P1 NASA ARSET - Introduction to Animal Tracking and Remote Sensing At NASA

0:00 Speaker 1: Sensors that are commonly used in animal tracking.

0:03 Speaker 2: Also to identify the types of remote sensing data and products that can be used for species distribution models and step selection functions. To recognize the process for integrating remote sensing and animal tracking data in species distribution models and step selection functions to facilitate an understanding of animal movements in relation to the environment. and also to recognize key takeaways from examples from terrestrial and marine applications that inform and characterize animal habitats.

0:36 Speaker 2: The only prerequisite that we have for this training is the fundamentals of remote sensing, or to have an equivalent experience in remote sensing.

0:50 Speaker 1: Like I said, this is part one.

0:52 Speaker 2: It's an introduction to animal tracking and remote sensing at NASA. And part two will be on May 22nd. And the homework for this training will open on May 22nd at the end of our second part training. And it will be open until June 5th. So you will have until June 5th to submit the homework to get eventually a certificate of completion if you attended both parts and also submitted the homework on time.

1:28 Speaker 1: This is us. Like I said, I'm Juan Torres Perez. I'm from NASA Ames.

1:33 Speaker 2: And I'm here with my colleagues Sativa Cruz and Justin Fein, also from NASA Ames and the Bay Area Environmental Research Institute. Now, during the length of this training, feel free to put your questions in the question box. And we will address them at the end of the webinar. Also, feel free to enter your questions as we go. We'll try to get to all the questions during the Q & A session of the webinar. If by any chance we don't get to your particular question, don't worry because eventually we will try to answer it on the Q & A document that will be available usually about a week after the training.

2:18 Speaker 2: Now, as I said at the beginning, we have two very special guest speakers today.

2:26 Speaker 1: Dr.

2:26 Speaker 2: Morgan Gilmore from NASA Ames and also Claire Terrelbaum from the USGS and the University of Georgia. And we're going to start with Morgan. So Morgan, thank you so much for being here today, as well as Claire.

2:42 Speaker 1: Take it away, Morgan. Thanks.

2:43 Speaker 3: All right.

2:46 Speaker 4: Thank you, Juan.

2:47 Speaker 3: So hi, I'm Morgan Gilmore. I'm a research scientist at Ames Research Center in the Bay Area of California. And I'm really excited to talk with you about animal tracking today.

3:01 Speaker 3: But first, I want to introduce Claire Teitelbaum as well.

3:03 Speaker 4: Hi, I'm Claire Teitelbaum, and I'm going to come in a little later to talk about the integration of animal tracking and remote sensing data. Thanks, Claire.

3:17 Speaker 3: Why track animals? Well, we are surrounded by many very diverse habitats and ecosystems, and many people are often charged with managing the habitats and resources within these different ecosystems. And because these ecosystems, as diverse as they are, are often interconnected, it's really important to have diverse integrated approaches and tools to be able to understand how they work together.

3:45 Speaker 4: And animals....

3:47 Speaker 3: Use all these ecosystems too and they often move between ecosystems and so understanding how animals are moving within and between the ecosystems can actually provide a really unique perspective on habitat use and help us understand how to better manage habitats and resources. And As you might be aware, animals play a critical role in many aspects of our daily lives, and this includes things like ecosystem services. So animals often provide seed dispersal and pollination services to us that.

4:23 Speaker 4: Contribute to our food production.

4:25 Speaker 3: And they also are involved in a lot of other daily activities, including tourism, spiritual and cultural events, and also

4:34 Speaker 3: maybe a little bit more adverse things like spreading disease and providing navigation hazards. And so it's really important to understand animal movement with respect to all of these different services and interactions with human infrastructure. And so if we do track animals, we can learn a lot. We can learn things about movement ecology. So asking questions like where do animals go and why are they going to those places? If we combine animal tracking and remote sensing data, we can start to get an idea of animal habitats and what the environment is like in those places.

5:09 Speaker 3: And due to advances in animal tracking technology, we can put ambient environmental sensors on animal tracking tags so that we can actually ask what's the environment like that they're actually experiencing and what can they tell us about their environment. And so as you're starting to plan a animal tracking project, it's good to think about the different types of data that we can get from animal tracking tags and sensors. And so this includes things like presence, so we know where the animals are going because we have the location data to show that.

5:42 Speaker 4: If we start to see.

5:43 Speaker 3: A lot of animals using the same general area, we can start to talk about hotspots of animal use of different types of habitats. We can use the movement data to calculate things like transit speeds, If the animal tracking tag stops moving for a while, we could use that as an indicator of mortality. And so this could be really helpful for some studies. If we were to overlay animal tracking data on maps of human infrastructure, so things like roads and buildings, and maybe like the urban wildland interface, we could start to understand how animals are actually interacting with different aspects.

6:17 Speaker 5: Of our environment as humans.

6:21 Speaker 3: If we... if we overlay animal tracking data on remote sensing layers like temperature or salinity or topography, we could start to understand how animals are interacting with those types of resources in the environment. And we can also use these data to understand how animals can respond to different changes in these variables and in their environment. And so, Another important aspect of planning an animal tracking project is to think about different types of animal tracking tags. Animal tracking tag technology has improved immensely in the last 15 to 20 years and so there's so many different options out there, but really there's four really major factors that we need to keep in mind that are really specific to your study species and your questions, your research questions.

7:18 Speaker 3: And so this includes things like tag size. You want to make sure to consider how tag size is relevant relative to your animal and also how that animal moves. So we want to make sure that the animal can not only carry the tag, so it has to be small enough so they can carry it, but also they can move with it without impacting their aerodynamics, if they're birds, their hydrodynamics, if they're swimming animals, or if they're animals that maybe burrow underground and the tag has the potential to get caught on underground tunnels or something.

7:51 Speaker 3: So we just want to really consider how the tag can interact with the animal's natural movements. Latency is also important parameter to think about with tracking tags, and so this latency means the time it takes between the data collection from the animal tracking tag and your ability to access this data. And so we're going to talk about this in the next slide, but some Sometimes you might deploy a tag, but not be able to get it back for six months. If you need to consider how that may help or hinder your research question and any other applications you want to use these data for, how long you can wait to get the data back.

8:35 Speaker 3: Location precision and accuracy is also something that's really important. is related to animals habitats, and so, if you have animals that spend a lot of time underwater it's going to be hard to transmit data from that tag through the water. If you have the animals that are they spend a lot of time in thick forest so there's a really dense canopy cover, then they might not be able to collect location data very well, and so that could affect the precision accuracy of their location and Going back to animal size, different tags have different data transmission capabilities.

