Can Prescribed Fire Help Forests Survive Drought in the Sierra Nevadas
This webinar was conducted as part of the Climate Change Science and Management Webinar Series, held in partnership by the USGS National Climate Change and Wildlife Science Center and the FWS National Conservation Training Center.
Webinar Summary: Prescribed fire is commonly used by managers in the western U.S. to remove potential wildfire fuel, such as small trees and shrubs. It is thought that this act of selectively removing trees helps remaining trees better survive drought events because there is less competition for water. Focusing on California’s central and southern Sierra Nevada Mountains, USGS supported researchers are working to identify whether current prescribed burning practices are sufficient to help forests survive drought, or whether these practices could be modified to meet this goal under increasingly frequent drought conditions. A variety of approaches are being used, including remote sensing, plot-based measures, and landscape models. Such information should enable managers to develop portfolio-based resource management decisions, such as where prescribed fire might be most effectively applied, and guide additional research and monitoring determine to what degree this method can be used to prepare forests for a warmer and drier future.
Image Dimensions: 1280 x 720
Location Taken: CA, US
James Thorne, University of California, Davis; John Ossanna, FWS National Conservation Training Center; Elda Varela Minder, Holly Padgett, Shawn Carter, USGS National Climate Change and Wildlife Science Center
John Ossanna: Welcome from the U.S. Fish & Wildlife Service National Conservation Training Center here in Shepherdstown, West Virginia. My name is John Ossanna. I'd like to welcome you to our webinar series held in partnership with the U.S. Geological Survey of National Climate Change and Wildlife Science Center.
Today's webinar is titled, "Can Prescribed Fire Help Forests Survive Drought in the Sierra Nevada Mountains?" Very pertinent obviously. Also, this is the last of the drought series, so be on the lookout for the next series we're going to be holding. I can't believe it's already through with this.
Let's get started. To start things off, please join me in welcoming Shawn Carter with the National Climate Change and Wildlife Science Center who will be introducing our speaker today. Shawn?
Shawn Carter: Great, thank you. Thanks, everyone, for joining us today. Yeah, it's with a little bit of a tight throat and tear in my eye that I welcome you to our last installment of this series on ecological drought.
We have a great talk lined up today. I'm happy to introduce Dr. James Thorne, who's adjunct faculty and research scientist at the Department of Environmental Science and Policy at UC Davis. He has expertise in biogeography, conservation, biology, and ecology. Some of his recent work is focused on the success of forest thinning and tree health.
Also, vulnerability assessments for mammals, trees, and vegetation types in California. He has his degrees in ecology from UC Davis and also degrees in geology from UC Santa Barbara and environmental studies from UC Santa Cruz. Without further ado, the floor is yours, James.
James Thorne: Thank you so much for the opportunity to present to the group. Welcome to everybody online. John, Shawn, Elda, thank you for your invitation.
We'll just dive right in. I need to recognize my co‑PIs and collaborators on this project, including Phil van Mantgem from the U.S. Geological Survey and a whole list of people that you see down there at the bottom of this particular slide, many of whom, this work would not have been possible without their collaboration and good help.
This presentation comes out of a project that's funded by the Southwest Climate Science Center. We wanted to try to take advantage of the drought in California to examine whether forest treatments led to better forest condition as the drought progressed.
As much as this is a results project, or presentation, it's probably going to be more about the project itself because we're still in the process of developing our final results. It's taking place in California, Sierra Nevada. I'm sitting right now at that little Red Star over in Davis, California.
We wanted to basically see if the canopy, if we could measure tree condition to be healthier via remote sensing in areas that had been treated versus not treated. As you know, California's drought was a very intensive one and has led to the direct or indirect mortality of many millions of trees.
One of the reasons that it was a more intensive drought than many of the droughts that are in our historical record, that we can observe, is because it was much warmer. Here you see minimum temperatures and how they really climbed during the 2013 to 2016 period, at the same time that we had remarkable decreases in the amount of precipitation.
To try to take advantage of this terrible thing, we wanted to do a large mashup of different types of information. This slide is really the "Here's The Progression of The Talk" slide. We have a remote sensing component. We have a plot‑based component. We have a leaf based component, and a GIS mashup. We want to try to combine all of those.
