 All right, we are one minute past the top of the hour. I see we have more attendees trickling in, but I'll get started with introduction so we can leave plenty of time for our speakers. My name is Samantha Weintraub-Leff. Thank you very much for attending this month's Science Seminar presented by the National Ecological Observatory Network, which is operated by Battelle. Our goal with this monthly series of talks is to build community at the intersection of ecology, environmental science, and neon. We are very excited to have Roland Kayes and Ariel Parsons with us to speak at this month's Science Seminar, but before we turn it over to the speakers, a few logistics. So first, we have enabled optional automated closed captioning for today's talk. If you would like to use this feature, please find the CC button in your own personal Zoom menu. The webinar will consist of a presentation. The speakers will both be presenting, followed by a Q&A. Now, as you think of questions throughout the talk, feel free to pose them in the Q&A box, and we'll facilitate a discussion at the end. There should also be opportunity to ask questions over audio. So when we get to the Q&A period, feel free to raise your hands, and we should have time to unmute and take a few questions. Neon welcomes contributions from everyone who shares our values of unity, creativity, collaboration, excellence, and appreciation. And this is outlined in the Neon Code of Conduct, which you can find linked on the bottom of our Science Seminars webpage, which I have also put in the chat, and I'll share again. So please check out our Code of Conduct. It does apply to all of the Neon staff, as well as anyone external to the Neon program who's participating in an event. And of course, we don't expect this from our beautiful community, but anyone who would violate our principles would, of course, this would be grounds for removal from the meeting. This talk will be recorded and made available for later viewing on the Neon Science Seminars webpage, which I'm showing here. So all of the talks when there is a recording, you can just click on that, and that will take you to the recorded talk on our YouTube page. To compliment these monthly Science Seminars that we are giving, we are also hosting related data skills webinars that teach people how to access and use Neon data. You can register for those on the same Science Seminars webpage. You just keep scrolling down. You can learn more about the data skills webinar. And the one coming up at the end of January is extremely relevant. Our own Sarah Paul, who you'll hear from in a moment, is going to go through a tutorial of introducing the Neon Small Mammal data. So we hope you can join us for that if you're interested. And lastly, we are soliciting nominations for the 2023-24 round of speakers for this very seminar series. So please consider nominating yourself or a colleague today by filling out this form, which is linked here near the top of the seminar series webpage. Okay, that's it for me. Now I'd like to turn it over to Sarah Paul to introduce today's speakers. All right, thanks, Samantha. Yeah, I'm very pleased to be introducing Dr. Ariel Parsons and Dr. Roland Kayes. And they're gonna be giving a joint presentation on Neon Mammals, large and small. So Dr. Ariel Parsons is a research scientist in the Alexander Center at the Lincoln Park Zoo. She's a quantitative ecologist and population biologist and her work focuses on wildlife population dynamics, movement and interspecific interactions. And she's worked extensively with large-scale ecological data sets and citizen science observations on a range of questions like global change impacts on mammal abundance and modeling factors that affect disease risk in humans. And then Dr. Kayes is a professor at North Carolina State University and director of the Biodiversity Lab at the North Carolina Museum of Natural Sciences. And his work focuses on ecology and conservation of mammals using camera traps and animal tracking to study animal movement. And he's also the co-founder of Camera Trapping Database Wildlife Insights and the Animal Tracking Database Move Bank. So with that, I'm very excited to turn it over to the speakers and I think Ariel is gonna start. Great, thank you so much, Sarah and Samantha for inviting us here. Roland and I are really excited about presenting our work. Let me get this started here. So yeah, so we're gonna be talking about some of our work looking at analyzing the neon data and also adding some of our supplementary camera trapping data to look at trends in populations of mammals both large and small. So we're living as everyone is no doubt aware in this age of relatively rapid change across the planet from rapidly increasing human populations which comes with it rapid habitat destruction deforestation and then of course climate change. And all of these changes, of course, exert pressures on ecosystems, on communities and on populations of plants and animals. And so to really kind of understand these pressures and how these pressures might be influencing population change over space and time, we need to collect a huge amounts of data across these really large spatial scales but also these really long temporal scales. So these kind of yearly or monthly data and have the ability to really analyze them on a variety of species and locations and scales to understand how these changes are influencing populations, communities and ecosystems and ultimately if there's some way that we can mitigate of course these changes. Now just because these changes are often detrimental to some species or communities or ecosystems there are of course some success stories. The classic would be that of the common raccoon but for every success story, especially when we're thinking about animal species and we're thinking about mammals, there are of course these stories of extinctions in the wild and increased extinction risk. And so there's a real need to understand of course how this is impacting animal populations, wildlife populations and we're seeing these trends of course of reductions in global biodiversity and the implications of course for those reductions on things like ecosystem services. And so getting these large scale data sets, large both spatial and temporal scale data sets allow us to really understand and monitor these changes, identify them as they're happening of course but ultimately understand what is driving them and what can we do to maybe halt or eliminate them. That's of course the dream. Now, NEON is going a long way of course to enabling us to do that here in the United States. So we've got this coverage of the continental US, we've got Alaska, Hawaii and Puerto Rico included and over these 81 field sites we're sampling in each of these 20 ecoclimatic domains and so we're getting of course that spatial coverage and we're