 Okay, today we have, it's a pleasure for me to be able to introduce Steve Garman. Steve is an assistant professor at Oregon State University for a science department. And Steve has been working in landscape ecology for a greater percentage of his career than most people. He is trained as a wildlife biologist. But fairly early on, he began looking at trying to understand a lot of life from a much bigger perspective. He is doing a lot of different kinds of research mostly on the west side, but he has been involved with so many side issues. And he tends to be associated with looking at vegetation dynamics and vertebrate habitat associations. And as opposed to a nutshell, you can say that what he likes to do is he likes to use remote sensing and GIS to look at disturbance agents. And so we thought it would be pretty appropriate to have him come over here and speak to us about the patterns and the processes that create them, especially considering the fact that all of us now in the PNW station are within that disturbance program. So we all have to be doing work on disturbance and understanding it. Okay, I wanted to also mention that I have three papers that are written co-authored by Steve here. And if anybody wants to, there's one on detecting fine-scale disturbance in forested ecosystems, as measured by large-scale landscape patterns. And there's one alternative to civil cultural practices and diversity of animal habitat in western Oregon. And another one alternative to civil culture regimes that he co-authored with Andy Hansen. So three papers here, and if you'd like those, just let me know and I can make copies right away and we can get them to you. Okay, with no further ado, I'll introduce Steve, titled this talk today, Landscape Analysis and Ecosystem Management, Modeling, Process and Pattern. Thank you. Thank you, Jim. Well, it's a pleasure to be here today. I don't get over this way much, so it's a little nice just to be here and check it out in general. As Jim mentioned, we're talking about looking at using pattern as a surrogate for disturbance and understanding the underlying disturbance mechanisms. So to start with, I want to set the context of this talk and begin talking about disturbance in general. Disturbance is a real important process across landscapes. And primarily, it functions to increase the structural composition or heterogeneity of the landscape. For instance, fire maintains these high elevation meadows in the Cascades. Wind throw creates structural diversity at its large spatial extent, ranging from a few couple meters to acres and hectares. Flooding, for instance, is an important disturbance process that maintains hardwood component in much of the western Cascades. This is an area in the H.J. Andrews that resulted from the 64 flood. And if left untouched, you'd expect conifers to be invading this area. The February 96 flood took care of that, however. This is the same spot today. And, of course, we have biotic pathogens such as pine beetle that create instructional compositional, heterogeneity, stand, and, of course, among stands or the landscape level. Now, the functional significance of these disturbance processes are, as we can imagine, very important for, say, wildlife species. It's important for a whole variety of ecosystem properties and processes. But being mostly a wildlife biologist that sort of want to stray, I like to always come back to critters and understanding the important effects of disturbance on the ecosystem. And real quick, this is a little summary bar chart that I put together reading. They came out of Kevin McGarregals and Bill McCombs' study to the BLM in 1993. This is the habitat associations for vertebrate species in western Oregon. And real quick, the red here indicates positive association with early cell stages. The blue means it's associated. The species tends to be associated with late serial stage. Again, real quick, this is to show that some species have adapted, have evolved in early serial habitat, and some have evolved in late serial habitat, meaning you need both on the landscape to maintain the indigenous population of vertebrate species. And I'm sure you can say the same about many different ecosystem processes and properties. Well, this whole natural disturbance and the historic range of variability of natural disturbance is sort of the underlying underpinning to ecosystem management and sustainability issues today. Bottom line here is if we can manage or design our land use management schemes that will fit within the frequencies of varying size of natural disturbance, then we'll probably produce a much healthier ecosystem over the long term. And a working concept of this, what we call HRV, Historic Range of Variability of Natural Disturbance. A working concept in practice is the Augusta Creek watershed analysis management system, I believe, This is about 7,600 hectare watershed since South Central Cascades. This is an area where they went out and they did a whole lot of looking at scarred stumps to come up with the fire regime history. And they went back 400 years and looking at the fire regime. And based on what they found in their interpretation, they came up with different zones. Areas with low fires, but when they did have fires, it was very hot. Areas with lots of fires and low severity. And they used this as a template upon which to base timber harvest schedules for the next two to 400 years. And so, yeah, it's a pretty long term. So, this is how we're using this as HRV concept in our land use management. And I'm sure, hopefully, or maybe I'm not telling you something you don't already know, but I thought I'd give a nice little context or set the context for today's talk by just doing that really quick review. Now, if we want to look at historic range of variability in other places and other types of ecosystems looking at other disturbance forces, we can all go out and spend three or four years like we've done in Augusta Creek doing a really detailed study. Let's say we want to, especially in the future, I don't want to downplay field work, but basically I think we're going to do less and less field work over much larger areas as we go into the future. One of the reasons is the expense. And another reason is the increase in GIS remote sensing technology. Let's face it, I like going into the field just as much as anyone else, but you can get a lot of information from satellites and satellite imagery and all sorts of other types of remotely sense imagery. And so I think the data models, which is essentially a veg map, they're increasingly important in ecological studies. Again, increase in technology, increase in, well gee whiz, how many people in this room ten years ago didn't know how to use a computer? And I'm sure we all use computers today. Well, ten years from now we'll all have our own little GIS sitting in our forest service truck or our own personal old truck as we're driving through the woods. So the technology is increasing to the point where we are foolish if we don't use it. Okay, so that's sort of my pitch for high-tech field ecology. With this increase in the GIS remote sensing technology to look at disturbance processes across the landscape, we're going to rely on using pattern as an indicator of process. And an example of this is this one meter ADAR