I'll walk you through that a bit. I'm going to then talk about the results from a number of different spatial scales. There are these two flight boxes in which hyper‑spectral data has been flown every year for now five years going on six years, starting in 2013.
Each box is about 13,000 square kilometers. One of them covers from the Central Valley of California up over the Sierras and the Lake Tahoe Basin. The other includes a transect that includes most of Yosemite National Park, as you can see there.
At those areas, a fixed‑wing aircraft flew and took AVIRIS plus data, which has just under 300 bands of record, and at 18 meters. Here, you can see them laid out against the national forest and national parks in the Sierras.
We wanted to try to use those data in conjunction with other types of land management that was brought in through GIS, and in some cases with LIDAR data, to look at differences in canopy water content in places that had been treated or not.
On this slide, you can see the yellow flight boxes. The red are locations where we know LIDAR was flown, at one point or another, that might be able to be used to look at structure. On the left inset, you can see some brown and orange polygons. Those would be representative of locations where prescribed fire or mechanical thinning had been applied.
Those locations would represent places where, perhaps, trees might be less impacted by the increasing ratcheting of the drought stress than in other places which had not been treated, and therefore had a continuation of the 70 years of fire suppression and denser stands, than we might expect otherwise.
I should jump back for just a second. I guess it doesn't show on that one. Three of these sites form an elevational transect that we looked at for a local‑level approach. I tried to break up the landscape and our analysis approach into three spatial scales ‑‑ a local level, or stand level, a mesoscale, which I'll show you in a minute, and then a full landscape scale.
At the stand level, we looked at these three locations, the San Joaquin Experimental Range, Soaproot Saddle which is mid‑elevation, and Teakettle Experimental Forest, which is getting up into the upper elevations of conifers in the Sierra Nevada.
At Teakettle, in the upper level where it's predominantly red fir/white fir forest, there's a 10‑year, ongoing experiment that is overseen by Malcolm North of the U.S. Forest Service. They have these 10‑acre blocks. The 10‑acre blocks represent, the different colors here, represent different treatments that have been applied.
They have three replicates for a control, or for over‑story thinning or understory thinning, or prescribed burn with understory thinning or over‑story thinning. We thought, "OK, here's a place where this is about as good as we can get. We know what the prescriptions have been.
We know pretty well where these locations are, and we have LIDAR for this location. Let's look at that in more detail." We used it as a location to develop some of the methods for the analyses. Here's an image of the LIDAR from those locations. Each of these boxes has three lines in it representing the three plots at that location.
As you go to the right on the x‑axis, you're going to increasing height. The vertical is the proportion of the canopy that is at that height. The upper left understory thinning, we can see there's actually quite a bit...Even though the understory was thinned, there's quite a bit of material to the left of the 15‑meter height line, which in this case, is shrubs that grew back.
We can see some differences. For example, the right‑hand side prescribed burn with over‑story thinning, you don't see a hump out to the right end, which would be where the larger tree canopies would be more frequent.
We can see that the different treatments showed a slightly different structural profile at this location. This image basically is showing the combinations of three colors, or the amount of NDWI, Normalized Difference Water Index, to NDVI. You can see that it ratchets down from 2013 to 2015, and then each column, or each collection of three, is a different plot with treatments.
The message that we got when we did this was that, "Yeah, we could see a loss of canopy water by this index." By the way, the NDWI is a unitless index. We'll talk a little bit more about that later on.
We could see it ratcheting down, but unfortunately there was almost as much noise within the treatments as there was among the treatments. It was a little bit unsatisfactory with regards to we were hoping that perhaps a prescribed burn with an understory thin would come out at losing less water overall than others. We weren't really able to show that.
We asked, "Well, why might that be?" We know that things have dried down. One of the answers might be visible in this series of images. The upper row is the location where the Teakettle Experimental Forest is located, in those red squares.
You can see that those three images in green, the green stays about the same level of saturation. It gets slightly paler through time, but that's telling us that this entire area is not drying down as much as we might expect under the drought.
By contrast, the lower row, which is at our lowest elevation site in the San Joaquin Experimental Station and it's in Oak Savannah, you can see the green is really paling out quite a bit. The intensity of the red in the right‑hand image is much heavier. That's showing us that there's a much greater impact of decline in this index of canopy water at that location than there was at the Teakettle location.
