https://cpw.cvlcollections.org/files/original/101f9cbf07b9fc853e93203168092bfd.pdf b3889042d1c57a905266f222a895c69a PDF Text Text One or more of the authors of this paper is a Colorado Parks and Wildlife employee. This is an open access journal article, archived on the CPW Digital Collections site as part of the CPW scholarly research archive. �Received: 19 May 2023 I Accepted: 27 November 2023 DOI: 10.1111/ 2041-210X.14274 Methods in Ecology and Evolution ~ RESEARCH ARTICLE &~,;:, .. s,cm A multi-property assessment of intensity of use provides a functional understanding of animal movement 12 G. Bastille-Rousseau • N . T. Gorman H. Manninen 3 4 e I 12 S. A. Crews • 12 I J.B. Pitman • 5 I S. Blake I I 12 E. B. Donovan • 12 I A. M . Weber • 12 M. W. Eichholz • e 12 M. E. Egan • E. M . Audia1•2 6 I E. Bergman I e I M . R. Larreur1•2 I N . D. Rayl7 'Cooperative Wildlife Research Laboratory, Southern Illinois University, Carbondale, Illinois, USA; 2 School of Biological Sciences, Southern Illinois University, Carbondale, Illinois, USA; 3 Department of Fish and Wildlife Conservation, Virginia Tech, Blacksburg, Virginia, USA; 4 Montana Cooperative Wildlife Research Unit, University of Montana, M issoula, Montana, USA; 5Department of Biology, St. Louis University, St. Louis, Missouri, USA; 6Colorado Parks and Wildlife, Fort Col lins, Colorado, USA and 7 Colorado Parks and Wildlife, Grand Junction, Colorado, USA Correspondence G. Bastille-Rousseau Email: gbr@siu.edu Funding i nformation IDNR-Wildlife Restoration Act, Grant/ Award Number: W213R1, #W13R23, #W135R22, #W87R45 and #W87R44; United States Fish and Wildlife Service Federal Aid Research Grant; Colorado Parks and Wildlife (CPW) Game Cash Funds; Rocky Mountain Elk Foundation; Pitkin County Open Space and Trails; National Science Foundation, Grant/ Aw ard Number: DEB 1258062; Max Planck Institute for Ornithology; National Geographic Society Committee for Research and Exploration; Galapagos Conservation Trust; Institute for Conservation Medicine of the Saint Louis Zoo; Woodspring Trust; Swiss Friends of Galapagos; SIU startup Fund Abstract 1. The intensity of use of a location is one of the most studied properties of animal movement, yet movement analyses generally focus on the overall use of a location without much consideration of how patterns in intensity of use emerge. Extracting properties related to intensity of use, such as the number of visits, the average and variation in time spent and the average and variation in time between visits, could help provide a more mechanistic understanding of how animals use landscape. Combining and synthesizing these properties into a single spatial representation could inform the role that a location plays for an animal. 2. We developed an R package named 'UseScape' that allows the extraction of these metrics and then clustered them using mixture modelling to create a spatial representation of the type of use an animal makes of the landscape. We illustrate applications of the approach using datasets of animal movement from four taxa and highlight species-specific and cross-species insights. 3. Our framework highlights properties that functionally differ in how animals use Handling Editor: Chris Sutherland them, contrasting, for example, heavily used locations that emerge because they are frequented for long durations, locations that are repeatedly and regularly visited for shorter durations of time or locations visited irregularly. We found that species generally had similar types of use, such as typical low, mid and high use, but there were also species-specific clusters that would have been ignored when only focusing on the overall intensity of use. 4. Our multi-system comparison highlighted how the framework provided novel insights that would not have been directly obtainable by currently available approaches. By making the framework available as an R package, these analyses can be easily applicable to a myriad of systems where relocation data are available. This is an open access article under the terms of the Creative Commons Attribution-Noncommercial License, w hich permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2023 The Authors. Methods in Ecology and Evolution published by John Wiley & Sons Ltd on behalf of British Ecological Society. Methods Ecol Evol. 2024;15:345- 357. w ileyonlinelibrary.com/journal/mee3 345 r.\... ~ �BASTILLE-ROUSSEAU 346 ET Al. Movement ecology as a field can strongly benefit from approaches that not just describe patterns in space use, but also highlight t he behavioural mechanisms leading to these emerging patterns. K EY W OR D S animal movement, clustering, GPS telemetry, intensity of use, mixture model, recursive movement, residency time, revisitat ion, space use 1 INTRODUCTION provide new insights into animal ecology (Brads et al., 2018; Freitas, Kovacs, Lydersen, et al., 2008; Riotte-Lambert eta I., 2013). However, Animal movement is a fundamental behaviour by which individ- it remains that none of the analyses mentioned above, when applied uals respond to their environment. As such, movement is often at alone, fu lly captures the ways in which an animal uses a location. the core of how an animal w ill access resources, avoid risks, inter- Ideally, information regarding the total time spent in that location, act with conspecifics and encounter disease (Lima & Zollner, 1996; the number of t imes the location was visited, the average duration Wittemyer et al., 2019). Given its role in a myriad of ecological pro- of each visit and the average duration of each interval between vis- cesses, understanding animal movement has become a predominant its should be combined to fully distinguish the diverse types of use. area of spatial and wildlife ecology, especially due to t he constant Quantifying the variation in the duration of each visit and the varia- development of technology to obtain relocations and ways of an- tion in the duration of each interval between visits to evaluate how alysing these data (Kays et al., 2015; Wilmers et al., 2015). Animal predictable/ regular the use of a location would provide additional movement data are also increasingly providing information that is information. Synthetizing all these different sources of information being applied in the management and conservat ion of species, nota- into a single representation of the use an animal makes of a location bly as a way of valuing the importance of specific areas to an animal in the landscape cou ld provide novel insights into its use of space. or a population (Wittemyer et al., 2019). For example, considering how frequently and regula rly locations are Although there are numerous ways to study animal movement, revisited could help better understand how animals patrol and main- most approaches can be grouped into broad categories that capture tain their territories (Graf et al., 2016). A lternatively, considering res- different aspects of the movement process. W ittemyer and co-authors idency time, the number of revisits and the predictability of such (2019) suggested that the existing plurality of approaches can be syn- visits could directly impact encounters between a predator and prey thesized into three main categories of movement analyses: intensity (Bastille-Rousseau et al., 2011; Mitchell & Lima, 2002). of use properties (e.g. definition of core range), movement path char- We developed a framework that characterizes the patterns in acteristics (e.g. speed and directionality) and the structural properties which an animal makes use of an area. In alignment with previous work of movement (e.g. landscape connectivity). Approaches related to (Bastille-Rousseau & Wittemyer, 2021), this framework relies on esti- intensity of use are the dominant analyses in the literature, perhaps mating several metrics that describe the intensity of use of an animal because they include approaches allowing estimation of home ranges at a location in terms of total time, number of visits, mean duration, and density isopleths (Kie et al., 2010; Powell & Mitchell, 2012), res- variat ion in duration, mean interval and variation in interval, and uses idency t ime analysis (Bastille-Rousseau et al., 2011; Freitas, Kovacs, machine learning to cluster that information to generate a landscape of lms, et al., 2008) and the w ide family of approaches estimating habitat use (i.e. a UseScape). We developed an R package named 'UseScape' selection functions (Boyce et al., 2003; Fortin et al., 2005). that extracts these metrics and then cl usters them using mixture mod- However, approaches that