Microhabitat use, daily activity pattern, and diet of Liolaemus etheridgei Laurent, 1998 (Reptilia: Liolaemidae) in the Andean Polylepis forests of Arequipa, Peru

Abstract This study describes the microhabitat use, daily activity pattern, and diet of Liolaemus etheridgei Laurent, 1998 in the El Simbral and Tuctumpaya Polylepis forests in Arequipa, Peru. El Simbral is a fragmented forest, whereas Tuctumpaya is unfragmented. Our results reveal that L. etheridgei shows no positive selection for any of the microhabitats we identified in Polylepis forests; on the contrary, it selects negatively against Polylepis trees and nonthorny bushes. The daily activity patterns indicate a bimodal pattern with peaks at 9:00–10:59 and 13:00–13:59 h. The diet of L. etheridgei consists mainly of plant material, and the most important animal prey category is Lygaeidae: Hemiptera, which is selected for positively. In particular, microhabitat selection varied for nonthorny bushes, which were selected negatively in the Tuctumpaya population but neither positively nor negatively in the El Simbral population. According to the proportions of plant material found, the L. etheridgei from El Simbral were found to be omnivorous, whereas the Tuctumpaya population was herbivorous. However, the percentage of plant material consumed in the El Simbral population was close to the critical value for herbivory–omnivory. We conclude that the three ecological aspects of L. etheridgei studied here are virtually identical in El Simbral and Tuctumpaya; therefore, this species is not affected significantly by the current fragmentation of forest.


Liolaemus etheridgei is distributed in the Arequipa and Moquegua
Regions in southern Peru (Laurent, 1998;Llanqui, 2020), occupying the scrubland, shrubland, and Polylepis woodland habitats found there. The Polylepis woodlands are present throughout the distribution of L. etheridgei, comprising small areas of relictual woodland.
However, the role of these woodlands in sustaining L. etheridgei populations is unclear (see Lloyd & Marsden, 2008). Of the 19 Polylepis species recognized in Peru (Mendoza & Cano, 2011), 13 (68%) have been included in a threatened category nationally taller trees. In this study, we aim to describe three aspects of the ecology of L. etheridgei in Polylepis forests: microhabitat use, daily activity pattern, and diet. We will also compare the differences in these aspects between the populations inhabiting fragmented and unfragmented forests. Differences in these aspects could indicate changes due to habitat deterioration. We also focus on these aspects as their divergence can lead to the ecological diversification of Liolaemus species (Jaksić et al., 1980). For microhabitat use, we expect L. etheridgei populations to select Polylepis trees positively as they have commonly been thought to be an essential resource for these lizards (Gutiérrez et al., 2010). For the differences in daily activity patterns, we expect L. etheridgei to be active for longer in El Simbral due to a lower tree cover enabling higher insolation, thus maintaining adequate temperatures in the substrate for lizards to be active for longer. Finally, we expect the diet to be more diverse in the Tuctumpaya population due to a diverse community of prey species in the unfragmented forest.  (Mendoza et al., 2010). We studied the lizard populations of two distinct areas, locally called "El Simbral" and "Tuctumpaya." Both of the areas are located within the Pichu Pichu volcano's lowland area, which has been subject to various anthropogenic pressures. El Simbral is located in the Chiguata District in Arequipa. It is comprised of Polylepis forest fragmented by a road passing through the Moquegua and Puno Regions of Southern Peru ( Figure 2). Tuctumpaya is located in the Pocsi District of Arequipa. In contrast, the Polylepis forest here is continuous, with a higher average tree height compared to El Simbral ( Figure 2). The two areas provide an ideal comparison for studying the effects of fragmentation on the populations inhabiting them.

