Site‐specific differences in the spatial ecology of northern cottonmouths

We conducted a yearlong telemetric study on 44 northern cottonmouths (Agkistrodon piscivorus) and recorded their weekly spatial position and use at two different and unconnected wetlands: Cooper Wildlife Management Area (Cooper WMA) and Texas AM 95% kernel density estimate (KDE) X¯ = 54.69 ha ± 9.36; 100% minimum convex polygon (MCP) X¯ = 22.9 ha ± 4.19) were significantly larger than those at the smaller Commerce WA (n = 10; 95% KDE X¯ = 19.88 ha ± 4.71; 100% MCP X¯ = 8.98 ha ± 2.27). The home ranges of males (n = 26; 95% KDE X¯ = 41.3 ha ± 6.84; 100% MCP X¯ = 17.55 ha ± 3.00) were also significantly larger than those of females (n = 7; 95% KDE X¯ = 7.04 ha ± 1.81; 100% MCP X¯ = 3.08 ha ± 0.82) at both sites. We found no relationship between body size and home range size at either study site. Average weekly distance traveled by males (X¯ = 110.07 m ± 10.80) was significantly greater than that of females (X¯ = 45.04 m ± 4.34), and male movement rates were bimodal in distribution with peaks of movement in the spring and fall. These peaks in movement rates coincided with the spring and fall migrations for mating, thermoregulation, parturition, and access to hibernacula and food resources. We found home ranges within these two‐dimensional wetlands to be larger than those reported for linear or one‐dimensional riparian sites. We discuss how wetland size and the size‐dependent dispersion of potential resources within similar wetlands may influence movement patterns and home range sizes. Such information may serve to evaluate effective wetland size for the conservation and management of wetland species.


INTRODUCTION
At large scales, spatial ecology is important for understanding stability, diversification, distribution, sympatry, connectivity, and the overall formation of biological patterns and processes (Tilman and Kareiva 1997, Moilanen and Hanski 2001, Moilanen and Nieminen 2002. However, at small scales, a population's spatial use reveals home range sizes, resource utilization, site fidelity, and possible intersexual or ontogenetic differences (Burt 1943, Reinert 1984a, Fieberg and Kochanny 2005. The spatial ecology of semiaquatic wetland species is often confounded as their survival frequently depends on the proximity and use of adjacent cover types (Semlitsch andBodie 2003, Simon et al. 2009). Understanding how semiaquatic species use space within a wetland landscape is important for maintaining population viability and providing appropriate management and conservation strategies dealing with edge effects, buffer zones, habitat management applications (e.g., when to burn or flood), and connectivity to neighboring satellite populations of a larger metapopulation (Law and Dickman 1998, Moilanen and Hanski 1998, Hanski 2001, Semlitsch and Bodie 2003, Trenham and Shaffer 2005. This could be even more critical for dominant semiaquatic wetland species such as the northern cottonmouth (Agkistrodon piscivorus; henceforth cottonmouth). Cottonmouths are opportunistic omni-carnivores that are also preyed upon by many species (Gloyd andConant 1990, Vincent et al. 2004). Their simultaneous niche as a wide-ranging predator and prey suggests that they are an important trophic link within wetland communities (Ford 2002). Although common in some areas, cottonmouth abundance is likely declining due to habitat loss and anthropogenic persecution (Ford 2002).
The use of space is directly tied to the distribution of necessary resources such as food, water, cover (e.g., foraging, parturition sites, thermoregulation), and mates (Gehrt and Fritzell 1998, Mitchell and Powell 2004, Willems and Hill 2009, Schradin et al. 2010, Wasko and Sasa 2012. Ecological theory posits that smaller home ranges are indicative of resources being in close proximity to one another, as individuals do not travel farther than needed because of the increased predation risk associated with movement (Stephens and Krebs 1986, Bonnet et al. 1999, Yoder et al. 2004. It is also generally accepted that larger individuals will have larger home ranges than smaller individuals of the same species because larger individuals require more food, which usually requires more movement to obtain (Harestad and Bunnel 1979, Lindstedt et al. 1986, Wasko and Sasa 2012. Few empirical studies have examined the relationship between wetland size, body size, and home range size of semiaquatic herpetofauna. Essential resources within larger wetlands could be more widely distributed than resources within smaller wetlands. Therefore, wetland size could directly influence home range size. When examining such questions, it is important that study sites are similar in terms of locality, structure, hydrology, vegetative composition, and available resources. The spatial ecology of cottonmouths is poorly studied when compared to other North American pit vipers such as timber rattlesnakes (Crotalus horridus) and massasauga rattlesnakes (Sistrurus catenatus). To date, no study on cottonmouths has quantified year-round movements. Existing studies, although multiyear in length, have focused solely on the summer season. These studies report relatively small home range sizes compared to other pit vipers of similar size (Roth 2005a, b). These studies also document cottonmouth use of wooded lowlands near streams (Roth 2005a, b). Furthermore, all existing studies to date have been conducted in riparian-riverine sites, resulting in linear home ranges associated with the river or stream channel (Roth 2005b).
We empirically examined how wetland size (and thereby resource distribution) influences the spatial use of cottonmouths through a yearlong telemetric study at two different wetlands that differed in size. We hypothesized that (1) the nonlinearity of our study sites will produce larger home ranges of cottonmouths compared to those reported for linear riparian areas, (2) individuals with larger body sizes will have larger home range sizes because of food requirements and therefore increased movement for foraging, (3) greater resource dispersion on our larger wetland site will produce larger home ranges than those on our smaller wetland site, (4) cottonmouths will utilize areas away from water for hibernation and parturition as in Roth (2005b), and (5) males will have larger home range sizes than females as in Roth (2005b).