focus on intensity of use typically focus elling to create a spatial representation of how an animal uses the land- on the total amount of time spent at a location (or the total number scape. We first detail the workflow behind the UseScape package and of relocations) w ithout considering how the resulting use pattern then illustrate its applications using animal movement datasets from emerges. For example, various foraging strategies and other move- four taxa: elk, giant tortoise, coyote and white-tailed deer. Finally, we ment behaviours can lead to the same overall intensity of use (Bastille- highlight some of the insights gained from this approach for each study Rousseau et al., 2010). An animal may stay in a patch until it is fully system, as well as insights from cross-system comparisons. depleted and not revisit this patch for the rest of the season, or it can revisit the same patch frequently for shorter periods of time (Benhamou & Riotte-Lambert, 2012). In addition, an animal might also visit these 2 MATERIALS AND METHODS patches on a regu lar basis or irregularly. Such behavioural differences can have strong implications for animal ecology, including predator- 2.1 Generating the UseScape prey interactions, disease transmission, optimal foraging strategies and the potential ecosystem processes that an animal might influence. Similarly to t he approach developed in Bastille-Rousseau and Approaches that investigate single residency time or revisitation Wittemyer (2021) and Bastille-Rousseau et al. (2018) and the rates (also called recursive movement) have been developed and can moveNT package (https://github.com/ BastilleRousseau/ moveNT), �11 1 1ii,&!iiffidffii,i-ii#iii,iffi@H■ BASTILLE-ROUSSEAU er AL 347 the first step of the workflow is to convert t he location dataset the entering and exiting step length that is in that cell (for example, into a movement trajectory (Figure 1a). The next step is to select when one-third of an entering step length is within a cell, the entry a cell size so that the trajectory can be overlaid on a movement time is adjusted by one-third of the duration of that step prior to grid. Selecting the cell size is a critical decision at the core of the the first relocation w ithin t hat cell). Once the timing history is ex- UseScape workflow and has the potent ial t o impact the insights tracted for each cell, several cell-level metrics can be calculated that gained from UseScape and their interpretability. We propose three summarize the type of use an animal is making of that cell. These ways of select ing the cell size and illustrate in our example appli- cell-level metrics include the tot al durat ion, the number of visits, the cations of its use (Figure 1b). The first approach is to test various average duration of each visit, the variation in the duration of each cell sizes and extract the residency time t hat the animal spent in visit (measured by the coefficient of variat ion of the duration of each each cell. Like previous work on residency and first passage time measure), the average interval between visits and the variation in (Fauchald & Tveraa, 2003), we suggest selecting the cell size that the average interval between visits (again, measured by the coeffi- maximizes the variance in residency time to dist inguish between cient of variat ion). Each of these metrics can then be exported into a encamped and exploratory movement s. The second approach is to raster (or stack) format (Figure 1d) and displayed within R. use the median step length of an animal estimated using relocation The second step to generate UseScape is to synthesize these data (Bastille-Rousseau et al., 2018). Lastly, when the above options metrics into a single representation of how an animal uses each cell provide cell size estimates that are below the GPS error, it might be (Figure 1e). We used Gaussian mixture models (Duda & Hart, 1973) preferable to use a cell size equal to or larger than the GPS error. to cluster cells based on the above metrics. Gaussian mixture models Once the trajectory of the movement is overlaid on a grid are a soft and unsupervised clustering approach where each point (Figure 1c), the third step is to extract each time an animal enters is given a probability of belonging t o a specific cluster. We based and exits a cell (hereaf ter t he timing history) for each cell within a our clustering on the mclust R package (Scrucca et al., 2016), which grid. The entry and exit time are adjusted based on the proportion of automatically compares different numbers of clusters and uses (a) Convert GPS locations to movement trajectory (b) Determine pixel size Met hod 1 • Variation in residency time 100 300 200 400 500 Pixel size Method 2 - Median step length (c) Overlay trajectory to grid f::,,, <..._ l's:: \ FIGURE 1 Worfkow of the UseScape approach. The workflow starts with (a) converting GPS locations into movement trajectory, (b) determining cell size based on one of three approaches (variance in residency time, median step lengths or GPS error), (c) overlaying t rajectory to grid, (d) extracting a variety of intensity of use metrics based on number of visits (frequency), t ime spent in a cell (total duration, average duration, coefficient of variation of duration [CV durat ion]), t ime between visits of a cell (average interva l and CV interval) and (e) clustering these met rics using machine learning to produce a synthetic representation of space use (Figure modified from Bastille-Rousseau et al., 2018). Ill ,\ '\ " ~ I 0 1) L " I\ .. \ --- l'I - ~ 1/ 100 200 300 400 500 600 S l&p length < " ~:, (d) Metrics Frequency Total Duration Avg. Duration CV Duration Avg. Interval CV Interval •••••• v (e) Machine learning {Unsupervised classification) Intensity of Use Categorization Legend High Use Regular Use Rest/Stopover site Excursion Low use System specific ■ ■ ■ ■ ■ ■ �348 I IliOIIM:/Mj,j,4i,j.1ffi@IHJ■ 141 BASTILLE-ROUSSEAU ET Al. BIC (Schwarz, 1978) to select their optimal number. Covariates are Information). We chose these 2years because 2014 temperatures centred and standardized before being clustered, but they can be were generally average, while 2015 exhibited notable warm periods re-transformed to ease biological interpretation. The UseScape (National Oceanic and Atmospheric Ad ministration (NOAA), 20 22). package provides a clustering f unction to cluster at the individual Our third example compared the use of the two primary ap- level and a function to synthesize at the population level, as pre- proaches to select the cell size. We used GPS data from two coyotes sented in Bastille-Rousseau and W ittemyer (2021). All functions and (Canis latrans). These two individuals exhibited different movement a vignette are available within the R UseScape package available on strategies; one coyote was a resident with a home range of 10.2 km 2, GitHub (https://github.com/ BastilleRousseau/UseScape). and the other exhibited exploratory behaviour, taking long excursions outside its home range (45 km2) and then returning (Gorman et al., 2023). Their collars w ere programmed wit h a 1.5-h GPS fix in- 2.2 Applications terval during 2020 and 2021. The study site is in central Illinois, USA, consisting of a patchwork of private properties, public land and small We applied the approach to fou r different systems and performed towns adjacent to a large reservoir. Most of t he land is used for row slight variations in the analyses to highlight different features of the crop corn and soybean agriculture interspersed with some forested approach. Our fi rst example examined t he movement of a migra- patches. We used the median step length and variation in residency tory adult female Rocky Mountain Elk (Cervus canadensis) in western time approaches to select cell size and selected the resolution that Colorado, USA, to evaluate the association between different types led to the most interpretable output for each individual. of use and human activities. The locations were t aken every 2 h from For our last application, we used resident species from the range May 2019 to June 2021, capturing two complete round-trip migra- to investigate the association between patterns of use and type of tions. Most elk in this herd are mid-distance elevational migrants, landcover. We used data from an adult female white-tailed deer with 15 km and several 100 m of elevation separating their winter (Odocoileus virginianus) tracked from January 2021 to January 2022 and summer ranges (Crews, 2023). Property development, outdoor in the same study area as the two coyotes above. GPS locations were recreation and human activity are expanding within the immediate recorded with a fixed interval of 30min. Female white-tailed deer vicinity of t his individual (Mao et al., 2013), and the w inter range generally establish home ranges that overlap close to those of their of this individual is located adjacent to a ski resort around 2300- genetic relatives. To determine whether if landcover differed be- 2500m. The migratory corridor and summer range of t his elk were tween the UseScape clusters, we calculated the proportion of cells mainly within the Maroon Bells Snowmass wilderness, minimally im- in each UseScape class that fell under forest, agricultural or urban pacted by human activity, with elevations ranging from 2500 to over cover, as def ined by the National Land Cover Database (NLCD). 