Framework for Evaluating Multi-Species Climate Change Vulnerability

Video Transcript
Download Video
Right-click and save to download

Detailed Description

This webinar was conducted as part of the Climate Change Science and Management Webinar series, hosted in partnership by the USGS National Climate Change and Wildlife Science Center and the FWS National Conservation Training Center. 

Webinar Summary: Frameworks for evaluating the vulnerability of multiple species to decline or extinction are increasingly needed by state and local agencies that are tasked with managing many species at once. USGS researchers in the Northwestern U.S. are looking at the “sensitivity” of wildlife species to climate change, which is a fundamental component of vulnerability, for freshwater fishes, amphibians, and reptiles native to the state of Oregon. They have evaluated species-level data across a large spectrum of geographic range sizes and climate sensitivity. 

Their results suggest that a combination of classifications based on species’ range sizes (the area they occupy) and their traits (e.g., body size, generation time, and investment in offspring) offer a promising foundation for regional multispecies conservation planning, particularly for species researchers know little about. Specifically, this framework can help identify focal species for monitoring and highlight priority species for which exposure to climate change and other threats should be assessed. 

Watch the webinar to hear from Meryl Mims about this project and the research findings! Meryl is a USGS Mendenhall Fellow contributing to science research at the Climate Science Centers. Meryl works closely with Jason Dunham, a researcher supported by the USGS National Climate Change and Wildlife Science Center.


Date Taken:

Length: 00:53:45

Location Taken: OR, US

Video Credits

Meryl Mims, USGS
Elda Varela Minder, USGS
Holly Padgett, USGS


John Ossanna:  How are we doing today, folks? I’d like to welcome you today to our broadcast. Welcome from the U.S. Fish and Wildlife Service's National Conservation Training Center in Shepherdstown, West Virginia.

My name is John Ossanna and I'd like to welcome you today to the USGS NCCWSC Climate Change webinar series. Today's webinar is titled “A framework for evaluating multiple species' vulnerability at regional scales”. We're joined today by Meryl Mims.

Meryl is a freshwater ecologist in a research integrated population community and landscape ecology to address fundamental and applied questions in ecology, evolution, and conservation.

She's worked in a range of systems and across many scales from local to continental. Her research focuses primarily on freshwater fishes, reptiles, and amphibians. Meryl got an undergrad degree at Georgia Tech in biology before pursuing both a master's and PhD at the University of Washington's School of Aquatic and Fishery Sciences.

She completed her PhD in June 2015 and joined the USGS as a Mendenhall postdoctoral research fellow shortly thereafter. She'll be joining the faculty of the biology department at Virginia Tech in January 2017. Meryl, at this time I will hand it over to you and I'll actually go reset your slides for you.

Meryl Mims:  Thanks, John. While John's getting us back to slide number one, I just want to start by thanking the National Climate Change and Wildlife Science Center for this opportunity and for making this webinar happen. I'd also like to begin by acknowledging my co‑authors, Jason Dunham and David Pilliod with USGS, and Dede Olson with the U.S. Forest Service.

The results I'm presenting today come from just over a year of work on this project. We are starting to wrap up this first phase of the project. I'm looking forward to sharing these with you. Thanks for listening in.

I want to start with a figure that may be familiar to many of you. This if from the intergovernmental panel on climate change in their 2014 report. It summarizes how we might expect climate change to affect both global temperatures and precipitation.

On the top, there is an image looking at the change of average surface temperature expected over the next century under low greenhouse gas emissions scenarios on the left, and high greenhouse gas emissions scenarios on the right. Then, on the bottom is the expected change in precipitation.

Although there remains considerable uncertainty in some regions regarding the magnitude and the nature of those changes, what we know from recent work is that anywhere from 8 to 16 percent of species globally are threatened with extinction, depending on what that trajectory looks like.

This presents a major challenge for conservation biologists, ecologists, and managers, because we simply don't have enough resources or time to fully assess and plan for all of these species individually.

This has led to a major question and challenge in ecology and conservation biology, which is this question of what are the major drivers of species vulnerability to a changing climate.

Are there generalities that we can make across species, and what are the limitations to those generalities? Where do these rules break down? How do we assess many species at one time, and how do we assess poorly understood species?

Before I get going into this, I just want to make sure that we're on the same page and that I define vulnerability, as I'll be spending a lot of time talking about this and the components of vulnerability during this webinar.

Vulnerability is often characterized as having these three dimensions of sensitivity, adaptive capacity, and exposure. What we know is that species attributes or characteristics can help inform both a species' sensitivity to a changing climate, as well as their potential adaptive capacity.