9:14 Speaker 3: And so if you have a small tag, it might not be able to transmit data very far, very fast, very frequently, I should say. And so that could also impact location precision and accuracy. And then because animal tracking tag technology has improved, we have the ability to include different sensors besides geolocation on these different tracking tags, which enable us to maybe measure ambient environmental variables, but also look at things like accelerometry, so we can get a really fine scale picture of how animals are actually moving within their respective habitats.

9:51 Speaker 3: So animal tracking tags are often defined by their data transmission modes, and so some of the more simple but perhaps more labor intensive animal tracking modes include radio tags. And so that requires a lot of movement to triangulate an animal's location.

10:09 Speaker 3: Really common tags right now are satellite tags. And so those tags collect data and transmit those data potentially in near real time. by satellite, so you can actually access this data pretty easily and frequently if you want. On the opposite end of that are archival tags, and so

10:33 Speaker 3: these tags collect data that are stored on the tag, but then you need to recapture the animals in order to download the data off of the tag. So depending on your study system, this may or may not be easy. There are combination archival and satellite tags. And so the way that these work, the shark on the right, for example, is carrying an archival satellite tag. And so this works really well for animals that spend a lot of time underwater, where the archival tag will record and store data. And then at a predetermined, pre-scheduled time, the tag will release from the animal, and then it'll start transmitting those data via satellite.

11:11 Speaker 3: And so this could be a good solution for some underwater species. There are in the last maybe five or 10 years, there are tags that actually transmit data through cell phone tower networks. And so that actually is really helpful for areas that have good cell phone coverage. The MODIS network is another type of transmission system that enables tags to ping it as animals transit past it. And so if you have a network of MODIS towers and you have an animal that pangs all its different towers, you know that it's on the move, which is pretty cool.

11:48 Speaker 3: And then finally, we have remote download tags, which work with archival tags to download the data via ultra high frequency radio, for example, so that you can access the base station that contains all the data instead of having to recapture the animal. And so that is really helpful if the base station is located next to a nest or in the area where an animal repeatedly returns. And so we've talked a little bit about size, weight, and power restrictions on different animals that you might be interested in tracking.

12:26 Speaker 3: And so you can see in the photo on the slide that there are lots of different types of animals that are tracked with lots of different size tags that have drastically different capabilities in storing and transmitting data. And so again, it's just really important to understand the locomotion method, the locomotion mode of your animal of interest and the research questions that you're trying to ask and how the data can be collected with these different pack options. It's also important to remember that we're only getting information on where the animals go.

13:00 Speaker 3: We don't actually know where they're not going and why they're not going to those places. And so we're working with presence-only information and it requires some statistical methods and approaches to understand the habitats that they are using with this presence only information and to be able to supplement that with absence information is something that we're going to talk about later today and then in session number two. We've talked a little bit about spatiotemporal resolutions, but again, depending on the size of the tag, it may only collect data infrequently to prolong the life of the tag and the battery power that it extends while collecting data and then transmitting those data.

13:49 Speaker 3: So you might have a situation where you can only record data once a day, for example, if you're trying to record data for an entire year. And so just thinking about the trade-offs of the spatio-temporal resolution in your sampling regime is also important in relation to the research questions you want to ask. And then finally, we can't possibly, as cool as it would be, we can't possibly track every single animal out there. And so we can only work with the data that we can get from a subgroup or subpopulation of any given species.

14:20 Speaker 3: And we can infer species level population data or movement data based on the subgroup of animals that we've tracked. And so Claire and I are, part of NASA's Internet of Animals project. And NASA has actually been tracking animals since 1970. You can see with this newspaper clipping photo on the right, this was Monique, the space elk. And they put radio collars on her and another individual to track the elk herd's migration through Yellowstone back in 1970. And ever since then, NASA has been a part of various missions to develop sensors and tracking tags to improve animal tracking methods over the years, including developing better communication networks like long-range radio to work with different types of tracking tags.

15:20 Speaker 3: NASA has a couple projects that use animal tracking to monitor biodiversity, and we are also developing ways that we can use in situ observations collected by animal ambient environmental sensors to contribute to models and validation and calibration of different satellites.

15:40 Speaker 4: And so NASA's Internet of Animals.

15:42 Speaker 3: Project is a project that's been going for a couple of years now, and we have four overarching goals. The first is to talk with the movement ecology community and understand why different people, different research groups are tracking animals. What kind of research questions are they asking? And we're also asking what they need to improve their animal telemetry research programs. And so we've spoken with academics and federal agencies to understand these needs and quantify how NASA potentially could help fill some of the gaps in the animal tracking.

16:20 Speaker 3: Our team is also trying to architect a next generation space-based animal tracking system. And so this includes things like

16:28 Speaker 3: developing a satellite system study to understand what tracking systems are already used and what could be improved upon, and also to develop a new type of tracking tag. Our team is also conducting a study of remote sensing of animal movement. And so again, going back to the animal tracking community and ask, and asking what the animal tracking community's remote sensing needs are, but we're also asking where and what gaps can animals fill remote sensing. And so we'll talk about this more in the next slide.

17:06 Speaker 3: And then finally, we have developed a couple of case studies to demonstrate how we can integrate animal tracking and remote sensing to characterize animal habitats and understand things, improving disease ecology research and quantifying efficacies of marine protected areas. And so, as I mentioned, animal tracking technology has improved a lot in the last 10 to 15 years. And so tracking tags are small enough that we can include ambient environmental sensors on them. So we can record in situ observations of the environment that the animals are moving through.

17:47 Speaker 3: And so one great example is this elephant seal here. And so on the left, we have a map of an elephant seal track moving through the ocean. And there was a salinity sensor on that tracking tag. And so you can see with the different colors that the animal moves through different water masses that have different salinities. You can see the dark and lighter colors indicating low and high salinity, respectively. And in addition to that, the plot on the right is a time series plot of water depth colored by water temperature.

18:21 Speaker 3: And so as the elephant seal was Moving around the southern ocean, it was collecting temperature data as it drove from the surface down to sometimes 1000 meters deep, and you can see the different color changes indicate again the different water masses that the elephant seal has moved through.

18:42 Speaker 3: And the second cool example of having animals as in situ sensors of the environment is actually the study that put video cameras on sea lions in South Australia, and so the pictures on the right were still shots from a video from these sea lions as they move through different habitats, and so the picture on the top is a sea lion moving through kelp, and then the middle picture is bare sand, and the bottom picture is an environment that had both sponges and sand in it. And by combining these types of pictures, images, with the geolocation data that the sea lions were also collecting, researchers could actually map the bathymetric habitats in South Australia, as you can see in the big map on the left.