3600 meters. For this application, w e used the median step length of We determined whether the proportion of each landcover class this individual as the cell resolution and focused our interpretation differed in each UseScape cluster using a chi-square test (Garson on a descriptive association of the different types of intensity of use & Moser, 1995). All animals handling followed best practice estab- in relation to anthropogenic land use. Our second example examined the definition of clusters in a longterm dataset. We used data from an adult female giant tortoise on lished by the American Society of Mammologists and were approved by SIU (IACUC 21-028 and 21-021), CPW (IACUC approval number 03-2020) and SUNY-ESF (IACUC approval number 121202). Espanola Island in the Galapagos (Chelenoidis hoodensis). Since tortoises are largely immobile at night, the GPS units logged locations hourly from 6:00 to 19:00 from December 2010 to December 2022. 3 RESULTS Espanola Island is a relatively flat (maximum elevation 206 m) and arid (annual ra infall 10- 600 mm; Gibbs et al., 2014) 60.5 km 2 island. Applying the approach to the four species highlighted several similar The island hosts a homogeneous distribution of herbaceous plants types of use as well as species-specific clusters (Figure 2). The opti- of the arid zone, but large, arboreal Opuntia cacti are found near the mal number of clusters detected ranged from three to five, w ith the central region of the island (Gibbs et al., 2014). C. hoodensis tends to resident coyote being the only example with fewer than five clusters be sedentary or nomadic (Bastille-Rousseau et al., 2017), especially (Figure 5). All individuals had a high-use group and at least one low- during the dry season when they rely on Opuntia cacti for shade use group, and all species except the coyote also had a moderate-use (Blake et al., 2021). During the rainy season, tortoises are less re- group (Figure 2). These t hree clusters were generally differentiated stricted to areas where shade is available and can wander for several based on the total amount of time spent in them and how frequently weeks to forage (Blake et al., 2021). Since the median step length they were reviewed (Figures 3 - 6). For all individuals except deer, was approximately 10 m and was below the GPS error for the device this gradient in intensity of use was associated with a longer resi- (Bastille-Rousseau, unpublished), we used a cell reso lution of 20m, dence time in those areas. Giant tortoise, coyote and deer all had a which corresponded roughly to the device error. To test for temporal stopover or resting cluster that corresponds to areas with low over- variation in intensity of use, we performed the analysis three times: all use but with a longer residency time when visited compared to once using a period of 10years (2011- 2021, Figure 4) and twice low-use clusters. There were also species-specific clusters found in using a period of 1 year (2014 and 2015, presented in Support ing one or two species, such as high-use areas emerging from frequent �BASTILLE-ROUSSEAU er AL FIGURE 2 GPS locations and landscape of use for each of the four study species. Rocky Mountain Elk (Cervus canadensis, '1' panels), Galapagos Giant Tortoise (Chelenoidis hoodensis, ' 2' panels), Coyote (Canis Jatrans, '3' panels) and White-tailed Deer (Odocoi!eus virginianus, '4' panels). 'A' panels show the raw GPS points used as inputs for the UseScape analysis. 'B' panels show t he raster outputs w ith f ive of the six colour-coded use classes. The system-specific cluster category represents the type of use that might differ across species. -.~r . . • - ~ f (1 b) "'r I' .: " · . ..... l ·) (2b) -~ (4b) (3b) Raster Legend revisits of short duration (elk), low-use emerging areas (incidental) ■ High use ■ Moderate use ■ Stopover or rest ■ System specific ■ Low use, infrequent ■ Low use, quick revisit (µ = 629.27) and an extremely high total t ime spent there (µ = 2141 h). frequently revisited (deer), passthrough areas (coyote and tortoise) The moderate group had an intermediate number of visits (µ = 97.8) or low-use areas quickly revisited (elk and coyote). and was adjacent to the high-use areas. The first of the three low- The elk analyses were carried out at a resolution of 93 m, cor- use classes, 'passthrough', was characterized by a low average num- respond ing to the median step length of our collared elk (Figure 3). ber of visits (µ = 24.6) and a mean duration (µ = 2.73h), along with a The analysis resulted in five distinct use classes. The first class repre- moderate mean interval between visits compared to the higher-use sented moderate use (µ = 6.47 visits,µ= 13.55 h total duration) with classes (µ = 144.5 days). Passthrough cells surrounded areas w ith long visit intervals (µ = 98.3days). These were areas with average- high concentrations of moderate-use cells. The second low -use but infrequent- use and were distributed primarily on the edges of class, 'stopover', had a longer mean duration than passthrough cells the winter range or adjacent to the core areas. The second class rep- (µ = 3.58h), despite having a lower number of visits (µ = 10.5) and a resented areas of low use that were quickly revisited and associated longer mean interval (µ = 226.8days). The low-use group was differ- with few visits(µ = 3.51 visits), above-average stays (µ = 2.43 h) and entiated by a high mean interval (µ = 449.0days), a low tot al duration very short visit intervals(µ = 7.44days). The third class represented (µ = 11.64hours) and a small number of visits (11 = 4.33), indicating low-use areas (µ = 3.57 visits, 11 = 7.39h total duration), with short that these cells are the least frequently visited. Comparing the long- stays (11 = 2.0h) and long visit intervals (11 = 168days). The fourth term classification with annual classifications revealed changing class represented high-use areas (µ= 18 .15 visits, µ=60.49h total temporal patterns in the areas used annually and, by extension, how duration), wit h long stays (µ=3 .69 h) and low to moderate visit inter- a given area is classified (Supporting Information). vals (µ = 42.8days). This class was found in the core area within the We utilized different methods to determine the best resolution w inter range, associated with high-quality undisturbed habitat. The at which to cluster the two coyotes' movements. For the resident fifth species-specific class was also a high-use category (11 = 12.23 coyote, we used the median step length (80 m). The graphical ap- visits, µ = 21.08h total duration), but with very short stays(µ = 1.74h) proach based on variation in residency time yielded a resolution and moderate visit intervals(µ = 59.18days). This class was primarily size that was too small for interpretability (18 m). For the explor- present in the w inter range, and its distribution largely fol lowed that atory coyote, we found that the graphical approach led to better of anthropogenic or forested areas. Low use was the most common resu lts with a cell size of 600m; the median step length (128ml class, comprising -31% of all cells, closely followed by moderate use yielded too small of a resolution (see Supporting Information). at -28%. High use (disturbed), low use quick revisit and high-use This indicates that the graphical approach to determining resolu- classes were less frequent, making up - 19%, - 16% and -4% of total tion may be more suitable for far-ranging individuals, while using cells, respectively. median step length may be a better alternative for individuals with The analysis of the giant tortoise individual resu lted in five use high home range fide lity (Supporting Information). Three clusters classes, including high use, moderate use and three distinct low-use w ere identified for the resident coyote and five for the explor- groups (Figure 4). The high-use cells were clustered near the centre atory coyote. The resident coyote presented only low-, moderate- of the geographic area and characterized by a high number of visits and high-use groups, which were most easily characterized by the �BASTILLE-ROUSSEAU 350 0.0 ....... 