We came around to noticing that the Teakettle location, it sits somewhat in a topographic bowl. There may be some groundwater dynamics that are going on at this location that buffer it, that essentially insulate it from the increasing effect of the drought.
Let's take a look at the mesoscale approach then. Here's our two flight boxes. The red and yellow little candy wrappers that are out there, since it's Halloween, those represent locations where we tried to match places that had had a prescribed burn or a mechanical thin that were available via GIS data with matching locations that had the same vegetation and slope that had not been treated.
Here's what that might look like. It's essentially a paired plot, or a paired polygon type of an analysis. Here, in the lower image, is a burn from 2008, a prescribed burn, and nearby, not terribly far away, a control of about the same area, same vegetation type, same elevation, similar slope aspect. We want to start to compare these things in some paired plot approach.
It gets a little complicated because you have prescribed burns in many different years, and you have the image years of 2013, '14, and '15. What we would expect to see, or our hypothesis, is illustrated in the right‑hand‑most column. There's 2011.
You can see that there's a burn and a control column there. We would hope that water canopy content would be...
The red shows the opposite of what we would expect. We would hope that the prescribed burn would have a higher water canopy content than the control.
For 2011, the control had a higher canopy water content than the burn. If you look at the far left, it's doing more of what we would expect. The burn has a slightly higher value than the control for each of the three years at that location.
What this suggests is that the timing, the historical timing by year of a treatment might have an effect on forests given the onset of a drought. If you did a prescribed burn and, that same summer, it went into a drought, the trees might have become somewhat stressed by the prescribed burn and be adversely impacted, whereas perhaps, if the burn was several years prior, it would be a better solution.
We were also not terribly happy with this initial set of results from burns and treatments because they seemed quite noisy as well. It may be also that, because we're going across a large elevational gradient, which we'll talk about in a moment, that the same treatments could have different health effects depending on the level of background stress that a location is subjected to.
Now, let's move up to the landscape approach. With the landscape approach, we learned that it takes a long time to process hyper‑spectral imagery to be at the point where you can look across pixels from year to year and have the same 18 meter pixels lining up across 13,000 square kilometers.
Indeed, you can see here, there are flight lines. There's about 11 or 12 flight lines in each box. We had to take each flight line, which was geo‑referenced by the Jet Propulsion Laboratory, but we had to further geo‑reference them by cutting each flight line into small pieces, locking it onto the landscape, and then locking the years from one to the next.
At the landscape scale, there's some things that I see that make me confident that we could use these data at this spatial scale. The first is on the left image here. You can see that down in the valley at the bottom of the box, the canopy water index for a single year is drier in browns and reds.
It moves into blue colors in our conifer zone, which is a more productive and moister zone. As you get to the other side of the Sierras and into Nevada, past Lake Tahoe, you can see that it dries down again.
The major moisture gradient of these mountains is captured by these images. On the right‑hand image, in the back behind those three flight lines, is the vegetation map for this area derived from other satellite imagery.
The patterns that we see in the hyper‑spectral imagery and the patterns of the vegetation line up very well. That also suggests to me that we could do things. We could look at this by different vegetation type, different forest type.
We go from oak savannas to mixed conifer hardwood to conifers, to sub‑alpine conifers. We could probably look at all of those in turn, and see what type of dynamics they display. Here's what one of our rendered images of the multiple flight lines looks like for, what I call, the Tahoe Box.
The nice thing about the NDWI is, although it's unitless, we've tried to address that, I'll present that in a minute, it does allow you to compare from one location to another and across years, because it's all on a single index.
Here's the change from 2013 to 2015. The tannish color shows the drying down. There's some speckling in that. The white locations are places where there was cloud or snow cover, and we weren’t able to calculate the differences for those particular combination of years.
Here's the same image again, now with some of the wildfires for this region overlaid on it. What you can see here is where the intense red is located on the left side of this image, is the King Fire. The King Fire took place during this sequence.
Some of the greatest drying down are within the footprint of that wildfire. We can start to bring in things like the wildfires, perhaps other disturbances, and look at them in relationship to these types of data. Now we switch over to the Yosemite Box. You can see Mono Lake in the upper right‑hand side.