When I talk about attributes, what I mean is anything from inherent traits or biological traits, such as body size or maybe trophic position, as well as geographical characteristics like range size.

We know that these attributes can help inform vulnerability. What that does for us is it offers us this pathway to start thinking about the vulnerability of many species at once. That's because these attributes, such as traits‑based attributes or spatial attributes, offer us a currency to compare vulnerability across many species at one time.

There's two current ways this is achieved. One is through traits‑based approaches, which evaluate more biological or innate attributes of species, such as their life histories, morphology, and behavior. There's also rarity‑based approaches, which focus more on the spatial attributes of species ‑‑ things like range size, habitat sensitivity, or local population size.

Based on decades of theoretical and empirical work and evidence across many scales and many different organisms, some of the generalizations we can make are that sensitive species tend to be later maturing, this can have longer generation time, have lower fecundity, so fewer offspring, and higher investment in those individual offspring.

That's trait‑based generalization. In the rarity side, we know that species that are more sensitive tend to be geographically rare, often constrained to one location or have few or small populations.

This is summed up really nicely in this 2014 paper by Pearson and colleagues. Here, the authors used a simulation‑based approach to look at factors that really contribute to extinction risks of species.

A nice figure from that paper delineates some of the major factors that contribute to extinction risks. What they found is the top eight factors contributing to extinction risks are driven by these spatial and demographic traits.

What this figure shows is that both of these life history characteristics or demographic characteristics, which are shown in the red, and the spatial characteristics shown in blue are major contributors to extinction risk. What I want you to definitely see here is that the occupied area which gets at range size is quite large.

What's also interesting that they found is that there are strong interactions between these things as well. In other words, they amplify one another to continue to contribute to that extinction risk. These trends are well known and are starting to get incorporated into literature and then to study its evaluating species' vulnerability to a changing climate.

What we know is the degree and nature of which they're incorporated is highly variable and can fall along this gradient of large extent, coarse grain studies that incorporate many species at once to studies that may be smaller in extent, have a very fine grain, detailed information, and focus on a single species.

If you think about these studies that fall towards the left side of that spectrum...Here's one example by Lawler and colleagues where they looked at the climate impacts on amphibians for the entire western hemisphere.

In this study, they looked at over 400 species and the grain size here was about 50 x 50 kilometer grid cell which in some cases can encompass even the entire range of some species. They modeled how those ranges of species might shift in response to changing climate, and looked at species turnover as well as in the other metrics.

For this study, this is one good example of a large‑scale study with a coarse grain with the primary metrics of range shifts or range loss. This kind of study really focuses on that exposure component of vulnerability.

On the opposite end of this gradient are studies that focus on a single species of a smaller extent, of finer grain size, and a framework for this type of study was proposed by Williams and colleagues in 2008. I put this figure out because the graph on the right is a nice, well‑done, comprehensive framework for assessing vulnerability of a species to climate change.

But you can tell just by looking, there's a lot of different components that go into this, everything from physiology to species' interaction, to genetic composition of populations, and the diversity of these populations within a species. This type of framework might be achievable for well‑studied species. But it's going to be more difficult to implement for species that are poorly understood where we don't have a lot of that information.

So this presents this logistical constraint, and that's based in this resource limitation and inhibits this from being applied to all the species we might want to, individually. Somewhere in the middle of that gradient, that's where there's an interest from the management side of what we can do with multi‑species vulnerability assessment.

There's a need for these regional studies that match the skill of management actions. Often, we can think about that as state‑level managers thinking about what to do and how to prioritize species.

These multi‑species studies give a lot of opportunities for things like facilitating regional prioritization, identifying challenges - so particular species that may be at risk for which we may not know much about, and highlighting opportunities, whether they're spatial or taxonomic species that are real opportunities for conservation.

One of the things that I'm interested in doing is thinking about not just one or the other of these two commonly applied approaches ‑‑ rarity‑based approaches and traits‑based approaches ‑‑ but combining them together for the application for these regional assessments.

That brings me to the objectives of this work. The first is to classify geographic rarity for the native ectothermic vertebrates of Oregon. Those are reptiles, amphibians, and freshwater fishes. What we're going to be looking at is the effects of study extent, grain size, and data type on those rarity classifications. I'll explain that in just a little bit.

We're then evaluating sensitivity by species trait, again, a second way to think about sensitivity. We're going to compare that to those geographic rarity metrics, how much overlap is there between this life history‑based sensitivity and geographic rarity.