1.5 3.0 Km • •• • •• • .. - -~-i':;: ~. ,. • 0.6 00 ET Al . 1-2 km • -■.. • •• ~ • • ✓• • I ) ...• • • Low use NV TD MD CVD Ml CVI Freq 3.57 7.39 2.00 53.99 168.05 119.21 31.5% Low use, Quick revisit 3.51 I 8.35 2.43 52.28 7.44 96.21 15.9% Moderate use High use 6.47 13.55 2.06 63.47 98.33 144.83 28.2% 18.15 60.49 3.69 72.13 42.8 184.73 4.4% High use, Disturbed 12.23 21.08 1.74 68.21 59.18 59.18 19.9% FIGURE 3 GPS locations and zoomed UseScape map of a Rocky Mountain elk. Left: The extent of our collared elk's re locations over the tracking period, with the winter range in the north and the summer range in the south. A dashed blue box represents the extent shown in the second panel, highlighting the winter range. Right: UseScape results on the winter range. Below: a matrix and legend showing the UseScape outputs for each class. All reported metrics are at the cell level: NV is the number of visits, TD is t he total duration in hours, MD is the mean duration in hours, CVD is the coefficient of variation of duration, Ml is the mean interval between visits in days and CVI is the coefficient of variation of intervals between visits. Values most important in assisting with cluster differentiation are bolded. Freq is the frequency (proportion) of each class as a percentage of all UseScape cells in the output. total time spent in each area (µ = 2.8 h for low use, µ = 9.42 h for characterized by a low number of visits (µ = 3.55), a mean dura- moderate use and µ=33.24h for high use), representing different tion of low to medium length average duration (µ = 3.08 h) and levels of use around the home range. Low use was the most fre- total duration visits (µ = 11.12 h) and low intervals between visits quent (47%), while high use was the rarest (23%), but no cluster (1, = 5.30 days). The low use quick revisit cluster was followed by a was extremely rare. The first group of exploratory coyote clus- traditional low-use cluster, and then a cluster consisting of areas ter corresponded to a 'low use areas w ith quick revisit ' (Figure 5), within the home ra nge that were used for travel, 'passthrough', �1ii,&!iiffidffii,i-ii#iii,iffi@H■ BASTILLE-ROUSSEAU er AL FIGURE 4 UseScape output maps for a fema le Galapagos giant tortoise (Chelonoidis hoodensis) on Espanola Island. UseScape is based on all data combined (2010- 2022). Below the map is a matrix containing the UseScape output s for each class. All reported metrics are at the cell level: NV is the number of visits, TD is the total duration in hours, MD is the mean duration in hours, CVD is the coefficient of variation of duration, Ml is the mean interval between visits in days and CVI is the coefficient of variation of intervals between visits. Values most important in assisting with cluster differentiation are bolded. Freq is t he frequency (proportion) of each class as a percentage of all UseScape cells in the output. 0 r-- 11 1 351 ■ M .--, I 0 I..(') r-M .--, I ■••• .. 0 00 M .--, I - 0 -89.685° 150 300 m -89.68° Pass !Moderate usel through High use NV 4 .33 10.50 24.60 97.80 629.27 TD 11.64 37.78 66.04 278.93 2,141.25 MD 2.70 3.58 2 .73 2.82 3.35 CVD 80.76 114.56 123.38 130.75 127.26 Ml 449.00 226.83 144.48 49.58 10.90 CVI 115.58 188.05 229.74 280.96 518.04 43% 19% 19% 14% 4% Freq which had high number of v isits (µ = 10.12) and high intervals common clusters, containing approximately 60% of all combined (µ = 15.06 days), but a low to moderate duration (µ = 2.19 h). The cells, followed by stopover (18%), passthrough (14%) and high use st opover sites used during the excursions were rarely visited (10%). (µ = 6.43) but had a longer duration for each visit(µ= 5.78 h). The We used the median step length (36 m) to generate UseScape for areas of high use were visited frequently(µ = 18.92) and for lon- the white-tailed deer. Five clusters were found. Cells in the low-use ger duration(µ = 5.8 h), resu lting in a long total time spent in these class were visited rarely(µ = 3.54)and fora short duration (µ = 0.25 h). areas (1, = 101 h). Low use and low use quick revisit were the most An 'incidental use' cluster represented cells visited more frequently �352 BASTILLE-ROUSSEAU ET Al. 39 s1•N 400"N Resident Exploratory 39 SO' N . 39.59.N L 3?.9'N •I •..~~J~.-:.t 39.8'N 'I I' 39.58'N • I 397'N ; . .: ~ . ... 39.57 N 39.5S N 39.G'N • 0.0 0.5 1.0 1.5 2.0km ■ l:~-1 .. ..-...= - or,: n .r.21rn J9.5'N 88.70-W ,._....., ...'W Low use, Low use Passthrough 4. 24 10.33 2.59 75.49 53.56 129.00 30% 10.13 22.73 2.19 69.78 15.06 186.27 14% quick revisit NV TD MD CVD Ml CVI Freq 3.46 2.82 0.80 74.53 35.87 94.33 47% 6.42 9.42 1.52 78.69 15.63 118 .84 30% 12.13 33.24 3.07 85.33 11.59 137.53 23% NV TD MD CVD Ml CVI Freq 3.55 11.12 3.08 66.16 5.30 98.35 29% 6.43 34.78 5.78 83.98 14.22 140.54 18% 18.92 101.04 5.80 84.79 5.51 274.67 9% FIGURE 5 UseScape output maps for the resident (left panel) and exploratory (right panel) coyotes. Corresponding tables located beneath each map contain the output data for each cluster. The resident coyote output had only three clusters, while the exploratory coyote output had five, ranging over a variety of movements. A zoomed representation of t he exploratory coyote core range is provided. All reported metrics are at the cell level: NV is the number of visits, TD is the total duration in hours, MD is the mean duration in hours, CVD is the coefficient of variation of duration, Ml is the mean interval between visits in days and CVI is the coefficient of variation of intervals between visits. Values most important in assisting w ith cluster differentiation are balded. Freq is the frequency (proportion) of each class as a percentage of all UseScape cells in the output. (µ = 8.01) for a similar average duration (µ = 0.30h). Stopover cells without much consideration of how these patterns emerged (Bastille- were less common than most other classes (10%), and while deer Rousseau et al., 2010). W e developed a novel analytical framework only visited this group a few times on average(µ = 5.377), each visit that specifically characterizes how use arises at a location and al- lasted longer (µ = 0.65 h). Moderate-use cells were the most common lows clustering of locations with similar types of use. In doing so, (30%) and were visited an intermediate number of times (1• = 15.61) our framework highlights areas that dif fer in how animals are using for an average amount of time (1, = 0.46). These areas were found them, contrasting heavily used locations that emerge because they near and around the centre of home range use and the core use area. are frequented for long durations versus locations that are repeat- The deer visited high-use cells the most frequently (µ=31.90) and edly and regularly visited for shorter durations of time. Applying the for a longer duration (µ = 0.65h). High-use cells were also common fra mework to individuals from four different taxa highlighted similar (30%) and mostly found in the core of the home range; however, types of use but also identified species-specific clusters. This com- some high-use cells were found throughout the home range. The parison between species illustrated that there are more subtleties proportion of points in each type of land cover differed according to in how animals use a location than the total t ime they spend there. the UseScape cluster (?