Study sites
Our study was conducted at two different constructed wetland complexes (Fig. 1) intermittent patches of grazed cattle pasture and prairie. Both sites are nearly identical in terms of age, vegetative composition ( Fig. 2), locality (14.18 km separation), structure, and hydrology. However, the total available area of wetland at each site differs considerable, with the Cooper WMA (200.76 ha) containing a much larger area of emergent herbaceous wetland (as defined in Homer et al. 2004; delineated using satellite imagery and ground truthing) than the Commerce WA (9.09 ha).

Surgery and telemetry
We captured cottonmouths through an ongoing community assessment project and opportunistic sightings at both study sites. Cottonmouths weighing more than 500 g were transported to the laboratory for surgical implantation of radio transmitters. Radio transmitters were never larger than 5% of the snake's body mass. We covered radio transmitters (LPI-2180-MVS; Wildlife Materials, Murphysboro, Illinois, USA) in a 1:1 mixture of paraffin and beeswax (see Lutterschmidt et al. 2012) and intraperitoneally implanted them following the procedures described in Reinert and Cundall (1982). Snakes were anesthetized using isoflurane and surgery commenced at the loss of the righting reflex. Surgical instruments were sanitized in an autoclave prior to surgery, and isopropyl was used to keep the surgical area and instruments sterile. We used a 4-0 synthetic absorbable suture (CP Medical, Norcross, Georgia, USA) subsequently combined with 3M Vetbond Tissue Adhesive (3M Animal Care Products, St. Paul, Minnesota, USA) to seal the incision site. Immediately following surgery, each individual was massed, sexed via cloacal probing, palpated to see whether gravid, and measured (Table 1). Snakes were given bedding, water ad libitum, monitored for 24 h in the laboratory, and then released at their site of capture. We relocated individuals at least once every week in the active season (March-November), and once a month in the inactive season, using an R1000 receiver (Communications Specialist, Orange, California, USA) and 3-pronged yagi antenna (Wildlife Materials, Murphysboro, Illinois, USA). Upon visual detection of a telemetered snake, we used a GPSMAP 64 global positioning system (Garmin, Olathe, Kansas, USA) to pinpoint each snake's location.

Movement, activity, and resource dispersion
We calculated the weekly distance traveled (m) using Geospatial Modelling Environment (GME; Beyer 2012) for each individual snake. We then calculated the mean weekly distance traveled for males and females throughout the entire active season (March-November). We tested for intersexual differences in mean weekly distance traveled using an independent two-sample paired t-test. We also recorded the snake's body position at each unique location as coiled, active, fossorial, or submerged in water.
In regard to our third hypothesis, we measured the distance from the wetland's edge to each snake's hibernaculum at both sites as a means to objectively quantify resource dispersion. We compared the distances from each site using an independent two-sample two-tailed Student's t-test.
Spatial details are only presented for individuals with over 20 unique locations; therefore, blank cells are only for individuals who did not have over 20 unique locations. Summer 95% MCPs are only presented for males with more than 10 consecutive summer locations. summertime spatial use to Roth's (2005b) results. We only present male-male comparisons with Roth (2005b) because of our female sample size for summer MCPs. Both 50% and 95% KDEs were generated using the least-squares crossvalidation bandwidth estimation (Seaman and Powell 1996) and quantified in ArcMap 10.2 (ESRI 2013). We examined intersexual home range differences for each method using an independent two-sample two-tailed Student's t-test. We tested for site effects (Cooper WMA vs. Commerce WA) on home range size using an independent two-sample two-tailed Student's t-test. Because of our site-specific female sample size discrepancies, we only included males in our site effect analysis. Individual snakes had to have a minimum of 20 locations for their home ranges to be calculated.
We performed a simple linear regression to evaluate the relationship between body size and home range size. We ran the regression between snout-vent length (SVL) and 95% KDEs. We performed this analysis separately for both study sites. All statistical analyses were performed in Program R (R Core Team 2017), means are presented AE 1 SE, and significance was set at an alpha level of 0.05.

