Self-made home: how and where does the anuran Rhinella dorbignyi build its retreat sites

. In this study, we observed that burrows of Rhinella dorbignyi (Duméril & Bibron, 1841) are distributed in a non-random manner in the habitat, suggesting a microhabitat selection for digging. This conclusion was based on a characterization of 36 burrows and surrounding micro-habitat. We established a 1 m x 1 m quadrat with the burrow in its central point (n=36) to measure the percentage (density) and the average heights of grasses, herbs, and shrubs. All measurements were repeated in two unused quadrats (without burrows) to evaluate the available microhabitat (n=72). The burrows are built in specific areas of the habitat with a higher percentage of grass, taller herbs, lower density of shrubs and low shaded sites than the founded at control sites. Based on three-dimensional models of the interior of the burrow (n=15), we observed that all of them were constructed with an elliptical opening that opens into a narrow channel perpendicular to the ground. Channels had a mean maximum diameter of 38 mm and a mean minimum diameter of 18 mm. The mean length of the burrows is 182 mm, and the mean volume is 95 mL.

Retreat sites must provide more stable temperature and humidity conditions than more exposed locations (Denton & Beebee, 1993) and increase survival in dry weather (Schwarzkopf & Alford 1996;Seebacher & Alford, 2002).The environment offers a series of available natural retreat sites, e.g., under the leaf litter, under rocks or in tree hollows (Gudynas & Gehrau, 1981;Peixoto, 1995;Kwet et al., 2010;Maneyro et al., 2017;Moser et al., 2017).On the other hand, many anurans "build" their retreat sites by burying themselves into the ground.Most burrowing anurans use the burrows temporarily, for example, during the hottest and driest periods of the day to decrease the rate of water evaporation through the skin, lose heat or rest outside the reach of predators (Wells, 2007).Some leptodactylids, for example, can also use burrows as reproductive sites (Prado et al., 2002).This behavior was observed in several species of Leptodactylus belonging to Leptodactylus fuscus group, as is the case of the whistling frog (Martins, 1988;Freitas et al., 2001).In this species, the burrow is used for egg laying and its construction is limited to reproductive behavior (Haddad & Prado, 2005).The South American Dorbigny's toad, Rhinella dorbignyi (Duméril & Bibron, 1841), uses its hind limbs to dig burrows (Langone, 1994;Wells, 2007;Maneyro et al., 2017) which, different from other species, it uses for several consecutive days, alternating its use with foraging periods outside (Sanchez & Busch, 2008;Maneyro et al., 2017, appendix, video A1).
However, details on the construction or selection of sites for use remain unknown (Achaval & Olmos, 1997).As digging is a time and energy-consuming behavior (Wells, 2007), it is plausible to consider that R. dorbignyi selects specific microhabitats to build them.In this study, we describe the shape of the burrows of R. dorbignyi and evaluate the microhabitat selection for their construction.
We studied the toads in a plain terrain of a coastal area of many small temporary ponds.In terms of soil classification, the substrate did not vary along the ESEC Taim (Cunha et al., 2018).According to the federal land survey service (available at https://dados.gov.br/dataset/cren_solos_5000/resource/4b1042a4-b88e-470d-8b2a-e716714fb3e5), all Taim's area is over a hydromorphic vertosol (Santos et al., 2006;EMBRAPA, 2013).Due to its high concentration of organic matter and silt, the Taim's soil is little permeable, favoring flooding and the formation of temporary ponds.Usually, the silt layer reaches up to 2 m in depth (Cunha et al., 2018).Because of flooding and drying events, the substrate is, in general, highly compacted (Cunha et al., 2018).