Finally, there's a thought exercise to compare current protected status at both the federal and state level to this rarity‑based index and to the traits of these species.

You may be asking, "Why Oregon?" To provide a little context for this study for anyone not familiar with the state, here it is on a map of the United States. Here we are in the Pacific Northwest.

Oregon is really interesting because it is topographically diverse, and it has a wide range of contemporary climate conditions with pretty high variability in annual temperature and precipitation.

Oregon also has quite a few native vertebrate ectotherms. Here are the counts. For reptiles, there's about 30, amphibians, 31, and fishes, around 75. We've got an estimate there because the taxonomy of many of Oregon's native fishes is not quite resolved. It's a shifting baseline there.

What's also neat about Oregon is that there is this applied interest in this type of categorization and then thinking about an objective and systematical way to compare these species to one another that avoids favoring or overlooking any particular taxa and puts them all on the same plane.

We're working with Oregon Department of Fish and Wildlife on this analysis. That's neat to have them involved in this, as well.

Our data types, just to bring you up to speed on what we're working with here. We use point‑occurrence data for our analyses. These primarily came from museum specimen records available from VertNet, which has proven to be a great resource.

VertNet is an NSF‑sponsored online database. It's free, it's really accessible, and it integrates over 170 museum collections throughout the world.

What we did is we first downloaded all data available from VertNet for freshwater fishes, reptiles, and amphibians in our study region. Then we went through this flowchart of data quality assurance. We first, of course, had to have records that could be mapped, so we needed locality information.

We then looked at records between the years of 1930 and 2002, because that is what we were able to evaluate with our climate data. This also ensured that we had at least 30 years of climate data prior to any given observation.

As you can imagine the taxonomy for many of these species has changed over that time period, so we needed to standardize and make sure that all the species were being called the same things and that those were up‑to‑date.

Finally, we then generated maps that were reviewed by experts to make sure that our data looked good and there were no really obvious outliers or issues. Finally, we only retained species for which we had at least 10 observations.

That's the VertNet record, which, you might imagine, museum records don't in any way account for all of the data that are available. What we were also fortunate to have and able to use were additional data sets, one from Oregon Department of Fish and Wildlife and one from OSU.

This is the combined data set for fishes. That is, again, primarily based on museum records to avoid any fish misidentifications which can come from unverified reporting. It was more targeted, so this one did have a little bit more information than the VertNet database.

For reptiles, we actually had this very comprehensive database assembled by USGS and U.S. Forest Service. There were many people involved in putting this together. That database incorporates not just museum records, but also federal and state records, records from herpetologists, all sorts of information. Both of these types of data represent a lot of work to assemble.

They are more comprehensive, but they also take a lot more time to put together. One of our major questions here is: how well does VertNet approximate rarity? If there's a lot of overlap between these, that's really promising for us in terms of this being applicable for many different species, and different regions, as well.

That was our data type, which is one thing that we will toggle. The next thing that I wanted to look at was the effect of study extent on these rarity metrics. We're interested in...

I see a couple of chats here. Let's see here. "Was BISON considered a source of occurrence data?" Yes, it was for those reptiles.

For study extent, this is the single state, which is Oregon only. For studies that might look at politically bounded regions, this is a lot of times the way that data is reported. It's maybe potentially the region that managers are tasked with, but it may not be as ecologically relevant.

We looked at one study extent with just the state. We then looked at a multistate extent for Oregon, Washington, and Idaho. Then we wanted to compare those to a more ecologically driven study extent, of looking at the ecoregional overlap within Oregon State.

To zoom in on that a little bit, here's the outline of Oregon, and here's these overlapping ecoregions. You can see how, especially towards the south and the southwestern corner of the state, and then also in the southeastern corner, if you use that political boundary, you're truncating quite a bit of those ecoregions.

These data for this ecoregional approach were all from VertNet. We were able to get these comprehensive records from VertNet for that study extent.

Here's what the data sets look like. This is for our three different taxonomic groups here, reptiles, amphibians, and fishes. The study extent is shown there. This, of course, is the full extent, ecoregional and tri‑state.

All the little points are our occurrence points. The summary information is at the bottom here, with number of species for which we had at least 10 occurrence records, the mean count per species, and the total records.

This slide shows the VertNet totals. Here you can see those comprehensive databases overlaid. There's a couple things to note here. Those totals are shown in green below the two different groups for which they were available.

For reptiles, for example, we have two additional species that were added. There's almost a fourfold increase in both the mean count for species and the total number of records. Those were available, those data were assembled for Oregon, Washington, and Idaho.