Here's the 2014 and 2015 images side by side. The stars are the field locations. We're going to come back to those in a minute. You can see that the band of blue above that heavy red in the left image is paling out and getting much less pronounced by 2015. That would be the increasing sequence of the drought.
Here's the change in canopy water content. We can see that almost the entire location is moving into drier condition, rather than into wetter condition. Again, I've put the prescribed fires on there with the crosshatch black locations. They match up pretty well with the heavy duty red emerging locations.
One of the things that we're planning to do with these data are to develop a topographic model. There's a scientist named Jenifer Cartwright who works for the U.S. Geological Survey. She is using these types of imagery along with topographic modeling to try to identify where refugia are located on the landscape within the drought.
We think we can do that here. We could either create a topographic model not using any of these data, but say, "Here's where we think the refuge, the least impacted locations will be for each vegetation type, and then look at the change in the canopy water index at those locations."
Or we could actually use the imagery to say where are the locations that had the least impact and go in and look at what conditions are at those locations. That's currently the major thrust of the research.
I've talked about the remote sensing and the three scales of analysis, but we've also tried to address the issue that the NDWI is unitless, and it would be nice to know that the millimeters of water in the canopy were.
While we were doing this project, we originally were only going to process three years, but we learned that NASA decided to fly the imagery again in 2016, and indeed they're actually going for another several years.
In 2016, we fielded a field campaign to get out on the ground and try to collect information about leaf water potential and field measurements of the spectral indices of trees so that we could come up with a canopy water content that was derived in the field at the same time that the planes were doing the over‑flights.
We could link the values from the field to the remotely sensed ones and that that would get us to getting to some actual measurement of the canopy water content.
Here are some of the folks who worked on that project. We collected from these four sites in June of 2016. The Blodgett Forest, the San Joaquin, Teakettle, and Soaproot. Notice all of the tree mortality at the Soaproot level elevation. Those are an elevational transect.
We collected leaf reflectants, leaf water content, leaf mass, leaf thickness, water potential from pre‑dawn and midday, and at the leaf level and at the canopy level, leaf area index, canopy cover or gap fractions, and species and tree data. We would lay out a nine unit plot. Within that, three plots would be selected and four dominant trees would be measured in each of those locations.
I'm going to talk a little bit about the three elevational locations, starting at the San Joaquin Experimental Range, which is the lowest elevation one. Here we have two species of oaks, the blue oak and valley oak, and the lowest elevation pine.
We can see that the water potential versus the leaf water content of the pine and the oaks are quite different. Also, the leaf water content against the normalized difference water index as measured in the field is also quite different.
This is the location at which our results are the most far along. Leaf water content against the NDWI is showing less of a distinct pattern. For the oaks, we can actually identify...our three different plots show us that topographic position is an important consideration for the condition of the canopy of those trees.
You can see these three groups are broken out here, and that they have different topographic position. We were able to use that with some repeat plot data to identify mortality that is associated with the drying down of the remote sensing.
The little red circles here are showing you the nine locations that we have repeated field measurements for, and then the imagery. There is a fairly good correspondence between the percentage of dead trees and the canopy water decrease that is shown in the statistical chart.
This is particularly of interest at the low elevation in these blue oaks, because blue oaks may have, they probably have, less pathogens and pests that are attacking them than is occurring at higher elevations. At higher elevations, the predominant cause of the mortality has been beetle outbreaks.
It's hard to get to a direct physiological link to mortality, but at this lowest elevation, in the hottest, driest places, we think that we might be closer to identifying the physiological tolerance of these trees.
We jump up to the middle elevation. Here, there's five different species, and they array themselves pretty well with regards to water potential and leaf water content. They also break out fairly well of leaf water content against the handheld NDWI. In this case, we might be able to go after dynamics with the individual trees.
However our 18 meter pixel of the over‑flight is still a little bit coarse for being able to extract out individual species from the canopy as has been done successfully in the leaf to landscape project that the Sequoia National Park and Professor Greg Aznar have done in the southern Sierras. We can identify that the oaks and the ponderosa pine are really quite different with their LWC and NDWI.
At the highest elevation, this is getting up into the Teakettle, you can see the species are quite mixed with their leaf water potential in NDWI. This is perhaps emblematic at the site level of some of the problems that we were running into with regards to the noise within the plots by treatment being as high as the noise between the plots on 10‑year old treatments.