getting coverage that will enable us to address questions about a lot of these large scale changes that are taking place across the country and across the planet and then continuing that through time, we've already gotten some cases nine years of data and counting which is super exciting and then continuing that through time is going to enable us to get that temporal kind of long temporal scale to be able to really understand these changes and these dynamics that although I'm calling them rapid, in reality it takes years and decades to perceive those changes and really understand them. So really exciting of course data set, I don't need to sell you guys on it but just kind of to give that context we of course have all of these data these biological data which include the mammal data that we'll be talking about today but of course across a number of different taxa, plants, animals, arthropods, microbes kind of runs up and down in terms of food web dynamics and entire communities and ecosystems and then we've got all sorts of abiotic data collection on different scales, other kind of habitat variables and that really again gives us the ability to be able to really model in terms of changes in some of these especially animal communities changes in these abundances over time and trying to identify what factors whether they be biotic factors or abiotic factors may be associated with these changes or even be driving them. So Roland and I are our mammologists, I would say so we will be focusing on mammal populations today. And when we think about changes in mammal populations over space and time or I should say over space this really happens kind of over these short time scales or over these more long-term time scales where we start to see and perceive some of these differences and be able to make predictions in terms of climate change, for example. And in both cases we can get these comparisons over time and space. We can start to make inference on drivers of population change over time and space but it's really these long-term trends that require very long-term data sets where we can start to see trends in populations over time and we can start to really understand the variation in terms of drivers of that change and be able to make the most robust predictions for example, how we might expect species to adapt or respond to climate change. For the purposes of the talk here and the results that we'll be presenting to you we're gonna be focusing on these short-term changes in abundance. So at each of these neon sites or I should say at a subset of neon sites there is a small mammal trapping. And so today we're gonna be talking about the small and large mammals. I will be starting with the small mammals and focusing on that neon box trapping data set. And then Roland's gonna follow up talking about some supplementary data analysis we did at neon sites where we put out camera traps to survey large mammals. And so he'll talk about some of those findings and relate them to the small mammal findings. So the small mammal box trapping data set most of you are probably aware of it but these are small mammals that are under 500 grams they're nocturnal, they forage above the ground they're non-volant. So those are the species that we're talking about when we talk about this data set. They're captured in Sherman live traps across 46 sites that had at least at one time had small mammal sampling. And this covers most of the echroclimatic domains neon. At each of these sites, there's anywhere from about three to eight small mammal trapping grids of a hundred Sherman traps that are set. And this happens several times a year. And then each sampling will be done for just a single night or for three consecutive nights. All within the same approximate time in the lunar cycle. And so from this effort, we get three main data sets. We get a large mark recaptured data set with individual IDs for each of the individuals that are captured. We get some DNA sequencing data so that we can correct the identifications for especially for those species that are difficult to identify in the field. And then we get some disease information which I'm not gonna be using and talking about is another kind of cool data set that's related to this small mammal box trapping data set. So when we get all these data and we get kind of these counts of species from the traps, we can develop maps based solely on these counts which we call an abundance index. Here's an example with perimuscus miniculatus. And this can kind of give us an idea of where the species might be relatively more abundant versus less abundant. We see the species relatively more abundant in the north and the west less abundant or in fact actually absent from a lot of these southeastern areas. So it can give us kind of that range map sort of idea. The problem with this index which is just solely based on the counts is that it doesn't always represent true abundance or at least if we're trying to follow these trends over these large spatial scales and over time, we might not be getting an accurate picture of the overall trends and the overall changes. So I'll belabor this point geekily for just a minute. So these rock counts that we use for abundance indices are actually the product of two different quantities. So these are the counts of individuals that we capture in traps during a trapping session at a trapping array. So one of those grid of 100 traps. The products is a product of the true abundance of the population of a given species on that grid and the probability of detecting individuals when we go out and capture. So that if P, the detection probability is one, then we would be capturing all of the individuals in a population every time we went out and trapped. But those of us who have done field biology know that that's almost never the case and usually P is less than one which means that we're just getting a subset of the population. Now that might be okay for the purposes of tracking trends and abundance over time and trying to determine what might be driving those changes but it also might not be okay. And I'll give you kind of some cartoon examples. So here is this index of abundance C and it's kind of trend over time or space. Got some trend. Now if detection probability is constant then in fact that index C really reflects the trend in true abundance over time or space and could be a reliable proxy and whatever inference we make on changes in that index over time and space would be valid because they would be replicating basically what we see in abundance. But what tends to happen sometimes or perhaps more often than we would like to think is that we see this trend in our counts and we also have a trend in our detection probability over time and space such that it could be that the trend that we're seeing in that index is actually a trend in detection probability and