In that case, we were able to compare the rarity classifications, just within the state of Oregon and at that tri‑state extent. For fishes, the additional database is only available within Oregon State.

The next thing that we had to do was assemble this geographic rarity index. To do this, we incorporated both area of occupancy, which is this metric of range size. We also looked at climate sensitivity. Species that were considered the most rare are those with the smallest area of occupancy and the highest climate sensitivity.

I'll explain both of those in just a minute. We'll walk through those methods. Again, we're really interested, not just in what those rarity metrics are and what that index value is for each species, but also how study extents, data type, and grain size affect that rarity index. I'll explain grain size now.

After establishing our occurrence datasets, what we did is we looked at the area of occupancy following the methods of Hartley and Kunin in 2003.

All you need to know, for the sake of this presentation, is that this is essentially a modified grided approach in which any cell of a grid if you divide the landscape up into a bunch of different grids is counted if a species is present, then it's not counted if the species is not present. All of the grids where a species is present or summed to get that area of occupancy.

We actually did a modified grid approach where we used buffered points as opposed to a set grid. This avoids any kind of edge issues with the grid itself. I'll show an image of that in a second. We used multiple grid sizes as recommended by Hartley and Kunin.

These different grain sizes or grid size can really give you different types of information about your area of occupancy. Here's an image of what that looks like.

In our case, we used four different grid sizes or grain size. It's the same thing for the purpose of this study. The first has this one kilometer diameter buffer. You can see at that scale, the actual area of occupancy closely approximates the count of data available. We went all the way up to a 20‑kilometer diameter point.

In those cases, points that are very close together are merged. That could be a really nice way to avoid bias from areas that are heavily sampled. For example in this case, this is the Lost River sucker. You can see that there are a lot of museum records collected along one reach of a river here.

By going to that larger grid cell, you help buffer the effect of that really heavy sampling in that area. The next thing that we considered is climate sensitivity, in addition to that area of occupancy.

For climate sensitivity, the main things we were interested in is the magnitude, seasonality, and variability of both temperature and precipitation that a given species might experience. Not just the actual values of those variables, but the range that the species might experience.

For example, if a species is found within a very narrow band of precipitation or temperature, you might consider it to be sensitive. Whereas if it's found in a wide‑range, under many different scenarios, it might be more of a climate generalist.

The data that we used are from Shafer’s work available through USGS. It's thirty‑second downscaled monthly historical data available from 1901 to 2002. There's a range of variables available there. Again, we picked ones which are to minimize collinearity, but also retain values to look at. Again, magnitude, seasonality, and variability of these key factors.

To look at our climate sensitivity, this aggregated index, we used this 102 years of climate data and the species area of occupancy. We're basically able to extract the full range that any given species might experience over that time period.

I didn't list all the variables here. I'm happy to take questions about that. We focused on temperature and precipitation. For two reasons, they're very important for this suite of animals that we're considering. They're also the most likely to change due to climate change.

Before I jump in to the results, just a quick reminder to bring everybody back to where we are. We looked at both area of occupancy, and then we looked at climate sensitivity of species. We aggregated those two to create this geographic rarity index. We're evaluating the effects of study extent, data type, and grain size on those results.

Here is the first of the results on this slide. This is geographic rarity shown for each species. It's a dot here on these different plots. They're scaled for rarity. The big blue dots are the most rare species. The smaller lighter dots are the more common species.

There's a couple of things to note here, climate sensitivity is shown on the Y‑axis. It's a relative metrics scale from zero to one, and area of occupancy is shown on the X‑axis. In this case, this reflects VertNet data at the ecoregional scale, and a grain size of 10 kilometers squared.

There's a few things that I noticed when I first looked at this which I find really interesting. The first is that climate sensitivity does appear to be constrained to some extent by area of occupancy. That makes sense. As you go towards a bigger area of occupancy, habitat heterogeneity tends to increase.

That's not necessarily a surprising result. One thing that is interesting here is that the spread of climate sensitivity that species with these lower area of occupancies experience can be quite large. It's certainly not a one‑to‑one relationship there.

This is again is just for one data type at one study extent - at that ecoregional extent - and for one grain size. But what was interesting is that the rarity metrics are really quite closely correlated across all of the different types of study design elements that we used.

This shows that here, and what these are...this is similar to what we just looked at, but what's shown, in addition here to the actual rarity points, is the standard deviation of an individual's position or species position along both of those axes.