19:31 Speaker 3: So mapping where there's kelp and bare sand and spongy, sponges and sands. This is a really great example of ways that we can work with animals to collect environmental and situ data. And now I will pass it off to Claire Teitelbaum who will go into the next section.

19:57 Speaker 4: Thanks, Morgan. So in this section, we're going to talk a little bit more about how we take all that wonderful animal tracking data and combine it with remote sensing data to get some of those really cool insights about the environment and animals' habitats. So as Morgan mentioned, location data from animal tracking tells us where animals go. But in order to understand what the conditions were like in those places, we have to combine animal tracking data with environmental data. And remote sensing is one of the most common sources for that environmental data.

20:31 Speaker 4: Two really common methods for combining animal tracking and remote sensing data to understand how animals are using their habitats are species distribution models and step selection functions. And we're going to talk about both of those in depth throughout the rest of the training. An example of what we can get out of both species distribution models and step selection functions are two sort of key outcomes. One is maps of probability, intensity of use, or relative use of different types of habitats.

21:06 Speaker 4: And the other is a more mechanistic understanding of how animals are using different types of environments. So in these figures on the bottom here, we have tracking data from green sea turtles. Those are the green points in the middle panel of the map. And the middle panel of the map shows an outcome from a species distribution model showing us the relative intensity that we would expect sea turtles to be using different areas of the Gulf. And you can see that we generally expect them to be using areas near the coastlines, but not right in the middle.

21:41 Speaker 4: And the right-hand panel explains, at least in part, why that is, using data on the depth of the Gulf at different locations, you can see that sea turtles are really strongly selecting for relatively low depths. Pretty much as soon as you get above about 10 meters, you're really unlikely to find a sea turtle. So by combining information on environmental conditions, in this case depth, as well as other variables that you're not seeing here, we can get information about the spatial locations where we would expect animals to be beyond the locations where we've tracked them.

22:17 Speaker 4: These distribution models are arguably the most common way that people go about characterizing these environment-movement relationships. And in general, they help us understand the likelihood of animals' presence in different habitat types. And the way they do this is by asking how the environment differs between where an animal went and an animal did not go, or between where a group of animals went and where they did not go. So to perform a species distribution model, the first step is going to be to gather some sort of information on animal presence.

22:52 Speaker 4: In this case, we're talking about animal tracking data, although species distribution modeling can also be performed with non-tracking data, for example, from surveys or other types of observations of animals. In this figure on the right, you can see in this top sort of slice these red points representing locations of the species. We then can generate what we call absence or pseudo-absence points, because as Morgan mentioned from animal tracking, we get information about where we know animals were, but we don't get information about where they weren't.

23:35 Speaker 4: So by taking information about where they were, we can infer locations where they weren't and compare the environmental conditions at these two types of locations. So in this example, we're looking at sea surface temperature, nitrate concentration, and salinity at both presence and absence locations. And in that 3D plot on the right-hand side, you can see the red points being locations where the animals were, the gray points being locations where the animals weren't. And then we can statistically characterize the space that is taken up by where the animals were compared to.

24:12 Speaker 5: Where they weren't.

24:13 Speaker 4: And then putting that back all together and mapping it out, we can get a habitat suitability map telling us the geographic locations.

24:21 Speaker 5: Where these animals were likely to be.

24:25 Speaker 4: Step selection functions are a similar way, but a specific way to characterize these relationships. Some key differences between step selection functions and species distribution models is primarily that step selection functions have a specific way of defining used and available locations, which we consider equivalent to presence and absence locations. So in the top panel here, you can see we have a track of a duck moving to three different places. These points would be locations given to us from a tracking tag.

25:01 Speaker 4: Based on the distances moved that we saw the animal move and the angles that we saw it moving at, we can simulate available locations on the landscape, and those are those gray points. those are locations where the animal did not go, but in theory it could have. It had the capacity to move that distance in that amount of time. Because we're using these step lengths and turning angles to characterize the available locations, step selection functions require that we use regularly sampled data, meaning that it has a relatively even time interval.

25:37 Speaker 4: It's also unique in that the analysis is at the level of an individual movement step. So we're comparing where an animal could have gone at any point in time compared to where it did go, rather than aggregating all of those locations at the same time. And you can see this in the middle panel here where we're comparing the value of, in this case, air temperature at the location where the duck went compared to the distribution of air temperatures at locations where it could have gone. The way that we statistically analyze this is using conditional logistic regression.

26:16 Speaker 4: Logistic regression is a very common method in some species distribution models. The conditional part here means that we're only comparing available steps to the focal used step rather than, again, aggregating across.

26:29 Speaker 5: All of them.

26:31 Speaker 4: And once again, we can get similar types of inferences out of step selection functions. So this bottom panel shows the relative strength of selection as a function of air temperature, and showing that, for example, this duck tends to go to places with relatively high air temperatures. For step selection functions, as I mentioned before, the data must be regularly sampled.

26:56 Speaker 4: depending on the type of tag that you're using, data may or may not be regularly sampled, meaning that this method may or may not be a good choice. However, there is some tolerance to this regular sampling. For example, if a tag is programmed to collect data every hour, it's probably not going to do so exactly on the hour every hour. So, for example, calculating step lengths that are between 55 and 65 minutes apart would still create a reasonable distribution from which to build a model. Gaps in data are also okay.

27:33 Speaker 4: It just means that we're going to take that very long step out of the data when building our model. So as I mentioned, in general, tags that are active, that are actively collecting and potentially also transmitting data, these are tags that are getting locations from GPS or Argos satellites are generally most appropriate for step selection functions. But tags that are passive and are only going to ping a location when they pass near a base station or a tower, so tags like MODIS or RFID tags, are generally going to provide locations at very irregular intervals, so they're not compatible with step selection functions.

28:20 Speaker 4: In general, step selection functions are performed with relatively high frequency data, but they can be performed with data that are up to daily or even less frequent. This is just going to determine the types of questions that we're.

28:36 Speaker 5: Able to answer.

28:38 Speaker 4: And this applies to species distribution models as well. So for example, if we're thinking about using daily data, We can infer information from a step selection function about local dispersal or relocation, but we're not going to get information about where an animal spends time at different times during the day. However, when we move to an hourly step, if we have hourly data, we can understand more about the function of movement. So where is this animal foraging and where is it sleeping, for example.

29:07 Speaker 4: If we have really high frequency data, things like every minute or every five minutes, you can start to understand science scale habitat use. So where is an animal going within its foraging patch, for example. It's also important to note that the temporal and spatial scale of your model are often going to be related. So if we have information at a daily scale, it might be important to understand the environmental conditions within a larger buffer area around the animal tracking locations. And finally, step selection functions are interesting because individuals can be analyzed separately or together.