= 168.16, p < 0.01). All classes were located Overall, our framework has broad applications in how we study ani- most frequently in forest cover and least frequently in urban cover. mal space use and is directly applicable across a variety of systems. However, t his deer was particularly unlikely to make stopovers in urban cover. Testing the framework w ith various taxa highlighted t he unique features of each system. In the case of Rocky Mountain Elk, our approach distinguished between two types of heavily used areas: one that emerges because of long visits and one that emerges be- 4 DIS CUS SIO N cause of visits of short duration. These two types of use could have ended up being lumped together w ith traditional approaches Although intensity of use has been one of the most studied prop- (but see Bastille-Rousseau et al., 2010; Brads et al., 2018; Riotte- erties of animal movement (Wittemyer et al., 2019), studies have Lambert et al., 2013). Likewise, habitat selection functions often generally focused on the overall use an animal makes of a location consider attributes associated with both types of use as selected �1ii,&!iiffidffii,i-ii#iii,iffi@H■ BASTILLE-ROUSSEAU er AL 11 1 353 J'l"34N J9033'30'N J9033'N Moderate High Use Use Urban Ag Forest 2.82 70.42 26.76 8.72 24.16 67.11 0.50 28.86 70.65 7.4 17.99 74.55 8.64 18.27 73.09 19.28 MD CVD Ml CVI Freq 882.12 89.39 14 % 15 o/c 0.65 0.63 86.12 477.00 78.49 82.20 236.06 137.36 174.98 196.81 30% 30% 10% FIGURE 6 UseScape output for white-tailed deer plotted against landcover classification. The top left shows the deer GPS locations overlaid on a resource layer categorized as forest (green), urban (red), agriculture (yellow) or water (blue). The top right shows the UseScape output. We determined the proportion of cells in each cluster located in each of three landcover types (urban, agriculture and forest). The bottom left table shows the percentage of points in each cluster belonging to each landcover type. The bottom right table shows the UseScape output values for each class type and each variable. All reported metrics are at the cell level: NV is the number of visits, TD is the total duration in hours, MD is the mean duration in hours, CVD is the coefficient of variation of duration, M l is the mean interval between visits in hours and CVI is the coefficient of variation of intervals between visits. Values most important in assisting with cluster differentiation are bolded. Freq is the frequency (percentage) of each cluster of all cells in the output. without providing context regarding the behaviour behind this in the area. The separation of two different high-use categories selection (Bastille-Rousseau et al., 2010). The high use of devel- may suggest differences in the resources provided in these areas, oped areas is more intuitive when given context regarding the be- resu lt ing in differing behaviours. For example, the presence and haviour of the elk in those locations. Elk have been shown to alter abundance of ponds that occupy a portion of this individual's win- their space use in human-modified landscapes and may display ter range may help explain frequent visitation with short stays, increased movement speed in more intensely developed areas, re- as they could serve as a re liable source of water. In addition, the ducing their use and selection for areas like the development pres- presence of a highway along the southern end of the range ap- ent in the western port ion of the winter range (Webb et al., 2011). peared spat ially associated with the disturbed high-use cluster However, when available, elk often make use of agricultural or and may contribute to the short duration of visits associated with less developed areas in the presence of additional resources (e.g. this cluster. forage) and demonstrate selection for water sources within their While the elk example highlighted different types of heavily home range (Barker et al., 2019; Brook, 2010). This elk appeared to used areas, the giant tortoise and coyote examples showcased dif- alternate its use between a forested area in the south of its winter ferent types of low-use areas. The giant tortoise had three types range and a large ranch separated by a sizable highway, with most of low-use areas: (1) low-use areas w ith rare short-duration v is- high-use clusters concentrated in this area. This ranch may act as its ('low use'), (2) low-use areas with rare visits but with a lon- a refuge from the more intense human activity and development ger stop ('stopover') and (3) low-use areas with more frequent �354 I IliOIIM:/Mj,j,4i,j.1ffi@IHJ■ 141 BASTILLE-ROUSSEAU ET Al. but short visits ('passthrough'). The high-use and moderat e-use types, but they can still differ in how they use these cover types clusters were easily distinguished from th e three low-use clus- (Hewitt, 2011). In such a scenario, naive habitat selection function ters by the steep increase in t he overall number of v isits and total analyses focusing on spatial covariates, such as landcover, might not duration spent in those cells. The use of space for giant seden- show patterns in the selection of specific landcover but ignore the tary tortoises on arid islands such as Espanola is strongly shaped functional role of landcover. by the need to remain near Opuntia cactus for food and shade We considered six metrics in our cluster definitions. For indi- (Gibbs et al., 2008), which could explain the extreme use of high- viduals from the four species considered, the clusters were more use pixels despite their overall low frequency (4%). However, our easily defined by a subsample of these metrics. Clusters were gen- resu lts highlighted that (at least for this one individual) there is erally defined by the number of visits, the total duration and the also some structure in the rarer movement away from th e cactus. average duration of those visits, while the elk, giant tortoise and We also noted that the tortoise data span more than a decade, exploratory coyote classifications also featured the average dura- and thus there may be nuance to how habitat is used temporally. tion of th e intervals between visits. Interestingly, our two met rics For example, throughout the study period, Oceanic Nio Indices on variation in duration of visits and variation in interval between have exhibited both warm and cool episodes (National Oceanic visits were not as informative for the selected species. Although and Atmospheric Administration (NOAA), 2022), and the climatic these results are not surprising, we expect these two metrics to be conditions of the Galapagos archipelago are particularly sensitive particularly useful for highlighting specific clusters in other spe- to these fluctuations (Wang & Fiedler, 2006). Figure 51 in the cies. For example, territorial animals might need to revisit specific Supporting Information shows the resu lts of single-year analyses locations on a regular and constant basis for scent marking or other in 2014 and 2015 and showed how habitat use and tortoise move- activities related to territory maintenance (Giuggioli et al., 2011). ment are not uniform over time. More work would need to be done Beavers are known for their propensity to repeatedly revisit and to directly link the variation in local climatic conditions w ith tor- refresh territorial markers, which serve as deterrents to neighbour- toise movement on Espanola Island. ing conspecifics (Rosell & Thomsen, 2006). In such a case, the vari- Similar to the giant tortoise, the exploratory coyote also had two ation in the interval between visits would be low. Relatedly, some infrequently visited clusters: low use and stopover, which differed in patches of habitat could only be used for very specific functions total duration and average duration. A third type (the passthrough such as resting or f eeding and might therefore be used for a consis- cluster) was revisited so much (around 10 t imes) that it became tent period of time, result ing in small variation in duration (Bakner moderately used (more t han 20h of total t ime spent on average). et al., 2022; Bista et al., 2022). Interestingly, the resident coyote clusters were much simpler (low-, One key consideration when applying our framework regards moderate- and high-use areas). The simple delineation of three res- the size of the cell to use. We proposed three approaches to select ident coyote clusters, in contrast w ith the greater variet y in the five the cell size, but our cross-species applications highlight that resu lts exploratory coyote clusters, highlights the variat ion in space use might be sensitive to the cell size used. We found that using the we were able to categorize between different individual movement variance in residency time approach versus the median step length strategies within a single species. Although it is often assumed that can provide fundamentally different estimates of pixel size, often resident coyotes exhibit more acute movement patterns than ex- ranging by an order of magnitude in difference. In our application ploratory or transient coyotes (Webster et al., 2022), our results, al- with coyote (Figure 5 and in Supporting Information), we found that though with a sample size of two, highlight t hat exploratory coyotes each approach highlights dif ferent types of clusters and might be can also display complex movement patterns. better suited for specific questions. The variance in residency time Our analysis using the white-tailed deer