AoR 25: Brady Allred & Matt Jones, the Rangeland Analysis Platform (RAP)

[ Music ]

>> [Background 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 guests today are Brady Allred and Matt Jones, a couple of range guys out of Montana who are largely responsible for building the Rangeland Analysis Platform. Brady and Matt, welcome to the show.

>> Thank you Tip.

>> Thanks for having us.

>> I'm really looking forward to talking about what I think is one of the coolest tools in range management that I've seen in a while, this Rangeland Analysis Platform. But before we get to that, I'd like listeners to know a little bit about who you are. Brady, you work for the University of Montana in Missoula?

>> Correct.

>> And what do you-- what do you do at the University of Montana and what was your pathway to becoming a range scientist?

>> So I'm a rangeland ecologist here at the University of Montana. I'm a professor in the College of Forestry and Conservation. I have been here since summer of 2013, so six years now. Absolutely love it here. It's a fantastic place to live and to work. My-- I guess my pathway or my background is, I grew up in southern New Mexico, in Las Cruces there in the Chihuahuan Desert. And really what pushed me into kind of rangeland ecology is I had opportunities while I was in high school to be a field assistant and to work as a field assistant for various kind of field biology jobs. And I worked a lot at the Jornada Experimental Range on the range itself and then while going to school there at New Mexico State University I worked at the Jornada Experimental Range and was employed by them. And so I was exposed to some fantastic people, some fantastic opportunities to do research, to learn about rangelands. And so that's really kind of what put me on the track I'm on now.

>> Good. Matt, you're also at the University of Montana I think and you're the technical guy. At least when I read your profile page on the university website I get lost in flux towers of vegetation optimal depth validation and optical infrared comparisons. So how did you end up in the numerical Terradynamic Simulation Group instead of the Starship Enterprise? [laughs]

>> Well, yeah there's far too many technical terms I guess on that page, I'll have to think about that now. So I started out really as more of a kind of-- I've always been fascinated by maps and interested in maps ever since I was a kid and I kind of-- I came out of college and basically got into the idea of I want to make maps and I want to learn about GIS software so that was sort of what started me down that path. And then I came out to Montana and started working for a group that was doing some research in Yellowstone National Park and worked with somebody who was a remote sensing-- I'll call him almost a guru-- early back in those days where he taught me so much about what you could get from airborne imagery, from satellite imagery, and that really just opened my eyes to this idea of wow, maps could be so much more and include so much more information using this data. So that interest basically led me to do some more research and to pursue a master's degree and a PhD here at the University of Montana within the numerical terodynamic simulation group which was really focused on incorporating remote sensing data to monitor and to develop vegetation products on global scales. And so that really sort of sent me down the path of really being a remote sensing sort of technician and a remote sensing scientist that I then thought, wow, I need to apply this in bigger, broader scales and so I went on to get my PhD in ecology. So it was the merging of those two things, of remote sensing and ecology that's really brought me to where I am now.

>> Yeah. I like that. I have some experience with on the ground monitoring on rangelands mostly using some of the methods developed by the Jornada-- Jeff Herrick and his team-- as well as some other methods. But you know mostly focused on trying to understand vegetation response to livestock grazing, whatever the response might be. And then we can try to call it good or bad. You know we've been saying for decades, at least with monitoring, that you can't measure everything because of, you know, because of the barriers of landscape size and heterogeneity which really characterizes I think most of western rangeland. So we give up on trying to measure specific things and just take pictures because a picture captures a lot. But it doesn't really quantify it. I think I read something from Ken Sanders, University of Idaho, a few years ago recommending more extensive rather than intensive monitoring precisely because the landscape is really variable and precision is difficult to achieve. And even if we're measuring, or monitoring intensively. But you know maybe that's not true anymore that we can't measure everything now that we've got tons of readily available satellite data. What do you guys think about that?