29:46 Speaker 4: So we can build a model for each individual to understand individual differences in habitat use or aggregate individuals together to understand on average how a population is using its environment. For both species distribution models and step selection functions, there are quite a few trade-offs that need to be considered in general when combining animal tracking and remote sensing data. The first is that both animal tracking and remote sensing data are characterized by both spatial and temporal resolutions.

30:22 Speaker 4: In general, the resolution in terms of both the accuracy and the frequency of animal telemetry data is going to be higher than the resolution of remote sensing data, especially that from satellites. So for example, many animal tracking tags can provide a location that is accurate to within a few meters, but there are very few remotely sensed products that have a pixel size that is as small as a few meters. For remote sensing data itself, we often see a trade-off between spatial and temporal resolutions.

30:58 Speaker 4: So if you are using a remote sensing product that has a very fine spatial resolution, it's often not going to be available very frequently. So when making decisions about which remote sensing products to use, it's important to consider both the characteristics of the animal tracking data, as well as the questions you're trying to answer in your study. On the right-hand side, you can see an example of

31:26 Speaker 4: climate product at two different spatial resolutions overlaid with a hypothetical movement path of an animal. You can see in the nine kilometer resolution product, you can guess that this animal is selecting for a pretty narrow temperature range. It's pretty much only moving in this green area. In the bottom panel, however, you see the animal moving into areas that are more different colors and the red area that the animal seems to have have been avoiding has disappeared because we don't have the resolution to sense it.

32:00 Speaker 4: So we might not get as strong a signal of this animal selecting for temperature with a 111 kilometer resolution product compared to a 9 kilometer resolution product. Another point to consider when selecting remote sensing variables is that we often measure proxies of the actual conditions experienced by animals. So for example, Remotely sensed data is readily available for sea surface temperature, but not all marine animals are existing at the sea surface. They might be two meters below the sea surface, and therefore, sea surface temperature is serving as a proxy for water temperature at two meters.

32:41 Speaker 4: Similarly, vegetation indices are commonly used as a proxy for food availability for terrestrial herbivores, and air temperature is often used as a proxy for water temperature of small terrestrial water bodies. And this is true for many, arguably most of the variables that we measure using remote sensing. And these are important factors to consider when thinking about whether a remote sensing product is appropriate for a given study compared to an in situ data product.

33:12 Speaker 5: And just how to interpret outcomes of these models.

33:16 Speaker 4: Finally, combining animal tracking and remote sensing data takes both computational power and time. So it's tempting to want more data, more animal tracking data more frequently, higher resolution remote sensing data more frequently. But higher resolutions and more variables mean more data. And there are trade-offs in how much more information we're going to get from a higher resolution data set compared.

33:44 Speaker 5: To the amount of time it will take to analyze it.

33:53 Speaker 4: Finally, as the last portion of today's training, we're going to talk a little bit about some of the technical aspects of working with animal tracking data in preparation for the next part of the training where we're going to go into a little more depth into how to perform species distribution modeling and step selection analysis. When working with animal tracking data, before we even get to the point of being able to analyze it, there are a lot of basic pre-processing steps. The first is cleaning data.

34:26 Speaker 4: This includes steps like removing missing values or points with high error estimates. Some tracking tags will provide error estimates. For example, Argos locations, provide location classes, which are based on the number of satellites that were used to determine the location, more satellites leading to a more accurate location. Some tags do not come with these error estimates, or some of them are incorrect. And one good way to clean data to remove erroneous points is to use a speed threshold. Essentially, if the distance that an animal appears to have moved in a given timeframe is much further than would be biologically realistic, remove that point.

35:21 Speaker 4: And then also in terms of the analysis that is going to follow this data cleaning, we want to consider whether these movements are important for the analysis that we're going to be doing. For example, if we're only interested in breeding season data, finding out what constitutes the breeding season, whether that's based on calendar dates, or an analysis of the movement data itself to figure out when an animal is breeding, and then filtering to only the relevant points can be really helpful, especially for reducing that computational load we just talked about.

35:57 Speaker 4: As with all data, it's really important to visualize tracking data before starting to analyze it or work with it in any other way. This allows for assessments of things like data gaps, it's also important for identifying these potentially erroneous points. For example, I've seen a terrestrial animal end up in the ocean.

36:21 Speaker 4: For step selection functions, as well as for other types of analysis, it can help to regularize the data to regular time intervals, meaning that the time lag in between consecutive locations is regular, This can also help smooth over short data gaps where data was missing because the tag was unable to collect location data, potentially because of the habitat that the animal was in or low power on the tag itself. Before performing any sort of formal analysis, it can also be helpful to quickly assess habitat use.

37:01 Speaker 4: To map things like locations overlaid on a map of habitat types, or to plot utilization distributions, which are the probability that an animal occurs in a given location. To get a sense of what that looks like, here's an example. This is a dataset that contains locations from a single red deer tracked over the course of a year in Germany. We're gonna go into a little bit more depth on this dataset and work through a step selection function in the next part of the training. But here, just to get you started, on the left, we have a plot of the eight to 900 or so locations of this year over the course of the year.

37:49 Speaker 4: And on the right, the output of a home range analysis that plots different percent utilization distribution. So these purple contours show the areas in which the deer spent, would be expected to spend more than 50% of its time over the course of this year. The blue shows 90% and the teal 99%. So looking at, for example, the 50% contours, we can see that these are generally centered around these darker green or forested areas. So that gives us a clue that, for example, we might want to go try and find remote sensing data that tells us about the locations of forests in this area when we start performing our step selection analysis.

38:48 Speaker 4: And now I'll pass it over to Juan for a summary of today, and we look forward to seeing you next time.

38:59 Speaker 2: Well, thank you so much, Morgan and Claire, for all that info about animal tracking and the Internet of Animals. Here's a summary of basically what you just said during your presentations. Animal tracking techniques can help understand animal behavior and species roles in ecosystem ecology. Biologging has its advantages but also has its limitations. While it can serve to monitor specific animals for long periods of time and distances, the limitations are that it only provides presence-only data, and size, weight, and power constraints also limit the number of species that can carry tracking tags.

39:43 Speaker 2: But the NASA's Internet of Animals project integrates animal tracking observations with remote sensing data for a better understanding of ecosystem environments and how it influences the presence of diverse animal species.

40:00 Speaker 1: As a reminder, there's only one homework for this training series. It opens.

40:07 Speaker 2: On May 22nd, so on the end of part two. And you can access this through the training webpage. It's basically a Google form where you submit all your answers. It's simple, multiple-choice homework. The important thing is that you need to submit it by June 5th to get credit for the webinar. So if you attended both training sessions and you completed the homework, submitted it on time, eventually you will receive a certificate of completion via email But consider that because of the large number of participants that we have in this training series, typically it takes about two months after the completion of the training to receive the certificate.