| Methods
All data were collected between July and November 2013, within the dry season in the SABNR. We followed the Type I design for resource selection studies to collect data on resource use and resource availability at the population level (Manly et al., 2002). For both El Simbral and Tuctumpaya, we surveyed 30 quadrats of 20 × 20 m, within a total area of ~475 Ha and between an elevation range of 3600 and 4200 m. These quadrats were randomly selected using QGIS version 1.8.0. In each quadrat, we determined the abundance of nine microhabitats: (1) Polylepis trees, (2) scrubland, (3) thorny bushes, (4) nonthorny bushes, (5) rocks (major diameter > 1 m), (6) small rocks (major diameter ≤ 1 m), (7) uncovered land, (8) dry organic matter, and (9) Nordenstamia longistyla. Nordenstamia longistyla is an Asteraceae shrub which was abundant in our study area. We estimated the coverage of each microhabitat in the quadrats using the Crown-Diameter method (Mueller-Dombois & Ellenberg, 1974) and calculated areas using the ellipse equation (see García-De la Peña et al., 2012). Uncovered land coverage was then calculated by subtracting the total area of the quadrat from the combined area of the microhabitats. In each area, the microhabitat coverages were averaged. The final means were considered to be estimates of the | 3 of 13 LLANQUI et al.
relative abundances of each resource within the forests and, therefore, a measure of their availability (Manly et al., 2002).
Liolaemus etheridgei individuals were searched for using the visual encounter survey (McDiarmid et al., 2012). Each quadrat was surveyed once, at a randomly assigned time between 8:00 and 17:00 h. In all quadrats, surveys were carried out by two observers who walked, one behind the other, in a zigzag pattern across the quadrat for 30 ± 5 min. We recorded the microhabitat of where each individual was encountered. We also recorded the time of encounter and the activity of the individual. In order to assess diet, several individuals were collected by hand, euthanized with Halatal, fixed in 10% formalin, and preserved in 70% ethanol. F I G U R E 1 Location of the Polylepis forests of "El Simbral" and "Tuctumpaya" in the buffer zone of the Salinas y Aguada Blanca National Reserve, Peru.
In addition, to study prey availability, we randomly installed six lines of pitfall traps (20 traps separated 10 m apart) in El Simbral and Tuctumpaya. The pitfall traps consisted of 1 L plastic containers filled with 0.5 L of water plus washing up liquid. Pitfall trap lines were left for 48 h; then, the specimens were collected and preserved in 70% ethanol. Specimens collected were deposited in the scientific collection of the Herpetology and Entomology Departments of the Museo de Historia Natural de la Universidad San Agustín de Arequipa (MUSA) in Peru.
Most of the stomach items were identified at family level using specialized bibliography (Borror et al., 1981) and by comparison to reference material collected in the pitfall traps. However, some were identified until a higher taxonomical level (see Table 2). Amongst the contents, plant material was also present. Thus, all taxonomical items plus the plant material were considered prey categories like in similar lizard diet studies (e.g. Semhan et al., 2013). Dry weight measurements were taken for each prey category using a digital scale (Anyload ES-203HA precision 200 g × 0.001 g).