not a trend in abundance. Maybe abundance is constant. This is just an example but it is a way to illustrate the importance of accounting and estimating detection probability so that we can get an accurate representation of the actual changes in abundance over time and space and what's actually happening in a population. To drive this home, this is some of our results from the NEON data where we estimate the detection probability for several species across different sites so different trapping grids across the NEON system. And we see that over those sites so these box plots are quite tall for some of these species, for most of these species which indicates that we have quite a lot of variation in detection probability across sites. And so if we simply use the count as an index we might be missing of the true trend and the true differences in abundance over those sites. So this was all a rather long-winded way of saying the catchery catcher data that we get from NEON is really valuable. And in particular the catchery catcher data for those trapping plots where we have at least three nights of successive sampling it is with those multiple nights of successive sampling usually three or more where we can get our most accurate estimates of detection probability that we use then to correct our counts and get estimates of abundance. And we can also use those multiple nights of sampling to assess factors that might affect detection probability and that can be interesting in and of itself. So using the catchery catcher data from NEON we were able to generate over 9,000 total abundance estimates for 47 species spanning 2014 to 2019 at over 300 of those trapping grids which represented 44 geographic sites. So a whole lot of data, 9,000 total abundance estimates it's a lot of abundance estimates. So really cool that we have that 47 species is really amazing. And we did use the genetic data available to correct any ID. So we had the most accurate data and the most data for some of these species that was possible. And so with these estimates we're able to see trends in abundance over time and space. So here's a look at trends in abundance over time. This is one trapping site. So this is abbey 002. This is one trapping array, one grid and the five species that were most commonly detected. So those five species for which we can get abundance estimates. And we see quite a bit of fluctuation for some of them over time. Here's another example at ORNL039 where three species we were able to get abundance estimates for three species there. And we see still some sub fluctuation maybe not quite as drastic these abundances are quite a bit lower. So we see kind of that difference across these sites and over time. And what's interesting is if we then kind of associate that and overlay that with some of the environmental data that we get for those same sites in those same time periods, we start to see some species that really track and are really well correlated with patterns in temperature or precipitation. So for example, the one that I think is really stark is at ORNL039 taniastriatus seems to really track that temperature quite a bit. So there's some clear associations that we can make between some of these trends in abundance over time and environmental factors, which can kind of lead us in terms of trying to tease apart what might be driving some of these changes. Of course, we can also plot these changes in abundance over space. Traditionally, we would do that with a map kind of like this where we can see where are we detecting a given species? This is perimuscus miniculatus. Where are they found at their highest abundances versus their lowest abundances? And so that's really useful in terms of range and understanding how ranges might shift and change and understanding how different environmental factors at these sites might be supporting or inhibiting potentially these high abundances. One other way that I like to look at that and that we've been looking at it with the neon data is with what we are calling niche plots. And so this is kind of a way to look at those spatial trends, those spatial patterns in terms of environmental factors. So here we're looking at the abundance of dipodemies or DI along the axes of temperature and precipitation. And so we're getting the frequency of their abundances. So weighted by their abundances, such that the yellow color in that plot is showing us where we are finding them at their highest abundances. And then the dark blue is showing us where we are finding them but at relatively low abundance, kind of the edges of their range and their tolerance or their preference, if you will. And so what we're seeing in the yellow is that this species is a high temperature species and a low precipitation species. And of course for your small mammal biologist out there, this is not news. This is of course an arid species, but we kind of can get that this pictorial idea of their range and of their habitat preferences of their environmental tolerances from these niche plots. And then over time, we could track any ships as environmental conditions change or as new species are introduced to the community. There are lots of different ways that we can kind of look at trends in these associations over time and space. So so far, I've talked, given you some population comparison examples of the neon data, this neon small mammal data, and I separated them to those over time and those over space. But the beauty of of course the neon data is that we have both. And we can make these simultaneous, this inference on simultaneous spatial and temporal patterns, which is really excited that, you know, again, like one of the real strengths of the dataset and of the continued monitoring of these sites. So here's kind of an example of what I mean. This is not from neon data, this is from eBird, but it's the same idea where in the red, we have areas where the woodthrush over time, woodthrush abundance has gone down. And in the blue are areas where woodthrush abundance has increased. And so trying to understand, of course, what is driving those changes is the key. You know, can we determine what's happening there and why this range is essentially shifting? Because of course then that helps us understand how these, you know, global changes are affecting wildlife populations and maybe what we can do. So to get that with the neon data, we used a modeling method, which is called G-JAM. It's generalized joint attribute modeling. And in this case, I'm looking at modeling changes in abundance over time as a function of three components. Okay, dispersal, not a big deal for small mammals. So we can, you know, more or less ignore that, but still, you know, component here of this model. The environment, so environmental factors, of course, like climate, habitat, you know, what are the preferences, what are the needs of each