resolution image of this H.J. Andrews. And if you look at this and you look at this, like this red area here, which happens to be deciduous, the green and the yellow are hardwoods. This is open, this is deciduous, the red. If you were to look at that, I mean it's a nice little picture and you can see a whole lot of the area there and it's fairly convenient to produce this. If you look at that red zone, however, we know it's hardwoods. I mean what's the first thing that comes to mind with respect to the processes that create that particular pattern of that patch? As you can see it's sort of linear in shape, probably some repairing zone. You probably have some type of repairing zone disturbance occurring that's creating that particular pattern. Now this is sort of a prime example of using more or less easily accessible, remotely sense imagery to look at patterns from which you come up with some idea of the processes involved. It's a very simple example. More complex issues say you want to look at the patterns created by fire. Fire or bugs or whatever. So I'm going to sort of put my foot in my mouth. I said that the remote sensing is real important. It's powerful. It enables us to look at pattern deduce process real quick. However, we're assuming that pattern is in fact related to process. So if you have a process, does it create a mutually exclusive pattern? That is the question. Now it probably seems pretty obvious that that's a real important question. And I would also say that there's probably a whole lot of times when people don't think about that question. They just pretty much are interested in the pretty pictures. They are interested in looking at the patterns and coming up with a deduction without really thinking anymore about it. And so a question that a colleague of mine and I, Gay Bradshaw, her name, we wanted to take a look at this pattern process interaction. We sort of wanted to double check to make sure that processes don't infect great mutually exclusive patterns. And we're also interested in, you know, our ability to statistically document pattern as it related to a particular process. And we wanted to play some games also. We wanted to switch the ordering of the processes to see if, you know, processes that are related if they produce patterns, if you put one before the other, basically just play a few games to look at this particular question of pattern process duality. So we put together a spatial landscape simulation system. It's a very simple thing. And we wanted to look at the relationship between the ecological and statistical significance of landscape pattern and process. And I will, this is pretty much the basis of the talk today. It's a very simple system and I'll mention right now that this is a theoretical assessment of this question that we'll get into more of the practical aspects of this concept in the closing remarks. As I mentioned, we were interested in simplifying a fairly complex problem into just very manageable questions. And one of those manageable questions, one part of the questions was looking at the order in which the disturbance occurred in the landscape. We're looking for that. Does a disturbance give you a mutually exclusive pattern on the landscape? And that was really the primary question that we wanted to address here. And to do that, we would just change the temporal order of our two disturbance processes. What are those two disturbance processes that we looked at in this simulation experiment? Well, the first one's my favorite and that's wildfire. And the second one was bark beetle, a Douglas Fair bark beetle infestation. These two are related through their associations with coarse woody debris. Yeah, but they're not totally interdependent. So it gave us sort of a nice experimental set of processes to examine in the study. I don't want to get too bogged down with describing the spatial simulation system, but I think it is important that you know what the components were really all about. Now, the vegetative dynamics model, well, actually let me back up. This spatial simulation system is comprised of this spatial, veg dynamics model, a fire model and a bark beetle model. The veg dynamics model, simply put, was a simplified variant of an individual-based process-based stand model that I work on. It's called Zeleg, which is actually named after Woody Allen character. I won't get into that any further. But with this model, we were able to simulate the stand structure, including snags and logs. That's a log. We're able to simulate the withstand conditions of 100 by 100.1 hectare cells, which is what that upper landscape is to represent. So it's a stochastic model, which means every single cell of that landscape can have a different species and size class distribution, a different amount of logs and snags. So it sort of more or less emulates the real world. Cell size, as I mentioned, was 0.1 hectares. What is that? 30 by 30 meters, roughly. 31. There's a couple of things that we did with this landscape. We simulated a 200-year-old landscape starting from bare ground. So that was our initial condition, a 200-year-old, old-growthy landscape. We evaluated the amount of log mass in each of the cells of our landscape. So we came up with mean amount of log mass. And for reasons I won't go into, we used that as a scaling factor in simulating the disturbances of wildfire and bark beetle. That's sort of a modeling technique. And I'll just leave it go with that. If you have any really detailed questions about that methodology, I'll be happy to address them later. So that's our veg model. Our wildfire model, again, it was quite simple. It's a special variant of the behave system, which comes from a variant that I developed during my dissertation work. What does it do? Number one, it spreads from cell to cell as a function of the amount of coarse woody debris, or logs. You randomly select a location in the landscape. The fire starts there, tends to blow, given the maximum direction of the rate of spread, which is randomly selected. You basically use flop of fire down there to tend to form an elliptical shape as a function of the amount of log mass. So there's a nice little interaction between the wildfire and the log mass. Overstory mortality is conditioned by the amount of log mass in a cell, for instance. The more log mass the higher the intensity, the more the overstory trees are killed. And we kill the young ones, the small ones first, the taller ones last. And that's a function of the amount of coarse woody debris in a cell. So that's how we did the wildfire. Very simple. For the bark beetle spread, the beetle would infest a cell if the cell exceeded a certain log mass criteria, which again was part of that scaling factor I mentioned, which I won't go into any further. And also you had to have a Douglas firs down greater than 60 centimeters. So we set up the interaction between the amount of coarse woody debris and the tendency for bark beetles to spread from that to live large dug firs and spread across the landscape as a function of the continuity of the log mass and the large dug firs. So we did use a little biology here. It wasn't just totally hokey pokey, but it is very far from being a very detailed biological model of the spread of bark beetle. But that's sort of the art of modeling, is to keep it