| Data analyses
We assessed the microhabitat use with selection ratios: where u i and a i are the proportions of use and availability of resource i , respectively. We also determined the 95% confidence intervals for each W i value using a conservative Bonferroni adjustment (Manly et al., 2002). If the confidence intervals of W i were >1, the resource use was greater than its availability, so it was interpreted as a positive selection. If they were <1, there was a negative selection for the respective microhabitat, so the resource was avoided. When the confidence intervals of W i included 1, there was neither a positive nor negative selection; thus, the individual was indifferent to the microhabitat. In addition, we used the log-likelihood of Pearson's statistic X 2 L to statistically test for the existence of a selective process in El Simbral and Tucutmpaya as well as for each particular microhabitat; the latter included a Bonferroni adjustment (Manly et al., 2002). Lastly, to test the selection process for microhabitats, we also estimated Manly's standardized selection index (B i ), which gives the probability that a resource is used when others are equally available (Manly et al., 2002). The above analyses were done using the adehabitatHS R package version 0.3.14 (Calenge, 2006).
Daily activity pattern was represented with frequency bar charts and kernel density charts considering two contrasting behaviors: active and inactive. We also tested for differences in the abundances of individuals from El Simbral and Tuctumpaya using the Wilcoxon signed-ranks test (Aho, 2014). This test was also used to compare the abundance of active and inactive individuals in El Simbral and Tuctumpaya throughout the day.
For each category of prey, we obtained the number of individuals (N), frequency of stomachs containing the prey (F), and weight (W).
However, for plant material, we only estimated the F and W; furthermore, this item was not included in any further analysis. We determined the percentage of each measurement (%N, %F, %W). With these measurements, we obtained the Relative Importance Index (IRI; Pinkas, 1971) but used the weight instead of the prey volume (Martin et al., 1996). Thus, the index is IRI = (%N + %W)*%F. The IRI is used mainly in fishery studies but is also recommended in other taxa, such as reptiles (Hart et al., 2002). We used the hierarchized %IRI values according to the categories proposed by Montori (1991), which have been used in diet studies in other Liolaemus species (see Semhan & Halloy, 2016). Thus, prey categories were ordered under the following dietary hierarchy: fundamental (F: %IRI > 75%), secondary (S: 75% > %IRI > 50%), accessory (A: 50% > %IRI > 25%), and accidental (a: 25% > %IRI). This hierarchy was originally proposed for values of the "Lambda segunda" dominance index (Montori, 1991); however, it has regularly been used with other indexes in dietary studies of Liolaemus species (see Cabrera & Scrocchi Manfrini, 2020; Olivera Jara & Aguilar, 2020; Semhan & Halloy, 2016). Trophic niche amplitude was estimated using the standardized Levin's index: where p i is the proportion of prey category i, and n is the total number of prey categories (Krebs, 1999).
We obtained the proportions under the "averaged" method as it reduced the bias of individuals that eat a large number of items (Zaccarelli et al., 2013). We excluded plant material in all the above analyses as it was not convenient to estimate N and %N. However, we re-estimated the %W to consider the plant content and then determined the type of diet according to the %W of plant material using the scale of Espinoza et al. (2004): insectivorous (0-10%), omnivorous (11-50%), and herbivorous (51-100%). Using data obtained from the pitfall traps, we determined the effective number of species (Jost, 2006) for both areas and the relative abundance of each prey category; the latter was considered a measure of prey availability in the forests. Then, we estimated the selection ratios considering each prey category as a food resource and the %N as a measure of use. In addition, we determined the standardized Manly's selection ratio with a type I design (Manly et al., 2002). To test for statistical differences in the diet of the two populations, we applied a permutational multivariate analysis of variance (Permanova) using a Bray-Curtis dissimilarity (Anderson, 2001) and a square root data transformation. As a Permanova assumes similar multivariate distribution (Anderson, 2001), we tested the multivariate homogeneity of group dispersions using the "betadisper" function from the "vegan" R package version 2.5-5 (Oksanen et al., 2013). Finally, we used nonmetric multidimensional scaling (Zuur et al., 2007) to order and represent the prey categories consumed by L. etheridgei in each Polylepis forest. We complemented the above analysis with a Pianka overlap where p i,j is the proportion of prey category i of the total of preys eaten by species j, p i,k is the proportion of prey category i of the total of prey eaten by species k, and n is the total number of prey categories (Krebs, 1999). All data analyses were done in R version 3.5.2 (R Core Team, 2019).
The microhabitats Nordenstamia longistyla and dry organic matter were not used by L. etheridgei, so they were excluded from further analysis. We found that L. etheridgei exhibited selection for microhabitats in general (X 2 L = 29.466, p < .01). Positive selection was not found for any microhabitats, whereas negative selection (i.e., avoidance) was found for Polylepis trees and nonthorny bushes. We did not find evidence of selection for uncovered land, small rocks, rocks, thorny bushes, or scrubland ( Table 1) In El Simbral, there was no positive selection (W i ± CI 95% > 1) for any of the microhabitats; however, there was negative selection for Polylepis trees (W i ± CI 95% < 1). We did not detect any selection processes for nonthorny bushes, rocks, scrubland, small rocks, thorny bushes, or uncovered land (Table 1). In Tuctumpaya, we did not detect a positive selection, yet there was negative selection for Polylepis trees and nonthorny bushes (W i ± CI 95% < 1). There was no selection for rocks, scrubland, small rocks, thorny bushes, or uncovered land (Table 1). Manly's standardized selection indices for small rocks, thorny bushes, Polylepis trees, and nonthorny bushes were higher in El Simbral, whereas the Tuctumpaya population showed a higher selection probability for uncovered land and rocks ( Figure 3).

| Daily activity pattern
We found that L. etheridgei had a bimodal daily activity pattern, with one activity spike between 9:00 and 10:59 h and another between

| Microhabitat use
Our results show that L. etheridgei does not select positively for any microhabitats found in the Polylepis forests; thus, this lizard can be considered a generalist in its habitat use (Vivas et al., 2019). The rocks, small rocks, and uncovered land have been associated with the thermoregulatory behavior of L. quilmes, L. ramirezae (Robles & Halloy, 2008), and L. tenuis (Victoriano et al., 2008). However, the absence of selective use for such resources may indicate that L. etheridgei does not use them actively for thermoregulation, which could support the idea that this lizard is a thermal generalist (Llanqui, 2020).
The microhabitats' thorny bushes and nonthorny bushes could serve as refuges, shelters, or foraging sites. For instance, L. multimaculatus prefers tussocks of grass for shelters (Kacoliris et al., 2010), L. quilmes has been found in rocks close to shrubs (Robles & Halloy, 2008), and L.
lemniscatus has been found in herbaceous patches (Jaksić et al., 1980). However, L. etheridgei was indifferent to scrubland, thorny bushes, and nonthorny bushes so that they could be used indistinctly as refuges, shelters, or foraging sites. This could be supported by our field observations of cavities used by L. etheridgei, which were encountered under different substrates such as rocks, thorny bushes, nonthorny bushes, and scrubland. Contrary to our expectations, Polylepis trees were avoided by L. etheridgei (negative selection). As there is no microhabitat positively selected by L. etheridgei, and this species may even avoid Polylepis trees, it is questionable as to why it inhabits these forests. Considering that L. etheridgei also inhabits sites where Polylepis trees are absent (Laurent, 1998), we suggest that L.
etheridgei is inhabiting Polylepis forests perhaps simply because they occur within its elevational range. It could also imply that L. etheridgei is able to tolerate variation in habitat structure, which is consistent with our conclusion that species is a generalist. Moreover, this could explain there being almost no differences in microhabitat selection in El Simbral and Tuctumpaya as well as the apparent tolerance to fragmentation of the Polylepis forests.