40:54 Speaker 2: So you will receive the certificate via email eventually.

41:03 Speaker 2: This is our contact information from us, Justin, Sativa, and myself, and also Morgan and Claire's contact info if you have any other questions regarding this topic. And again, as I mentioned at the beginning, make sure that you go to the ARCET website and look for additional trainings that are coming up. Also to follow us on Twitter, on X, and there's also the ARCET YouTube channel as well. where you can find presentations from previous training and webinar series. And also, if you're an undergrad or grad student or a recent grad, we recommend you to visit the Develop webpage.

41:47 Speaker 1: That has opportunities.

41:48 Speaker 2: There to do an in-person or an online teamwork with us on different topics regarding ecological conservation. Okay, so with that, let's just go to the Q & A session. Thank you again for being here with us, and let's go to the Q & A.

42:12 Speaker 1: All right, here's one again.

42:23 Speaker 2: I just want to confirm with the team that everyone can hear me fine, and thanks, Natasha, for sharing your your screen there. Of course, we have Morgan and Claire here with us, and Justin and Sativa as well. And let's go straight into the Q & A. Some of you have been posting some of your questions, and we've been trying to answer some of them. And let's start. So question number one. Is there any guidance regarding the size, weight of the animal, and size, weight of the transmitter, or rule of thumb ratio for good practice?

43:11 Speaker 5: Yes. A good rule of thumb for birds is usually 3% of their body weight. I honestly don't know what the rule of thumb is for mammals, but it's usually the permitting agency. And also, I saw somebody in the chat say IACUC, and that's also another good one to check out for restrictions on the size of the tag for any given species.

43:44 Speaker 1: All right. Thanks, Morgan.

43:48 Speaker 2: Question number two, it says, specifically in the sea lion study that you presented, what do the different colors in the graph or plot, what do they mean?

44:02 Speaker 5: So the colors represented the habitat that the researchers were able to map based on the video camera data and the dive depth data and the geolocation from the sea lion tags. And so when they had the If they had a video camera image of bare sand, they knew where that sea lion was and how deep it was, and they could map that habitat as bare sand. And so those colors represent those different habitat types. Cool.

44:32 Speaker 1: All right.

44:33 Speaker 2: Question number three there. Which sensors does NASA use to locate the exact location of a particular animal? For example, if it's the smallest bird that it rarely found. I don't know if this was Claire or you, Morgan, who answered.

44:49 Speaker 4: Yeah, I can take this one, Juan. So this is a complicated question. It's a big challenge in animal tracking is basically to locate and capture an animal first because an animal has to be located and caught before a tag can be deployed on it. So for some animals, there are some sorts of sensors, for example, acoustic recorders or camera traps that might help researchers Locate a species and know where it tends to come back to, where would be a good location to set a trap or do some other sort of capture method.

45:25 Speaker 4: For other species, it might be important to rely on local knowledge if people know where that species is. Once a tag is deployed, there are a lot of different types of systems, so satellite tags, MODIS tags, and so on.

45:42 Speaker 5: That can be used to locate the precise location of the tagged animal. So, it's only really after.

45:50 Speaker 4: Capture that these sorts of systems can be used to locate animals.

45:57 Speaker 1: All right. Thanks, Claire. Question number four.

46:01 Speaker 2: There are several questions here. Someone posted in the chat that they're facing issues with road kills for wildlife.

46:12 Speaker 1: So the person.

46:13 Speaker 2: Is asking, where can they purchase the tracking devices and where can we get the methods to install the device on the animals? Also, is fixing the device will affect the animal's daily life? And also, can this device provide real-time data signal to any device that can be installed on the road area, which can show a warning light or signal at the area for the road users to reduce the speed stop.

46:46 Speaker 1: Or potential hitting of some of these animals?

46:50 Speaker 5: Yeah, those are all really good questions. conducting literature search on Google Scholar or Web of Science or something else to understand the types of tags that are used for either your focal species or related species or similar species. And those papers will also help you understand how those researchers attach those different types of tags.

47:12 Speaker 4: And so the literature is really a good place to start.

47:15 Speaker 5: And you can probably contact the researchers and ask them specific questions. Most researchers are more than happy to talk to people about how they conduct their work. And so along with that, people work really hard. I mean, we are really excited to track animals, but we also recognize we have a responsibility to not impinge their daily movements. So people work really hard to make sure that tags are attached in a way that they don't bother the animals too much. And so that also goes back to the permitting process and the agency's are also very aware of the different attachment methods for different types of animals, and they'll make sure that you are aware of the most up-to-date methods that enable you to track the animal without bothering it too much.

48:03 Speaker 5: And the idea of devices on the road that maybe send a warning signal to cars and animals nearby is really cool, but I don't know of any studies that are currently doing that. So that might be a good area of research for somebody.

48:16 Speaker 1: Yeah, indeed.

48:21 Speaker 2: All right, Natasha, if you can go there to question number five on the Google Doc. Let's see. Just wait for her. Can you move the document up?

48:46 Speaker 4: Okay, I can.

48:50 Speaker 1: Okay, let's see. I'll have it here. All right, number five.

48:56 Speaker 2: It says, I've noticed that some tags seem quite large relative to the size of the animal, like the one I saw on a bat. Has it been shown that some tags can affect the animal's behavior and prevent it from acting naturally?

49:15 Speaker 4: Yeah, so as Morgan just said, this is a really important point in this field of animal tracking because we do know that some tags can affect animal behavior and that includes some tags that are allowed to be deployed as long as we don't think they're affecting behavior too much to sort of inhibit the animal's survival or affect the study. it can actually be permissible to deploy a tag that does have some effect on an animal's behavior. But we have a relatively poor understanding of exactly when and how and for which species these effects occur because it can be difficult to compare behavior between tagged and untagged animals because the untagged animals are hard to observe.

50:00 Speaker 4: But some of the effects we see are reduced mobility, increased preening in birds or picking or rubbing at the tag in other animals. And sometimes we can see changes in social interactions where members of the individual who's tagged members of their social group might behave differently around them because they look different. So it's always important to understand these implications as much as possible when designing a study. And that will be part of these approvals, such as IACUC or agency approvals for working with live animals.

50:36 Speaker 4: And I also want to note, I also agree the picture of the bat that we showed, that tag looked very large. But the weight of a tag and its size are not always directly related. Some might have denser batteries in them, for example. So a tag that appears large might still have been light enough for that bat to carry.

50:58 Speaker 1: Yeah, good point. Thanks, Claire.