data highlighted the typ- approach seems to provide a larger cell size than the median step ical gradient of intensity of use (low, moderate and high) with addi- length and seems especially useful for an animal displaying large- t ional types of low-use clusters (incidental and stopover). Although scale movement outside of a core range. This approach provides all clusters were t he most common in forest cover, some clusters more interpretable clusters in the exploratory coyote example. On occurred less frequently in some types of land cover. Given that the other hand, the median step length was better at highlighting intensity of use is directly related to habitat selection behaviour variation in use within a core area. It created a more intuitive out- (Freitas, Kovacs, lms, et al., 2008), differences in the proportion of put for the resident coyote. The third approach based on GPS error landcover t ypes in each cluster appeared to be less associated with should be used when the median step length is less than the es- intensity of use t han w ith the role of the landscape. Specifically, t imated GPS error. To summarize, we suggest that the variance in deer showed relatively consistent but low use of urban cover other residency time approach be used when the interest is in understand- than for stopovers that were nearly never located near roads. Since ing variation in intensity of use at a larger scale, while one of the urban cover often represented roads in the study area, it is unlikely other approaches should be used when examining w ithin-core area that they would opportunistically stop while moving through these patterns. lastly, alternative approaches could also be considered. areas. Across the clusters, the proportion of cells in each cover t ype One approach, inspired by Alavi et al. (2022), uses autocorrelation in was similar to the proportion of t hat cover type in the study area. the velocity of movement data to determine an optimal grid size for Deer are habitat generalists that may not avoid any of these cover identifying independent directional changes in animal movement. �1ii,&!iiffidffii,i-ii#iii,iffi@H■ BASTILLE- ROUSSEAU er AL 11 1 355 Although we kept the application of our approach simple, several assistance with collaring white-tailed deer and coyote. Fredy extensions are possible. First, w e only performed the analyses on sep- Cabrera for assistance with tagging the giant tortoise and retrieving arate individuals, but it is also possible to synthesize clustering from the data. The Galapagos National Park Directorate (GNPD) and the mult iple individuals into population-level results. Bastille-Rousseau Charles Darwin Foundation (CDF) provided critical technical, logis- & W ittemyer (2021) proposed a two-step clustering process t o pro- tical, administrative and political support. Coyote and white-tailed vide those inferences. This f unctionality is already integrated into the deer movement data were supported through an IDNR-Wildlife UseScape package. The deer analysis also offers a simple example of Restoration Act grant #W87R44 and #W87R45. Elk movement data how clustering can be combined w ith additional analyses to evalu- were supported by a United States Fish and Wildlife Service Federal ate how the landscape might influence different types of use. More Aid Research Grant, Colorado Parks and Wildlife (CPW) Game Cash complex analyses could be performed, for example, by using multi- Funds, auction and raffle grants administered by CPW, t he Rocky nomial regressions to evaluate how different landscape features are Mountain Elk Foundation and Pitkin County Open Space and Trails. more likely to be associated w ith specific clusters. Such a regression Giant tortoises' movement data were supported by the National framework can also allow for proper consideration of spatial autocor- Science Foundation (DEB 1258062), the Max Planck Institute relation Bastille-Rousseau & Wittemyer (2021). Similarly, association for Ornithology (Radolfzell, Germany), the National Geographic analysis to evaluate how conspecifics might overlap in space can be Society Committee for Research and Exploration, the Galapagos highly informative. For example, understanding whether the areas of Conservation Trust, the Institute for Conservation Medicine of the high use of some individuals overlap with the areas of high use of Saint Louis Zoo, the Woodspring Trust and the Swiss Friends of other individuals is key to understanding many aspects of a species' Galapagos. NTG, AMW and ME were supported by IDNR-Wildlife biology, as it can provide information on intraspecific interactions Restoration Act grants #W87R44 and W87R45, STC through the such as territoriality (Giuggioli & Kenkre, 2014) as well as the poten- SIU Klimstra fe llowship, JBP and EBD through the SIU startup Fund, tial for disease spread (Dougherty et al., 2018). Likewise, new insights HM and M L through IDNR-Wildlife Restoration Act #W135R22 into interspecific competition and predator- prey interactions could and #W13R23 and EA through IDNR-Wildlife Restoration grant be gained by expanding the approach to compare overlap in the type W213R1. of use between multiple species. We developed a new analytical framework to characterize the mechanisms that lead to the intensity of use of a location by an an- CONFLICT OF I NTEREST STATEMENT None to declare. imal or animals. Our multi-system comparison highlighted how this straightforward framework can provide novel insights that would PEER RE V IEW not have been possible with currently available approaches. By The peer review history for this article is available at https://www. making the framework available as an R package, these analyses can webofscience.com/ api/gateway/ wos/ peer-review/ 10.1111/ 2041- be directly applied to numerous systems where location data are 210X.14274. available. Movement ecology as a field can strongly benefit from approaches that not just describe patterns in space use, but also high- DATA AVA ILABILITY STATEM ENT light the movement mechanisms leading to these emerging patterns. A R package and vignette containing and explaining all f unctions These mechanisms are often critical to better understand a variety developed here are available at https://github.com/ BastilleRous- of ecological processes, such as intra- and inter-specific interactions seau/ UseScape, and the first release of the R package is archived (competition and predation), social interactions, disease spread and in a Zenodo Digital Repository at https://zenodo.org/doi/ 10.5281/ ecosystem functions. zenodo.10277762 (Bastille-Rousseau et al., 2024). AUTHOR CONTRI BUTIONS ETHICS APPROVAL AND CONSENT TO PART I CI PATE G. Bastille-Rousseau, E. M. Audia, S. A. Crews, E. B. Donovan, M. All animals handling followed best practice established by the E. Egan, N. T. Gorman, M. R. Larreur, H. Manninen, J. B. Pitman and American Society of Mammologists and were approved by SIU A. M. Weber developed the original idea and structure of the manu- (IACUC 21-028 and 21-021), CPW (IACUC approval number 03- script. G. Bastille-Rousseau developed the original functions with 2020) and SUNY-ESF (IACUC approval number 121202). assistance from J. B. Pitman and S. A. Crews; E. B. Donovan, M . E. Egan and N. T. Gorman analysed each species. G. Bastille-Rousseau, O RCID M. Eichholz, S. Blake, E. Bergman and N. D. Rayl secured fund ing for G. Bastille-Rousseau $ https://orcid.org/ 0000-0001-6799-639X animal tracking. G. Bastille-Rousseau led the writing of the manu- M. E. Egan $ https:// orcid.org/ 0000-0002-9952-5602 script, with input from all co-authors. M. W. Eichholz $ https://orcid.org/ 0000-0003-4592-8809 A C KN O W LE DG EM ENTS REFERENCES We thank Michele Vasquez, Lane Jeakle, Sujay Singh, Blake Hoffman, Alavi, S. E., Vining, A. Q., Caillaued, D., Hirsch, B. T., Havmoller, R. W., Havmoller, L. W., Kays, R., & Crofoot, M. C. (2022). A quantitative Megan Diamond, Kameron Kulbertson and Joshua Sunberg for �356 I IliOIIM:/Mj,j,4i,j.1ffi@IHJ■ 141 BASTILLE-ROUSSEAU ET Al. framework for identifying patterns of route-use in anima l movement dat a. Frontiers in Ecology and Evolution, 9, 1- 16 https://doi.org/ 10.3389/ fevo.2021.743014 Bakner, N. W., Cohen, 8. S., Collier, 8. A., & Chamberlain, M. J. (2022). Recursive movements of eastern wild Turkey broods in the southeastern United States. Wildlife Society Bulletin, 46(2), e1274. ht tps:// doi.org/ 10.1002/ w sb.1274 Barker, K. J., M itchell, M . S., Proffitt, K. M ., & DeVoe, J. D. (2019). Land management alters traditional nutrit ional benefits of migrat ion for elk. The Journal of Wildlife Management, 83(1), 167- 174. https://doi. org/10.1002/ jwmg.21564 Bastille-Rousseau, G., Crews, S., Donovan, E. B., Egan, M ., Gorman, N. T., Pitman, J. B., Weber, A., Audia, E., Larreur, M., Manninen, H., Blake, S., Eichholz, M., Bergman, E., & Rayl, N. D. (2024). Data from: A multi-property assessment of intensity of use provides a funct ional understa nding of anima l movement. Zenodo Digital Repository https:/ /Zenodo.org/ doi/ 10.5281/ zenodo.10277762 Bastille-Rousseau, G., Douglas-Hamilt on, I., Blake, S., Northrup, J. M., & Wit temyer, G. (2018). Applying network theory to animal movements to identify properties of landscape space use. Ecological Applications, 28(3), 854- 864. https:// doi.o rg/10.1002/ eap.1697 Bastille-Rousseau, G., Fort in, D .. & Dussault, C. (2010). Inf erence from habitat-selectio n ana lysis depends on foraging strat egies. Journal of Animal Ecology, 79(6), 1157- 1163. https://d oi.org/ 10.1111/ j.13 652656.2010.01737.x Bastille-Rousseau, G., Fortin, D., Dussault, C., Courtois, R., & Ouellet, J. P. (2011). Foraging strategies by omn ivores: Are black bears act ively searchi ng for u ngulate neonates or are they simply opportunistic predators? Ecography, 34(4), 588- 596. https://doi.org/ 10. 