We do see that the incense cedar, the CADE, is perhaps the most differentiated, those purple dots going off to the left.
While the individual site results, to my mind, leave something to be desired, when we put them all together, we do find that the leaf water content versus the spectral indices of the species from the San Joaquin and the Soaproot and the Teakettle can track fairly well between a leaf water content of grams per cm squared and the NDWI values.
This does revert back to the concept that at the landscape scale, we have a fairly robust measure. If we had more data, we might be able to get better estimates of the actual leaf water content in the canopy. That's perhaps another step in future research here.
While I've mentioned a number of things that are still in progress, Phil Van Mantgem, who is a co‑PI on this project, has some results from long‑term demographic data that I would like to bring to your attention.
These are locations that have been measured, I think, quite a few times, and pre and post fire. Here's what the stand density looks like when you stratify by those values, and the probability of death being lower in burn stands after accounting for tree size and groups.
You can see that the probability declines for the pines after the prescribed fire in these demographic plots. Contrary to the results from Teakettle, here we're seeing that perhaps these fires are beneficial in trying to insulate the pines from the ongoing drought.
Jumping back now to the GIS data for just a few more ideas of our next steps.
We have stand treatments, we have climate data and models, we have environmental data, and building those into a landscape model and using the remote sensing to try to validate that, or using the remote sensing to build the model and then trying to translate it over to perhaps landsat are steps that we would like to take next.
There's other data. Here's a view of the 2016 tree mortality surveys, the aerial surveys, the ADS surveys, which also occur in Oregon and Washington and many parts of the United States flown by the Forest Service. In California, they’re a combined state and federal initiative.
Here in the background is a change in climatic water deficit. This is a GIS model that we produced here. The change from a standard 30‑year average to the average of the 2013 to 2015 drought, that's the red to blue in the background, and then the black is the level of tree mortality, conifer mortality that has occurred.
You can see that there are, in the Southern Sierra, down at the bottom of the image, the black is occurring in some of the reddest locations, and the black just south of lake Tahoe is occurring in less intensively drought‑stricken areas, and perhaps, there may be a difference in the proportion of trees killed by beetles versus by direct physiological impacts between those two areas.
This could be a way that we could start to identify different types of impacts in different locations.
Finally, here's that same climatic water deficit change in the background for the whole Sierra Nevada, and here are our two flight lines with the change in canopy water content between 2013 and 2015.
Looking to see how we might be able to link, or how good is the correspondence between the loss of canopy water index as measured through our hyper‑spectral, and a model of landscape hydrology and the change in the background hydrology could be another way that we might be able to get into predictive modeling about future levels of physiological stress on these different tree species or on these different vegetation types, I should say.
Now, a lot of this was censored on the hypothesis that if we are able to somehow reintroduce a land management techniques like prescribed burning to large areas, that those areas would become more resilient to these types of perturbations.
Of course, there will be barriers to those that can include the funding, and in California, air quality is a big issue because state air quality agencies regulate how much federal forest lands are permitted to burn. The timing of the burning, the site accessibility, rough terrain is often…cannot be treated.
Prescribed fire may not be sufficiently severe and hotter the droughts may produce stresses that exceed the potential management responses. These are some of the management challenges and questions.
I'd like to leave you then with, perhaps, a framework for how we're thinking about all of these, that we have forest plots with the water potential in remote sensing by experimental sites or remote sensing in landscape samples, landscape levels, GIS integrations. I talked a bit about those.
There's fire suppression, no treatment, prescribed burns, mechanical fitting, are there other techniques that we might be able to use? What trend information is critical? What condition information is important? and what future projections would be needed to pull those together?
Finally, the bottom box of some other questions, what experimental treatments, and at what scale, and what monitoring would ratchet forward our understanding so that we don't merely look at the response of perturbations or the projections of future impact.
We try to integrate those so that we have a way to advance our understanding even when management techniques might not do what we would hope they would do. With that, I'll thank you for your attention and take any questions that may be out there.
John: Thank you, Jim. I see a few people typing away. I just want to say real quick, thank you for the presentation. I also like to thank USGS for continuing these webinars series with us in the last 10 months with this particular drought‑specific series.