51:00 Speaker 2: All right, question number six. The person is asking, I'm still a little unsure about how available in quotes the locations or their spatial bounds as for SDMs. I'm guessing it refers to species distribution models.

51:19 Speaker 1: Are defined.

51:20 Speaker 2: If it can be clarified a little bit more, it seems clear for use in quotes locations.

51:31 Speaker 4: Yeah, this is definitely a complicated question, and it's an important part of designing any of these types of analyses. So we're going to walk through examples of going through both a species distribution model and a step selection function in the next session. So we'll talk in more detail about exactly how these available locations are selected in each case. But as a general overview in step selection functions, available steps are simulated starting at a known used location. And the distance of that simulated movement is drawn from the distribution of observed movement distances in the tracking data.

52:14 Speaker 4: And the angle that the animal moves away from that used location is also drawn from the distribution of observed turning angles in the tracking data. In species distribution models, there are a lot of different ways to define the available area. So it could, for example, be based on the species known range, so within the species range, comparing locations of an individual to the rest of the species range. It could be comparing it to the known range of another similar species. It could be using a buffered area around the observed locations.

52:51 Speaker 4: or a number of other ways of defining that available area. And that definition is going to depend on the question that you want to answer. So, for example, using the known range of another species might allow you to ask how two species differ in their distributions or habitat use.

53:10 Speaker 1: Great.

53:11 Speaker 4: Okay.

53:12 Speaker 2: Number seven, what is the ideal time interval for identifying behavioral states and what is the acceptable time an acceptable time interval, let's say for large or medium-sized mammals?

53:28 Speaker 4: Yeah, I'll take this one again. So this answer depends on the types of behaviors you're trying to identify because different types of behaviors occur at different time scales. So for an animal that's migratory and you're trying to distinguish a migration and a non-migration state, if the animal is migrating for a week or more, You might get away with daily locations to identify these behavioral states, although you would only be able to identify that transition point with an error of about a day.

54:00 Speaker 4: But for an animal that transitions between behaviors very quickly, so if you think of a small mammal that's going out and searching for food and then running back to cache it, you would need much more frequent data, maybe minute-to-minute data or even more frequent than that. A good rule of thumb would be to have about 10 data points for each behavioral state, but this is going to depend on how distinct the states are. So, if they're pretty similar, you're going to need more data than that to be able to really carefully understand the distributions of how animals are moving in those different states.

54:39 Speaker 1: All right. Thanks, Claire.

54:43 Speaker 2: Number eight, is it possible to use satellite tags on small mammals?

54:47 Speaker 1: And if so, what kinds are available?

54:52 Speaker 4: So, yes, it's definitely possible.

54:53 Speaker 5: It definitely depends on your species size and how they move. And, again, I direct you to a literature search to help you understand the types of tags that are available for small mammals of your focal species size and the habitats that they move through. I'm not sure if Claire has anything else to add to that.

55:14 Speaker 4: I think that's a great answer, Morgan.

55:17 Speaker 1: All right.

55:19 Speaker 2: Number nine, is a step selection function better, more precise than species distribution models? Should step selection function be preferably used when appropriate data is available?

55:35 Speaker 4: Yeah, so not necessarily. I can see why you might think so, because step selection functions provide this more constrained and movement-specific definition of available habitat. So you're using more of the movement-specific data that you have. But what this means is that they're best for asking questions about how individual animals move and how animals are making decisions at sort of relatively fine timescales. But because those available locations are relatively constrained, step selection functions can't be used to ask some other types of questions about animal distributions.

56:14 Speaker 4: So with species distribution models, you can ask questions about how an entire species might be distributed and can also incorporate other types of data. So if you want to integrate movement data with field observations, observations from citizen science, or other types of data on animal locations, species distribution models would be a better choice. Okay.

56:43 Speaker 1: All right, number 10.

56:45 Speaker 2: For slide 31 specifically, there was a mention of a threshold used for cleaning data.

56:53 Speaker 1: Can you give an example of.

56:55 Speaker 2: How species-specific speed threshold is calculated?

57:01 Speaker 5: Yeah. There are a couple different ways you can think about this. One way is to just take all of your movement points and calculate a mean standard deviation across those points and then just get rid of everything that's one standard deviation away from that. Another way is to just think about the context of your species moving in their environment. So I work a lot with seabirds and seabirds often can fly really fast because they catch wind. And so the wind makes them move faster than maybe you would normally expect.

57:32 Speaker 5: But those bursts of speed are due to the wind and they're totally real points. And so that's another thing to keep in mind. Habitat specific, you might want to allow for bursts of speed because of something like that. So you really have to consider your focal species and the habitats that they're moving in. But those are a couple different ways that you can think about how to select kind of a maximum threshold of movement.

57:59 Speaker 1: All right. Question 11. Are there any commonly used tools.

58:05 Speaker 2: Or packages that are suggested for animal tracking analysis?

58:13 Speaker 4: There are a lot of packages and tools out there. Many of them are really great. And we're going to give some examples of our packages and tools in the second session on Thursday. But that's by no means an exhaustive list. Depending on the type of analysis that you're working on, the type of data that you have, it's almost impossible to quantify exactly how many packages and tools are available. But we'll add a link to a review paper that at this point is a few years old that lists some of these animal movement R packages in the Q & A document.

58:50 Speaker 1: Awesome. Thanks, Claire. Yeah, everyone, stay tuned for part two this Thursday. Okay, number 12.

59:01 Speaker 2: Hi, I'm studying the spatiotemporal ecology of yellow mongooses using only VHF corals.

59:10 Speaker 1: So no GPS units on the animals.

59:12 Speaker 2: I'm currently relying on triangulation from multiple observer points and any recommended complementary methods, such as passive remote sensing or autonomous receivers or others. that could enhance the data resolution without the need of GPS collars?

59:33 Speaker 5: It kind of depends on your research question and how frequently or how large of an area the animal is traveling through every day or monthly or something, and also how frequently you want those data. to understand their movement patterns. But there are definitely, the thing that came to mind when I saw this question was RFID tags. And so those are just little things. You can kind of think about it, not only as probably Bluetooth options these days, the RFID is really simple technology that just pings a sensor as the animal walks past it.

1:00:05 Speaker 5: And so if you have a network of these sensors, you have a timestamp of whenever that animal passed that sensor.

1:00:10 Speaker 4: And so that might be one way.

1:00:12 Speaker 5: That you could monitor the mongoose activity through, if you know, it always hangs out in a certain area. And there are, that's probably the most common, honestly, it's a reliable traditional technology. And so those types of tags are probably pretty cheap, which is also helpful if you're thinking about designing a study on a budget. And yeah, GPS callers would definitely provide you with more fine scale movement data. But again, it depends on your research question.