1111/ j.1600-0587.2010.06517.x Bastille-Rousseau, G., Gibbs, J. P., Yackulic, C. 8., Frair, J. L., Cabrera, F., Rousseau, L. P., Wikelski, M., Kummeth, F., & Blake, S. (2017). Animal movement i n the absence of predation: Environmental drivers of movement strategies in a partial migration system. Oikos, 126(7), 1004- 1019. https:// doi.org/ 10.1111/ oik.03928 Bastille-Rousseau, G., & W ittemyer, G. (2021). Characterizing the landscape of movement to identify critical wildlif e habitat and corridors. Conservation Biology, 35(1), 346 - 359. https://doi.org/10. 1111/ cobi.13519 Benhamou, S., & Riotte-Lambert, L. (2012). Beyond the utilization distribution: Identifying home range areas t hat are intensively exploited or repeatedly visited. Ecological Modelling, 227, 112- 116. htt ps:// do i.org/10.1016/ j.ecolmodel.2011.12.015 Bista, D., Baxter, G. S., Hudson, N. J., Lama, S. T., Weerman, J., & Murray, P. J. (2022). Space use, i nteraction and recursion in a solitary specialized herbivore: A red panda case study. Endangered Species Research, 47, 131- 143. https://doi.org/ 10.3354/ esr01171 Blake, S., Yackulic, C. 8., Cabrera, F., Deem, S. L., Ellis-Sot o, D., Gibbs, J. P., Kummeth, F., Wikelski, M ., & Bastille- Rousseau, G. (2021). Movement ecology. In Galapagos Giant Tortoises (pp. 261- 279). Dougherty, E. R., Seidel, D. P., Carlson, C. J., Spiegel, 0., & Getz, W . M. (2018). Going t hrough the motions: Incorporating movement analyses into d isease research. Ecology Letters, 21(4), 588- 604. https:// doi.org/10.1111/ele.12917 Duda, R. 0 ., & Hart, P. E. (1973). Pattern classification and scene analysis (Vol. 3, pp. 731- 739). Wiley https://www.svms.org/ classificat ion/ DuHS95.pdf Fauchald, P., & Tveraa, T. (2003). Using first-passage time in the analysis of area-restricted search and habitat selection. Ecology, 84(2), 282- 288. htt ps://doi.org/ 10.1890/ 0012-9658(2003)084[0282: UFPTIT]2.0.C0;2 Fortin, D., Beyer, H. L., Boyce, M. S., Smith, D. W., Duchesne, T., & Mao, J. S. (2005). Wolves influence elk movements: Behavior shapes a trophic cascade in Yellowstone National Park. Ecology, 86(5), 13201330. https://doi.org/ 10.1890/ 04-0953 Freit as, C., Kovacs, K. M ., lms, R. A., & Lydersen, C. (2008). Predicting habitat use by ringed seals (Phoca hispida) in a warming Arctic. Ecological Modelling, 217(1- 2), 19- 32. https://doi.org/ 10.1016/ j. ecolmodel.2008.05.014 Freit as, C., Kovacs, K. M., Lydersen, C., & lms, R. A. (2008). A novel met hod for quantifying habitat selection and predicting habitat use. Journal of Applied Ecology, 45(4), 1213- 1220. https://doi.org/ 10.1111/j.1365 -2664.2008.0150 5.x Garson, G. I., & Moser, E. 8. (1995). Aggregation and the Pearson chisquare statistic for homogeneous proportions and distribut ions in ecology. Ecology, 76(7), 2258- 2269. https:// do i.o rg/ 10.2307/ 1941700 Gibbs, J. P., Hunter, E. A., Shoemaker, K. T., Tapia, W. H., & Cayot, L. J. (2014). Demographic outcomes and ecosystem implications of giant tortoise reintroduction to Espanola Island, Galapagos. PLoS One, 9(10), e110742. https://doi.org/ 10.1371/ journal.pone.0110742 Gibbs, J. P., Marquez, C., & Sterling, E. J. (2008). The role of endangered species reintroduction in ecosystem restoration: Tortoisecactus interactions on Espanola Island, Galapagos. Restoration Ecology, 16(1), 88- 93. https://doi.org/ 10.1111/ j.1526 -100X. 2007.0026 5.x Giuggio li, L., & Kenkre, V. M. (2014). Consequences of animal interactions on their dynamics: Emergence of home ranges and territoriality. Movement Ecology, 2, 1- 22. https:// doi.o rg/ 10.1186/ s4046 2-014-0020 -7 Giuggio li, L., Potts, J. R., & Harris, S. (2011). Anima l interactions and the emergence of territoriality. PLoS Computational Biology, 7(3), e1002008. https://doi.org/ 10.1371/ journal.pcbi.1002008 Gorman, N., Eichholz, M., Skinner, D., Schlichting, P., & Bastille-Rousseau, G. (2023). Using a novel f ramework of animal space-use behaviors revea ls a gradient of responses to human modification. https://doi. org/10.22541/ au.167542206.6591760 7/Vl Graf, P. M., Mayer, M., Zedrosser, A., Hacklander, K., & Rosell, F. (2016). Territory size and age explain movement patterns in the Eurasian beaver. Mammalian Biology, 81(6), 587- 594. https://do i.o rg/ 10. Academic Press. https://doi.org/10.10 16/ B978-0-12-817554-5. 00018-6 Boyce, M. S., Mao, J. S., Merrill, E. H., Fortin, D., Turner, M. G., Fryxell, J., & Turchin, P. (2003). Scale and heterogeneity in habitat selection by elk in Yellowstone National Park. Ecoscience, 10(4), 421-431. https://do i.org/10.1080/ 11956860.2003.11682790 Brads, C., Bildstein, K. L., & Mueller, T. (2018). Revisitation analysis uncovers spatio -tempora l patterns in animal movement data. Ecography, 41 (11), 1801- 1811. https:/ / doi.org/ 10.1111/ ecog.03618 Brook, R. K. (2010). Habitat selection by parturient elk (Cervus elaphus) i n agricu ltura l and forest ed landscapes. Canadian Journal of Zoology, 88(10), 968- 976. https://doi.org/ 10.1139/ Z10 -061 Crews, S. (2023). Strategy-specific differentiation in resource responses and drivers of spring migration phenology in Rocky Mountain Elk (M.S. Thesis). Southern Illinois University. 1016/ j.mambio.2016.07.046 Hewitt, D. G. (2011). Biology and management of white-tailed deer. CRC Press. Kays, R., Crofoot, M. C., Jetz, W., & Wikelski, M . (2015). Terrestrial animal tracki ng as an eye on life and planet. Science, 348(6240), aaa2478. https://doi .org/ 10.1126/ science.aaa2478 Kie, J. G., Matthiopoulos, J., Fieberg, J., Powell, R. A ., Cagnacci, F., Mitchell, M. S., Gaillard, J.M., & Moorcroft, P.R. (2010). The homerange concept: Are traditional estimators still relevant w ith modern telemetry technology? Philosophical Transactions of the Royal Society, B: Biological Sciences, 365(1550), 2221- 2231. ht t ps://doi. org/10.1098/ rstb.2010.0093 Lima, S. L., & Zollner, P.A. (1996). Towards a behaviora l ecology of ecological landscapes. Trends in Ecology & Evolution, 11(3), 131- 135. https://doi .o rg/ 10.1016/ 0169-5347(96)81094-9 �BASTILLE- ROUSSEAU er AL Mao, J., Groves, J., Wright, K., & Cacho, D. (2013). Avalanche Creek Elk Herd E-15 data analysis unit plan. ht t ps://cpw.state.co.us/ Docum ents/ Hunting/ BigGame/ DAU/ Elk/ E15_AvalancheCreek.pdf Mitchell, W. A., & Lima, S. L. (2002). Predator-prey shell games: Largescale movement and its implications for decision-making by prey. Oikos, 99(2), 249- 259. https://doi.org/10.1034/ j.1600-0706.2002. 