>> Well I think you hit it kind of right on the head that, you know, absolutely ground based monitoring doesn't scale in space or time. There's really just-- there's not enough people out there to do it. There's not enough resources to get it done everywhere that we need to get it done. And you just can't do it through time. And so I think what we're doing now is we are entering a kind of a new era and this era is really unprecedented and it's not strictly limited to rangelands. We have this idea now, or we live in this world where we have more satellites up in space than we've ever had. Some of those satellites take very fine resolution imagery. Some courser resolution imagery. We live in a day where we have so much data sets that describe the world around us, whether that's weather data sets or topographic data sets or, you name it, we just, we live in a very data rich community now, time period. And I think that's starting to bleed into rangelands and rangeland science, rangeland ecology, and specifically rangeland management and conservation. In addition to the things I described earlier, we live in a time where computational power is abundant. Cloud computing and the ability to process massive, massive data sets, whether that's stacks of satellite imagery or weather data sets or whatever it may be, we can do that now where just five or 10 years ago that was a little bit more of a dream. Or cost prohibitive. Or just we could do it, it would just take years for the analysis to run. Or years to get the data to your computer--

>> Right.

>> so you can do the analysis. Now all those barriers, those hurdles, are not entirely done away but are largely removed to some degree. And so I think with this new era we're in, we're seeing it come into rangeland monitoring, rangeland management and conservation, and I think we're going to see a lot of changes happen. We're going to continue to do things the way we've always done them but we're going to build new tools, build new processes, build new-- have new frameworks, new ways of thinking, that are going to incorporate all these new things we have into our everyday decision making regarding rangelands.

>> Yeah Jason Carl [assumed spelling] and I talked about this a little while ago on a podcast episode and he mentioned that one of the things that-- one of the things that we're getting with some of this big data is that there are, you know, there are known unknowns and there are unknown unknowns and, you know, we're getting, in some cases, answers to questions that we didn't even know we needed to ask. And some of those things can be a pretty big deal. I don't have a specific example but, you know, when you're able to analyze larger data sets like that, it may reveal answers to questions that we weren't even asking. So how did-- if, for somebody who has not yet heard of the Rangeland Analysis Platform, can you, how would you describe that to somebody who's never-- who does, has no idea what this is? Well I'd say, Tip, the wrap is essentially it's an easy to use online tool that allows anyone to monitor rangeland vegetation really over any area they choose. And that was-- the goal of that essentially was to make this very accessible to people, whether you're a land manager, whether you're a rancher that really spends almost not time on your computer, right, except in the evenings when you're coming in after being all day out on the ranch, to have that rancher still have the ability to get on this website and have a real quick method to have a look at their ranch through time. And so that was really the goal was to make it this easy to use online tool that's available to everybody.

>> One of the things we were really interested in was bridging that last gap between science and application. I think a lot of the times that gap remains there, remains unfilled. Not all the time of course. You know lots of people are good at bridging that gap and do a fantastic job doing it, but a lot of the time science gets done and it stays at the university or it stays on the shelf or on our computers in our offices. And the Rangeland Analysis Platform, you can kind of view it as two things. You can view it as there's the data that goes into the platform. But then there's the platform itself, the web application. And right now there's only one data set on there. We plan to release future data sets in the coming months. But we can bridge that gap by, again just exactly as Matt was saying, someone can get on and use these data sets where before your normal, everyday person, whether it's a range hand, whether it's a landowner, or even maybe it's even a graduate student or a scientist, they might not be able to know exactly how to interact with these data. And in some cases these data sets are massive and just getting them to your computer, wrangling them, can be a tough job.

>> Right. Even for people that maybe have a GIS application at their workstation, which is fewer and fewer of us. I played with that for a little while. I came to the University of Idaho, graduated in 2000 and I was probably the last class that did not have, you know, basic GIS as their undergraduate requirement, and so I took a GIS class at Central Washington University-- I live in Ellensburg, Washington-- and because I like mapping and like data and like, you know, bridging that gap between science and application, but I pretty quickly found out that unless I'm going to spend most of my time, you know, fiddling with GIS applications, it's not quite worth the cost to maintain a license. And so you have somebody else do that. And I think that largely describes most people's situations. Of course, and somebody like a rancher would have no access to something like a GIS application. So I think this is-- I really think it's a pretty big deal. What's been the reception so far?

>> The reception's been great. We've heard from the full gamut of folks out there that are using it that appreciate it and that find it easy to use and they get back to exactly what you said where they didn't really have a means to look at the data that might be out there. They didn't have a means to especially sort of analyze it through time. And now they have sort of this quick and easy ability to do just that regardless of skill level. So, which is, really a critical step. And I can speak to that-- it's a bit of what Brady was saying before of this sort of new era where now we've entered a stage where, as someone who's worked with remote sensing data for, you know, 20 years or so, I spent you know 80 % of my time preprocessing that data, getting it ready to do analysis on it, and then I'd spend that last little bit of time doing the actual analysis and have some results which left almost very little time for the delivery of that to a user that might actually implement it. And this new era--

>> Right.