I want to thank everybody there, Holly, Kate, and Shawn and everyone who's participated in this and help put this on.
We will be taking a two‑month break. Be on lookout for future webinars. We have another series that we're planning right now.
We got our first question from Johnny. "Did you see any interaction with times since burn and drought effects in terms of mortality response?"
James: We mostly have been trying to look at the flip side of that, and the flip side would be the idea of resilience or locations that had less impact. There are some publications that are out there for sure. I'm going to back up.
Here's the mortality as measured across the state. Greg Asner has a nice paper in PNAS where they made projections about mortality, and these aerial detection surveys from the forest service, the annual tree mortality that the 2016 says 102 million trees. It’s probably surpassed that now. The great majority of those died by insect.
I would say, we don't have a good measure of that, but we recognize that we need it because the NDWI can be saturated by, or can be affected by the number of trees that are actually in a pixel or the proportion of a pixel that is in a green, similar to NDVI. We recognize that that is a potentially confounding factor in these measurements.
John: If anybody’s on the phone line, if you like to throw out a question, you can press *6 on your phone and that should unmute you, if you have any questions.
Toni Lyn: Hey, James. This is Toni Lyn. Great to hear this work. I was just wondering if you could talk a little bit more about any signs of fire refugia that you found. There's really neat work happening in the Pacific Northwest on this. Maybe you just know about, in general, what is being seen in the state even not just from your work.
James: Fire refugia?
Toni Lyn: Yeah. Thinking about not just managing places so they don't burn as much, but places that are naturally not burning as much. Do we put our prescribed burns there or think about putting them elsewhere in fact?
James: Yeah. Great question. Good to hear your voice. We're interested in the work that is here, these resources.
One of the things we learned last thanksgiving, almost a year ago now, from a stakeholder meeting was that, "Hey, wow, if you can just create the entire flight box and give us four years of consecutive June values that we might be able to explore various different applications of these hyper‑spectral data."
That's been one of the main focuses, and going after refugia in that location is something that we're interested to do starting from a climatic perspective. From a perspective of are there places that seem to have just had less canopy moisture loss than others, and what are the physical characteristics of those places because that will be, of course, interacting with the management of those locations.
If we can do that by veg type, then we might be able to identify some fire refugia. Fire is keenly on the minds of people in California. It's quite smokey outside here in Davis today even though the coastal fires have died down now a bit.
The combination of four years of drought and then double precip that brought the thatch up in many of our more arid systems, so there was a huge annual fuel load that was out there and then very dry conditions have led to some of the sweeping wildfires that have been in the news.
Those types of combinations in California, it makes me wonder whether, perhaps, we're moving out of our models of gradual climate change and the increasing stress that that brings, and into some of the more stochastic events that are broadly predicted under climate change.
Four years of drought and then double rain and then an extremely dry and fire season and high winds, those things within a five, six year period seem to be pointing us to those greater extremes.
One thing that I am interested to do is to try to look for spatial correspondence between measures of, say, future climate exposure like some of the climate exposure maps that we've developed in California, lay those over the annual maps and, say, what was the 2016, 2017 year, what did that look like relative to our projections into the future of increasing stress?
Was 2016-17 the equivalent of a mean condition in 2050 or something like that. We might be able to start to get a handle on what some of these outlier conditions bring to us. I guess, you could say that the places that retain the most canopy moisture might become the fire refugia, but we could go around on that one in a couple of different directions.
Toni Lyn: Totally. Thanks. There's just…Meg and some others up in the Northwest, they're thinking about this in a way that's very different than I think about climate change refugia, but taking into account topography and then looking at historical data to see places that have remained unburned, and then what's the pattern there.
It's cool. Maybe, we can all join forces or something. [laughs]
James: That would be great. I’d love to work with Meg and with you. Thanks so much for the question.
Toni Lyn: Thanks, Jim.
John: Thanks, Tony. Alright, I don't see questions in the chat box.
Once again, I'd like to thank Mr. Thorne for your presentation and, like I said, it will be available shortly on the USGS website. Be on the lookout for that if you guys know somebody that didn't have the chance to view this.
Thank you very much for participating. Have a good day.
James: Great, thank you, sir.