AoR 110: Lots of Grass but Little Fire (Yet); 2023 Fuelcasting with Matt Reeves

[ Music ]

>> Welcome to the Art of Range, a podcast focused on rangelands and the people who manage them. I'm your host, Tip Hudson, Range and Livestock Specialist with Washington State University Extension. The goal of this podcast is education and conservation through conversation. Find us online at ArtofRange.com.

[ Music ]

My guest today on the Art of Range is Matt Reeves. Matt is a researcher with the Forest Service out of Missoula, Montana, and he runs the Rangeland Production Monitoring Service and the related product Fuelcast. Matt is a repeat offender here, and I'm glad to have him back today. Matt, welcome.

>> Thank you, Tip. It's nice to be here.

>> Well, I know we've visited a number of times about RPMS, but because the purpose of this program is to track forage production and fuel loads over time, every year, there's new things to talk about. And I think there's also still, at least in my own mind, a lack of knowledge about where the data comes from. And there is at this point beginning to be an increasing number and diversity of remotely since data products that are, I think, increasingly available to land managers. More people are aware of things like RPMS and the Rangelands analysis platform. I think most people are using that for cover data. But that's becoming increasingly available to people like private land managers and not just agencies, people who wouldn't otherwise have access to some of the agency-specific data. The RPMS is one that maybe has not had quite as much airtime or has gotten out there as much as the wrap outside of the agency world. But I wanted to take a few minutes to talk about where those data come from before we get into talking about what are the actual results and what do you see around the country in terms of forage and fuel conditions. So tell me again what is the basis for the Rangeland Production Monitoring Service again, which, as I understand, is estimating the annual net primary production or what some people would call annual forage yield, you know, for some remotely sensed pixel on the ground. What is the basis for RPMS?

>> Well, I should mention first before we get into the basis that it's not as you -- We commingle a couple of ideas there. It's not just the forage component, but it represents the production of all life forms, whether or not they would be useful for forage. For example, the production of creosote bush. If it were estimated, say at 200 pounds per acre, this would not be tantamount to 200 pounds per acre of forage from the creosote bush. So we need to make sure people understand it's about productivity and it could be forage, but not necessarily.

>> Right. So moving to the basis for the production monitor. It all revolves around the Landsat Thematic Mapper Archive. And that archive is most widely available from 1984 to present day. You could go back a little further, but it gets very messy and inconsistent. And so, we mine that archive using the Google Cloud platform and other tools to mine the depository of Landsat archive satellite images and create the Normalized Difference Vegetation Index from those images. And we allow that Normalized Difference Vegetation Index, or NDVI, to estimate the annual production. And the way you do that is by calibrating that NDVI response to a wide number of observations of productivity. And the Rangeland Production Monitor has data from 1984 to 2022, as I mentioned. And we usually publish the year-end results about November 15th. And there's a reason for that. You may have some growth in some warmer climates that we don't get at that point, right? Because if we're calling it the end of the season, say in the middle of November, any growth that may occur, say in some of the warmer ecosystems, would not be captured. But that's usually negligible. In most cases, it's negligible. And the reason we like to get that product out, the annual production for the year in mid-November, is that we like to work with the Farm Services Agency to aid them in their determination of forage losses. So that's why we do that.

>> Yeah, forgive the dumb question. So are you generating NDVI values from the raw Landsat data, or are you accessing an NDVI data product that's already been produced?

>> No.

>> And maybe the question doesn't even make sense.

>> Well, it does make sense. Somebody has to produce it, whether or not somebody somewhere produces it and stores it or we do it ourselves.

>> Right.

>> Yeah. So there are several versions of thematic map or imagery floating around in the Ethosphere. You know, Google has one, planetary computer has one, USGS has one, AWS has one. And so there are many depositories of this large dataset and many different kinds of products floating around, but we create our own NDVI.

>> Okay. So NDVI is Normalized Difference Vegetation Index. Is that a greenness index, greenness over time? What exactly is NDVI?