as simple as possible, yet as useful as possible. The other part I'd like to mention about the beetle infestation is we sort of impose this thing which we call drought. And that conditioned the ability of the beetle to spread across the landscape. And all that really did was under the no drought situation, the beetle would spread, the threshold was higher for the beetle to infest a particular cell, a threshold being the amount of log mass. Under what we call the severe drought, the log mass was less. So a bug could infest the cell that had less log mass under the severe drought condition. And again, it was based on our simulated mean log mass of our initial landscape. So again, a little modeling technique. So with that, that's our system. It was fairly simple. It was fairly simple. And we designed this nice little simple experiment in which we had two different levels of fire in beetle initiation. And these levels corresponded to like a single wildfire burning 10% of the landscape. Single multiple wildfires, three burning each burning 30% of the landscape. Beetle attack with low would be the number of initiations meaning you would randomly select an area where beetle infestation would begin. And it would move across the landscape as a function of all the log mass criteria that I mentioned. And you would have one initiation or five initiations. So you had multiple numbers of initiations of each of these disturbances. On top of that, so that's actually a total of four. One, two, three, four different experiments. On top of that, we had two levels of drought which really affected the ability of the beetle to spread across the landscape. It either increased or lowered the threshold values. So the bottom line is we end up with eight different experiments. Now again, as I mentioned, we're interested in looking at the order. So you would have fire would occur first, then you would have a beetle outbreak. So that would be the first part of a paired simulation. You'd then have beetle outbreak followed by fire. Given that the initial locations of each of these disturbances would remain constant in this pair-wise set of simulations. So we control the spatial aspects for a pair-wise set of simulations in which we were changing these temporal order of the disturbance. And again, this was the primary thing we wanted to look at. It does the particular order of the disturbance actually result in a unique pattern on the landscape. A unique pattern that we can detect statistically using some various landscape metrics. So that was a primary objective of this particular assessment. Now to further complicate things, we had five replications of each paired set of simulations. So I forget what I have lost track. I think that's 16 times 5. I think that we had 16 times 5 simulations. So the number doesn't matter. The important thing to remember, even if you don't remember the fire low-high, beetle low-high, drought low-high, is we have a big disturbance that is occurring either once or multiple times. We have a disturbance that tends to propagate in a linear fashion. And that occurs under really non-severe conditions. And the other end of the gradient is very severe conditions, meaning the bugs won't take much for the bugs to propagate across the landscape. So again, I want to emphasize that the property of these disturbances is what we're emphasizing. We're not trying to say that we are truly simulating welfare in bark beetles. For the detection of the pattern, we used a couple of very simple metrics. One, two of which were spatial in nature. And the one we used was nearest neighbor distance. And I'll explain that in the next slide a little better. And the other metric was total edge, which is simply the amount of circumference of a patch. Now, those are fairly easy metrics that you can get from like the FragStats software package, which if you haven't heard of, ask me a question about it later. I'll tell you more about that one. We also, through our remote sensing, we, being people like Warren Cohen at OSU with the Forest Service, it's developed the ability to actually come up with estimates of basal area from remote sensing. So you could total up the amount of basal area of the landscape. Well, that's a non-spatial metric, not the most landscape level. So we're interested in looking at the ability of spatial and non-spatial statistics to differentiate between ecologically different processes, more specifically between the patterning of the process, I mean the temporal ordering of the disturbance processes. So these were our metrics, nearest neighbor, total edge, and total mean basal area. This is our simulated landscape. The blue, now you keep in mind that this is the pixel size here is 0.1 hectares. So the blue represents greater than five square meters, or you put it in a per hectare basis, 50 square meters per hectare, meaning it's all growth. The red represents one to five square meters, and the yellow represents less than one square meter. Now this is the initial landscape with a low intensity wildfire. And that's what this collection of color represents. So a fire was randomly selected to occur here. Since it's at an edge, it does some funny things. It just sits there and burns and burns and burns until 10% of the area is consumed. So the yellow is where the fire burned real hot because the log mass was great. And again, all that was simulated from bare ground. So there's an underlying heterogeneity of basal area and log mass and snags, which you can't really see just using these three simple colors. So that's the yellow. The redacorus means the... Well, the fire was hot, but it wasn't real hot. It was an intermediate severity of fire. Again, it's a function of the amount of, of course, woody debris or log mass. Now, I keep talking about all these patch metrics. What we did was we base our patch metrics on the three colors that you see here, the three patch types. And the patch types are a function of the amount of underlying basal area. And forming a patch, a patch is a set of adjacent or diagonal cells of the same type. So we would simulate this condition and then go through and form our patches and then perform our spatial... use our spatial metrics to describe the pattern of this. If you take a look at the patch, it is red. For instance, you'll see that it's a very convoluted patch. But it is fairly extensive. And we have various software programs that performs all this patch assessment for ourselves. It's really easier than it looks. So this is what a single wildfire on the landscape looks like. This is what a bark beetle outbreak looks like, known as the sort of spaghetti light running from left to right. Very different type of disturbance. Of course, I do want to emphasize that this is all simulated. I mean, it's obvious to me it's simulated. Halfway through some of these talks, I always get people that raise their hands and say, that doesn't look like a real pattern. I've never seen that. Well, it's because it's simulated. This is a beetle infestation with wildfire overlaid on top of it. Now, again, the position of the wildfire in the beetle outbreak is the same as what I showed you previously. Now we're just putting the disturbances down together. Now, if you recall, there's this prominent feature of the outbreak of the bark beetle. And you