| Daily activity pattern
The activity peak of L etheridgei: from 9:00 to 10:59 h, is identical to that recorded for L. qalaywa (Chaparro et al., 2020), a species found near the distribution of L. etheridgei. Therefore, it is likely that both species would have a similar degree of solar radiation exposure and, accordingly, show a similar behavior pattern. The second activity peak occurs after midday: 13:00-13:59 h and is similar to that of L. evaristoi, a sister species of L. etheridgei (Gutiérrez et al., 2018). The drop in activity between 11:00 and 11:59 h supports a bimodal pattern for L. etheridgei. Bimodal activity patterns have also been recorded in L. darwinii (Videla & Puig, 1994), L. koslowski (Belver et al., 2010), L. multimaculatus (Vega et al., 2000), and L. occipitalis (Bujes & Verrastro, 2006.
In particular, a bimodal activity with a drop around midday, like in L.
etheridgei, has been recorded for L. arambaresis (Liz et al., 2019), L. constanzae, and L. nigroroseus (Labra et al., 2001), and this is thought to be a strategy in avoiding overheating. Throughout the day, there was no significant difference between the number of active and inactive individuals. Thus, there were no particular hours where active individuals were more abundant than inactive or vice versa, which is probably not valid in the late afternoon when the number of inactive lizards is logically greater than active ones. However, the actual number of inactive individuals could have been underestimated by the fact that the lizards would tend to seek shelter during periods of inactivity and thus would be harder to detect. This would likely have been near-surface shelter between periods of high activity; however, deeper refuges may have been sought later in the day, further reducing detectability. The lack of differences between active and inactive individuals would suggest that the variation between the Polylepis forests, particularly the abundance cover, appears not to affect the activity patterns of L. etheridgei significantly. The range of daily activity of L. etheridgei was shorter in El Simbral than in Tuctumpaya, where the former began activity later (9:00 h) than in Tuctumpaya (8:00 h), and ended earlier (16:00 vs. 17:59 h). This result contradicts our expectation that the Tuctumpaya population would have a shorter daily activity range due to a higher cover of trees than in El Simbral, limiting access to solar radiation.
Polylepis trees occupy rough terrains with areas where the sunlight reaches late in the day, while windy conditions can affect the body temperature of lizards (see Bujes & Verrastro, 2008); thus, both could be involved in the variability of the starting hour of activity.

| Diet
Lygaeidae: Hemiptera was found to be the most important dietary component for L. etheridgei, and this is the only item that was selected for positively in both Polylepis forests. Even though Hemiptera does not seem to be a typical choice in the diet of Liolaemus (Halloy et al., 2006), it has been recorded as an important item in the diet of L. eleodori (Astudillo et al., 2015), L. quilmes, L. ramirezae (Halloy et al., 2006;Robles & Halloy, 2008), and L. ruibali (Villavicencio et al., 2005). Formicidae has been recorded as an important component in the diet of several Liolaemus species (Halloy et al., 2006) such as L.