1:00:45 Speaker 1: All right.

1:00:51 Speaker 2: 13 here, and almost like a capacity building kind of a question here.

1:00:58 Speaker 1: How did you get into this field of study?

1:01:01 Speaker 2: And what are some tools or skills that a student or any early career researcher can learn on their own to get into the field of animal tracking or monitoring? Are there any open source tools? You already mentioned some in a previous question. Or initiatives that one can take part in? And also, how can someone start a project from scratch with minimal resources?

1:01:31 Speaker 5: I'll go first and let Claire answer this also. I did field work with mainly birds for a really long time, and several of the researchers I worked with in different field camps over time were doing different animal tracking projects, and I eventually got involved with their tracking projects, and I went to grad school to do that. That was one way that, you know, that was kind of my introduction into animal tracking. The first animal tracking project I saw was with albatrosses, and I just thought it was the coolest thing.

1:02:02 Speaker 5: And so I've managed somehow to keep doing that over these years, which is really amazing to me still. And so I would recommend getting any firsthand experience if you can. If you know of a research group that's doing something you're interested in, definitely try to See if there are any volunteer opportunities or field technician opportunities with them. Get out in the field with them and really understand what's going on out there. And I would also recommend a lot of the data analysis requires coding skills, and I would definitely recommend learning how to code in some language and just familiarize yourself with different things you can do with that.

1:02:41 Speaker 5: Because it's one thing to collect tracking data, but if you want to analyze it, you really need some coding skills. And yeah, as we've been saying, we will provide more resources in the slides today and on Thursday.

1:02:57 Speaker 4: And I'll pass it over to Claire. Yeah, thanks, Morgan. So I got into this field of study, I always say, because I loved maps. I loved looking at maps. I loved flying in an airplane, looking out the window and seeing what sorts of patterns I could see just looking at the ground. And then Being able to map animals was just a bonus. So unlike Morgan, I have relatively little experience going out in the field. My background is more from the statistics and math and GIS side. And from that perspective, I think a great way that I got started was working with people who had already collected data for one purpose and thinking about ways to use that data for asking and answering other questions.

1:03:56 Speaker 4: It's really important because the fieldwork is so labor-intensive, the data can be relatively expensive to collect. Tracking data can be relatively expensive to collect. to really try and make the most out of the data that has already been collected. So figuring out what data are out there and might have a little bit more in them to be able to understand about the world can be really valuable. And luckily, we're in an age where more tools, trainings, and even data are available publicly so that those can be a great place to get.

1:04:40 Speaker 2: Started thanks both uh morgan and claire those are great answers and i would uh probably add to that uh that uh a lot of times there's a lot of citizen science based you know projects uh in different places and that could be something that you can you know start with just uh start working as a statistician and collecting some data for some of these researchers. And eventually one thing will lead to another.

1:05:12 Speaker 1: All right. Question 14.

1:05:15 Speaker 2: Are there any public repositories of animal tracking data, or do you have suggestions from networking with folks that have such data.

1:05:25 Speaker 1: Who are interested in working with data scientists? Kind of similar to what we just answered, but go ahead.

1:05:33 Speaker 5: Yeah, some options include the MODIS network. And so MODIS is a receiver network that a lot of people tap into. And so a lot of different animals carry MODIS tags. And then when they fly past the receiver, that pings it and you have a record or timestamp of the animal passing through.

1:05:52 Speaker 4: And so there's a lot.

1:05:53 Speaker 5: Of those kinds of projects, which are all over the country, especially along the east and west coasts. And so that might be something worth looking into. Movebank.org also is a really big resource. They have a lot of data sets from lots of different types of animals.

1:06:09 Speaker 4: And so that's really.

1:06:09 Speaker 5: Cool to just kind of poke around and see all the tracks from all the different species around the world. Animal Telemetry Network is another one that's available. And again, we'll have links and slides from today and Thursday that offer some more of these resources.

1:06:26 Speaker 1: All right, cool.

1:06:30 Speaker 2: 15, then, could you also shed some light on using machine learning or deep learning algorithms for this purpose?

1:06:40 Speaker 1: I mean, really, computational complexities exist.

1:06:47 Speaker 4: Yeah, machine learning and other sort of similar computational methods can really help with the analysis of animal tracking data. Also very well known for analysis of remote sensing data. And they can also be used to integrate the two. So in session two, Morgan is going to show an example of using boosted regression trees, which are a machine learning method, to perform a species distribution model. Machine learning can also be used to process remote sensing data to get biological information that might be relevant for inputs into these models.

1:07:21 Speaker 4: And those are just a couple of examples of how these methods could be used, animal tracking data,

1:07:29 Speaker 4: is approaching sort of big data types of approaches to understand things like behavioral states or other sorts of behavioral or biological information that might be embedded in those data that are hard to find statistically.

1:07:51 Speaker 1: Okay.

1:07:51 Speaker 3: 16.

1:07:56 Speaker 2: How were the outliers handled in the aggregation of groups in the same species? This person in particular is thinking about horses in this question.

1:08:09 Speaker 5: Yeah, so typically you'd just remove outliers. And I assume that you're talking about an animal track that maybe has a wonky point that's, you know, 50 kilometers away from all the surrounding points. And typically you just remove those and those are removed either to speed thresholds.

1:08:25 Speaker 4: That we talked about earlier or some other.

1:08:26 Speaker 5: Like geofence essentially where like the animal can't go over there because of habitat, like the animal, the deer, but you know, that points in the ocean or something like that. So typically you'd remove those, but if you, were to keep them in, I would imagine that they would just kind of average out across, you know, your whole sample size, and they probably wouldn't affect your data too much if you have a large sample. But I would recommend just removing outliers.

1:08:54 Speaker 2: Okay, 17, those using drones, surveying drones, is it in into animal tracking? And can you give me some insight regarding drones and animal tracking, drone surveying used for these purposes?

1:09:10 Speaker 5: Yeah.

1:09:12 Speaker 4: Yeah.

1:09:13 Speaker 5: Yeah, so I am aware that drones are used to either find where animals are hanging out, like hot spots, you know, where a bunch of sea lions might be hauled out on the beach, for example. And they're also used to make counts of animals. And so this happens in, again, for a marine example, like on a beach where a lot of birds are hanging out or, you know, for sea lions or elephant seals, for example. But I'm not aware of drones that really work with the tracking tags themselves, but that's kind of a cool idea.

1:09:45 Speaker 5: And so that might be a good avenue of research for somebody listening today.

1:09:49 Speaker 4: Okay.

1:09:53 Speaker 5: Okay.