These bars in gray indicate the standard deviation you might see along climate sensitivity, or that relative area of occupancy metric for different study extent, grain sizes, and data types.

What you can see is that these do appear to be fairly well‑constrained. If we look at different correlations across these different types of studies, what we find is that the Spearman's rank correlation coefficient for rarity values is around 0.7 or 0.8, up to above 0.99 for different taxa.

The take‑home message from this really is that grain size, you can see it on the bottom, really highly‑correlated across different grain sizes, so a very little effect there. Study extent had only a modest or small effect anywhere from 0.9 up to 0.94. Data type can only be evaluated for fishes where we had that overlap with the State of Oregon.

For reptiles, we were able to do that both within the tate of Oregon and across all three states. Data type had a slightly stronger effect there on reptiles with a 0.77. These are really pretty closely correlated which is exciting because it means that VertNet might be something we could really employ for these types of approaches.

This next slide shows how these species stack up. These all of our species on one slide, and they range from the most rare, at the top left, to the most common, to the bottom right, within those major taxonomic group.

Again, you can see the error bars for any given species that show how rarity might change depending on how the study is actually setup. Again, how those three different components affect rarity.

What you can see here is that for the most part, that ranking of the most rare to the most common is quite constrained. That's really exciting. Although I'm sure you are probably already noticing that there are a couple of outliers. I just want to focus on a few of those first before moving on.

Here, these are two species that really illustrate the effects of study extent. In this case, it's two fishes, the California roach and the Sacramento sucker. What happened here is that when you look just within the state of Oregon, you're picking up only really the very northern extent of the distribution of both of these species.

By truncating at that political boundary, these species look much more rare than they do if you incorporate the ecoregional extent. That's what contributes to the error bars here. A good example of how study extent might affect individual species.

There actually was one example of grain size. This is for the sand roller. This is a species where there's been a lot of museum specimens that were actually collected. What happened again is that effect that at very small grain sizes, the species might actually look more common than they do when you go to that larger grain extent where those individual records that are very close to one another become merged at that larger grain size.

Here are the effects of data source. Of course from this case, the best examples come from reptiles, which is where we saw that effect for the most part. This is for the aquatic garter snake and the western pond turtle.

What happened here...just a quick reminder that comprehensive data sets from USGS and Forest Service incorporated many different studies and sampling efforts. That includes a lot of samples that were taken for the western pond turtle that were incorporated into this.

For aquatic garter snake, it's hard to say if they are over represented in VertNet, or underrepresented in the comprehensive database. The VertNet does tend to put them as more common in that comprehensive database to shift them towards more rare.

In this case, I see a question here about the scale. That's a good question. Thank you. These are scaled. This is relative to one another. It's across all study extent to grain sizes and data types. It is relative. It's relative to one another, but it's in numeric scale.

I want to go over to some of our take‑home messages here. It's a lot of information. What we basically found is that rarity rankings are generally robust to grain size study extent and data source. There are some variation there though. That's why plotting them and looking at where those error bars are can help you identify species where there may be bias.

We also found that incorporating multiple study extents and grain sizes might help identify species where that bias exists. This is also good evidence that VertNet might be a promising tool for these types of analyses, or in regions where these comprehensive databases either don't exist yet or there may not be the funds or support to assemble them.

VertNet may offer a nice surrogate. That said, there's one really important caveat. That is that even with VertNet, there are species that fall through the cracks.

In this case, I put up lists of amphibians, reptiles, and fishes for which there were only partial representation from the data. That means that at one or more study extents, or from one or more data types, these species, that were not at the minimum number of 10 observations. 10 is pretty liberal. That's really very few observations, so that's interesting.

For fishes, there were quite a few species where there was no representation at all. Thinking about that, moving forward, there's a couple of things that are worth noting, the greatest number of species were represented in the ecoregional study extent.

I would argue that that is an important thing to consider if you're looking to employ this type of approach at a state scale to really go for more ecologically bounded study extents rather than the political boundaries because it is ecologically meaningful.

In our case, it was the greatest number of species that were represented, and those two were correlated to each other. The other thing is that some species whether they are recently described or for some other reasons just have few museum records, are poorly represented or aren't represented at all.

This is going to be something to keep in mind because what we found...I didn't show the results here. But the number of museum records generally and the number of individual specimens that are getting deposited into museums is decreasing through time.

Thinking about what other databases are out there that can help fill that gap is going to be really important. That brings me to the next part of this which is looking at the life histories of some of these species, and thinking about whether or not vulnerability or sensitivity, that's based in these life history strategies, is related to geographic rarity.