see that it extends up into where the wildfire occurred. Now, let me back that up. Up here where that extends upward, the wildfire came through, and that particular area burned hotter because there was more log mass on the ground. Log mass originated from the beetles. So that's that area burned harder. What you see is this is just the wildfire. You see that there's quite a bit of a mosaic of patch types. Well, that mosaic is retained in this wildfire outside of where the beetle occurred first. This is the opposite order. This is where the wildfire occurred first, and then the beetle occurred. Interesting point. The wildfire occurred first. There was more log mass on the ground. The beetle came across. In fact, at this area, it not only further reduced the amount of basilaria going from red to the yellow, but it also the beetle, because we're using a propagation or contagion model, which more or less is how the beetles do spread. The beetle was able to spread out even further than before. This is just the beetle alone. You can see the pattern here, or this edge. And when the wildfire is down first and the beetle comes along, you can see that it extends out even further. So it's doing something, this particular order not only affects the intensity of the processes within a cell, but also is affecting the distribution of the processes across the landscape. This is just another example of the low severity. The example I just showed you was the low severity class, where you had one event of the fire, one event of the beetle. You were under no drought condition, meaning the propagation threshold were set so that it was hard for a beetle to propagate across the landscape. This is an example of the same condition. This is actually another one of our replicates. But what this shows is the effect of the temporal ordering on the connectivity of the landscape. And I'll backtrack. This is a beetle. It happened first. Wildfire came along. You can see where the tends to overlap, tends to be a little hotter, hotter meaning more of the basilaria is removed and the resulting pixel color is yellow. So this is when you have a wildfire and then the beetle. So you have predisposed this area to propagation by this other disturbance process, i.e. the beetle. It comes along, hits that, and in fact is able to move through and pick up on the other side and traverse across the other side of the landscape. This is an example under the same set of conditions, but this is another replicate. This indicates or attempts to show that in some situations it really doesn't matter what the temporal ordering is because the processes simply don't overlap. Notice the difference between these two slides. There is no. The temporal ordering has changed, but just physically they don't overlap so it just doesn't matter. Again, it's sort of an obvious and intuitive result. But all these past three sets of slides, which demonstrate that you can come up with a variety of combinations, given everything else the same. Everything else meaning the disturbance frequency intensity and the disturbance types. Yet the resulting landscape pattern can be very different. This is just a couple of examples showing what happens when you increase the frequency of bark beetle infestation. This is a single wildfire with five infestations of bark beetle. Again, this is bark beetle than fire, wildfire than bark beetle. The difference, again, is the homogenization effect of the wildfire patch. If you look up here in the right-hand corner, you'll see when the wildfire occurs first that the bark beetle is able to extend the amount of area that it affects because it's traveling through that patch that the wildfire burned. And these are sort of the bizarre pictures. This is the most intensive combination of disturbances we looked at. This is where you have a lot of fire, a lot of bark beetle initiation, and under what we are calling the severe drought, meaning there's least amount of resistance to the propagation of the bark beetle disturbance. So, as you can see, the whole landscape pretty much gets as white-cleaned and rearranged. A very interesting pattern. Again, this is just one of the replicates. So, what does it show us? What it shows us is, number one, ecologically similar disturbance. And in the same type, same intensity, same frequency, same ordering, can give us some visually very different landscapes. This is an example. Everything is held constant here except for spatial location in these two slides. And as you can see, they look very different. Second set of conclusions or results, because they were a little complex, I put them on an overhead so I didn't have to go flipping through the slides. Okay, well, oh wow, cool. It helps when you focus things. I'm sorry? Say what? Okay. You mean, like, put it on the screen? It's a very demanding group. Okay. So, what do we have here? First conclusion or result is that you can have the same ecological processes happening and that's something that looks different in the landscape. So the second general conclusion is, if you notice, this is total basal area. This is total edge. Every time in the fire, you probably can't read that, but this, the stripe bars is when fire occurs first. The open bars is when bark beetle occurs first. You basically, when the fire occurs first, if you recall, the area where the two disturbances would overlap would be homogenized. It had this great big yellow ball. Okay, so what you're doing is you're increasing the amount of coarse wooded debris. You're predisposing the landscape such that the bark beetle disturbance is able to propagate more so. And in doing so, it reduces the amount of basal area in the landscape. It also creates a big ball which has less total edge than many small balls. And you're also increasing the distance between patches of similar type. So instead of having a whole bunch of little patches going right across an area, you have a few patches off to the side and this great big ball of a different type sitting in the middle. What that effectively does is increase the distance between nearest neighbors. Nearest neighbors being patches of similar type. So that's just a real general effect that we've noticed and it's consistent regardless of the frequency of fires, single fires, multiple fires. If you look at the statistical significance of our results, which are based on the five replications, you notice that under the single fire, and the fire here in this particular in our simulation seem to be a very dominant, a very dominating disturbance, under multiple occurrences of this dominating disturbance, our metrics could significantly distinguish between the temporal ordering of the disturbance. That is the bark beetle and the wildfire disturbances. Regardless of the number of initiations of the bark beetle, if you can't see it, the bottom parts of these graphs are the five initiations of bark beetle. And this side of this graph is the severe drought, corresponds to severe drought. So you have multiple or single initiations of the bark beetle. You're constraining or encouraging or increasing propagation of the linear disturbance. And all of these cases under the multiple dominating disturbance, the temporal ordering is distinguished based on our very simple pattern assessment. Under the lowest severe conditions, which