>> computing with loads of data and loads of interfaces that you can access online, it's flipped it on its head really. Where I now just spend 10 %-- 20 % of my time on the preprocessing and then I can dedicate the rest of that time to analysis and delivery which has really created that new era that we're in.

>> It really-- yeah it really lets us focus on the question.

>> Right.

>> And ultimately getting an answer to that question as opposed to focusing on the process and the procedures and the protocols and everything else that went along with that before. We can now devote most of our time to answer the question.

>> What kind of data is on the RAP? For people that have not seen this and aren't familiar with it, if somebody goes on and, you know, outlines an area, describe what they're going to get back in terms of data that they could then do something with.

>> Yeah so I should probably step back and say the RAP or the Rangeland Analysis Platform, it's, if you go to the website rangelands.app, that's dot A-P-P, that's where you'll be able to find it. And the-- when you first go there you'll land on a landing page and there's some information-- there's a fact sheet on there. We have some recorded webinars. There's actually also a very detailed user guide that you can read and go through that explains step by step not only how to use it but also how, you know, what is it used best for and what are its limitations and everything else regarding the RAP. Now the data set that's on there right now currently is a vegetation cover data set and it's specifically cover data %-- % cover for the common rangeland functional groups-- perennial forbs and grasses, annual forbs and grasses, shrubs, trees, and bare ground. The geography that this data set covers is from the Great Plains all the way to the Pacific coast. So it's a very large geography. And then the time period, we produce annual estimates of these functional groups starting from 1984 all the way to present. So our last one was 2018. And going forward into the future we'll continue to update this particular data set annually. There'll be one for 2019 as we roll into 2020, probably January of February we'll put that on up there. And so that's the data on there right now. So when you go there, someone can visualize that data on a map. It's built on a Google Maps interface. They can zoom in to their specific area of interest or their location and they can look at the spatial variability of these functional groups in that area. And then they can look at the timber variability as well. They can see how things have changed through time and across space. So that's kind of-- that's kind of the neat aspect of the RAP is that it allows you to one, easily visualize that information without downloading a thing.

>> Yeah that is really interesting. One of the things that stuck out to me from Nathan Sayre's excellent book on the history of rangeland science was the extent to which environmental variables like precipitation and inter annual variation in precipitation drive vegetation changes on rangeland maybe even more than management. And on some landscapes-- large landscapes-- that I'm familiar with in central Washington, I've pulled data, you know side by side data from a grazed and ungrazed piece of ground and find that the patterns of vegetation change over time are almost identical responding almost entirely to precipitation and not to any grazing variables. I'm curious if that is something that you guys see in other places.

>> Yeah, definitely. We-- the climate you know is obviously just this very large scale primary driver of what's happening with the vegetation and the landscape. And when we were developing this data set to try to essentially predict these % cover values for all of these functional types, it was very apparent from just what you were saying and through other things throughout the literature that precip was going to be a pretty key factor. And so we do incorporate that into the model where the prediction of % cover is partially driven by precipitation and multiple sort of divisions of that data. So saying the amount of precipitation in the spring, summer, and fall, the total sum of the precipitation, and then sort of a water deficit measure even too that basically says how much precipitation versus evapotransporation was happening across the land's surface. And those are key drivers to determine what's happening across the land, especially in our model.

>> And where does your precipitation data come from? The reason I'm wondering is I know some of these-- there aren't that many, you know, rain gauges out there in place and some of that has to be modeled in some of the landscapes that I'm familiar with. The precipitation I think is relatively accurate in terms of, you know, say there's a precip-- say there's a rain gauge in Ellensburg at the airport and Ellensburg receives eight inches that year, but you know at 3000 feet elevation 25 miles away, precipitation might have been 14 inches but there's no rain gauge there and I think your model is showing relatively accurately how much precipitation was there. How does it get to that?

>> Well we can't take much credit for that. That's developed-- that's the [inaudible] data set that's behind it--

>> Okay.

>> that we bring in and use and that comes out of the University of Idaho. John Abatzoglou I think is how you pronounce his last name.

>> Yup. Yup.