species within a community? And then we can make this, this is a community level model. So we can actually incorporate other members of the community and assess species interactions. And in the small mammal community, we're talking competition. So how do these three components ultimately affect or contribute to changes in abundance over time for each member of the small and animal community that we sample or at least that we're able to estimate abundance for from the neon data? And so this is a dynamic version. So it allows us to look at this over time. And the input is those true abundance estimates taken from the capture recapture from the neon dataset. And this all comes from Jim Clark at Duke, if anyone's interested in this paper. And so when we run this model on the neon abundance estimates that we generated, we can rank what is the relative importance, what is the relative sensitivity of changes in abundance over time to those three components, the environment, dispersal or species interactions. And so we've got here 16 species from the neon dataset. And we can see the environment in the orange ranks significantly higher than the other two components for almost every species. Meaning that the environment is more important to in terms of being a potential driver of changes of abundance over time. It's more important than dispersal or species interactions for most members of this community. And so that environment encompasses several things. And we did include covariates for food availability, so mast, grass seeds, so those measurements that come from the neon data, understory, leaf litter, other measurements that are part of the neon dataset. But above all, it ended up being climate that was the most important environmental factor in terms of changes in abundance over time for these species, in particular temperature and precipitation, ranked much more highly even than mast availability or grass seed availability. Now, just because the environment ranked higher than species interaction for almost all of the species in the community, it doesn't mean that we weren't able to identify some seemingly important interactions between species, so competitive interactions in this community. So here we are looking at the effect of the density of species one on the y-axis on the growth rate of species two on the x-axis. And the darker colors represent pairwise competitive interactions that were particularly strong. Where species one, in this case, either perimuscus manipulatus or perimuscus meucopus seem to be exerting an effect in terms of their density at higher densities in suppressing the growth rate of these four other species down below. So this is an indication that in some cases there are some strong interactions potentially in these small mammal communities. Now, whether or not they're having a big effect is unclear and deserves some more study, but we did find a couple interesting things. I'll just give you a couple of examples here as I wrap up. The first, so these were two species, perimuscus manipulatus and perimus flavus, which we identified as having a strong interaction or a relatively strong interaction in this community. And we do see a pretty clear negative relationship in abundances where at high abundances for perimuscus, we tend to get low abundances for perimus. So that certainly supports some sort of competitive interaction. What was more interesting to me was this relationship kind of played out along gradients of environmental factors. So in this case, I'm gonna show the example of temperature. So here we are looking at that frequency-weighted abundance, I should say abundance-weighted frequency for perimuscus manipulatus. So kind of again showing where is their temperature tolerance, where are they found at the highest abundances and that's at that peak of the curve there. This is where we find perimuscus. And the purple here is that same abundance-weighted frequency of perinatus in the absence. So at sites and during times when perimuscus is absent. And so we see quite a bit of overlap with along that frequency, abundance-weighted frequency of perimuscus. So they're using similar temperature kind of ranges. But here's what it looks like in the yellow when perimuscus is present. And we see that perinatus is either not present or present at very, very low abundances at sites and during times when perimuscus is present and at the optimum of perimuscus. So this is kind of again an indication that we have this mediation of this competitive interaction potentially by some of these environmental factors. So this is really interesting and something that we're gonna be studying more and find something that was made possible by the neon data and the neon data collection. So these comparisons of small mammal populations over time and space base suggest some drivers, potential drivers of population change. Climate, habitat and competition with climate seemingly ranking at more important to the small mammal community than habitat or competition for most species. Now whether this holds for large mammals, so we'll have to let Roland tell us. So I will hand it over to him. All right, hi everybody. Thanks Arielle, I'm gonna jump in now and go on to talk about some of the large mammal stuff that we did with camera traps. Some similar questions, some different questions. Unfortunately not getting into competition, but... So camera traps are great because they record the presence and the relative abundance of larger species without any bait, without any interaction. We're just putting these cameras out on the trees and seeing how often animals walk by. And our goals of this were to, for one, let's just see what animals are out there at these neon sites and what are some of the factors that drive their abundance and distribution. So some similar questions to what Arielle talked about. And then also look, can we start to map out these mammal communities across the continent? Do we start to see differences where are those differences? And can we use these data to actually create these maps? And in particular, we also are focusing in on the importance of mast, the seeds and fruits that trees make. Cause this is mammal food. And this was part of what Arielle talked about. That was part of our habitat measurement, but I wanna go into this a little bit more detail. This comes from our, because here's some mammals eating mast, right? Obviously important for a lot of the squirrels, but also one of the larger mammals here, you can see a gray fox with some seeds in his mouth. And this data comes from our colleague, Jim Clark, down the street in Duke University. And this also is made in part with neon data and other data, but basically lots of these seed traps that allow him to fit this model that says for a given kind of tree of a given size and given conditions, how many seeds will they make on a typical