Eyes on Earth Episode 102 – LANDFIRE 2022 Update

Inga La Puma

In only seven months, we were able to incorporate both those submitted disturbance events and the remote sensing of change into the vegetation and fuel layers and have them available for those fire and land managers prior to the Southwest fire season. So this is huge. This is a momentous occasion, and we're really proud of the team for getting that done in such a short amount of time.

Jane Lawson

Hello, everyone, and welcome to another episode of Eyes on Earth, a podcast produced at the USGS EROS Center, which celebrates its 50th anniversary this year. 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 LANDFIRE Program's 2022 update, which is being released incrementally this year. LANDFIRE, which is short for Landscape Fire and Resource Management Planning Tools, is a key national data source for the management of wildfires, management of the plant materials that fuel fires and planning for prescribed fires across all 50 states and the U.S. territories. The U.S. Department of the Interior and the U.S. Forest Service provide resource support and business leadership for LANDFIRE, with data products generated at EROS. The Nature Conservancy is a partner as well. For the first time, the LANDFIRE 2022 Update contains disturbances from as recently as the year before. This represents a significant step in the nearly 20-year-old program's evolution to annual releases. Our guest today is here to talk about this exciting development. Fire scientist Inga La Puma, who earned her Ph.D. in ecology from Rutgers University, is one of LANDFIRE's technical leads. Her background includes research related to carbon and wildfire in forests, as well as landscape disturbances and fire science communications. Welcome, Inga, to Eyes on Earth. 

La Puma

Thanks, Jane. I'm excited to be here. 

Lawson

Can you first tell us briefly about the data layers included in the new LANDFIRE release and how they are useful?

La Puma

So LANDFIRE has over 30 spatial data layers, and these include annual disturbance layers with information on disturbance types, which include things like wildfire, windthrow and new development, along with its severity, or the level of effect on the landscape. So we have vegetation layers with up to 700 different modeled natural vegetation classes based on a classification called Ecological Systems. Our vegetation layers also include things like the percent cover and the height of the dominant life form, so like tree, shrub or herb. LANDFIRE's fuel layers include information on surface and canopy fuels that support the potential fire behavior analyses, including things like the rate of spread and flame length based on the vegetation type. LANDFIRE also provides information on historical fire regimes and a measure of the vegetation departure, so that's like an indicator of how much the vegetation and the fire regime have changed from pre-European colonization conditions. We also provide seamless topography layers that fire behavior modelers rely on, like slope, aspect and elevation. We derive our disturbance data from reported areas of disturbance such as harvest or prescribed fire, along with remote sensing of spectral change. We also use submitted field plot data and lidar, which uses laser beams to detect and measure vegetation for training machine learning models in our vegetation mapping. We use things like Landsat and Harmonized Landsat Sentinel satellite imagery to help extrapolate across the landscape. Although the impetus for LANDFIRE layers was for fire management when it started 20 years ago, there's so many spatial data layers produced along the way to understanding those fuels that we've actually become a resource for habitat modelers, climate change scientists and countless more applications. Recently, I even saw a paper where LANDFIRE vegetation layers were used to help find where septic systems might be failing. So the diversity of the applications for our layers surprises even me.

Lawson

In the past, updates would cover multiple years of disturbances, and this update includes two years: 2021 and then through September 2022. So why is LANDFIRE trying to get to the point of annual releases with disturbances from the prior year?

La Puma

We've been releasing updates annually for the past three years. So first we had 2019 L, which was a limited version; then LF 2020, and now we're releasing LF 2022. But not only do we want to release more frequently, we also want to get the information about what's changed over the last year into our annual releases. So that means making sure that as many recent disturbances, like fire, harvests or other changes, are accounted for in the LANDFIRE vegetation and fuels products. So LF 2022 is the first LANDFIRE release to include disturbances from the previous year. And it's really a major milestone. We've had people that have been with LANDFIRE since its inception, and they never thought this would be possible. I mean, it took eight years to create the very first national LANDFIRE datasets that did not include recent disturbance. LANDFIRE 2022, it actually represents the first release that's essentially up to date with those disturbances. It accounts for everything. Every release, we update our agriculture and our building footprint in developed areas as well. Essentially, we want LANDFIRE products to be applicable and usable in the current field or fire season. So in the past, the users actually had to update LANDFIRE vegetation and fuel layers on their own with the understanding that those recent disturbances probably weren't in there or accounted for. But most users won't have to do that anymore if they use the most recent LANDFIRE releases. So when you think about something like the 2021 Dixie Fire, for example, almost a million acres burned. That's a lot of change to fuels that people had to account for before, where, you know, they won't have to now.