>> And he has produced this data set and essentially what it does is it's used all of those weather stations you're talking about-- and I'm not an expert on climate modeling here by any degree but I'll do my best-- but he basically uses those weather stations all throughout the U.S. and is able to sort of model the variability in between those weather stations by bringing in all sorts of other variables where he then gives us this gridded product that for every four kilometer pixel on the ground you have an estimate of daily precipitation. And he also produces a host of other climate variables as well for that gridded data set. So that's where really he's the more of the climate modeling side that we're taking advantage of their expertise and bringing in their data to inform our model and then also provide it to folks via the web-- via the RAP.

>> I see. Yeah it's pretty impressive. So how do you get cover values? What are the satellite data sets that are going into that and how do you convert it into, you know, % perennial grass or % annual? Is this some kind of a greenness index you know where annuals are green for a shorter period of time? That obviously requires, you know, several different pieces of satellite data to put that together. How do you get to those values?

>> That's an excellent question, and there's lots of things that go into it. I think first we have to state that the ground based monitoring data is actually the cornerstone of the data set on the RAP. Without it we would be unable to build that data set. And when I talk about ground based monitoring, you know, rangeland scientists, range cons, ecologists, managers, have been collecting range data forever. And you know you can pull into any field office or extension office or scientist office and they got a pile of, you know, data sitting in their drawer or on their computer. We've just been doing that for a long time. It's something that we are actually the best at. Our discipline has spent so much time and effort in training people in plant identification, in training people how to collect data and, you know, using Trans X and the various methods and how to analyze that data. We do that very, very well. A couple of years ago-- well, a ha- many years ago now there was this idea that, you know, we need to collect standardized data sets across all rangelands using standardized protocols and methods. Using the same methods. And so we can get an assessment of these rangelands. How are they doing? And out of those conversations many years ago, various kind of national monitoring efforts were implemented. And the two that we use exclusively is the NRCS NRI data collection and the BLM AIM data collection. I think most of the listeners will be familiar with those two programs, two monitoring frameworks. They're very similar. Most of the listeners probably actually participated and helped collect those data. And if you have, Matt and I will just say right now, thank you very much.

>> Thank you.

>> Those data are invaluable. And so we've taken those data sets and again they span that same geography from the Great Plains to the coast. So kind of all western rangelands. And there's more or less, now there's about 40,000 plots that have been collected over the last 15 to 20 years. And so it's the most extensive rangeland monitoring data set out there. It started in 2004 and is ongoing. And so again that's the cornerstone to what the data-- the cover data is built on. And I'll let Matt explain kind of how you take the jump from on the ground data to these larger data sets.

>> Thanks, Brady. So really, yeah it comes down to you've collected data over a plot in one location and that's giving you an estimate of the ground cover in that location. And then let's say you're always in a pasture or something and you move another-- you know you move to the 10,000 acres away because you're going to the other corner of that large pasture and you take a plot measurement there and that gives you an estimate of the cover in those two places. But the goal here was to try to figure out well how do we get an estimate of what's going on in between those two plots that are, you know, spaced out apart, and then how do we get that through time? And so basically that was the idea of building what they call a model, right? Where it basically says well, we can take a bunch of satellite imagery and climate data and other things over that location where we've measured that plot and figure out how those satellite measures and how that climate relates to the plot we've measured. And then we can take that relationship and apply it now to all those spaces in between those two plots. And so that's essentially what we did is we said-- we had this great collection of over 31,000 plots, field plots, collected via the NRI and the AIM protocols. And those are fantastically distributed across the western U.S. So then we used the real power of this new era that we've been talking about of access to unprecedented satellite imagery and cloud computing platforms where I basically said, okay let's just use every possible predictor that we can to see if we can predict what's over these plots, what's been measured there. And so I used what's called a random forest algorithm which is essentially like a machine learning algorithm that iteratively just does its best to determine how it can use those predictor variables to predict what's on that plot. And from that you get a model that you can then apply to sort of every pixel or every location across the west. And so that's the sort of real sort of basic version of that is you're using all of these plots to really build a model that takes advantage of the full extent of all of this data that's available to predict what's on the land's surface.

>> I would guess then once you've used the model to make that transformation from satellite data into cover values then you go back out and ground truth that in places where you didn't have NRI plots? Absolutely. And there's a couple different ways of doing that. One of the more common ways is just as you're building that model you can hold out data that doesn't go into that model. And then you test on that data that was held out. And so we were able to do that. And that gives you a measure of the error of the model because you have to remember--

>> Right.