year? So this is not year to year variation necessarily, but once he does that, he can then make predictions for other places based on the forest inventory plot. So for example, if you have a FIA forest analysis, an inventory plot where you know how many trees there are and what size they are and what the conditions were like, he can predict on a typical year, how many kilos of seeds will be made for each one of those. So, basically he's saying, if you tell us your forest composition, I'll tell you how many seeds that we'll have for a given species on a typical year and he's working on year to year predictions in the future. So this is some of the data that we have in here that turns out to be pretty important. So camera traps are not part of the standard neon protocol. And so we had to run them ourselves as part of this project. So we visited all these sites. There's a couple of non neon sites here as well. We tried to get to all the major regions that had trees because of this focus on this tree mast, some of the desert or predominantly grassland sites we did not visit. And so I'll show you around a little bit as we go in more detail. And our basic idea was to get a representative sample of the site. So we put out 50 cameras, about 300 to 500 meter spacing on a grid. We leave them out for a month or a little bit later. Generally we would set them and then some of our gyms group, the plant crew that was going out and checking these seed traps, they would pick them up, send it back to us and we process the data. So I'll show you a little bit. It was really fun to get a chance to see some of these different places, lots of different habitats. We teamed up with lots of different camera trappers all across the country. We had the joy of getting stuck, cars stuck on sandy roads, setting, when you set them on these trees that had just burnt, you'd get all covered in soot from setting cameras on the trees. Here's a camera in Ordoa Swisher that they burned while we were there and the burn crew was actually nice enough to defend our tree they stood there. You can see the camera with the tree of the camera on there not burnt. It actually saved our camera for which we're very grateful. On up to Talladega, some great pine forest, a lot hillier than I was ever expecting. Tromping around through their setting cameras was a lot more exhausting than I was expecting. And we've had some in the Southeast got a number of these invasive feral hogs looking quite fecund there. Also some beautiful bobcats looking out for the piglets. On up to the kind of Mid-Atlantic, we hit a number of sites here, great cool forest and understory vegetation in Mountain Lake biological station in Virginia, nice open places, lots of places to set cameras. Are these bears dancing or fighting? They were fighting and they actually, we didn't put bear boxes on our camera unfortunately and we got some of our cameras chewed up at this site. Here's a picture from Duke Forest where there is a neighborhood nearby and a neighborhood cat wandered in where he shouldn't have I think. On up to the Northeast to Harvard Forest and up to a site in New Hampshire in the White Mountains as well. Now this wasn't actually a real, we brought a couple of extra cameras just to set in really dramatic places to try to get good video and this was one of those. We did get some moose up there, which is super great. Up to the Great Lakes, we hit Wisconsin and Michigan got some nice fisher in this forest. On to Kansas, got to set some camera traps with Bob Tim, a famous mammologist from Kansas. Now, when we got to Kansas, it's mostly grassland but they have some of these forests that they've been protecting from fire and they warned us that there's a lot of poison ivy and there's a lot of ticks and we're like, yeah, yeah, we're field biologists. We've seen poison ivy and ticks before. The holy cow, here you can see Bob Tim not amused at the condition of this forest and the number of ticks. And here, if you look at the back of this deer, you can see all the ticks on its ear. I've got a lot more, it's pretty dire actually. So this was my least favorite forest and I do hope a fire finds its way to this forest at some point. And then out to the west, the mountain research station where we ran out of trees and had to strap to some rocks, some nice views there and Yellowstone where sometimes your random point on a map gets you climbing up a rock face but the pictures are worth it. The data is really cool. We get some large terrestrial birds as well, of course, turkeys and grouse. And then finally out to the west where actually the soap site burned a couple of weeks before we were supposed to go there which is I guess a better than a couple of weeks after we went there because I know we had to burn up all our cameras. So we had to shift and get a little bit nearby to a forest nearby that wasn't officially part of the NAM protocol but still representative. And you can see as the drone kind of zooms out some of the, a lot of the freestanding dead trees that are just a matter of super dry just a matter of time before they burn up at some point. So really interesting to see some of these landscapes and really fortunate to get a chance to go out and visit so many of these neon sites. Of course, out the west, you also have the joy of knowing that you could have a mountain line following behind you at any moment. So for each of these sites, we created a summary page, shared the pictures with the sites and the kind of overall, which species we did, did we detect where? And I'm gonna present some results here now where we're using the detection rate from a camera trap as a measure of relative abundance. And as Ariel just explained, indices of abundance are not always a best. And in this case, we're using an index because we don't have actual density because we don't have these animals marked. But in this case, I think we're doing better than with the beta traps because it's not baited. And so it's just animals walking in front of the camera and we get the picture, we're not luring them in. And so the rate at which that happens is reflective of their relative abundance. So that's the measure that, and I'm happy to talk more about this if there's some questions about that. So this is what we found overall across all these sites. The size of the pie chart represents the relative abundance of all of these terrestrial mammals. The color shows some of the different sort of general ecological types. And so you can see there in Maryland and Virginia had overall the highest abundance of sites. I was a little bit surprised that the circ site in Maryland is kind of mixed agricultural and forest. We ran all our cameras in the forest, but fragmented, not quite suburban, but wow, there were a lot of animals there. Raccoons and deer and squirrels and red foxes, the typical things, but really