>> The NDVI is the ratio of the red band to the near-infrared band. So in the satellite sensor world, you know, there are different portions of the spectrum that's captured by the satellite. And so, we have the red, the green, the blue, the near-infrared, mid-infrared, far-infrared. There's some thermal stuff usually. And what we're leveraging there is the fact that plants preferentially absorb red light. And they scatter or reflect light in the near-infrared portion of the spectrum. So by differencing the red and near-infrared, or taking a ratio of those, we can understand the amount of greenness that is occurring. So it can be related to a lot of things, including rate of photosynthesis. It can be greenness. It can be phenology. So you really have to qualify it and tell people what you're linking that NDVI response to because it could be a lot of things to a lot of people.

>> You've mentioned that because of that, NDVI data can be misused. In other words, the same NDVI value in different years could mean different things. How is the RPMS? How are you interpreting NDVI?

>> Well, we interpret NDVI as annual production. So it depends on what you calibrate it against. So you have to be careful with y-variable or the left-hand side of the equation. In other words, what are you relating that NDVI response to? It could be phenology. It could be photosynthesis. It could be Leaf Area Index. In our case, it is annual production. And since we calibrated it against annual production, for a given increase in NDVI, we see a given increase in productivity. But the relationship is not linear for the entire spectrum of productivity across the United States. For example, above say 2,500 per acre, the NDVI response begins to get saturated, we call it. And so, for a given increase in NDVI, we no longer see a corresponding increase in production and vice versa. So there is an asymptotic feature of that relationship that exists above say 2,500 pounds. And this is where some of the other remotely sensed vegetation indices can be useful. And there are dozens of them. But we choose NDVI because it's simple, it's widely understood, and it has stand the test of time. And the good news is only about 25% of U.S. rangelands exceed the 2,500 pounds per acre on an annual basis.

>> Right. So it would be applicable in nearly all of the semi-arid and arid west and would only not be useful once you get into the eastern half of the tallgrass prairie transitioning into the deciduous forests.

>> I think that's plus or minus, correct?

>> Yeah. And am I right that there are multiple remotely sensed entries, for lack of a better term, over the course of the year, and that's all being synthesized into the end-of-year product?

>> Yeah. There is a nominal, we call it a nominal overpass of roughly 16 days in the past anyway. Meaning that plus or minus every 16 days, the satellite or the image is capturing the same piece of land. So about every 16 days, it's looking at the same spot. But since we now have multiple instruments to query, you know, you've got Landsat-8. There's Sentinel, a different satellite, and some others. We now can have a repeat frequency, in other words, the parcel of land is being viewed, maybe say every six, seven, eight days, something like that. So our repeat time has been greatly increased. Now, as far as how we use that information, what we do is you have to have multiple overpasses in a given year. Think about the monsoon season in the Southwest. You might go two months without seeing through the clouds, right? So this begs the question now, what are we supposed to do? And that's why having multiple repeats to catch that sunny day is so useful. And the way that we use this information is not to aggregate it all. But we look for the cleanest observation that we can find out of that stack of observations at every given pixel and we choose the maximum value. And this is widely criticized because people say, well, You know, you're missing all the tail ends of the growth curve. Yes, but when we have done a relationship between the maximum and, say, a sum of all the values, you know, the r squareds are very impressive. They're usually north of 0. 93, let's say. So there's a very tight correlation between the peak, the maximum value, and the sum of the values. It's not one-to-one, but it's pretty close. The advantage of using the peak is it's simple, and it tends to work pretty well. So we just find the peak and that represents the annual production that can occur. One of the things you have to be careful with if you're playing that game is what to do in bimodal systems. In other words, systems that can have two growing seasons. I have found, yeah, and I have found after looking at the entire archive of Landsat data that the notion of bimodal systems is a little bit oversold, in my opinion. In other words, there's only about 18% of our rangelands that exhibit that character. And generally, especially more since the year 2000, generally speaking, one leg of that two-bumped curve, let's call it, one of the bumps is going to be much higher. Yeah. It's the majority. So you kind of got to take it with a grain of salt if you're using the maximum to characterize this. But especially increasingly, we see one bump dominating.