>> this is a model, right? And even though we are in this new era of big data and everything else, we still have error associated with these models. And in fact anytime you collect vegetation, whether it's on the ground, you know, with a tape and a clipboard, or whether it's with a satellite, there's actually error involved. And so it's important to account for that error. And so yeah, we can account for that error by holding out certain, a certain number of those plots and then comparing the predicted result versus the actual result. And those errors are on the web page for the RAP and you can view them. And they range from anywhere from, you know, 4% to up to 10%. So you kind of have this range of confidence between 4% and 10% depending on which functional group that you're looking at. In addition to doing that, we also collected or collaborated with people who have collected data in various areas across the west. And these were independent data sets. And they were, in some cases they were collected using different methodologies than say the NRI or the AIM programs. And we compared the outputs of the model to their collection. And again they kind of fell within the same range of those error estimates from, you know, 5 to 10%. And that gave us, you know, good confidence, yes this thing is generally working well. It's giving us the answers that we would expect.

>> Um hum. What exactly is a satellite data set that gets you to, say, perennial grass cover?

>> So--

>> What's the satellite measuring?

>> Well that's-- so there's, there's a host of things that go into that essentially but from the satellite specific perspective, say the land set sensors that we're using, we allow the model to sort of parse through these over 250 different variables to determine what's working best. So in the case of the satellite imagery, we're giving it just simple surface reflectants. So the reflectants of the red band, the green band, the blue band, the near infrared off the surface. We're then calculating standard vegetation indices from that data which is a lot of listeners would probably be familiar with say NDVI, the Normalized Difference Vegetation Index as sort of a measure of greenness. And we give it a huge host of those vegetation indices as well. And then there's also something called-- we won't get too specific here-- but called tasseled cap indices. And these are all different ways that remote sensing scientists over time have developed to try to measure the vegetation on the land surface. And so what we do is we say let's take this entire suite of satellite imagery of data and feed it into the model and let the sort of the model and the machine figure out which is the best predictors for each land cover type. So say for your question about perennials, well we basically determine there's a set of 40 variables that best predict perennials. And then there might be-- there's the set that best predicts annuals and it's probably some of the same ones. There's also different ones in there. So there's another set of 40 that predict annuals better than perennials. And then there's another set of 40 that better predict shrubs versus annuals etcetera etcetera. So you end up with a mix of different variables predicting each functional type.

>> And those variables include, you know, say multiple images for lack of a better-- multiple sets of data within the same season? Because like if you, you've got an image or a set of data from April 15th I would guess that it would not very accurately separate out annual from perennial but if you had data from August or, you know, maybe June 15th where annual plants have turned color and perennials have not, is that how that works?

>> Yeah very similar to what you're describing. What we bring in, say for a given year where we have a plot that's measured on the ground and we're trying to predict what's on that plot, we bring in sort of this data in seasonal segments. So we bring in that--

>> Got it.

>> the NDVI for spring, summer, fall, and winter and we bring in the reflectants of those different bands for each season as well so that way we're capturing that phenology that you're talking about right there and that-- yeah to predict annuals you might see that, you know, the spring indices are actually some of the more powerful indicators while predicting perennials you might see that the summer values are stronger predictors of that as well. So you bring in that seasonal component to track the phenology of those different vegetation types.

>> Right. I think you-- I think it says on the website maybe that the pixel size is 30 meters. Is that currently the limit of the spatial resolution of the model?

>> It is. yes, and the analogy that I've used over and over again that people probably are sick and tired of hearing from me is that 30 meters is about the size of a baseball diamond.

>> Yeah.

>> More or less. And, you know, I actually counted it up the other day and when we do these predictions for the covered data set, we are predicting on eight billion 42 million pixels across the western United States. And that's just for a single year. So if you multiply that by 34 or 35 years, it comes out to a very large number in the billions. And so when we're talking about big data, you know, this is big data and it's only possible through these, you know, computational and technological innovations that have happened in the last couple years.

>> Yeah. Is that spatial resolution a function of the data sets that are most useful like the land set imagery or is that an optimal spatial resolution based on the model's ability to come up with cover values?

>> The former that you said. It's based on the land set imagery.