is represented by this column, you can see with only one exception that our pattern metrics cannot distinguish between the temporal ordering of the disturbances. So that's when the disturbances are occurring, let's say, they're mild disturbances, the least severe of all. You can't seem to distinguish between the temporal ordering of these disturbances. When you have multiple events of this dominating force, you have pattern metrics are able to distinguish between the temporal ordering, only when the other disturbance, i.e. bark beetle, the propagation of the bark beetle is in fact constrained. When you have unconstrained or lower threshold for propagation inertia, you have... Our metrics are able sometimes to tell us something about the temporal ordering. But not all the metrics, for instance, our basilaria metric cannot distinguish between temporal ordering in that latter situation. So what we actually... This is low intensity, this is most severe intensity. And what we seem to have is somewhere in between these extremes, our pattern metrics can tell us something about the temporal ordering of the disturbances. And what that pretty much says is there are some thresholds within which, well, outside of which, we can't tell the difference between what happens on the landscape, what occurred and when it occurred and what ordering it occurred in, relative to another disturbance. Within a certain window, our pattern metrics, in fact, seem to indicate that we can say something about the underlying processes. So that's sort of the long-winded result of our very simple little simulation study. The short-winded, or I should say the short answer to this... The short answer to this, can we distinguish ecological significance is statistical significance and ecological significance equal? Sort of the pessimistic answer is no, it doesn't always work. There are some instances where it does, some places where it doesn't. And to be quite honest, I've explained our simulation, I think, hopefully well enough for you to be able to grasp it. I think you can understand or appreciate its simplicity, its non-realism, but its usefulness. And so we pretty much have just taken a sort of a quick and dirty look at a very complex problem and we've simplified it. And we've come up with a conclusion that more or less is a little more than a hypothesis. Our hypothesis pretty much says that there are conditions and situations in which very simple pattern metrics can tell us something about underlying disturbances, but not always. So not always. Well, under what conditions can you and can't you use simple pattern metrics to look at the disturbances? And that's pretty interesting because I have no answer. I don't know. What I can suggest or indicate is that we need to know. We need to go out and figure that stuff out. I will return to how we do that. I only have two more points I'd like to make regarding this type of pattern process evaluation. Now, imagine if you will, that this is the type of monthly sense information that you pick up from your local corner GIS store. And you are looking at some particular process. Well, this happens to be 25 meter resolution. You may recognize this slide. I had a slide of this earlier only, different res. This was produced by Earl Echley and Warren Cohen at OSU. If you were to pick up this information and to look at the pattern and use this pattern assessment to come up with disturbance processes, at this particular resolution you, for instance, may come up to you may, I don't know what you conclude. You'd look at this, let's say let's concentrate on that patch. You may conclude that that's just an on-growth patch and there hasn't been any disturbance there at all. There's no disturbance happening there. If you were to look at a 15 meter resolution of this site, you in fact see that, well, it's not quite a big patch of oak growth. There's some hardwoods that again are young conifer which is what the yellow represents. So this may give you a slightly different answer to your pattern assessment and the resulting deduction regarding the type of disturbance that has or possibly will occur in that particular spot. Now, you take that down to 5 meter resolution. This is all 8R data, I might add. 5 meter resolution and, you know, you look at that and you go, well, gee whiz, that's not a really on-growth stand. That's sort of broken up. So your pattern metrics would give you a different answer and you may end up with a different conclusion using this resolution of information. If we take it down to 1 meter and you take a look at that spot again, you see in fact that there tends to be a whole lot of young conifer in this area and it's just sort of a mishmash. And again, the type of pattern metrics that you would generate for that image is very different from what you would get with that image. Now, these are the identical images. The real data is the 1 meter and the 25 meter is just using an aggregate algorithm. But the bottom line here is the resolution at which you go out and look at your particular process or at which you collect your remotely sensitive image. It's real important to make sure that you're collecting information at the correct resolution, correct being the resolution that matches the pattern and the processes that you're really interested in, especially the disturbance process that you're really interested in looking at. And the intent of showing these four slides is to show you potentially the different answer you would get as a function of the resolution of the information you were looking at. The second sort of take home here or the last point I'd like to make is I talked about these pattern metrics and I went over that real quick. I said nearest neighbor and total edge and total basal area. I won't go into why we use those metrics. The one thing I want to point out is there are quite a few patch and minescape metrics out there. Again, you can pick up at your corner there's quite a few. You have you can sort of break them down into three groups and those groups are metrics related to shape, size and extent and connectivity. What I've just shown here is just really a handful of metrics. The problem is you really have to be careful in choosing the metrics you use for your assessment. The reason for that which I will show by example these are two simulated landscapes at the same point in time. This was simulated using an aggregated harvest strategy and that was simulated using a dispersed harvest strategy. You can take a look here real quick. The red, the yellows, the aquas are hardwoods and just looking at the aqua and the reds you can really see it. There's quite a difference in the patterning. If we were to use just these six landscape metrics you can see that some of the metrics are in fact quite different and some in fact are pretty much the same. The aggregated, the metrics associated with that image on the left-hand side of the aggregated harvest image is in red and the other one is in the yellow here. You can see the number of patches per unit area and the inside area and the amount of circumference is very different. Differences between patches of the same similar type aren't quite that different and then you end up with these things like patch fractal dimension. Jim and I were talking about fractals before the talk. I