To look at this, we evaluated those relationships both qualitatively by looking at key life history traits and comparing that to rarity using a multi‑varied approach or principal coordinate analysis. I'll show you those results in just a minute.

We also combined all taxa together to increase our statistical power and performed a classification and regression tree to look at how well, not just the life history traits predict geographic rarity, but how well primary habitat association and taxonomy play in as well.

Our goal with this part of the study was to really think about measurable, comparable traits that reduced the need for relative or categorical metrics across species.

The use of traits to assess species' vulnerability or sensitivity to climate change has gained much attention in the literature. There's certainly a lot of studies that are coming out looking at this. Many of them do rely on these expert opinion‑based categorical metrics.

In this case, we were aiming for traits that can be directly measured and directly compared allowing us to look at all three of these major taxonomic groups together.

The traits we included are agent maturity, fecundity, and investment for offspring where the later maturing individuals with lower fecundity and again higher investment per offspring are hypothesized to be sensitive. Those three traits were key. We also looked at body size and longevity.

Not only because those are often linked to these other key life history traits, but because all of these traits together, these five key traits, not only might tell us something about sensitivity, but they also speak to some extent to adaptive capacities.

You can imagine that body size is often, in many studies, related to mobility and the ability of species to migrate in response to a changing climate. Longevity speaks to the ability of species to wade out bad conditions. As maybe they have a really harsh year, this allows them to skip generations.

This traits‑based approach may actually be addressing two components of vulnerability. Here's one of the first results filed here from the traits‑based work. Just to make sure everybody is on the same page, this is for reptiles. Each of these points is a species.

This is a principal coordinate analysis. What you see here are two axes that explain the most variation of these five different traits. The traits are shown here as the vectors. All you need to know is the points that are closer to the end of those vectors are associated with that trait.

For example, those that are in the top end of the plot where we see this big arrow for fecundity, those are species that tend to be more fecund. Those towards the bottom portion of the plot would be less fecund.

Looking at where these species are, we could hypothesize that these species in the red highlighted area that are longer generation time and have lower fecundity, we can hypothesize that those might be sensitive to a changing climate.

The next question is how well does that overlay with rarity. If there's a close overlap, we might expect that the most rare species would occur in that sensitive trait space. But in fact, what we see is much more of a shotgun pattern of seeing rare species throughout this plot as well as common species just spread throughout this life history space.

For amphibians, this is again the same idea here, same type of plot. You can see your different traits with the vectors. Again, if we hypothesize sensitivity here, we have this group here of later maturing individuals with somewhat lower fecundity, a higher investment for offspring which is shown by egg size.

This might be where we can look for those overlap with rarity. In this case, we do still have that shotgun pattern, but you can see that there are some species that are clustering there. For anyone who's interested, those are actually primarily plethodontid salamanders.

Here, maybe a little bit of overlap but you can see that common species also are spread throughout the trait space, and there are other rare species that are not falling within that same hypothesized sensitive life history space.

Finally, for the fishes. This one is a little more tricky, because for fish life history, there's this really well‑established relationship between age at maturity and fecundity. We don't have that nice gradient that makes for really obvious sensitive life history space.

But most of the literature including that Pearson paper that I showed earlier indicates that generation time is a major driver in the sensitive life history. If we go with that and put a question mark here to highlight that this is maybe a less strong relationship, this might be the area where we expect to see more sensitive species.

Again, here with rarity, for the fishes in particular, just really all over the map for geographic rarity. That was the life history approach and a little more qualitative. Now, what we did is we combined all taxa together at the ecoregional extent. We evaluated not just those five life history traits but we also looked at habitat associations which in this case are really basic.

It was whether or not a species was aquatic, semi‑aquatic, or terrestrial. And we looked at order of the species to get at taxonomy. We evaluated this again with that classification regression tree. What we found is that there was absolutely no support for any relationship between any of those traits and geographic rarity.

For anyone who's interested, the way that we knew that was that our cross validated standard error increased with three sides for every combination of variables that we examined.

Here's to visualize that. You can get a look at it for yourself. Here is that taxonomy component. The order on the bottom across the X‑axis, and rarity on the Y‑axis. You can see here there is some variation between some of the groups, but none of this was statistically significant and didn't come out as a good predictor of rarity.

Finally, here is that really simple habitat metric here of aquatic, semi‑aquatic, terrestrial, and rarity. You can see that not only are there no real strong significant difference here, but there's quite a range of rarity values for some of these different groups.