>> Okay. Yeah, what are some limitations of the technology? Like if somebody wants to go out and make land management decisions based on vegetation change that they are seeing evident in the RAP data, any cautions with that? Limitations? Technology? You know weaknesses that you know about that people should be aware of in applying this?

>> Yes. And you know it's interesting, when we were building this, we really built this to empower the user or to empower the decision maker with more information. It-- we wanted to be able to sort of fill in those gaps like Matt said, right? If you have a management unit that's 20,000 acres and you've got a handful of plots out there through time, you're really only measuring, you know, less than .005% of the landscape. And so we wanted to fill in that rest of that 99.995% of the landscape. And so we-- when we built this it's meant to be another tool in the toolbox, it is not meant to be the end all answer for everything. And really it's meant to help the decision maker, the user, the range con, the rancher, whoever is using it, to think more critically about the question they are asking, to think more critically about how do I solve this problem or how do I improve this management, or how do I change this to get that? It is not meant to just be an end all answer to everything. And as such it's not intended to be used in isolation. You know the user has a responsibility--

>> Right.

>> to examine the data and the results that they get from the simple analysis on the RAP and say does that, you know, correspond to what I know is out there? Does it correspond to the data I have? Many times it will. I'm not going to say all the time because there are some cases in which it won't. And so again we're hoping that people take this and they think more about the problem that they have or the solution that they're trying to find. They think more critically and think how can I use this information to fill in the gaps in what I don't know.

>> Right.

>> And I will add one thing to that though, Brady's point of, you know, it's critical of having some responsibility of the user to examine that data and say is this working, you know, how I think it should be working in this location? But at the same time, there's this sort of intrinsic difference between what we're providing which is aerial cover which is the cover from, say, a plane or from a satellite looking down on that land surface versus someone standing out there in that rangeland and looking out across the land and having sort of their estimate of what the cover might be for shrubs or trees or whatever it might be they're looking at. So there is a bit of that, I'd say maybe education with, that provided in the user guide even we have these points outlined that say yeah from the perspective of standing on the land and looking out it might seem like there's loads of trees or there's loads of shrubs and the cover's very high. And that's something I like to call the savannah effect where if you're standing there and there's a handful of trees, they tend to look like they're covering a lot more area than if you were to look at them down from the-- like a plane or a satellite perspective. So there's that combination of--

>> Right.

>> does it give you what you're expecting? And then if it doesn't, think a bit about your expectation and does, you know, does your expectation match what would be viewed from a satellite or an aerial look?

>> I'm interested-- I think one of the things that people would like to use monitoring for, even people that aren't doing anything right now-- and again, you know, in my opinion one of the main limitations with ground based monitoring is, you know, you measure three spots and extrapolate that to a 10,000 acre pasture. One of the things that's used for is to be sort of a canary in the coal mine, you know, to pick up some early warning signs that there might be impending degradation that's going to happen in response to say a loss of perennial grasses. And one of the things you mentioned in the paper about this is detection of susceptibility of invasive plants. And something like an increase in bare ground might be a leading indicator, rather than a lagging indicator, of the possibility of invasives moving in. How accurate is the bare ground increase and are there any other measurement s that might be used as a leading indicator of damage rather than just something that's being measured after the fact? I realize the intent with this is not to be necessarily that kind of monitoring, but if there's landscape scale information that could be used for that purpose, or at least in combination with, you know, with other information that people are collecting on the ground, what would those things be?