high. And then if you just look at the most Northern sites in Bartlett, New Hampshire, up in Northern Michigan, up in Washington state, those have the small circles. Those had the fewest mammals overall by quite a dramatic amount. And now if we just do a PCA on these mammal communities, the principle component one there on the X axis is sort of overall abundance. Basically you're seeing the sites on the right had overall more animals and the sites on the left had fewer. And the next index, the next sort of axis of the next most important amount of variation are the sites at the top that had more rodents and rabbits and the sites at the bottom that had more terrestrial birds. So Florida had quite a lot of turkeys and Mountain Research Station and so California had quite a lot of rodents and rabbits. Of course, we can break down any of these into more detail and look at the species composition across these different sites. I'll show you just a couple sort of fun comparisons. Here's who had the most coyotes, right? We can sort of look at that with this site. Talladega, Alabama had just a ton of coyotes followed by Florida and Wyoming. So I thought this was interesting because coyotes are not native to the Eastern United States that recently colonized there but yet they're still very abundant there. Black bears, Virginia and California had the most black bears. Raccoons, again, Virginia, Maryland, Kansas, lots of raccoons. Gray squirrels, there's our Maryland site again, Virginia, Massachusetts, lots of gray squirrels. So the problem is here, there's still a lot of gaps on these neon sites. And so during the same time, we worked with Bill McShay and Mike Cove to do a project called Snapshot USA where we looked for mammologists to run camera traps in a standard way all across the country, running their own camera traps. And so here you can see the first year in 2019 we ended up, we were able to recruit people across all 50 states to run cameras. And so they would run sort of 10 to 20 cameras in a standardized protocol near where they live. And you can see here all the sites we were able to get. So really starting to fill in some of the gaps. And this is very sustainable. We've been able, in 2020 in the COVID year we were able to keep it going. This year we actually have a new grant from actually have a grant from National Science Foundation to support this. The first couple of years we were just shoe stringing it and doing the best we could. And so now we're starting to ramp up and hopefully get even more this year we have over 130 arrays across the country. So we can start to look at trends over time from these sites. Many sites are surveying the exact same sites over years. This is just two years. You can see some places, Bobcats going up some places are going down. We've also now got colleagues in Europe doing a similar project. We're in the second year of Snapshot Europe. So we see this as a potential for a large scale sustainable sort of a big science way to survey mammals. And it's similar to citizen science. We've done some citizen science before but scientists are a lot easier to work with, right? You scientists, you have your own equipment you don't need as much training you get better data we get higher sample size per participant. And so it's really been I think a good experience. And the payoff is we publish the data together and so we're all co-authors on the paper. It gives good projects for students to be involved with. They get early access to the data. We make the data public but the collaborators get early access. And we have this network of people that are working together analyzing the data for their own papers. And sometimes we just need an excuse to go out and set cameras in the field. And so this offers a good excuse every fall to go check out your favorite place, set some cameras to see how the animals are doing. So I wanna mention one of the challenge of doing these big picture, a big camera trap project is you end up with all these pictures and how are you gonna deal with that? And I'll just briefly mention we use the wildlife insights platform that we helped develop which has artificial intelligence to help identify the species as you're uploading the pictures and it has some automated analytics that are just about to be released hopefully soon the next month or two that help you get some insights from your data as well. So this has been really great, really helped us here you can see all the data and wildlife insights but it really made the Snapshot USA project possible. And it also offers the opportunity to start to share globally data for conservation although it's not required. So by combining the neon data and the snapshot data we now have data from all these different points. And we use the GJM model that Arielle mentioned and we used 25 most common mammal species for this community analysis. We wanted to see how are these different covariates some habitat and some human and some climate factors. But then also we have the mass data and we broke the mast into four different kinds. The big nuts which are walnuts and hickory that most mammals can't get into those. So we separated those out. Then we've got the pines, the other hard mass like acorns and the soft mass which are fruiting trees now not raspberries or blueberries or bushes but just fruiting trees. And so here's the sensitivity analysis across all species for these models the top factors were, four of the top five factors were climate and then also terrain which is kind of also an abiotic thing. So climate is as Arielle found with the small mammals with the big mammals really important. These forest mass values came in to be pretty important as well for different amounts for different species. And then human population and agriculture a little bit less important which was a little bit surprising to me. There's kind of some fun things we can do with this with these results, these model results looking at the for example, here if you're above zero it's positive impact on a species and if you're below zero it's a negative impact. So you can see population size is great for eastern gray squirrels, right? They do great in developed areas. Not so much for the Fox squirrel but the Fox squirrel does great when there's agriculture around. So you can start to see some of these differences we can do this with the seed production too. If you look at the big nuts, squirrels of both kinds, eastern gray and Fox do great but if you look at the hard mass, just the acorns that's really important for the gray squirrel but the Fox squirrel, not so much. Okay, I wanna finish up, wrap up by talking about some of the community stuff we did. We're all familiar with these maps of ecoregions or the communities, the biological communities which are made from plant data. They're made from maps of plant data and the relative abundance of those