have yet to figure out what a fractal is but I know what a fractal dimension is. I haven't quite figured out why we'd want to use them. You will see it in the landscape literature and again take a look at these two landscapes. Now, do they look different to you? They do to me. Well, based on Shannon's diversity index that fractal dimension are not different. So here the take-home message here is be careful in choosing your metrics for any type of landscaping assessment. In fact, it's really cool to do what I've done here. I've been meaning to do this with a whole lot of metrics is generate a couple of really different landscapes and then start looking at how the metrics react to those landscapes. That's the only way you really get a good feel for... It's the only way I think you can really believe or trust in your powder metrics for instance. So I am sure you're also terribly fascinated by the scintillating display of simulation and colors and all that sort of super groovy neat stuff. I'm also sure you're sitting there saying what the heck does this have to do with my stuff? Because I sure as if I were sitting in your shoes I sure would be saying that too. In fact, I sometimes ask myself that. Well, let's say you want to go out and you want to take a look at where disturbance doesn't happen. The old ponderosa pine being I'm sorry, fire suppression ponderosa pine systems and the fur coming in underneath. You may want to go out and take a look at where fire suppression has resulted in these types of stems. You may want to go out and take a look at where some prescribed fires or natural prescribed fires have gone through your system. You want to look at the frequency intensity. You want to look at the underlying disturbance processes. Like I said my intent here today wasn't to give you the answers but to give you some potential insights into how you would go about using some of the remote sensing. GIS technologies to use and using pattern as a surrogate for disturbance processes. My wife tells me people only remember five points from a talk. I'm going to be a nice guy and I'll only give you four points you have to remember from my talk. The first one is hmm the second one is I'll be done. The first one, what is the first one? Boy, I had these memorized. I really have to apologize. I do forget it. Let's go to the second one. Maybe the first one will come back to me. The second point is related to the first one so I can't remember that one either. Well, three and four are pretty easy. The third one has to do with the scaling issue. Choose a scale that your process represents. Oh, that's it. Wow. The first, really is under some conditions you can use this pattern stuff, pattern metrics to really come up with some estimate of the type of disturbance or ordering of disturbances that are going on out in this landscape. What we need to do is we do with building habitat models or doing classification of remotely sense imagery. You always have a test data set and you develop your models based on the test data set and even with that model, you go out and you test it against some data that we're not used to generate that model. If you want to go out and you want to look at these fine scale processes that are occurring here in the west side, for instance, go out and collect some information on the ground collect some information from the sky, so to speak, i.e. overlaid on top of your remotely sense imagery do an assessment of looking for correlates between your remotely sense imagery and what you see in the ground and see how well you can fit the disturbance processes with particular pattern. That's sort of easier said than done because you have a whole lot of different things happening out there in the real world. That's why we did a simulation study 20 years and millions of dollars on extensive field studies. So maybe you go out and you look at a couple of the real important disturbance processes that you're interested in. The big ones that are really affecting the system that you're working in and basically develop your correlates with your pattern using empirical data and then test it, meaning go someplace else and try it out and see if it works. If it does, then maybe you have a halfway decent management tool that you can use to do a very broad scale assessment of disturbance processes. That's really point number one. How could I have forgotten that one? I don't know. But I still forget the second point so I'll be darned if I can't give you that one. Third point has to do with the scale and going out and doing your assessment out in the woods and doing your assessment using the remotely sense imagery. How to use the information at a resolution that is really that means the most to the process that you're looking at that reflects the correspondence to the level at which your disturbance process is affecting the ecosystem. And the fourth point is be careful in choosing your landscape metrics and the patch metrics that provide the pattern on the ground. As again, as hopefully I've showed you here some of the metrics make some sense seems to correspond to what you would think just by looking at two very different images and there are sometimes where some of those metrics are pretty much the same for very different visually different images or landscapes. If I happen to think of the second point I'll let you know. I have an email address and I can email you. It was a really good one but I'll be darned if I can remember it. What should the second point be? Any suggestions? What's a take home other than the three and the missing fourth one point that I mentioned? I have a question as to whether or not this satellite imagery can really be used on the east side that are scales that require much higher resolution and we have so many different disturbance agents that are all interacting. You just showed two and we've got half a dozen that are all operating there at different scales and sequences. I really wonder whether or not this technology is going to be all that useful for the district. Well, that's a really good question. I really don't know the answer. It really depends on the question you're asking. This is something, I mean it's totally unrelated to this pattern process stuff. It's just reality events. You sort of, we always make compromises in the resolution at which we look at something whether it's out there measuring trees or doing simulation modeling or using remote sensing to look at the pattern. Really, to answer your question you'd have to understand the objective. I mean a particular objective has to be identified and described. In some cases, the remote sensing probably would do a terrible job because you're right. It's tough getting down to individual species. The remote sensing image, the one meter is 5, 15, 25 meter radar image. There's only four class types. There's nothing there about three species, individual species. And that's one meter resolution. And we have quite an extensive remote sensing shop. And they find it very, very tough to identify individual species. And a very simple system relative down here. So you've tried to use the spectral frequencies to identify species? No, I've been pushing for it. It isn't all that easy. And a remote sensing person is sort of resistant to do that kind of research. Because on the west side