Artificial shelters provide suitable thermal habitat for a cold-blooded animal

Human activities such as urbanization often has negative affects wildlife. However, urbanization can also be beneficial to some animals by providing suitable microhabitats. To test the impact of urbanization on cold-blooded animals, we first conducted a snake survey at a national nature reserve (Xianghai natural reserve) and an adjacent tourist bird park (Red-crowned Crane Park). We show high presence of Elaphe dione in the tourist park even with high human activities and predator population (the endangered, red-crowned crane, Grus japonensis). We then radio-tracked 20 individuals of E. dione, set seven camera traps, and recorded the temperature of the snakes and artificial structures in Crane Park to document their space use, activity, and thermal preference, respectively. Our results show E. dione preferred to use artificial facilities to shelter from their predators and for thermoregulation. The high number of rats from the camera traps indicate abundant prey items. Overall, E. dione appears to be adapted to modified habitats and may expand population size at the current study site.

Human disturbances such as urbanization and deforestation have accelerated over the last century 1 , affecting biodiversity at a global scale 2,3 . Human activities have shown significant impact on wildlife, such as influencing their behaviours 4 , feeding schedules 5,6 , habitat use 7 , and time spent on nursing young 8,9 . Although urbanization generally has a negative effect on wildlife [10][11][12][13] , it has also been shown to provide animals living space in city habitat 14 . For example, urbanization may provide advantages for invasive amphibian populations and bolster amphibian ability to establish and spread within novel landscapes 15 . Urbanization has also benefited some insectivore bats by increasing access to insect prey and foraging habitat 16 . Therefore, it is important to evaluate the impact of urbanization and human activities on urban wildlife to help with conservation and green urban planning.
Reptiles are generally sensitive to habitat change 17 because of their life history (low dispersal capacity and home range) and physiological dependency on environmental conditions such as temperature and water availability 18 . Therefore, reptiles are more prone to risks associated with human activities than other vertebrates [19][20][21][22][23] . Artificial constructions and infrastructure, a dominant landscape change in urban areas, can alter the surrounding microclimate [24][25][26] , which can directly affect the reptile's ability to physiologically and behaviorally regulate 18 . For example, human recreational activities can change both nesting and basking behaviour of map turtles (Graptemys flavimaculata) 27 . Reptiles are commonly killed on manmade roads during mating migrations and searching for oviposition sites 28,29 . However, studies have also shown artificial constructions in suburban areas have provided suitable shelters for urban adapted lizards such as the blue-tongued lizards Tiliqua scincoides and their commensal prey species 30 .
Snakes are an important part of the food web and are widely distributed in various habitats, including grasslands, wetlands, forests, agriculture fields, around the residential areas, scrublands, deserts and sea 31 . As habitat destruction, environmental pollution, introduction of invasive predators, and road kills can negatively impact snake populations and health 32,33 , it is important to understand the impact of human activity and habitat modification on the activity and behavior of snakes to better inform habitat management strategies. The Xianghai National Nature Reserve of Jilin Province in China provides an interesting case study for the effects of human activity on snake abundance and activity, because a preliminary survey showed more snakes (e.g. Elaphe dione) were present in the Red-crowned Crane Park than in the adjacent protected area. The Red-crowned Crane Park is heavily used by tourists for bird watching, and there is high abundance of predatory birds in the park. In this Body and environmental temperature. Temperature data recorded during our radio tracking period showed daily substrate temperature patterns followed a unimodal pattern, with temperature rising steeply to reach a peak of 35-40 °C between 07:00 and 13:00 h, and then falling steeply between 15:00 and 18:00 h and dropping to 20 °C at 18:00 (Fig. 3). Daily shelter inside variation in temperature was small, with temperature gently fluctuating between 21 °C and 27 °C. Substrate surface temperature was significantly higher than that inside of the shelter during midday where substrate temperature and shelter inside temperature were 33.5 ± 1.1 °C and 25.5 ± 0.6 °C (P < 0.001), respectively. Tracked snakes' body surface temperature presented two rising periods, 13:00 to 12:00 h and 14:00 to 16:00 h. None of snakes were sighted between 12:00 to 14:00 h and 17:00 to 18:00 h respectively during tracking, and thus no body surface temperature was recorded during these periods.   Fig. S2a,b,c). Rodents (38 events) were the most common animals captured by camera traps (Fig. S2d). All monitored sites had at least three images of snakes visit during the deployment period, with 97 visitation events from snakes recorded; 80 (82.5%) of these visitation events were made by E. dione, nine (9.3%) were made by O. spinalis, and eight (8.2%) were made by R. tigrinu. Elaphe dione, O. spinalis and R. tigrinus were active throughout the sampled time points between 4:00-21:00, 11:00-17:00 and 07:00-16:00. However, all three species showed a strong bi-modal pattern of activity, with a peak in activity in the morning between 9:00 and 11:00, and again in the afternoon between 14:00 and 16:00 (Fig. 4). The air temperature followed a unimodal pattern, with temperature rising to a peak of 45 °C from 07:00 to 12:00 h, and then falling steeply to 20 °C between 12:00 and 17:00 h (Fig. 4).