What this tells us is if we're thinking about these rarity‑based approaches and traits‑based approaches, one of the major questions is how much overlap is there between these two approaches, are they telling us the same things, or different things?

What we're really starting to see is that for these organisms at this regional scale, they're much more complementary than they are overlapping or redundant.

Although we can't use one to predict the other, they may be giving us more information combined, which again is very much in line with what Pearson and colleagues found, that these two interact with each other. That these approaches really provide complementary information, and may be describing different components of vulnerability.

The last thing that I want to show you before we wrap this up and I take any questions is just looking at how these rarity metrics, particularly geographic rarity stacks up to federal status and state status. I'm going to come back to a plot. Again, what we're trying to do here is highlight geographically rare species that might be getting overlooked for any number of reasons.

It's important to note here, before I show you these results, this is really more about sensitivity. This does not include exposure. There's only one part of the puzzle here. But there's still important information to be gained particularly, again, for those species that are geographically really rare and that are currently not protected.

First, I want to start with the federal list. These are under ESA. You can see here at the top. I'm going to show by species whether they're endangered, threatened, or a species of concern. There's also going to be a marker for species for which subspecies variation is considered. These could include distinct population segments or evolutionary significant units. Here's what this look like.

A couple of things to note here. We can see there's quite a bit of spread. There are geographically rare species that are not listed across all three of these major taxonomic groups particularly reptiles. There are quite a few common species that do have some protection.

But again, as you get further down that list, you're really seeing these species, particularly obvious for the fishes that these are, at that subspecies level, that protection is being granted. This is for the federal level. Here, I want to show you the state level.

These are listed either as critical or vulnerable by ODFW. Here again is that list. You can see a wide range here, and further down the list you go, for most of these species, there is some subspecies variation that starts to come in. That's one way to look at that. This is on the individual species level.

I also wanted to look at overall. This is the rarity on the Y‑axis, and this is the federal or the state level. This is for each one of those categories on the Y‑axis. The species that, just based on this broad approach that we may want to first look at, are those that are falling in this area, high rarity but not listed for whatever the reason might be.

To think about that and put this stuff in the context, one thing that we did see just by this thought exercise here is that it does appear there are some geographically rare species that may be overlooked or not listed for any number of reasons. We'll get to those in a second. There's also some common species with conservation status that reflect subspecies level variation.

Those tend to be fishes, and they tend to be salmonids or other economically or culturally valuable species which makes sense. But it's important to keep in mind that that level of variation is not evaluated for many of these species. There is maybe a taxonomic bias there for that actual evaluation.

Again though, just to keep all this in mind, this is sensitivity only and does not incorporate exposure. For some of those species that appear to be more common, but they perhaps are listed, that could be because there's an exposure component that's not getting picked up in the sensitivity piece of the vulnerability.

If we think about this a little bit further too, let's say we have that list of species that are geographically rare, but that are currently either poorly understood or not listed, some important questions to consider, it doesn't necessarily confer that all these should be protected immediately.

There are some important questions to think about first. One is the distribution of species outside of Oregon or the Pacific Northwest. This is a regional analysis, what do those species look like outside of this region? Even though we took this ecoregional approach, is there some reason why we're only picking up a portion of their distribution. That one is important to consider.

Another is the potential for overlooked genetic or subspecies variation for many of these species. That is just something that, for many of these, we may not know even though they are species that might be doing well throughout the book of their range, they could potentially either be genetically distinct or ecologically distinct from other portions of their range within Oregon and Oregon’s eco regions.

Finally, the next obvious piece here is to look at whether or not those life histories might tell us something about sensitivity as well. For those geographically rare species, do they also have a life history that could indicate some sensitivity or limited adaptive capacity as well.

The real goal here ultimately is to really think about standardizing the approach and putting all these species on a level playing field, and considering them in that way.

To wrap this up, to summarize here across all the different things we looked at today. This approach of combining traits‑based approaches and rarity‑based approaches potentially is promising for regional multi species assessments. Both for identifying focal species for monitoring and potentially prioritizing species for which we assess exposure to climate change and in addition to sensitivity. That's the next phase of this project.

Today, I really talked about those lower two components and in particular sensitivity. The next thing we're interested in doing is evaluating exposure for a range of species. Incorporated in that exposure will be things like climate velocity within species ranges, land use, and habitat connectivity to think about movement corridors.

What we're doing at this point is trying to think about which species we will be able to evaluate this for. We would like to look at a range of life histories and geographic rarity. We're also going to be somewhat constrained on the logistics front as exposure requires much more data and higher quality data to assess then sensitivity and adaptive capacity alone.