if you're a structural stage it's clear cut. That's what they do. It's just a mosaic of clear cuts. So as trees grow they get bigger. The structural size classes you say maybe four of them then you can do this kind of stuff. Over here we don't clear cut that much. There's some of it but a lot of it's selective cutting. So what does that do to the spectral frequency you see from space? Yeah, you're doing selective cutting. What's your spacing like? What's your spacing between your big canopy trees? It's a real variable. I mean we have a lot of people who've done that work. Some people do these real seed tree kind of things. Some people basically take an overstock stand and thin it to a point where it begins to have a little bit of growing room. But the point is that let me real quick before you move on here. My understanding is that you're doing this selective cutting in ponderosa pine systems where the trees are much more spaced out than say an old growth deck first system and in fact there may be even a greater chance, depending on the spacing but there may be even a greater chance of seeing the effects of selective cutting in ponderosa pine systems because they're more open. The trees are more widely spaced and so you have a satellite coming along it's... You want to make a bet? I think I can do it. My luck. At least thanks for telling me I didn't stand there all day and just rip it out. Hopefully... Well, you might have to increase your research budget to get some time. Okay, so my point here and I'll do it with graphics. This is ponderosa pine system and let's say this crown here. I mean the stuff that I was showing was one meter, ADAR, which is pretty intense. Everyone in the brothers and sisters doesn't have ADAR data sitting around. It's like TM, which is at 25 meters. So if you're... If the spacing between your big canopies are greater than 25 meters or at about 25 meters you probably have a better chance of picking up the fact that you had some selective cutting. Especially of course at the one meter. Well, or five meters. Actually that series that I showed you people told me is the five meter resolution is really the best to work with. I mean the one meter gets just... There's too much stuff. And a five meter is a lot easier to work with. There's less information. And when you start working at resolution less than the canopy widths you're asking for some potential problems. Yeah. The thing, because the satellite was looking down to those widely spaced remnant older trees and picking up the under the second canopy and calling it a young stamp where some of these some of these hands actually put a map in the stamp. So they're ejected and they can choose from our original... That was originally working with the 25 meter? I think it was... TM? We had so little confidence in that that we just pitched it. Well, it sounds like you needed to use something a course of resolution. Find it? You think find it? The problem was it was so fine that you were looking... You were coming down, you were looking... You saw too much. You saw too much and you were calling this old gross stamp a young stamp because you were picking up the young stamp. So of course a resolution would potentially wipe that out. I don't know. Good research project. Okay, so that's okay. That's anyone else? That's it. One thing that appears to me is that these type of modeling systems work if you've got a homogeneous setting. And it looks like you're considering basically uniform stands of similar species. And I would guess similar substrates. And things like precipitation levels all pretty similar. Out here in the Northeast Oregon you get very complex settings where you might have completely different species on one side of a North slope than you do on a South slope. It seems like you'd have to make sure that if you're going to use this type of process that you stayed within homogeneous areas and apply the technique within those areas. Are there ways of doing that? Do you take a big area and apply the techniques selectively for areas within that big area? By the techniques you're talking about the pattern assessment and relating process. Or the simulation modeling. Well, the pattern assessment you did on the simulation. Okay, the pattern assessment. I just wanted to make clear. The simulation was to generate the pattern assessment as something that you could do on any place. So that's what you mean. Sure. GIS is a magic wand. You can do anything. Sure, you can separate your if you know the underlying environmental conditions you can separate out an area that had similar substrates or similar conditions and just sort of cookie cut that piece out and just perform an assessment. In fact it's a sensible thing to do if you can imagine of an elevational gradient in which fire frequency would change of that gradient and it would make sense to look at pattern on top and down at the bottom and coming up with an overall assessment of fire frequency. Just a matter of tinkering with the GIS. Just cutting out the pieces. Actually, yeah, that's the easy part. The more difficult part would be finding the domain of the environmental region. I'm sure you can imagine or know what soil series maps are like their quality, their resolution pre-zip temperature information. Again, the same thing. There tend to be very coarse level information but you could use that to delimit your environmental region of interest. Also I can make maps like that. I just want to help if they could be applied. Sure. The question I have, this is only my second time on the ground. I get over to the east side. It's usually Metolia, some Pringle Falls. So I don't get way east here. I don't get up to the blues that much. Actually, I've only been through them once so I got a speeding ticket doing it. I'll be darned if I go back. The question, what am I leading up to? The question I'm leading up to is what's the most significant disturbance process of interest to the managers? Now I understand me. I know about the east side issues. I've heard about the east side issues. We're possibly way outside the historic range of variability. I guess my question is what factor, what disturbance process do we need to implement to sort of bring the system back into historic range of variability? Is there just one single? Is it just fire? Is that the major problem over here? We love to have things over here too. That's a severe disturbance. So how do we swing the system back to where we think it should be or where we think it used to be? Log mass? The other management concern on the east side big time is thermal cover. We raise out. So those are the two big things. So you don't what I'm hearing is it's We have to try people to really attend to your answers. So what's your answer? My answer is I think fire is component but I think we're going to have problems in trying to figure out how we can use fire without some of the massive complications also the air quality issues and how we can produce it. And I think actually I think one of the big problems that we have is getting education and acceptance of the fact that we've got a problem that's been brewing for their part of the century and it probably isn't a very gentle way to reverse that. Yes sir. I'll give you an example of a real problem. A university in Montana student wrote to me