Lawson

So how is LANDFIRE adapting to try to produce these products more quickly? I would imagine there is some effort involved.

La Puma

Yeah. To produce the LANDFIRE products faster, we really did have to focus on speeding up our disturbance mapping. And so we managed to map seven years of disturbance in just three years as part of those past three releases. So part of that, speeding that up, is asking our data contributors to submit their polygons of disturbance, such as where a prescribed fire occurred or a thinning, by the end of October of the same year. And then shifting our disturbance year from a calendar year to sort of a fiscal year time period. Then that gives us time to incorporate that information into our products. So additionally, we're using the Denali high-performance computer, a supercomputer at EROS, to create seasonal image composites from Harmonized Landsat Sentinel imagery for our remote sensing of land cover change as soon as that imagery is available. So that's now only a few weeks after it's collected. So it's really helpful in speeding up the process. But we've also been steadily improving LANDFIRE's ability to automate disturbance detection as well. We have a creation of new algorithms which actually make it easier for our image analysts to kind of home in on areas that need additional review on whether a disturbance occurred or not. But yeah, in general, disturbance layers tell us that there was a change to the vegetation and fuel, and then we end up using expert informed rule sets to help identify what the new vegetation and fuels would be after that disturbance. And so that includes things like severity, time since disturbance and the type of disturbance. With the introduction of last year's disturbances, so those expert informed rulesets that we've been using to adjust our vegetation and fuels needed to actually include a new time step as well. So in prior updates, we never really needed the rules for, like, that 0 to 1 year time period, that most recent year, because we were always at least two years behind. So we've added that into our change. One of the things that we wanted to do was really kind of focus on releasing the data as soon as possible. So we've already begun releasing the LF 2022 product in increments, and we started with the Southwest on May 1st, mainly because this region is such a high priority for fire management. It includes California, for example. This means that in only seven months, we were able to incorporate both those submitted disturbance events and the remote sensing of change into the vegetation and fuel layers and have them available for those fire and land managers prior to the Southwest fire season. So this is huge. This is a momentous occasion, and we're really proud of the team for getting that done in such a short amount of time.

Lawson

Who uses LANDFIRE data, and how do they use the data, and how will they benefit now that you are updating disturbances more frequently?

La Puma

I always try to put myself in the situation of our users. So if I'm the incident commander on a large fire, for example, I want to know that the predictions I'm seeing for that projected fire behavior on the wildfire are accounting for prior large fires in the area, or prescribed fires, or fuel treatments that were completed last year, or a harvest that just occurred on a neighboring timber company's land. So I want to make sure that all of that is accounted for when I'm doing my fire behavior analyses. So when areas like these with reduced fuels are in the path of a fire, it will behave differently in terms of how fast it spreads, in what direction, how intense it will be. So incident commanders ask questions such as, Can I use these reduced fuel areas to help slow down the fire so that people can be evacuated? Can I use these reduced fuel areas on either side of a road as control points to better fight the fire? You know, these types of decisions in the landscape context need to be made fast with the best and the latest data so that more informed decisions can be made. For those folks trying to actually put together prescribed fire management plans, having current data is also important. It helps them decide which areas to prioritize and which areas might be more difficult to burn, for example. But besides the on-the-ground operations, you know, LANDFIRE data is used nationally for wildfire risk assessments. You can imagine, you know, if you're doing an assessment like this that having the latest data is important for large landscape hazard models, some of which are used to help prioritize government resources. Likewise, if I'm, you know, a researcher trying to understand things like post-fire impacts, or which invasive species thrive from disturbance versus those that don't, or if I'm charged with delineating suitable habitat for the conservation of a threatened species, or I'm trying to understand why a species of concern prefers a certain habitat, or if I'm estimating the carbon stored or lost to be able to enact, you know, some sort of mitigation efforts, current data is key, right? We need that for those kinds of analyses. LANDFIRE also just strives to create lots of ways for users to access the data to get it quickly, so via download or streaming to make it easier to use. All LANDFIRE releases are posted to all of our delivery platforms as soon as they are complete.

Lawson

What does the future look like for LANDFIRE from here? 