>> Is that incremental set of readings also the way that you can distinguish annual production from perennial production, I mean, the annual plants from perennial plants?

>> Sort of. There's a timing of perennials versus annuals that can be diagnostic. Let's take, for example, cheatgrass, which is generally considered to be a cheater. It comes out of the blocks pretty quick in the spring. And you leverage that knowledge to estimate the amount of annuals, but there can be some head-fakes there. For example, Sandberg Bluegrass, as you know, can co-occur pretty tightly with cheatgrass in terms of phenology. So there's definitely some head-fakes that occur. But we like to leverage not just the pattern that's seen this year in terms of whether or not it's annual or perennial, but you can allow your algorithm that's trying to sort this out to look at past years as well and relate that to the field data that basically is telling the computer, let's say, that this signal right here on these sites comes from annuals and not perennials. So it's not perfect, but there are ways of teasing it out.

>> Yeah. I'm curious about the current year. Last year, we had significantly above-average, late spring precipitation across quite a bit of at least the Intermountain West and I think the Great Basin. And a relatively cool, in some places a record cool, spring. And there was a lot of vegetation generated, and a lot of that carried over standing dead into this year. This year, we also had in much of the west, in some places, a pretty extreme wet spring, which was wonderful for a lot of locations, like California, and above average precipitation in some places. But in the meantime, it seems like Canada's on fire and we're not in the Western U.S, largely. And you and I haven't talked about this in a while. Where do you see the current situation and the trends and annual net primary production this year and what we had left over from last year?

>> Well, the hotspots that were on our radar from last year were, of course, the Columbia Basin, as you mentioned, with the tremendous production, especially cheatgrass. But I think the fire season got pretty compressed. So anytime you have those cool springs, things stay green for a long time.

>> It shortens the fire season.

>> It shortens the fire season because you have to remove all of that, let's say, wet energy, let's call it, for lack of a better term, from the system for large wildfires to occur. So the same thing happened in the Southwestern U.S. We had record yields in a lot of cases. In fact, in Northern New Mexico, there were places that got, you know, 30 inches of rainfall in just the three or four-month monsoon period. And that translated to some really wolfy yields, oftentimes in excess of 4,000 pounds. And so, we came into this --

>> Wow.

>> Yeah, I know. It was really remarkable. It blew everyone's mind. And we came into this year with a lot of that fuel in place. And, of course, the fire season has gotten off to a very slow start, as you mentioned. June saw the lowest acres of wildfire burning since the year 2000. So this June has been the slowest on record since 2000 in terms of area burned. Meanwhile, 400 locations in the United States, coterminous U. S, in the month of June, posted record temperatures. And I'm told that, last night, Albuquerque broke some kind of record in terms of the low temperatures, you know. It basically wasn't cooling off at night. Yeah. And so, some of the predictive service folks have been telling me about these weird phenomena. So this begs the question, if it's so hot in the south and we got all the fuel, why aren't there more fires? And I think there are two primary reasons we've seen that have limited wildfire expansion to date. One would be they had so much snow in the high country, in, say, the Colorado Plateau, that again, it was very much delayed. The fire season was very much delayed. And it took a long time for the snow to melt. And things are still pretty green up in the high mountains, not so in the lower elevations, but it's still greener than normal. In fact, if you were to visit Climate Engine right now and you were to zoom in to Northern New Mexico and you asked the computer to show you the greenness as a percent of normal, it would be, you know, 30, 40, 50, 60% greener than norm. So it just hasn't dried out enough yet in the high country there. That's probably going to change with the upcoming nonsoon [phonetic] as they're calling it. Predictive services has been talking about the fact that we are not going to get a monsoon this year, or at least the likelihood is low. So we may have some late-season fires that finally crop up. The other point about the wildfires and the lack of fire this year, I think can be -- We can point to the fact that when there's not very many fires, the firefighting apparatus can muster a lot of resources very quickly to any starts that do occur and catch them before they get, say, out of control. And so, I think there's a human factor at play here. We've got environmental factors that would permit large wildfires. But because there's, to date, not very much going on, they're able to muster a lot of resources and take care of it.