RESULTS
We telemetered 44 cottonmouths (Commerce WA = 24, male = 13, female = 11; Cooper WMA = 20, male = 18, female = 2) from May 2017 to June 2018 for a total of 1246 unique locations (Table 1). Twenty-eight snakes survived the entire study: Four were depredated (two via known predators: coyote (Canis latrans; Delisle et al. 2018) and redtailed hawk (Buteo jamaicensis)), four lost their signals (two experienced malfunctioning radio transmitters), and six transmitters were found with no signs of depredation (n f = 4, n m = 2). Individuals had an average of 27.6 unique locations. All telemetered females were gravid upon capture, as every female we captured that was more than 500 g was gravid.

Movement, activity, and resource dispersion
Males' average weekly distance traveled ( X = 110.07 m AE 10.80) was significantly longer than those of females ( X = 45.04 m AE 4.34; P = <0.001; Table 1). Males' mean weekly distance traveled was bimodal, with peaks in the spring and fall (Fig. 3). Females' mean weekly distance traveled peaked in early spring, but overall was more constant than males throughout the year (Fig. 3).
We commonly found cottonmouths either coiled or active from May to September (Fig. 4). From November to February, we found cottonmouths mainly underground, but periodically, we found individuals above ground on warm days near their hibernacula (Fig. 4). In several instances, overwintering occurred communally, although solitary instances also occurred. Nearly all individuals used small mammal burrows as hibernacula, aside from one that took refuge for the winter under a pile of hay bales and another that took refuge under a refrigerator that was displaced during a flood. During March and April, we observed individuals submerged in water ❖ www.esajournals.org more often than in other times of the year (Fig. 4). The distances from wetland edge to hibernacula at the Cooper WMA ( X = 395.36 m AE 60.60) were significantly longer than those at the Commerce WA ( X = 187.18 m AE 47.98; t = À2.7114, df = 29, P = 0.01114).

Home range and body size
Male home ranges were significantly larger than female home ranges, and the home ranges from the Cooper WMA were significantly larger than those at the Commerce WA (Table 2). We found no relationship between SVL and 95% KDE for cottonmouths at the Cooper WMA (R 2 adj = <0.001, F = 0.999, P = 0.333) or the Commerce WA (R 2 adj = À0.067, F = 0.056, P = 0.816; Fig. 5). The average summer 95% MCP for male cottonmouths was 6.118 ha (Table 2).