1:09:54 Speaker 2: What type of lamp cover can be mapped using remote sensing? And for this, we would like to We recommend to our participants that there's actually several outset trainings that already cover the topic of monitoring land cover changes using remote sensing.

1:10:17 Speaker 1: We can definitely.

1:10:19 Speaker 2: Recommend for you to visit or to go to the hyperspectral data for land and coastal systems. There's also remote sensing for conservation and biodiversity, and there's one specifically that it's land cover classification using satellite imagery. So feel free to go into the ARCET webpage and do a search for those. As always, with all the ARCET webinars, all the materials are there for you to download for your benefit. Okay, there's a question about R and RStudio. Is it needed for the homework? No.

1:11:05 Speaker 2: No, R is not needed to follow along the examples in session two. It's just to follow along the examples, but you don't need it for the homework. The homework will be posted on the training webpage on Thursday at the end of the training. And it's practically a multiple choice kind of homework.

1:11:32 Speaker 1: So no worries about that.

1:11:33 Speaker 2: All right. Is it feasible to study the spatial distribution of wild animals in protected areas and estimate their future distribution inside of it?

1:11:44 Speaker 4: Yes, this is a great question.

1:11:46 Speaker 5: I will be showing an example of this on Thursday in the second session. But yeah, you can definitely combine animal tracking and remote sensing data and projected future environmental scenarios in species distribution models to predict how animals might use habitats in the future within protected areas.

1:12:06 Speaker 1: Okay.

1:12:07 Speaker 2: 21, where and when will you be posting the results of our Internet of Animals report where you gather input on needs and recommendations from the community expert?

1:12:21 Speaker 5: Yeah, so our team, our Internet of Animals team, is actually working on a manuscript about this right now. And so I expect it will be published within the next year or so.

1:12:30 Speaker 4: And so I don't know.

1:12:32 Speaker 5: If we can post training information a year from now provide links to papers, but it's something that will be published in the peer-reviewed literature eventually.

1:12:43 Speaker 1: All right, cool. Stay tuned.

1:12:47 Speaker 2: So question 22, when someone wants to study the movement patterns of tracked animals and how they select habitat in response to habitat features, what are the best covariates they should consider? what would be the best model approach if they want to study at the individual variability level?

1:13:09 Speaker 5: Yeah, so we'll talk about this on Thursday quite a bit, but my overall big recommendation is just to do a literature search and that will help you start thinking about your specific species and also your study region because remote sensing data kind of varies between regions based on your habitat. So in the tropics, there's a lot of cloud cover and that can affect the different remote sensing variables that you can use efficiently. and effectively in your different models. And so it's really important to check the literature and see what other people have done for your specific region and research questions.

1:13:44 Speaker 1: All right.

1:13:45 Speaker 2: On 23, we'll probably do a couple more just because of the time limitations. But we'll make sure to answer all your questions in the final Q & A document.

1:13:58 Speaker 2: If you have tracking data or a suspected ontogenic shift in relation to environmental covariates, can the panelists recommend some methods that allow the relationship of the covariates to the movement data to change over time?

1:14:16 Speaker 4: Yeah. Step selection functions are one modeling framework that could be really appropriate for this question. In step selection functions and other sorts of resource selection or habitat selection frameworks, you could, for example, fit an interaction term between your ontogenic state or simply time and your environmental variable, which would allow you to see whether your animal's response to that focal environmental variable changes depending.

1:14:50 Speaker 5: On time or state.

1:14:55 Speaker 2: Okay, kind of related to that as well, when should SSF be used rather than RSF?

1:15:05 Speaker 4: So the answer to this really, it depends on a lot of things. There's also not necessarily a single right choice in any given case. So it's possible that you could use both methods and get the same or different answers to your questions. But I would say primarily this choice is going to depend on first the data that you have available and second the question you want to answer. So step selection functions are really good for understanding third order habitat selection, which is the selection of specific habitat patches, resource types, or locations within an animal's home range.

1:15:44 Speaker 4: But it's not really as flexible as other resource selection function or species distribution modeling methods. So it can't really answer questions about where a species range is located on the globe, for example, because it's really only considering available habitat where an individual could have moved to during a specific time window. It's not considering any environmental data outside of those locations. So, you might consider a resource selection function or species distribution model if you're interested in first or second order habitat selection questions.

1:16:25 Speaker 4: And also, it does depend on the data you have available. So, as mentioned in the presentation, step selection functions require regularly sampled or close to regularly sampled data. And if that's not the type of data that your tag is providing, then these other methods would be more appropriate.

1:16:45 Speaker 5: All right.

1:16:46 Speaker 1: Thanks, Claire.

1:16:48 Speaker 2: Are there any suggestions for what workflow or tidy apps or other resources to use when interested in doing analysis of both the individual and the population levels? The person has data collected as five-minute intervals for 50 moves and want to do both. But it's a ton of data before even bringing in the remotely sensed data as well.

1:17:15 Speaker 5: Yeah, so I started doing this, talking about how MoveBank actually has tools where you can integrate different things, and that'll provide you some workflows. There are other resources that we'll talk about in session two about this as well.

1:17:31 Speaker 1: Okay.

1:17:31 Speaker 2: All right, let's do one more, which is number 26, GNSS on a Herbie board, for example, a cow.

1:17:43 Speaker 1: In a mountain.

1:17:44 Speaker 2: Pasture is supposed to give us its position. The question is whether we can use the accelerometer to also infer weather through the movement.

1:17:56 Speaker 1: Of its head. Is it eating? How can this information be analyzed?

1:18:02 Speaker 5: Yeah, so this is definitely possible as long as the tag is, I assume a cow would be.

1:18:08 Speaker 4: A collar tag.

1:18:09 Speaker 5: So you just have to calibrate it so that you would know when the cow cow's head is moving, and so you could recognize that in the patterns in the data, what those would look like. But yeah, this is definitely possible. There are tools and R packages to process accelerometry data specifically. One that I've used recently is called MoveACC, but there are other ones, and there's also a commercial package. There's a commercial software, and the name of it is eluding me in this moment, but there are many, many resources out there to process accelerometry data.

1:18:43 Speaker 1: All right, cool. Thanks, Morgan.

1:18:45 Speaker 2: Okay, so we're going to make sure that eventually we're going to answer all the remaining questions, just for the sake of time. We're going to stay here today. Thank you again all for being with us today in this training. We look forward to seeing you or having you on Thursday for the second part.

1:19:07 Speaker 1: Thanks, Morgan and Claire.

1:19:10 Speaker 2: Also for all the contributions and for helping putting all of this together. And yes, with that, we will see you on Thursday then.

1:19:21 Speaker 1: Have a great day. Thanks, everyone.