>> You may not be prepared to answer this, but is that a change in policy at least with the Federal agencies, or is that just a reflection of, you know, semi-local, interregional fire response agencies that are just choosing not to let stuff go? Because it seems like we've heard in the past that groups like the Forest Service and the BLM have, in some cases, chosen to let fires burn in the interest of, you know, returning to some natural fire frequency. But then because it's been so long since many places have burned, you end up burning five counties instead of a 500-acre, nice, you know, controlled burn.

>> Yeah. I don't know. I know very little about policy. But I do know that if you look right now on NC web, you will find the pass fire in southwestern New Mexico in Gila. You will see that it's, I don't know, plus or minus 65,000 acres or something. And it's been burning, I don't know, for a few months. And I think that fire, they've been allowing it just to kind of do its thing for the most part. They haven't kept up with all the day-to-day activities, but that would be an example of where the conditions, you know, the officials felt the conditions were just right that let the fire go and --

>> Because it's not a threat to 100,000 homes.

>> Well, I don't know that for sure, but it doesn't seem like it. It's in Gila, so it's a pretty remote area.

>> Yeah. That's interesting. Any, any ideas on -- You mentioned a bit of, you know, forage, fuel casting for parts of the areas of the country that receive, normally receive, some kind of a monsoon. Any ideas on the -- Do you have any prognostication for the rest of the West?

>> Well, I know in places like the Northern Plains and into especially Southeastern Montana, there has been some fantastic forage responses. You know, the Black Hills recently has really come online, and we were forecasting a pretty good year, especially on the north end of the Black Hills type of area, you know, from the Belle Fourche country to the Black Hills in that spot. It started a little bit dry. But for some reason, you know, the Fuelcast suggestion was that it was going to green up, and it sure did. So it's looking pretty good in that neck of the woods. And I think the Southwest is a bit of a conundrum. You know, from what the meteorologists are suggesting, our chances for monsoons are very low. And it doesn't seem to me like the conditions are setting up for good activity. So we'd see probably some muted growth there. And I know the central plains, especially last year, as you know, really had a hard time. It really struggled. Reductions were well in excess of a 50% drop in forage, say, in the country like western Kansas, northern Oklahoma, and those places. They're getting a little bit of rebound, but it's too little, too late I think overall. So I don't think the conditions are very well primed on the Southern Plains as they are on the Northern Plains.

>> Well, that's a fairly encouraging report. Lots of grass, not a lot of fire.

>> Well, there's not a lot of fire at the moment and though we tend to just evaluate. I'm a non-forest scientist. But I keep tabs on, you know, forested regions because I live in one. And when we look at long-term drought and when we look at, you know, the number of days say above 90 degrees and when we look at humidity recovery and those sorts of indicators, you know, the Northern Rockies, especially up towards the Canadian border, is extremely dry. They had a very bad snowpack. In fact, you might have seen on the news that our governor was working with some other folks to ask that more water be let out of Hungry Horse Dam to fill Flathead Lake because it's really in a drought. And they wanted more water. It seems like so far the request was denied in part because The Middle Fork of the Flathead River was at record low. So this part of the Northern Rockies, you know, say north of Missoula, is very dry. And I would be surprised if they didn't have some kind of fire activity, you know, sometime soon. Time will tell.

>> Yeah.

>> But it's very dry out there, as is southwestern Oregon. You know, there's the fire that recently took off there. And it's been -- If you look at any drought map, that neck of the woods, you know, just near the California border, has been incredibly dry and windy. And so, we see a result in the wildfire pattern that we see down there today.