With that, I'm going to wrap up. I just want to thank everyone so much for listening and for tuning in. I want to, again, acknowledge my collaborators and thank them for their hard work on this project, as well as a couple of great undergrad volunteers, Mike and Abby.

Thanks for listening. I think we do have time for questions. My contact information is here as well, so please free to follow up with me with any questions or comments. Thank you.

John:  I have a...Jake has a question. Could your term climate sensitivity be more simply described as an envelope which avoids suggesting causality?

Meryl:  Thanks, Jake. I think so. In this case, you're right. This is sensitivity in the sense of these range of factors that through these correlative methods, that we see these species experiencing. I think, you're right that we want to avoid that language that might imply causality. We'll be really careful with that for sure. Envelope would work there.

John:  Stephen has another question. Is the numeric scale for rare to common, or is it purely qualitative?

Meryl:  It is numeric and it is based on that area of occupancy metrics, smallest to largest. Those are standardized from zero to one for comparability. It is basically the average of the area of occupancy index with the climate sensitivity to get that composite rarity index. It is numeric.

John:  Another question, from Jake. Could you expand on the potential role of the adaptive capacity component of vulnerability in terms of this either study or more broadly the taxa?

Meryl:  Absolutely. I got you. Great question. In this case, you'll notice any time that I put up the vulnerability figure, the adaptive capacity piece and the arrow to it was a little bit smaller and maybe gray. That's because that piece is really difficult to characterize. With these traits that we are considering, they're really only one part of the story there.

Another big one is mobility. Maybe body size is related to that. We can see that for a lot of taxa, but a lot of that is also are behavioral. Some of it is going to be the tendency or the ability of species to move. Then other pieces of adaptive capacity. Behavior is going to be really a major one. That's a hard one for us to get at. The other thing to consider is physiology.

There's a lot of studies looking at this now and thinking about what that realized versus fundamental niche incorporates, whether or not we're doing a good enough job at getting at the fundamental niche and what we're missing. I would say that we are just barely scratching the surface on that.

And that adaptive capacity is difficult. When we think about that range of the large scale multi-species studies to those that are species specific, that's one of those things where those more detailed studies tend to do a much better job at capturing that.

John:  We have a question coming in, Jonathan asks: what advice would you give a listening biologist about how to efficiently evaluate species for listing based on your work?

Meryl:  Again, a really great question. I think, one thing to keep in mind with this type of approach is this is really designed and meant to be a first step. This helps highlight species that really should at least be on the radar. This work is not going to be the first and last step in terms of defining which species should be listed.

But it would help identify species where you might want to dig in a little more to see what other data are available to assess things like range size and other important components or attributes that go into that listing decision. The other thing I would say that I think is telling from this work, again, we're still looking at how this stacks up with the listings.

There is an element of taxonomic bias there in terms of things like population structure and subspecies variation. It's important to remember that for some of these species that we have evaluated, the range size of the actual species at the species level may be smaller than some of the subspecies either DPS or ESU populations that are protected.

Approaching this with even a course or broad scale approach like this is still really useful in getting everybody on the same playing field.

John:  Ok and from Darren, how can you actively evaluate geographic rarity in a fragmented landscape? Generally, what are your thoughts on that in the context of exposure?

Meryl:  That's really important. What we did here, of course our metric is just that kind of footprint on the landscape area of occupancy. There are many other metrics you can incorporate to look at things like fragmentation. We have not done that. At least not yet for the sensitivity piece.

I agree that it's really important for exposure. Two important things to think about is how the landscape has changed through time. Is it becoming more or less fragmented and making sure that the data you’re using accurately reflects that or captures that.

In the context of exposure, my colleagues and I are thinking about how we're going to try to approach this for this exposure piece. Of course for aquatic species, for fishes in particular, thinking about things like barriers which are really obvious components of fragmentation will be absolutely critical.

For terrestrial species, there are all kinds of approaches like thinking about landscape resistance or other major factors that are known to inhibit movement.

That's going to be really key when evaluating how ranges might shift in where they might be limited. In terms of fragmentation, we did not explicitly incorporate that. With the data that we have, we try to keep a broad, as assumption‑free approach as we could. It's important, so thanks. Thanks Darren.

John:  Thank you. At this time, there's no more questions. I'd like to thank, Meryl. I'd like to thank the USGS as well for their participation in this. I'd like to thank everyone that was on the line with us today.

Meryl:  All right. Thanks, John. Thanks, everyone.