Four people searched for the toads' burrows during daytime via visual search.We only sampled burrows with indications that they were being used until recently by the toads (without signs of dryness or spiders inside).When a burrow was detected, its interior was inspected with the aid of a flashlight to check for the presence of the "resident" amphibian.Recently abandoned burrows that were not currently occupied by a toad were used to build threedimensional casts to characterize the internal design.To minimize the stress of animals, only burrows without toads were modeled.To create the models, we diluted Plaster of Paris in water until it acquired a pasty consistency.Afterward, the mixture was poured into the opening of the hole until filling it.After 20 minutes, the model was hard enough to be removed (Fig. 2).To describe the internal design of the burrows, we performed the following measurements from models: chamber length, straight from the top to the bottom of the cast; maximum channel diameter, and minimum channel diameter, measured in the burrow opening.The  linear measurements were made by an electronic caliper with a precision of 0.01 mm.To obtain the total volume of the model, we submerged it into a graduated container filled with water and measured the displacement of the liquid.
To characterize the micro-habitat where each in-use burrow was built, we established a 1 m x 1 m quadrat with the in-use burrow in its central point.Within this quadrat, the percentage (density) of each vegetation type and exposed soil was visually estimated.For each site, we estimated the percentage of the quadrat that would be shaded at midday.The height of grasses, herbs, shrubs was measured from the highest plant origin in soil to its highest branch.All described measurements were repeated in two additional quadrats, with no burrow, situated one quadrat in the north and the other in the south direction, with its nearest border 2 m from the burrow.We chose this distance because distances smaller than 2 m include very spatially dependent areas, having a similar vegetation cover.At the same time, greater distances include very different points in terms of water accumulation, prey availability, and other elements that we did not assess.This decision is somewhat arbitrary but, according to literature, the toads are not expected to move very much (Oliveira et al., 2017;Moser et al., 2019).Therefore, supported by previous data about this species, we believe that the chosen distance is compatible with the area that could be chosen by the toads at the time of site selection for the construction of the burrow.The measured characteristics of the used micro-habitat were compared to the micro-habitat available around each burrow.
To test whether the sites where in-use burrows were found were selected by anurans considering the measured habitat characteristics, we performed a Permutation Multivariate Analyses of Variance (PermANOVA) with Randomization Tests based on a matrix of Euclidian distances between sites described by the habitat variables, previously centered and normalized within variables (Pillar & Orlóci, 1996).The test statistic for main effects was the sum of squares between groups (Qb), while the pseudo-F ratio was the test statistic for the interaction term (Pillar, 2013).This numerical analysis was run in MULTIV version 3.55b (Pillar, 2006).We then performed a Principal Component Analysis (PCA) based on a correlation matrix between the microhabitat variables using PAST3 (Hammer et al., 2001), transforming percentage values of vegetation cover in its arcsin for normalization.We selected the ordination axes that explained 75% of the variance in the site's characteristics for interpretation.Using a standard linear (least-squares) regression, we tested the proportionality between the average length and maximum diameter of the burrow and the length of the burrow.For this analysis, we used PAST3 (Hammer et al., 2001).