>> Yeah. And places like California that had tremendous growth in the spring are just now beginning to get hot enough and dry enough that it's the beginning of a fire season probably.

>> You'd sure think so. You know, it's July 21st. And who knows what the true fire season will be there? But you're right. I think, you know, looking at the data, the data suggests it's finally starting to cure pretty well. And they'll have a shortened fire season, for sure. I mean, it's July 21st. And normally, by this time, it would have already been, let's say, cured. The grasses would have been cured for at least a month by now in a lot of the Northern California area. But, you know, fires can still continue into September and even into October at times. So we'll see.

>> For those that haven't listened to your reading the tea leaves or haven't listened to previous podcast episodes on the stuff that you're doing, what are some tools and links for people that don't work for an agency that are interested in trying to understand forage production and fire risks? Yeah. For the average Joe, where do you go to learn more about this?

>> I tell people, if there is one tool to learn and to use, it would be Climate Engine because it has weather from about 40 different sources. It has climate. So it has the climate normals, and you can see what's normal. And then you can map the deviation from normal; for example, how wet was this June compared to the 40-year average in terms of percent or, you know, millimeters or whatever. It also has remotely sensed data in there. It's extremely fast to use. I mean, I can make a map of greenness departure in there in two minutes. And you can look at a trend. So it has graphing abilities. It has mapping abilities, and it has just about every variable you'd need as far as a one-stop shop is concerned. I usually send people to Climate Engine. But we also like to look at the Crop-CASMA Soil Moisture Viewer because soil moisture, as you know, drives a lot of our patterns that we see. And so, you can go to the Crop-CASMA Soil Moisture Viewer and just look and see what's the soil moisture look like. And that's coming from a satellite. So it's a real observation of moisture. And we use that to a pretty great effect, especially early in the season, to look at the trend of soil moisture. And if it's going down, like in 2022, in February and March, we were reporting on abysmal conditions, especially in Central New Mexico. And lo and behold, we had, you know, record wildfires there. So you could see the trend in the soil moisture and almost say this is something we've never seen before. It's dry. So those are two tools that I like to send people to. There are dozens. In fact, we have interacted with more than 40. But I think overall, Climate Engine's the best.

>> Okay. Yeah. I made a note of that when you mentioned it the first time because I don't think I looked at it before. Who produces the Climate engine?

>> Well, the Climate Engine is a product from University of Nevada, Reno, and I hope I'm not missing anyone there. And if I am, I apologize, but I believe it's linked to the UNR. And it's on-demand cloud computing. And you'll have to have a Gmail account before you can use it. And why do you need one? You need one because if you're going to download anything, it sends it to your G-drive. So to use it, you're going to have to have a Gmail account. There's also a pay-to-play system there that I've never done anything with. You'd have to call them and find out what's on there. But in my experience, I've been doing this kind of work probably 25 years or so, it's one of the most impressive tools I've ever seen.

>> Yeah, I'll have to take a look at that. And who is the target audience for your range production monitoring and the Fuelcast?