>> Yeah I think that this is an excellent discussion and I think one of the things we haven't talked too much about so far is the application of these data and, you know, what can they be used for? And I'm going to story tell a little bit here I guess. I think what's neat is these data are multi scale. And so we have these cover estimates across entire western rangelands and it allows us to look at cover at multiple scales. And what I mean by that is you can look at the cover of these function groups, say at a national scale or at the western rangeland scale. And by doing that you can see trends or you can see dynamics or interactions that you wouldn't be able to get if you were just looking at a local regional watershed or even a state scale. And one of the things we like to talk about is when we created this data set we started doing that. We started looking around and say, okay what's happening out there on the rangeland at a very broad, you know, national [inaudible] scale? And one of the things we notice, which isn't surprising until you actually calculate the numbers, is the amount of increase of woody plants across the Great Plains. It is astounding that-- how much woody plants have increased over the last 30 years, particularly in the southern Great Plains, in Texas, Oklahoma, and Kansas. There is a threat there where we are losing our herbaceous rangeland vegetation to invasive woody plants. We don't attempt to, you know, identify a mechanism or anything like that, we're simply just looking at the changes in the data, but it's astounding. And it's important that you kind of step back and you do that because at this national level, at the national scale, decisions are made. Where to allocate resources, where to launch initiatives, where do we need to improve our rangeland management and conservation at the national scale. And data sets like these can help us to identify those areas. It can help us to target where do we need to put on the ground conservation? What are the most impacted areas? And so you can zoom down in and you can do the same thing at a regional scale. A couple of months ago we were contacted by Bloomberg and they were doing a story about cheatgrass in the great basin. Everyone knows about cheatgrass. And they asked us to produce a map showing the increase in annuals across the great basins from the early 90's 'til today. And we did that. And of course annuals include other things than cheatgrass, obviously. But cheatgrass is the primary, you know, annual invasive out there. And when we, when you look at that map you can see, again, just the astounding amount of increase in cheatgrass going from low to high across the last 20 or 30 years in the great basin. And these are things that we know. I mean we talk about it all the time. But to be able to quantify it is important because, again, decisions are made at that, say that regional scale. We have to decide where are we going to allocate our resource? Where are we going to focus our work? Where are we going to try to improve conservation of these rangelands? And then also where are the areas that, you know, they're too far gone. It's not worth our time or our effort to really focus on them. Having data sets like this allows us for that spatial targeting and helps us to make decisions not just at the national scale but also this regional scale. And then it goes down to the management. You know it goes down the management scale. You can use these data to better understand the management or a pasture or an allotment or any type of management unit. And that really, you know, that user again, coming back to the idea where we empower the decision maker, they, that individual has knowledge that we will never have sitting here in our offices in Missoula, Montana. But we can provide them with information that fills in those gaps in space and time. They can combine it with the local knowledge and the local data that they have to make the best decision going forward. Speaking of making decisions, one of the things you mentioned in the 2019 paper on patterns of rangeland productivity land ownership is the possibility of having a data set that includes a forage yield of net primary production probably because that's one of the indicators of rangeland health and also because it's a driver of lots of ecological processes. What-- is forage yield or NPP available from the same data sets that you're already using just requiring a, you know, a different set of model calibrations or is that a new data set that you're working on?

>> Well it's actually both. So we do have the current NPP data set that's available but it's not yet on the RAP. So these will now be available via the RAP in the coming months. But we have that first level that is the NPP data set that is driven largely by the same sort of input variables that help predict the RAP.

>> Okay.

>> Predict the land cover. You know, NDVI from the land state sensor, climate data-- those are all brought in to give estimates of primary productivity on the ground. So that is one data set that will be coming available on the RAP in the next few months. And then the real sort of advancement is coming where we're going to be, or we have I should say, combined that land cover data set that's giving you a percent cover value of a functional type with that algorithm that gives us the estimate of primary productivity where we're now going to have estimates on the land surface. For every one of those baseball diamonds Brady was talking about, we're going to provide an estimate of how much of that forage is attributable to annuals, how much is attributable to perennials, and how much is attributable to shrubs. So we'll be able to see in these places say of where we have cheatgrass really invading the landscape how much of an effect is that having on productivity as a whole, but then also is that really having an effect on the perennial productivity as well? Or is it simply sort of a peak in annual productivity that's happening? So we'll be able to examine that parsed out productivity across the landscape. And I think for rangelands it's going to be a really helpful tool and data set to examine what's happening and make decisions.

>> And I think as scientists we like to talk in terms of, you know, net primary productivity, gross primary productivity, and what that means is, you know, we'll be having measures of actual forage, like you said Tip. We are able to convert those estimates to measures of aboveground biomass. So traditional range production data pounds per acre, forage.

>> Right. Are there any other expected future changes to the tool besides incorporating forage yield?

>> We-- so yes we will be incorporating the production data. There's a few other data sets that are, will be, they will be on there in the coming months as well. The actual tool itself, the platform we hope to redesign a little bit and make a little bit more user friendly and allow for a little bit more bells and whistles to do a few things. One of the things when we built this is we did not want it to be a kind of a one stop shop for everything. And we also didn't want to make it horribly complex to use. And so the intent has always been able to keep it simple, keep it very usable from a broad perspective of users. So they'll be a little, you know, changes here and there, but the core functionality will remain the same.