RESULTS
We evaluated the location of 36 in-use burrows of Rhinella dorbignyi and built 15 models of recently abandoned burrows in plaster to describe their shape.All burrows had a single entrance (opening) that is slightly elliptical and connects to a channel.The channel is perpendicular to the ground and ends in a single blind bottom.The mean length of the burrows was 182 mm.The average maximum diameter of the channel was 38 mm, and the average minimum diameter was 18 mm.Both diameters (maximum and minimum) showed a small range of variation, the average total volume of burrows was 95 ml (Tab.I; appendix, table A1).
We observed a positive relationship between the maximum diameter and the length of the channels (linear regression; r = 0.590; t = 2.638; p = 0.021; appendix, fig.A1).The microhabitat where toads built their burrows was significantly different from the available microhabitat in terms of vegetation cover (Q = 0.21853; P = 0.024).The two first PCA axes explained 76% of the variance in the micro-habitat characteristics on each site.The first axis loadings (56% variance explained) were positively related to shrubs height, shrubs percentage and shaded area, and negatively related to the grass percentage and herbaceous height.The second axis (22% variance explained) loadings were positively related to the percentage of herbs and shaded areas and negatively related to the percentage of grass (Tab.II; Fig. 3).

DISCUSSION
We found a consistent pattern in the construction of all the observed burrows, as they all had a well-defined single entrance connected to a channel perpendicular to the surface.There was less variation in the height and width of the burrows' opening than the variation found in burrow length (Tab.I), which could suggest a tendency of individuals to build burrows adjusted to their size, which does not show much variation for adults in this species (Sanchez & Busch, 2008).Constructing a smaller burrow, adjusted to the animal's body size, could minimize the energetic cost of excavation and reduce the thermal exchange, with surface thermal stability (Schwarzkopf & Alford, 1996).Adjusted burrows may also be more efficient in preventing predators from entering since only animals with the size equal to or smaller than the resident of the burrow will be able to access the shelter.Unfortunately, we were unable to crossmatch measurement data from toads and their burrow since we sampled abandoned burrows exclusively.We assumed this weakness in our sampling (modelling only recently abandoned burrows) not to affect toad survival since our burrow modelling process implicated in the destruction of the burrows (see methods for more details).
The channel's length was the linear component with the greatest coefficient of variation (CV% = 46.48) in comparison with the maximum diameter (CV% = 31.48)and minimum diameter (CV% = 30.59).As the channel is arranged perpendicular to the ground, we presume that the longer the channel, the greater the thermal gradient between the opening (surface) and the bottom.By adjusting their position along the channel, the toads possibly could obtain the appropriate conditions of temperature and humidity, even when external conditions are harsh (Székely et al., 2018).In addition, considering the correlation between the maximum diameter and length of the burrows, it is possible that larger individuals, with greater energy capacity, can build deeper burrows.Studies in other locations with R. dorbignyi (Pereyra et al., 2021) found that the length of the burrows corresponds to about three times the size (SVL) of the inhabiting toad (Gallardo, 1957;1969).As amphibians can rehydrate by absorbing water from the soil (Booth, 2006), the level of soil moisture may influence the extent of the dug channels.Thus, it is plausible to think that, during hot sunny days, a long channel would offer a wide range of conditions, from hot/dry to cold/moist as the toads move from the entrance to the bottom of the burrows.
In addition to the design of the burrows, the choice of the location for their construction also proved to be an important factor.Burrow sites were not randomly distributed between the nearby available microhabitats.As showed in our results, the toads selected sites that were less shadowed than the average burrow surrounding microhabitats (see our PCA).They also selected sites with a higher percentage of herbaceous vegetation and a smaller shrub extent.It seems plausible to suppose that the configuration of the selected sites helps toads improve the control of body temperature.When selecting little shaded and more "open" microhabitats, the toads enable solar irradiance over the burrow opening.This would be an important strategy considering the harsh winter in the study area compared to the tropical pattern.
We emphasize that our study sheds light on the possibility of the burying behaviour contributing to active thermoregulation in toads.This increases the relevance of the data we present and is also important for observing behavioural aspects of ectotherms.

Fig. 1 .
Fig. 1.Representation of the vegetation cover from Taim Ecological Station (ESEC Taim), a federal protected area and Ramsar site in Brazil's extreme south.In the first plain, a grassy microhabitat (where we can find Rhinella dorbigbyi's burrows), and a shrub area in the back (Photo: A. M. Tozetti).

Fig. 2 .
Fig. 2. Characterization of the shape of the Rhinella dorbignyi (Duméril & Bibron, 1841) burrow opening, three-dimensional model in plaster of the burrow and the representation of a longitudinal view of the excavation that forms the burrow (Photo: L. K. Schuck).

Fig. 3 .
Fig. 3. Microhabitat ordination calculated using principal component analysis based on Euclidean distances.Colors define the use of the site to the construction of Rhinella dorbignyi's burrows.Used sites are in blue (n = 36) and available sites in orange (n = 72).Large circles indicate centroids for each group.The points represent each evaluated site, and the lines connect each sample to the group centroid.
Tab. I. Means and standard deviation for cast measurements of Rhinella dorbignyi toad's burrows.