990205.x National Oceanic and Atmospheric Admi nistration (NOAA). (2022). NOAA's climate prediction center. Cold & Warm Episodes by Season. https://o rigin.cpc. ncep.noaa.gov/ products/ ana lysis_monitoring/ ensostuff/ ONl_v5.php Powell, R. A., & Mitchell, M. S. (2012). What is a home range? Journal of Mammalogy, 93(4), 948- 958. https://doi.org/ 10.1644/11-MAMMS-177.1 Riotte-Lambert, l., Benhamou, S., & Chamaille-J ammes, S. (2013). Periodicity ana lysis of movement recursions. Journal of Theoretical Biology, 317, 238- 243. https://doi.org/10.1016/ j.jtbi.2012.10.026 Rosell, F., & Thomsen, L. R. (2006). Sexual di morphism in territorial scent marking by adult Eurasian beavers (Castor fiber). Journal of Chemical Ecology, 32, 1301- 1315. https://doi.org/ 10.1007/ s1088 6 -006-9087-y Schwarz, G. (1978). Estimating the dimension of a model. The Annals of Statistics, 6, 461- 464. http://www.jstor.org/ stable/ 2958889 Scrucca, l., Fop, M., Murphy, T. B., & Raftery, A. E. (2016). mclust 5: Clustering, classification and density estimation using Gaussian finite mixture models. The R Journal, 8(1), 289. Wang, C., & Fiedler, P. C. (2006). ENSO variability and the eastern tropical Pacific: A review. Progress in Oceanography, 69(2- 4), 239- 266. Webb, S. l., Dzialak, M. R., Harju, S. M., Hayden-Wing, L. D., & Winstead, J. B. (2011). Influence of land development on home range use dynamics of female elk. Wildlife Research, 38(2), 163- 167. https://doi. org/10.1071/ WR10101 Webster, S. C., Beasley, J.C., Hinton, J. W., & Chamberlain, M. J. (2022). Resident and transient coyotes exhibit differential patterns of movement behavior across heterogeneous landscapes in the southeastern United States. Ecology and Evolution, 12(3), e8725. https:// doi.org/ 10.1002/ ece3.8725 W ilmers, C. C., Nickel, B., Bryce, C. M., Smith, J. A., Wheat, R. E., & Yovovich, V. (2015). The golden age of bio-logging: How animalborne sensors are advanci ng the frontiers of ecology. Ecology, 96(7), 1741- 1753. https:// doi.org/10.1890/ 14-1401.1 W ittemyer, G., Northrup, J. M., & Bastille-Rousseau, G. (2019). Behavioural valuation of landscapes using movement data. Philosophical Transactions of the Royal Society B, 374(1781), 20180046. https://doi .org/10.1098/rstb.2018.0046 1ii,&!iiffidffii,i-ii#iii,iffi@H■ 11 1 different periods of time. UseScape is based on all data combined (2010- 2022, panel A) and two single years of data, 2014 (panel B) and 2015, a warmer than usual year (panel C). Below each panel is a matrix containing the UseScape outputs for each class. All reported metrics are at the cell level: NV is the number of visits, TD is the total durat ion in hours, MD is the mean duration in hours, CVD is the coefficient of variation of duration, M l is the mean interval between visits in days and CVI is the coefficient of variation of intervals between visits. Freq is the frequency (proportion) of each class as a decimal percentage of all UseScape cells in the output. Many of the lower-use class cells are not visited annually. One explanation for the variation in habitat use could be changes in prevailing climatic conditions. If local conditions were warmer overall in 2015 than in 2014, as suggested by data from the National Oceanic and Atmospheric Administration (NOAA, 2022), this could help explain the dramatic increase in movement in 2015 to seek food and shade. However, we have not directly linked the movements of tortoises on Espanola Island to variation in prevailing climatic conditions. Figure S2. UseScape output map for the resident coyote at a resolution of 18 m based on the graphical approach based on variation in residency time. The corresponding table contains the output data for each cluster. This resolution was too small for the map to be easily interpreted, and with only two clusters, the cluster characteristics did not reflect any other category found in the outputs of the main text. Figure S3. UseScape output map for the exploratory coyote at a resolution of 128 m based on the median step length. The corresponding table contains the output data for each cluster. This resolution was too small for the map to be easily interpreted, and the passthrough cluster was lost at this resolution. How to cite t his article: Bastille-Rousseau, G., Crews, S. A., Donovan, E. B., Egan, M. E., Gorman, N. T., Pitman, J. B., Weber, A. M., Audia, E. M., Larreur, M. R., Manninen, H., Blake, S., Eichholz, M. W., Bergman, E., & Rayl, N. D. (2024). A SUPPORTI N G IN FORM ATION Additional supporting information can be found online in the Supporting Information section at the end of this article. Figure Sl. UseScape output maps for a female Galapagos giant tortoise (Chelonoidis hoodensis) on Espanola Island across three 357 multi-property assessment of intensity of use provides a functional understanding of animal movement. Methods in Ecology and Evolution, 15, 345- 357. https://doi. org/ 10.1111/ 2041-210X.14274 �Copyright of Methods in Ecology & Evolution is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal Articles Description An account of the resource CPW peer-reviewed journal publications Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource A multi‐property assessment of intensity of use provides a functional understanding of animal movement Description An account of the resource <p><strong>Abstract</strong></p> <ol start="1" class=""> <li>The intensity of use of a location is one of the most studied properties of animal movement, yet movement analyses generally focus on the overall use of a location without much consideration of how patterns in intensity of use emerge. Extracting properties related to intensity of use, such as the number of visits, the average and variation in time spent and the average and variation in time between visits, could help provide a more mechanistic understanding of how animals use landscape. Combining and synthesizing these properties into a single spatial representation could inform the role that a location plays for an animal.</li> <li>We developed an R package named ‘UseScape’ that allows the extraction of these metrics and then clustered them using mixture modelling to create a spatial representation of the type of use an animal makes of the landscape. We illustrate applications of the approach using datasets of animal movement from four taxa and highlight species-specific and cross-species insights.</li> <li>Our framework highlights properties that functionally differ in how animals use them, contrasting, for example, heavily used locations that emerge because they are frequented for long durations, locations that are repeatedly and regularly visited for shorter durations of time or locations visited irregularly. We found that species generally had similar types of use, such as typical low, mid and high use, but there were also species-specific clusters that would have been ignored when only focusing on the overall intensity of use.</li> <li>Our multi-system comparison highlighted how the framework provided novel insights that would not have been directly obtainable by currently available approaches. By making the framework available as an R package, these analyses can be easily applicable to a myriad of systems where relocation data are available. Movement ecology as a field can strongly benefit from approaches that not just describe patterns in space use, but also highlight the behavioural mechanisms leading to these emerging patterns.</li> </ol> Bibliographic Citation A bibliographic reference for the resource. Recommended practice is to include sufficient bibliographic detail to identify the resource as unambiguously as possible. Bastille‐Rousseau, G., S. Crews, E. Donovan, M. Egan, N. Gorman, J. Pitman, A. Weber, E. Audia, M. Larreur, and H. Manninen. 2024. A multi‐property assessment of intensity of use provides a functional understanding of animal movement. Methods in Ecology and Evolution 15:345-357. Creator An entity primarily responsible for making the resource Bastille-Rousseau, G. Crews, S. A. Donovan, E. B. Egan, M. E. Bergman, Eric J. Rayl, Nathaniel D. Subject The topic of the resource Animal movement Clustering GPS telemetry Intensity of use Mixture model Extent The size or duration of the resource. 14 pages Rights Information about rights held in and over the resource <a href="http://rightsstatements.org/vocab/InC-NC/1.0/">IN COPYRIGHT - NON-COMMERCIAL USE PERMITTED</a> Type The nature or genre of the resource Text Format The file format, physical medium, or dimensions of the resource application/pdf Language A language of the resource English Is Part Of A related resource in which the described resource is physically or logically included. Methods in Ecology and Evolution Rights Holder A person or organization owning or managing rights over the resource. This is an open access ar ticle under the terms of the Creative Commons Attribution-Noncommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. Date Accepted Date of acceptance of the resource. Examples of resources to which a Date Accepted may be relevant are a thesis (accepted by a university department) or an article (accepted by a journal). 11/27/2023 Date Issued Date of formal issuance (e.g., publication) of the resource. 02/06/2024 Date Submitted Date of submission of the resource. Examples of resources to which a Date Submitted may be relevant are a thesis (submitted to a university department) or an article (submitted to a journal). 05/19/2023