La Puma

LANDFIRE's at a crossroads in how we're thinking about updating our data. Traditionally, we have used those expert informed rule sets to update or transition the vegetation and fuels data using that disturbance information. So the rule sets use the disturbance type, severity and the time since disturbance to apply a transition rule to each disturbed pixel. And those are all expertly informed across the country with workshops and things like that. So the methods that we used for the LANDFIRE 2016 Remap, which was considered a base map, was actually different. So those methods essentially were machine learning models that were trained with plot and lidar data on satellite imagery, and then, you know, extrapolated the relationship across an area of interest. But Remap took five years to create. So emulating the parts of the 2016 Remap methods, namely mapping lifeform, cover and height, are actually now within reach on an annual basis for the entire country. We use the most recent satellite data to map disturbance every year, so why shouldn't we be able to use it to model our vegetation and fuels on an annual basis, too? So that's kind of what that idea is. But the rule sets are great. They've worked really well for a really long time, but there's things they don't consider like interacting disturbances or, you know, unusual site conditions where things grow back differently. And the satellite imagery and more recent training data can actually help account for those. So the number of scenes available now and the easy access to the recent imagery, it's actually making the prospect of using machine learning on the imagery every year for updates really enticing. And we're currently prototyping how we might be able to do this for future LANDFIRE releases. But, you know, like one issue with using satellite imagery for updating fuels is that surface fuels aren't always visible from space in the forest, you know, when there is an overlapping canopy layer. So using lidar to supplement the mapping of fuels is another area we are talking about expanding on in LANDFIRE. There is these new physics-based fire behavior models that are highly reliant on the amount and the distribution of fuels in that three dimensional space because they use things like fluid dynamics and heat transfer properties to predict fire spread and intensity. So these types of models can be coupled with atmospheric models, and LANDFIRE's working with researchers to kind of figure out the best organizational structure to provide the fuel data needed to run those kinds of physics-based models. On this front, we've been working to calculate cover and height for large amounts of recent lidar from 3DEP, the national lidar data repository at USGS, on the Tallgrass HPC, which is another USGS supercomputer, and on the AWS cloud computing platform. This type of infrastructure, setting things up like this, I think will provide the basis we need to sort of slice and dice the lidar point clouds in whatever way we need to in order to help create the structural data needed for 3D fuels in the future.

Lawson

Can you describe a little bit about the LANDFIRE team and what you find most rewarding or most exciting about being part of it?

La Puma

Yeah, you know, I love my team. I love the entire LANDFIRE program, but the best word I can use to describe the LANDFIRE team is dedicated. I mean, there's a level of commitment needed to produce a data set that so many people rely on, and we're fortunate to have people that really get that. As a multi-agency program funded and overseen by the DOI and the U.S. Forest Service, we have our partnership with The Nature Conservancy, and the production conducted at EROS by contractors managed by the USGS, it takes some work to balance the priorities of over 20 people across different agencies, but I actually think that the diversity of voices that we have is really valuable. I was super impressed with the willingness of our production and operations team to try to meet the goals of LANDFIRE 2022. It takes guts to say, Yeah, I think we can do this and then actually follow through. So I like how we help each other out. You know, if someone's encountering a challenge, they aren't on their own. We support each other, help try to find a solution. And there is actually a really good balance of talents on the team. So basically, yeah, the definition of a great team. I mean, I'm a spatial fire ecology nerd, so I enjoy the daily technical challenges, and luckily I enjoy communicating about the science and helping people understand what we have to offer in presentations and in writing, because I do both of those things quite a bit. You know, one thing that is kind of cool is that many folks on our team are truly remote, and they don't live in Sioux Falls. For example, I live in New Jersey, and I think the farthest from me right now is Charlie, our fuels lead, who lives in Oregon. We have folks in Idaho, New York, Virginia and Maine right now. I think it's great because sometimes you just like need someone to look out the window and tell you what they see. But now, you know, I mean, really having folks with ecological and fire knowledge across the country, it's something we definitely use to our advantage.

Lawson

Inga, do you have any closing thoughts about LANDFIRE's latest update or fire science in general?

La Puma

One thing I always encourage folks to do is check out our viewer, our LANDFIRE viewer, and zoom in on an area you know well. If you see something off, please let us know. It's not easy to get everything right, but we sure do try, and we want to know if it's wrong so that we can work to make improvements in future releases. And especially now with the annual releases, there's even more opportunity to make improvements in a sort of more timely fashion. LF 2022 is just the beginning for a new level of currency within the LANDFIRE products, and we hope that will help all our users on so many levels. You know, it's important to remember that the science and data behind the mapping models are constantly being improved, and having folks in fire science who are highly experienced with LANDFIRE is invaluable. There are so many people that we can reach out to across the country that are willing to help.

Lawson

Thank you, Inga, for joining us for this episode of Eyes on Earth, where we talked about LANDFIRE's 2022 update. And thank you to the listeners. Check out our Facebook and Twitter pages to watch for our newest episodes. And you can also subscribe to us on Apple and Google podcasts. 

Various voices

This podcast, this podcast, this podcast, this podcast, this podcast is a product of the U.S. Geological Survey, Department of Interior.