Discussion
Our results indicate that E. dione has a strong preference for the artificial area with a relatively small activity range indicating high site fidelity. During our tracking period, we observed E. dione preferred to bask around their shelters such as concrete crack and space between bricks (Fig. S1b). When E. dione noticed a predator approaching, they rapidly move in these hiding sites, which may explain why E. dione live in an area with high crane population. We expected snakes may behave differently in the reserve site because they may rely on rodent holes or tree cracks. Future studies are required to compare snakes' behaviours and site choices between natural and artificial conditions. Shelters provided protection against predators and buffer environmental conditions, which can be especially true for non-burrowing ectotherms, including most snake species, which rely on existing shelters 34,35 . The habitat at the Xianghai National Nature Reserve is mainly grassland with limit shelters for snakes relative to the Red-Crowned Crane Park. This may be the reason why the snake abundance is high at the high human active park in this study. However, due to the short survey efforts (5 days), our results may not accurately capture snake abundance in the Xianghai National Nature Reserve given the large reserve size. Nevertheless, there was high abundance observed in Red-Crowned Crane Park within a given area of survey.
Our radio tracking showed that both male-male and male-female E. dione home ranges overlap to a large degree, indicating individuals of this species do not exclude conspecifics from their home ranges. In addition, the three snake species (Fig. S2) observed shared a same shelter by our camera traps, implying tolerance to intraspecifics. Detailed examination of date and time records of GPS locations of individuals showed that vast individuals were found at same hiding area (98 encounters in 75 days of tracking). Such site-fidelic behaviour can be found in garter snakes Thamnophis sirtalis sirtalis 36 , tiger snakes Notechis scutatus 37 and Boa constrictor 38 . Mutual attraction and aggregation may serve multiple functions such as mating, food finding, and group vigilance against predators 39,40 . However, more field observations are required to test these potential aggregation functions in E. dione. It is possible that E. dione use their excellent sense of smell to detect and remember where the aggregation area is 41 . Given the exposed nature of the open artificial habitat, which results in direct exposure solar radiation, we expected these snakes would use their artificial shelter (e.g. concrete crack) for thermoregulation from the extreme heat during the middle of the day as reported in other reptiles that live in hot habitats [42][43][44][45] . Similar to these expectations, we did not observe that E. dione outside of their shelters during the hottest part of the day (12:00 to 14:00 h) from our radio tracking and camera traps. Hiding inside of the concrete or brick shelters resulted in shelter temperatures between 24-27 °C during mid-day, which were 9-18 °C lower than the surrounding substrate surface temperature (33-45 °C). The hottest part of the day often exceeds the physiological thermal limits of most terrestrial ectotherms 42,46 , and many ectotherms, including snakes, behaviorally thermoregulate by seeking cooler parts of their microhabitats to maintain optimal body temperatures or avoid reaching their upper thermal tolerance limit 47 . We therefore interpret E. dione sheltering in cooler environments during the hottest period of a day. In the context of climate warming, such artificial structures may play an important in behaviorally adapting to climate extremes such as heat waves. Based on the current study, E. dione always retreated to their shelter in the presence of cranes, and thus cranes may not easily capture them even though their activity area overlapped. Snakes often use other animal burrows to shelter from predators 48 . Thus, conservation strategies are often designed by providing artificial refugia. Zappalorti and Reinert 49 constructed subterranean refugia and successful attracted corn snakes Elaphe guttata and pine snake Pituophis melanpleucus for shelter, ovipositing, courtship, shedding and basking. For the threatened lizard, Timon lepidus, the artificial refuges were rapidly occupied by lizards, suggesting that this technique was successful to improve lizard habitat 30 . Therefore, further conservation strategies for endangered snake species are encouraged to provide artificial shelters to against predation and microhabitat loss.
