may then shift to aspen, birch and/or poplar for a period of years. Ultimately, more shade-tolerant but slow-growing spruce will again dominate, but this may take 50-100 years. This pattern of recovery is termed “relay succession”. The moisture available at the site and pre-fire species composition also influence recovery, as illustrated by Johnstone, Hollingsworth and Chapin in the Key for Predicting Postfire Successional Trajectories in Black Spruce that they prepared for managers in 2008 (below).
Now, here’s the punch line: Alaska ecologists have long been asking themselves what percent of the time, over the whole landscape, does self-replacement vs. relay succession occur? For many modeling efforts to date (LANDFIRE, for example) we had only our best guesses. At a scientific meeting in December 2020 (American Geophysical Union) Jennifer Baltzer, a Canadian forest ecologist, related the findings of a large ecological study with over 1,538 burn study plots across Alaska and western Canada (1,140 of the plots were black spruce forest). Her team demonstrated that about 62% of the time burned black spruce forest recovers by self-replacement (73% for all conifer forest types), while approximately 20% of the plots were headed for a relay type succession. Another 20% or so of the plots were showing little forest regeneration (regeneration “failure”)—which seemed to be more likely after repeat burns in a fairly short time period. The exciting thing about this research is that it provides—for the first time–quantitative estimates for these competing post-fire pathways. The research is being reviewed for publication now–we’ll let you know!