and their class is working on a program on the health and national recreation area and the question for me was what do you know about these areas? The Martin has almost never gets more than about five meters from a river. And so here we had the canal fire and the twin lakes and the fire that burned together separating the Hywel Laos which would be your source habitat from the habitat on the national recreation area and of course there's a lot of little spots of green stuff. But how are they ever going to come up with any kind of plan or thought to get the interconnected corridors and things like that so that the habitats on the national recreation area are not the same habitats where you're going to lose the property. That's why you don't know when you're going to fix the mountains. But it's going to be the mountains that are in evidence. Well, sir is there an interest in looking at the historic range going back four hundred years in time and trying to figure out the fire record? Because of mother nature takes its course but let us keep the raining keep having elk hear from these levels and you know it's let's have a commodity and you can do it but nobody wants that that's the hard part this is a display we're probably going to have circumstances in any way if there's no doing Well, there's an uplifting thought The problem with the law is that users have to define what role of a relationship they walk with fire and that's going to be extremely difficult what we've done and what we continue to do is to delay the night we're delaying every time we delay we have bars, stands, places, events and they're going to happen that is the most required regime anybody who's stood in front of these types of fires the last ten years you're going to see this in the future but if you're a habitat connectivity there's no amount of money in the world that will stop the plumes from laying fire huh, wow this is sort of bleak I'm depressed one gentleman mentioned that the logging that takes place and I would simply suggest that if we move some of the logging that's taking place from healthy forests to unhealthy forests meaning ones that are overloaded with fuel that might make quite some gains that's what we're trying to do with our local groups the problems that our companies will build and also economics the big problem is what someone sees as a healthy forest is not necessarily a healthy forest based on risk and that's the whole point what's your risk assessment and if you can couple something like this with a risk assessment that's what this is but it's really tough because when I look at a stand of trees and I have someone from the public standing next to me if it's green it's healthy and there may be a pine forest with a duck fur, a grand white fur under story and I'm just going this is the way it's loaded and they're looking at me like how can you think that way and it's very tough to convey this you've got a public that's far removed from what happens in nature most of them live in cities where the closest they get to nature is Albertsons they're safe hey they have a good produce section even out here I don't mean to imply that we're out here in the boonies there's a lot of people in California really? even over here there's a lot of people moving in with families there's a lot of retirees to buy as soon as you retire it seems like your values have gone totally to you for the next time we ever other years I mean I'm getting old enough but I can appreciate those that's what's really good people can appreciate or can't understand the dynamism of these systems over here that's what I mean our tendency is to prevent change and it's true that the old growth is so limited now that every single patch we have we want to protect well we're not really going to be able to protect them they're going to change eventually well we're going to set our thumbs up for that so in a month or two we're going to smoke again so maybe what? did you say you think we should smoke them? no no no well that's what I thought you meant by smoke them that's cool wow in the sense of smoke smoke still has its strong message to play is that if the fire is bearing down on your house because it's an unwanted fire it's a human ignition we have people moving in smoke still has a role to play but it's such a complex role where you can't do it it's only good PR that's the only good thing we've got going for us these days is that it's an interesting character he's got a neat history and it's about the only plus we've got there's a lot of poppies in this well I'll have to go back to Corvallis and report that there's a group of people up here that want to shoot smokies up there now I understand I'm going to agree with your following I think you got a real good thing going and I was keeping very vigorous notes for up half way and then it dawned on me it didn't look like it was going to say what I wanted you to say oh shucks there was really more happening over on the east side of the forest that would be adaptable to what you have because it's a wide range of resource variations we burn up a lot of fire and I thought wow I can get you to give us the same program on just the North Fork of the John and Day and it would answer a lot of our questions and then you got to the last when you said no I don't think it does answer the west side I think you're saying it would answer problems that we have on this side I think let me just reiterate so that I'm clear I think the pattern process interactions being able to detect processes from pattern through our little simulation study suggests that in some cases you can in some cases you can't and that's really the take home maybe that's the second one good Lauren I can't believe I did that yeah that was the second one and the other one is which I think gets back to the first take home and that's we sort of need to figure out where you can and where you can't and this simple, very simple little study simulations that I showed really don't do that again all they've we've done is generated hypothesis and added a little bit of fodder to the fire encouraging us to basically to look for it in fact to be quite honest very very honest it has been a little while since we wrote the study or did the study and I sort of forgot how neat it was until I had the opportunity to come over and talk about it and the one thing you know how it is I mean you always get busy you get off at something else that will take you a week a year later you're still on it but then going back over this I sort of realized that there's some really cool opportunities to actually get out in the field and start looking at this you know hey take some of these hypotheses and get out there and start looking for examples and I was sort of encouraged to potentially do that Have you been watching over the shoulder of the Guadalajara Eastside study because they use an awful lot of what we have there and it keeps you really fascinated with what they're doing in the last couple of years No I haven't been keeping track Okay thanks Great Are there any other like more general questions I realize it's getting late and people want to go So Steve will be around he's made a commitment to be around for a while so if anybody wants to talk to you later you can answer more questions Thank you all for coming Thank you