plant data. So we wondered, could we do this with the mammals using this data that we have? And so I'll fly through this a little bit because I'm running a little behind but basically we have these maps of the relative abundance of these species and then we annotate points across the continent and then identify the communities from that. And so this is our, it's kind of preliminary map of mammal communities across the United States and you can see the first break is the orange and the green, right? If you look at the hierarchy there on the right which is east and west, which is kind of interesting. And then as you go in the Eastern United States from north to south, you start to get different communities and in the mountains, it's less of a north and south and it's more of an elevational factor that breaks that down. So you can see how these play out. And for each of these, we can also look then what are the animals that are there, right? And how common are they? And so on this graph, this is showing the relative abundance of all the different species. Those four sites to the right, those four equal or mammal regions are the Eastern sites. You can see they're all more common. They especially have a lot of white-tailed deer, a lot of raccoons, a lot of Eastern gray squirrels. And then for each of these regions you can see which species sort of characterize it in terms of them being present at all or in terms of their relative abundance. And we can dive a little deeper into this hierarchy as well and make 16 communities and start to see how that changes as well. So this is brand new stuff. We're still figuring out what it all means and still sort of start to compare it. But I think it's exciting. It shows the kinds of things you can do when you do have broad-scale data across large areas. So to kind of wrap up, we had the fortune of working with neon data for small mammals, for the seed traps and for using the sites for our camera trapping. You know, really excited to see what the long-term trends are that come out of this, which will take a few more years before we can kind of get there. But just right now doing these comparisons over space and time has allowed us to look at what are some of the drivers of population change, finding that in particular for the small mammals, climate is really the most important thing and competition only seems to play in for a few species. And for larger mammals, climate is also very important and mast being important for some of these. And it's interesting, right? Because the mast isn't just about forest composition, but it's also about the climate, the conditions, right? The rainfall that those trees have as well. And so, you know, the motivation for this is that starting to understand these relationships and having these kinds of models to make some of these predictions will help us figure out how can we manage and conserve these mammals as the planet warms up, as the forest change, as the places dry up or get wetter or get hotter? How will that affect our mammal community? So that's it, I'll stop here. And I think Arielle and I are both are happy to take any questions. Great, thank you so much. Can everybody can give a round of applause for our maybe virtual speaker? That was phenomenal. Thank you so much. Really, really fun to hear. So yeah, we'll open it up to questions. I am trying to find the Q and A here. Oh, there we go. It looks like there are a few, and then I've got one of my own. But I'll go ahead and read off one from Lila Hernandez. She says, Dr. Parsons, thank you for your talk. Very interesting. I'm wondering if the seasonal signal is what is providing a strong correlation to the abundance while environmental and interactions among species factors might not have a clear behavior over time. So Arielle, I'll open that up for you if you wanna. Oh, yeah, yeah, no. Yes, I think that these correlations with precipitation and temperature, especially that in a lot of cases that these are those seasonal kind of fluctuations that we know that small mammals are really sensitive to. And so the fact that we've got those correlations with abundance for some species, those really seemingly strong correlations is not super surprising. I still think it can kind of shed light on the dynamics and the relationships between changes in abundance over time for those species and temperature and precipitation that we could kind of carry through in terms of making predictions about climate change. But what's gonna be really interesting is thinking is more data, as Roland mentioned, in those long-term trends and those longer-term data, and seeing how some of those seasonal cycles in these abiotic factors actually change and then what the responses are from the small mammal communities, small mammal populations. Roland, I don't know if you had anything. Yeah, no, I just would, a little bit more detail on the analysis is that each plot month is an estimate of abundance. And so that's our data point. And then so we have the climate from that, I remember, do we use the exact three days around that or the month? I can't remember exactly what we, was it the three days? Yeah, we were exact, yeah. Yeah, so it was the temperature at that point. And so as obviously is, and then most of the, all the neon sites are just sampling sort of this few months during the warm years. And so we don't have any data from the winter. So we had this nice data series from the summer and then no data for the winter. And we looked at a couple of different ways to try to add some winter factors in there and none of them worked out. So we don't have that, but we do have the weather at the time when it was sampled. And so that's kind of very seasonally, but also a lot between sites, which I think is going to be even more important. Oh, wonderful. Before we move on to one question, may I just suggest Roland, you mentioned a lot of interesting projects and collaborations and websites. If you felt like dropping any of those in the chat through the Q and A, just if people wanted to learn more about the USA snapshot or the different projects mentioned, that'd be great. Yeah, great idea, Samantha. Great, I'm going to throw a question out there that I had that came up actually from Dr. Parsons. So did you do any analyses of those factors that are influencing the variation in detection probability that you showed kind of at the very beginning of your talk? And specifically, I'm kind of wondering from a neon perspective, is detection probability varying with temperature and precipitation in a way that would make some of the patterns you're finding harder to find without the capture capture? Yeah, exactly. I mean, that's the idea, right? That if it is, then we can capture that and control for it. So this is, yeah, I'm going to try and simplify this explanation as much as possible. So the way what we decided to