>> The production monitor is so far almost exclusively used for our capacity modeling through the Forest Service. We use it to support -- We have seven right now, seven current projects looking at the productivity and what it means when we look at the distance from water and the slope and those sorts of hollow check components, let's call them. So we use it quite, quite extensively for the purposes of NEPA, to look at differences in, you know, if we want to change some kind of water distribution, what kind of forage would be availed if they did that. And so, we use the production monitor to understand what changes in management might look like. But we also like to use it to look backwards for the Farm Services Agency. And we develop reports for them that aid them in determining forage losses. So we've done that in California, in Utah, in Nevada, in Arizona quite extensively, Colorado at times. So that's what the primary uses so far has been for the production monitor. For the, the Fuelcast work, we work almost exclusively with the fire and fuels folks. And the Fuelcast looks forward four months. And it uses the daily precipitation. It uses a plus or minus daily drought metric. And it uses that Landsat archive and the current Landsat imagery to, you know, kind of coach the computer on what it thinks those conditions might mean for future yield. For example, when we've had two inches of rain in May in the past, where did we end up in terms of yield. And the computer learns from that. And we run it about once a month. This year, we took a pause. But usually, we run it about once a month. And that information is used. I condense it for predictive services quite often. And we talk about the growth of fine fuels, especially in the Southwestern U.S. because it's one of the first places to green up, as you know. And we also use -- There's one more component. We also use the production monitor, and we take from that and create a standing dead product, or it's a leftover, let's call it, the brown stuff in the spring. And we use that. That's the nest vegetation. We use that and quantify that for fuels mapping purposes. So if we think there's a thousand pounds left over, then we'll adjust the fuel models accordingly.

>> Some of that sounds a bit like the Grass-Cast. And I can't remember whether or not we've talked about that before. Is Grass-Cast a product of, excuse me, University of Nebraska in Lincoln, and are they using your data, or are they using some other dataset to generate that? Because it sounds a bit like this retrospective forecasting.

>> Yeah. The Grass-Cast is a different type of situation. The Fuelcast is a deployed artificial intelligence application that learns as new data comes in and then decides what that means for future yield. Grass-Cast is based on a fundamental model of plant growth and soil moisture and things of that nature. It is not a product of Nebraska Lincoln. Initially through Colorado State University and the Climate Hubs, it's now a widely dispersed or widely viewed dataset by a number of those organizations. And the Grass-Cast uses the forecast weather to determine, okay, based on where we're at right now, what would an above-average rainfall situation give us in terms of yield, let's say. So it's a fundamental model that, let's say, grows the plants in a way versus a strictly empirical approach, like Fuelcast, that doesn't care about any of that. It just wants to know what the yield is and what the effect of the weather has been in the past. And one of the interesting things about the artificial intelligence method is climate is changing and land use, land cover changes. So if those two things are not stationary, then what two inches of rain in May meant 40 years ago for yield is not the same.

>> It's different.

>> It's different. And I think that this is getting complicated enough that human beings are having a hard time understanding all the intricacies, which is why we decided to leverage a computer and the artificial intelligence to learn about all that and then make a prediction. So two totally different approaches to [inaudible]. Somewhat of a similar response.

>> So Grass-Cast is predicting a forage yield based on these long-range outlooks from NOAA or a variety of predictors?

>> It does. And it uses, you know, things like the soil properties, you know, sand, silt, clay, and other things. So it's really a fundamental model. But like I said, you put the water in the ground and then the grass grows. Yeah, it could use forecasts from NOAA or just about any other organization I think. I'm not really sure which data pipe they use, but it certainly could be NOAA.

>> Well, my guest again today was Matt Reeves with the Forest Service talking about Rangeland Production Monitoring Service, Fuelcast, and current conditions. And I look forward to doing this again and maybe having some more discussions about these decision support tools in the near future. Thanks, Matt.

>> Thank you, Tip. It was good to talk with you again.

>> Thank you for listening to the Art of Range podcast. You can subscribe to and review the show through iTunes or your favorite podcasting app so you never miss an episode. Just search for Art of Range. If you have questions or comments for us to address in a future episode, send an email to show@artofrange.com. For articles and links to resources mentioned in the podcast, please see the show notes at ArtofRange.com. Listener feedback is important to the success of our mission, empowering rangeland managers. Please take a moment to fill out a brief survey at ArtofRange.com. This podcast is produced by Connors Communications in the College of Agricultural, Human and Natural Resource Sciences at Washington State University. The project is supported by the University of Arizona and funded by the Western Center for Risk Management Education through the USDA National Institute of Food and Agriculture.

>> The views, thoughts, and opinions expressed by guests of this podcast are their own and does not imply Washington State University's endorsement.