DISCUSSION
The average male summertime 95% MCPs from both of our study sites were larger than those found in previously published studies, which supports our first hypothesis. Roth (2005b) reported much smaller male MCP home ranges during the summer ( X = 1.862 ha), as did Wharton (1969) and Tinkle (1959). However, Roth (2005b) did acknowledge that his small home ranges were potentially influenced by linear habitat structure associated with a stream riparian zone. Cottonmouths inhabiting riparian areas can simply move up or down the stream to find resources (e.g., food, irradiant exposure, cover). This causes a one-dimensional resource distribution which results in linear, smaller home range sizes. Our study sites were non-linear and offered a more two-dimensional distribution of resources. The small home ranges presented in Wharton (1969) and Tinkle (1959) could simply be a function of their non-telemetric mark-recapture methods, as Weatherhead and Hoysak (1989) found home range estimates derived from recapture data to significantly underestimate home range sizes.
We found cottonmouths to be less active during the summer than in spring and fall (Fig. 3). Several individuals remained within an area less than one hectare during the summer, even though their year-round home ranges were far larger. This summertime sedentary behavior was exhibited at both study sites. Additionally, we failed to find a relationship between body size and home range at both of our study sites, in contrast to our second hypothesis (Fig. 5). Both summertime sedentary behavior and the lack of correlation between body size and home range  † P value from independent two-sample two-tailed Student's t-test comparing male and female cottonmouths. ‡ P value from independent two-sample two-tailed Student's t-test comparing male home ranges from the Commerce WA and the Cooper WMA (only male home ranges because of female sample size discrepancy across study sites). size suggest that the distribution of wetland food resources was not the underlying factor for the larger home ranges at the Cooper WMA. Food resources within both study sites were likely equally abundant, and therefore summer spatial use was similar (Cooper WMA summer 95% MCP X = 5.150 ha, Commerce WA summer 95% MCP X = 7.472 ha; Student's t-test, t = 0.93, df = 10, P = 0.3758). Similarly, when Wasko and Sasa (2012) experimentally lowered food availability, they found a significant decrease in the activity of other pit viper species. In our study, cover type distribution (e.g., wetland and forest) and, more specifically, the resources that each cover type offers (e.g., parturition sites and hibernacula) were influences on the larger home range sizes at the Cooper WMA. Because of the Cooper WMA's larger size, individual snakes had to travel farther to find essential seasonal resources such as hibernacula, which supports our third hypothesis.
All of our telemetered cottonmouths (minus one individual) made migrations out of the wetland during the late summer/early fall, which supports our fourth hypothesis (Fig. 6). Possible reasons for making this fall migration include parturition, mating opportunities, and locating suitable hibernacula. Females usually made migrations to wetland-neighboring forests just before parturition in mid-August to mid-September. Males exhibited their highest peak in movement rate just after females made their parturient fall migrations (Fig. 3). We observed multiple courtship, copulatory instances, and male-male agonistic behaviors (i.e., combative dance) in the fall near parturition sites (~15 m; see Delisle et al. in press). Unlike other studies that observed bisexual pairing year-round and/or mating in the spring, we observed these behaviors only during  the fall (Wharton 1966, Schuett 1992, Siegel and Sever 2008. Therefore, mating seems to be responsible for this increase in male movement during the fall. After parturition and breeding, both males and females (minus one individual) hibernated in wetland-neighboring forest, usually in small mammal burrows. Following hibernation, males and females exhibited a peak in movement during late March and early April, coincident with return migrations back to the wetland (Fig. 3). We believe this return migration is made for the more available food and thermoregulatory resources found in the wetland. Several amphibian species (e.g., southern leopard frog, Lithobates sphenocephalus, and northern cricket frog, Acris crepitans) make spring migrations to water for breeding and oviposition (Sinsch 1990). We found cottonmouths to predominately utilize these spring breeding amphibians. One female cottonmouth (non-telemetered and likely parturient the previous fall) regurgitated 16 southern leopard frogs at the Commerce WA in April 2017 during a snake dietary study (unpublished data). Female movement rates were highest during the spring, suggesting that food was especially important for postpartum females after leaving hibernacula (Fig. 3). Gravid females of other viviparous pit vipers, and likely gravid female cottonmouths as well, fast late in their gestation and therefore are emaciated after parturition (Madsen andShine 1993, Graham et al. 2011). It is likely that postpartum females at our study sites do not eat until the following spring when they return to the wetland for amphibian prey. Wasko and Sasa (2012) also found other pit vipers to utilize the abundant food resources within wetlands in response to the lack of forestdwelling prey. Secondly, cottonmouths may also make this return migration for the direct sunlight exposure which was less available on the forest floors where they hibernated. Upon reaching the wetland in the spring, cottonmouths were more likely to be submerged in water than in other times of the year (Fig. 4). Cottonmouths could be using water to aid their thermoregulation, as the water temperature in the wetland was likely warmer than the air temperature during the cold nights from March to early April.
Males had significantly larger home ranges than females, which supports our fifth hypothesis. This finding is similar to previous studies on cottonmouths and other pit viper species (e.g., Roth 2005b, Wastell and Mackessy 2011, Shipley et al. 2013, Sutton et al. 2017. However, all of the females we telemetered were gravid. Gravid females in many snake species alter their spatial requirements and increase their thermoregulatory behavior to aid embryonic gestation (Reinert 1984b, Reinert and Zappalorti 1988, Tu and Hutchison 1994, Charland and Gregory 1995. To that end, Roth (2005b) found gravid female cottonmouths to have larger home range sizes than non-gravid females. More research is needed to determine the relationship between fecundity and cottonmouth spatial ecology within non-riparian areas.

ACKNOWLEDGMENTS
We thank Mike Crowell and Phillip Aaron for allowing us to track snakes on their property. We also thank Texas Parks and Wildlife for giving us access to the Cooper WMA. Our research was funded by the Texas A&M-Commerce 2017 FY Unit Strategic Initiative Funding. Research was conducted under the Texas A&M-Commerce Institute's Animal Care and Use protocol P17-036.