do is try and maximize the amount of abundance information that we could take from the neon data, which means we actually ended up doing a modeling kind of framework that would allow us to harness the three nights and the one night surveys together. So that maximized the number of unassessments that we could get, but it meant that we could not model variation in detection probability over space and time, except the fact that it was allowed to vary between sites and between months at a site, but we couldn't model it as a function of covariates, right? So if we went back and we just took out those one night surveys and we just focused on the three night surveys, that could get us to that, but we haven't done that yet. But yes, conceivably you could do that and that would give you just even more precise abundance estimates, I would say. Yeah, fantastic, thank you. If you do do that, please let me know. I would love to hear. Great, so let's see, I'll open it up to the chatter. You can also raise your hand as well if something else comes up. Yeah, I'm seeing one here from Eric. Eric, would you like to just unmute maybe and share your question? Or maybe, is that a not a possibility with this format? I can also read it. No, no, that's that. If we could unmute Eric Sokol and Eric can ask his question over audio, that would be wonderful. Thanks, yeah, I just was unmuted. Can you hear me? We can. Okay, yeah, my question was for Dr. Case and kind of following up on the detection probability thing. And I was just curious if detection, well, how you handle detection probability with these passive traps, for example, like I know there's work on how personalities might differ among populations or animals with different dispersal abilities might be more likely or less likely to get captured in passive traps. This could interact with landscape type, like if you're in a dispersal corridor or not. So I know that can get complicated, but I'm wondering if just your replication is so high that in your placement was random that it just washes out or just kind of would be curious about your. Yeah, that's the idea is we don't go out and try to find the best place to run a camera trap. We go out to the GPS point and then that's it. And so we're trying to get a representative sample of the species that are out there. And so for sure, sometimes a camera might be in a corridor where animals are cruising a lot, but then there's gonna be another camera that's in a non corridor. And so we're hoping that it just washed out. We did an analysis a couple of years ago. I can drop the link in the chat where Arielle was on this paper. We looked at a bunch of existing camera traps to try to figure out studies, to try to figure out how many sites do you need for that to work to kind of have the variation level off. And it was sort of in the 40 or more was a typical, it's gonna vary from place to place, but 40 or more. So we're hopeful that our 50 is enough for that. Well, thanks for that. Really interesting. Great. And there's another question here. It says, how do you overcome the challenge of comparing relative abundances of mammals between sites when cameras are not set up at the same time at each neon site throughout the US? Yeah, so we try to hit it at the same season. We looked at growing degree days across the sites. And so we hit Florida and Alabama a lot earlier in the year and Colorado and other places later. So the seasonality can map obviously, some animals hibernate. So you run a camera then you're not gonna get them at all. And there's some other seasonal aspects to it. So we tried to, for the neon sites, we tried to make that as consistent as we could. Obviously there could be some year to year variation as well. In our snapshot USA stuff, we find the spatial variation is way, way higher than the year to year variation at the same site. But that's kind of what we're focusing on in this analysis is what is the spatial variation without, we don't have the time series to do what Arielle did to look at the temporal variation as well at the moment, but hopefully we will eventually. Fantastic. Samantha, do we have time for one more? You think or should we start? We could do, let's do one more quick question. Okay, great. There's one from Courtney as well about just how neon samples tend to dominate, sample in the dominant habitat types. And so it's just kind of getting your sense for how more rare habitat types, if they were sampled would affect the results. So this is coming from a plant perspective, I think where rare habitats hold a lot of richness at the landscape scale. And so how do rare habitats and small mammal abundances? Really? Yeah, that's really interesting. It's not something that we've looked at a heck of a lot in terms of, with our analysis of the small mammals, we were pulling kind of proxies, we were correlating proxies of habitat, mass production, grass availability, understory, kind of thickness, a lot of these measurements that neon takes at these sites, but not anything more specific about the habitat types themselves. But obviously, if you have low representation of some of those variables among all the samples that you're collecting, then it's gonna wash out any contribution that has to kind of the abundances. So, yeah, it's hard, because you wanna sample the dominant habitat types or you wanna kind of stratify amongst the habitat types that are available really to get the most kind of representative sample in my mind. Yeah, I think to just add, this wasn't really a diversity study, we weren't trying to find all the species that were there, in which case we would do that. But both Arielle's analysis and mine focus on the most abundant species. There were mice that only show up a couple of times and you can't do a density estimate on that and you can't add them to this kind of analysis. So I think that there's lots of interesting stuff you could do with them, but that's kind of doesn't fit within the these large scale comparisons that you're doing, you gotta have enough data to work with. So on both these, we're sort of just focusing on the more abundant, more common, which are ecologically more important, but certainly there's a lot of importance to the diversity and the species that are rare as well, that just wasn't our focus. Well, thank you all so much. We have reached the end of our hour. I know there were more questions, but this was a fantastic presentation. Thank you so much for coming. Everyone, if you're interested in Neon Small Mammal Data on January 31st, we're gonna have a webinar on how you can access and work with the Neon Small Mammal Data hosted by Sara. So please consider joining as well as we have our next iteration of the monthly seminar series in February. I've got the link in the chat for you. Thanks again, everyone. We really appreciate having you and we'll see you next time. Bye.