In addition to predator avoidance and thermoregulation, our data from camera traps also show these manmade constructions supported a large number of rodents residing in the park which may provide additional prey items for E. dione. The presence of human waste and crane food leftovers may attract rodents or small sized bird species foraging in this area, result in increasing the prey abundance for E. dione. Traditional snake conservation strategies include translocations, reintroductions, habitat manipulations and the creation of artificial hibernation sites 50,51 , whereas, studies testing the effectiveness of these management strategies are generally limited. Our study highlights the inclusion of predator-prey information and thermal data for conservation strategies 52 . Future studies are required to examine appropriate artificial shelter could promote endangered snake species' survival.
Our study provided important insights into the spatial use and behavior in predator avoidance and thermoregulation of E. dione. First, E. dione preferred to use artificial habitat as shelter in the presence of high predator (red-crowned crane) population. Second, E. dione used their shelter to avoid high ambient temperatures during the middle of the day and were not active during this period. The manmade constructions experienced relatively constant internal temperatures during the middle of the day, and the lack of predation suggests that E. dione are able to persist in urbanized areas with high human activity and predators. Study area. This study was conducted at the Xianghai National Nature Reserve of Jilin Province (Fig. 1a), which is located on the Northeast of mainland China (45.0158-45.1390 N, 122.2568-122.3900 W; GPS datum ¼ WGS84; Fig. 1a) which covers 20,218.8 ha and the adjacent Red-crowned Crane Park (Fig. 1b)  Red-Crowned Crane Park were captured by hand. All snakes captured did not have obvious food items in their stomach. Once caught, the snout-to-vent length (SVL) and total length were measured with a flexible fiberglass measuring tape (± 0.1 cm), mass was recorded by a digital scale (± 0.1 g), and the sex determined by squeezing the tail base to evert the hemipenis of males or examining the thickness and scales at the tail base 56 . If hemipenes could not be everted, we inserted a blunt probe into the cloaca and probed in a posterior direction towards the tip of the tail to confirm sex. Probe depth of males is generally longer than that of females in reptiles 57 . A radio transmitter (Model BD-2, HOLOHIL Canada, 1.6 g, which was, 0.9% of average E. dione body mass; average mass of E. dione with transmitter 180.0 ± 13.6 g) was then attached to the dorsal surface of the base of the tail using subdermal stitch (Fig. S1a)  Temperature measuring. Due body surface/scale temperature of a snake is strongly correlated with its body temperature 60 , each time the tracked snake was sighted, the body surface temperature of snakes and the substrate surface temperature immediately adjacent to the snake (within a radius of 3-5 cm) were measured with infrared thermometers (BD380, Brady, China accurate to ± 1.0 °C). If located snake was hiding under a shelter, both the outside temperature was measured by infrared thermometers (BD380, Brady, China; accurate to ± 1.0 °C), and the temperature 10-15 cm inside of shelter was measured with a thermocouple thermometer (Omega 871A type K; accurate to ± 0.5 °C).

Methods
Camera traps. Camera traps were used to monitor activity levels of snakes and their prey around known aggregation sites to test for activity correlations with the ambient air temperature and predator avoidance if a predator appears. Seven camera traps (LTL-6210 Plus series, LTL ACORN, UK) with inbuilt temperature sensors were set up to capture images of snakes adjacent to a sample of seven snake aggregation sites (e.g. concrete crack, rat holes and building roof) at crane park between 21 May 2021 and 24 July 2021. Camera traps were set at each location for 30 days. All camera traps were triggered by motion sensors and could be triggered 24 h per day. The triggers were set as sensitive to be able to detect slow moving snakes. Camera traps were positioned 30 cm face the selected location. Each camera trap had a 0.5 m 2 field of view over the nest ensuring that any snake activity was recorded. This enabled information on the frequency, time of day, ambient temperature and species to be collected.