>> One, I guess a technical or logistics question, I know a lot of ranchers use Google Earth to do some ranch mapping and that's a fairly accessible and easy to use tool for lack of a better term-- application. And it's a, my own experience with trying to pull boundaries for properties into either the [inaudible] survey or into the RAP is that you need to somehow convert say a KML file from Google Earth into a shapefile so that you can import the shapefile because trying to redraw something on these, you know, on these maps inside the tool is just murder. Do you have any recommendations or a better way that if people already have some kind of a ranch map in Google Earth or some other application, a way to get that data converted to a shapefile and pulled into the RAP?

>> Well I-- honestly I think you just gave us a better way and I think with your suggestion we're going to, you know, we'll look into allowing the RAP to use KML files directly as opposed to a shapefile so you don't have to do-- you know jump through that hoop and you can upload a shapefile or a KML file. I'll look into that.

>> Yeah I've used a-- I think the website is geoconverter or something like that where you can convert a KML file into a shapefile and then pull that in because I don't-- it's not easy for me to make a shapefile.

>> Yeah, there's a handful of websites out there like you mentioned that can do that. And honestly sometime they work and sometimes they don't.

>> Right.

>> Both the shapefile format and the KML format are kind of--

>> Cumbersome.

>> cumbersome is a good word to describe them--

>> Right.

>> and so yes, I think we will definitely look into supporting KML files directly.

>> Great. Well, I've been really impressed with this tool and have used it on several large grazing projects that I'm familiar with or places where I have done quite a bit of on the ground monitoring and have really found it tremendously useful and I want to thank you guys for putting it together.

>> You're welcome. Thanks for using it.

>> Yeah, thank you.

>> I don't think I asked when we started, but what was the original impetus and or funding source that got this launched? I know you said your goal was to kind of bridge the gap between science and big data and management applications but, you know, why you and why now and what was the project or funding source behind this pretty large undertaking?

>> Yeah that's an excellent question and I think it's important to recognize that we are obviously not the first to think about this problem and we are not going to be the last. And indeed there are many other groups working on very similar data sets not just for the United States but for other countries. Again, this is-- this whole idea of rangeland monitoring is changing. We're kind of moving beyond just the inventories that we've collected for so long and we're going to a more, a true monitoring framework. It's funny because in 1976 there was a paper written by Dr. Eugene Maxwell in "The Journal of Range Management" and the title is, "A Remote Rangeland Analysis System". And so that was over 40 years ago.

>> Yeah.

>> And so people have been talking about this for a long time. It's just now we can actually pull it off because of the imagery archives we have and because of the advances in technology. To get at your question of kind of how we got started doing this, Matt and I work for a fantastic group of conservationists. We work with the Working Lands for Wildlife out of the NRCS. We are part of their science team. There's a handful of scientists that we work with across the nation focused on providing science to support NRCS conservation programs. And largely that science support comes in the form of spatial targeting tools and outcome evaluations. And so you asked about funding. The NRCS in conjunction with the BLM, they were the primary funders of this project.

>> Okay.

>> And how it got started I-- I don't know, Matt. [laughs] I-- we'd been talking about it for a long time and we thought you know we can do this, let's take a stab at it.

>> Right. Somebody ought to.

>> Exactly. And so we decided to do it. And again, there are other products and there are other platforms coming online and there's plenty of room in the sandbox for everyone to play. And I think as these methodologies and as these things become more available, we're going to really find out how useful they are and they're really going to help us answer questions that we weren't able to answer before.

>> Right. Well I appreciate your time. If there was-- maybe restate for the record where people can go to take a look at this and try it out and we will also post the website, the rangeland.app correct?

>> Yes. So it's the Rangeland Analysis Platform and the URL is rangelands.app. And if you just Google Rangeland Analysis Platform, more than likely it'll come up as one of the first hits, so feel free to get on, to give it a test drive, play with it. Feel free to reach out to Matt or I if we can, you know, help answer any questions or if you have any trouble we'd be happy to help out.

>> [Background music] Very good. We will put on the show notes which show up in the episode description in iTunes and Stitcher both the website for the RAP as well as any of the open source journal articles that have been published about the methodology and the system. Yeah, Brady and Matt, thank you very much for your time.

>> Thank you for having us Tip.

>> Thank you Tip.

>> 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 Connor's 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.

[ Music ]