F I G U R E 7
Unconstrained ordination of diet data based on nonmetric multidimensional scaling (stress = 0.194).
Formicidae was avoided by L. etheridgei and seems not to be an essential prey for this lizard. Interestingly, Mella et al. (2010) (Lobo et al., 2010), so this could be a case of ecological convergence. The avoidance of Cicadelidae, Drosophilidae and Tephritidae could be due to their high mobility which makes predation difficult despite their abundance, but more comprehensive studies are needed to assess this.
Based on the percentage of plant material found, L. etheridgei can be classified as herbivorous, which contradicts the classification of this lizard as mainly arthropophagous by Olivera-Jara and . L. etheridgei would therefore belong to the 9.5% of species within the Eulaemus subgenus, considered herbivores (Cabrera & Scrocchi Manfrini, 2020). Analysis of diet within each forest found L. etheridgei to be herbivorous in Tuctumpaya and omnivorous in El Simbral, although the latter showed a percentage of plant consumption close to the critical value that separates omnivorous from herbivorous ( Figure 5). Intraspecific variation in plant consumption has also been recorded for L. elongatus (Quatrini et al., 2001). Although we recommend additional surveys to confirm the diet of L. etheridgei, we think that herbivory or an intermediate consumption of plant material is evident.
There are several Liolaemus species with an intermediate consumption of plants: L. alticolor, L. jamesi, and L. signifier (Valencia et al., 1982;cited in Semhan et al., 2013), L. eleodori (Astudillo et al., 2015), L. pictus (Vidal & Sabat, 2010), L. polystictus (Olivera Jara & Aguilar, 2020), amongst others (see Abdala et al., 2012). Semhan et al. (2013) point out the existence of a continuum between carnivory-omnivory-herbivory; accordingly, we suggest that L. etheridgei is not in any extreme but biased toward herbivory. Regarding the tendency for herbivory, L. etheridgei would be a "widely foraging" species (Huey & Pianka, 1981) as it would need to move in order to reach plants. We also noticed a sitand-wait foraging strategy, which is suitable for catching highly mobile preys (e.g., Diptera), and this aligns with the foraging behaviors observed by Pearson (1954) in Liolaemus populations from Arequipa, Peru. Therefore, we suggest that L. etheridgei is an intermediate forager with a tendency to sit-and-wait, as has previously been proposed for L. wiegmanni (Aun et al., 1999). This led us to believe that the current fragmentation of these forests does not affect the diet of L. etheridgei significantly.
The three ecological aspects evaluated in this study are not independent, so it is important to interpret them jointly. The absence of selection for microhabitats in Polylepis forests suggests that L. etheridgei is a generalist and does not choose specific sites for heating or foraging. In lizards, microhabitat use is affected by environmental temperature, which varies across the day (Carothers et al., 1998). Thus, the frequent use of uncovered land in the morning (Appendix 1) seems to relate to the heliothermic strategy of L. etheridgei (Llanqui, 2020), which would gain heat by direct exposure to the sun until it achieves adequate temperatures to be active. Likewise, the bimodal pattern opens the question of whether microhabitats are used in the same way in the morning and afternoon activity peaks. However, more data and design type III (Manly et al., 2002) are necessary to test this hypothesis. L. etheridgei reaches its maximum body temperature (~30 °C) between 12:00 and 13:00 h (Llanqui, 2020), coinciding with the increase in activity in the afternoon. We think L. etheridgei is likely physiologically operative during these hours. The absence of microhabitat selection and the selective consumption of Lygaeidae: Hemiptera can be explained considering that this bug is not restricted to specific plants but rather dispersed in vegetation. Thus, L. etheridgei would not need to select foraging sites, and the prey selection process would occur at a fine spatial scale.

| Final remarks
Given the complexity of studying the multiple ecological aspects of L etheridgei, we outline some variables that could affect our results.
Therefore, we believe that variations in the ecology of the L. etheridgei inhabiting Polylepis forests would likely occur through different seasons as the weather conditions change dramatically from summer to winter. Secondly, it is also plausible that juvenile individuals show different patterns in microhabitat selection, daily activity patterns (Cabrera & Scrocchi, 2014;Carothers et al., 1998), and diet (see Duarte Rocha, 1998;Rocha, 1999). Finally, reproductive state (see Llanqui, 2017) and sex have also been recognized to potentially lead to differences in activity patterns (Cabrera & Scrocchi, 2014;Duarte Rocha, 1998) and diet (Semhan et al., 2013).
We would encourage future studies to consider the above factors to improve our understanding of L. etheridgei and Liolaemus species in general.
We conclude that L. etheridgei is a generalist species with little selection for most of the resources we identified in Polylepis forests.
The absence of significant differences in the use of microhabitats, daily activity patterns, and diet between the populations of the fragmented and non-fragmented forest also supports a generalist character that is unaltered by the variation of the habitat resources. We predict this generalist tendency is also present in populations living outside Polylepis forests, such as grasslands, thickets, and even near cities. Finally, we point out that L. etheridgei in the Polylepis forests represents an opportunity to ensure the conservation of the species in southern Peru by protecting the Polylepis forests of the RNSAB and its buffer zone.  Villavicencio and José Perez Z. fort their useful comments.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Raw data: Dryad, https://doi.org/10.5061/dryad.31zcr jdnj.