Data analysis.
Space use estimation. Space use was defined as an area used by an animal throughout the study duration 61 . We calculated variograms, fit movement models, and estimated spatial probability density functions via the ctmm package 62 for each individual via an area-corrected, autocorrelated kernel density estima- www.nature.com/scientificreports/ tion (AKDE C ) that was optimally weighted to determine the distribution of occupancy using the continuoustime movement model. The AKDE C fits the telemetry data a continuous time stochastic process movement model to fit an appropriate kernel bandwidth to account for spatiotemporal autocorrelation, particularly at small sample sizes 63 . Space use and movement parameters for each individual was estimated visually via variograms 64 and an automated model selection via the perturbative hybrid residual maximal likelihood (pHREML 65 ) which considers a suite of potential movement models with either an anisotropic (asymmetrical diffusion) and isotropic (symmetrical diffusion) version of the Ornstein-Uhlenbeck motion model (UO 66 ), the Ornstein-Uhlenbeck motion model with foraging-process (OUF; Gaussian estimates of the root mean speed 64 ), a special case of the OUF model (OUf) where the velocity autocorrelation timescale and the position autocorrelation timescale are equal, and the independent and identically distributed model (IID). Model selection was based on Akaike's Information Criterion adjusted for small sample sizes (AIC C ; Table S2). For individuals with high pHREML estimate bias (1 /N 2 > 2%), where N is the effective sample size from the initial pHREML estimate (Table S2), we used parametric bootstrapping pHREML (bpHREML) to reduce effective sample size bias in the model estimates 65 .
We estimated the 95% space use and the area encompassed by the 50% AKDE C isopleth, the proportion of the total (95%) space use area contained by the 50% core activity (Fig. S4).
Biometric correlates of space use. We tested if body size (SVL and mass) or sex predicted the variation in space use by constructing Bayesian regression models via the 'brms' function from the brms package 67 . We constructed four chains of 5,000 iterations per chain, including 2,000-step warm-up periods so a total of 10,000 steps were retained to estimate posterior distributions (i.e. (5,000 -2,500) × 4 = 10,000). Adapt delta was set at 0.99 to decrease the number of divergent transitions, and the max tree depth was set to 15 when the depth of tree evaluated in each iteration was exceeded. We conducted model fitting with either the natural logarithm-transformed SVL or body mass with sex as a covariate. All models were assigned default priors, and the degree of convergence of the model was deemed as achieved when the Gelman-Rubin statistics, R (Gelman & Rubin, 1992) was 1.
Results were presented as mean Bayesian estimates ± 95% credible intervals.
Individual space use overlap. Range overlap was quantified using the Bhattacharyya coefficient 68,69 on the 95% AKDE C estimates via the 'overlap' function in the ctmm package. The Bhattacharyya coefficient calculates the relative overlap between two probability distributions, and therefore allows calculation of uncertainty in the AKDE C estimates with a range between 0 (no overlap) and 1 (complete overlap). Range overlap was presented as percentage overlap ± 95% confidence interval.
Temperature profile. Differences between the temperature profile of the substrate surface, shelter outside and inside temperature across the day were compared with an analysis of variance (ANOVA) in R (R × 64 3.1.3) 70 .
Comparisons were reported as mean ± S.E. and P < 0.05 was considered to be significant.
Habitat preferences. To determine if E. dione had a preference for a particular habitat type, fixes were first plotted on landscape maps by combined visual mapping of habitat type with data imported from Google maps using the adehabitatHR package 71 in R, so that each fix could be assigned to a habitat type. A habitat selection index (Ŵ) was calculated for each of the habitat types (sandy plain, bushland, rainforest, mangrove and walkway) using the Eq. 72 : where O i = the proportion of the population sampled in each habitat type, and π i = the proportion of total study area each habitat type covers. Then the standardized habitat selection index (B i ) for each habitat type was calculated using the Eq. 72 :