Introduction

Climate change has been identified as one of the main threats to highland ecosystems in the tropics (^{Epstein 2000}; ^{Hughes 2000}; ^{Hilbert et al. 2004}; ^{Rull and Vegas-Vilarrúbia 2006}; ^{Laurance et al. 2011}). In Costa Rica, the negative effects of this global phenomenon have been modeled for highland amphibians, reptiles, and birds (^{Pounds et al. 1999}, 2006; ^{Karmalkar et al. 2008}). As a consequence of climate change, the temperature, dry season length, and number of dry days have increased in tropical highlands (^{Pounds et al. 1999}).

These climatic changes have affected the distribution and interaction of species along tropical altitudinal gradients. Many middle elevation species have recently expanded the upper limit of their altitudinal distribution, moving into the habitat of those species that occur at highest elevations (^{Morales-Betancourt and Estévez-Varón 2006}; ^{Dirnböck et al. 2010}; ^{Bellard et al. 2012}; ^{Ripple et al. 2014}). However, species that inhabit the summit of tropical mountains are trapped in sky islands as their habitat contracts and competition with lowland invading species increases. The synergistic effect of these factors imposes a serious extinction threat on endemic and highland-restricted species (^{Thomas et al. 2004}; ^{Malcom et al. 2006}; ^{Urban 2015}; ^{Pyšek et al. 2017}).

In Costa Rica, the highlands have relatively low diversity but a high percentage of endemic species from different taxonomic groups (^{Barrantes 2005}; ^{Vargas and Sánchez 2005}; ^{Barrantes et al. 2019}). For instance, 22 (73 %) mammal species are endemic at middle and high elevations (Ramírez-Fernández et al. in press), and in the highlands many of these species accomplish an important ecological role, such as the dispersal of seed and fungal spores (^{Lacher et al. 2016}).

Most Costa Rican mammal research has centered on lowland habitats. Thus, information on the highlands is scarce and fragmented (^{Carrillo et al. 2005}). Considering that rodents are an abundant component of the highland fauna (^{González-Maya et al. 2015}), and that their role as seed and arthropod predators, fungal spore and seed dispersers, and prey for birds and larger mammals is important, we sought to quantify the richness and abundance of the mice community in the high montane forest and the paramo in the Talamanca mountain range. This is the largest and highest Costa Rican mountain range, and it is recognized as the terrestrial region with the highest endemism in Central America (^{Holz and Gradstein 2005}; ^{Powell et al. 2022}).

Materials and methods

Study site. To conduct this research, we selected two study sites in the highlands of the Costa Rican Talamanca mountain range: the Cerro de la Muerte Biological Station (CMBS) and the paramo (Figure 1). The CMBS is located in the high montane oak forest at an elevation of 3,100 m (9° 33´N, -83° 44´ W) and the paramo at 3,400 m elevation (9º 33´ 20″ N, -83º 45´ 41″ W). The two sites are separated by a distance of 2.5 km. The mean annual precipitation is 2,500 mm, with a relatively dry period between December and April, with a mean annual temperature 11° C for the CMBS and 7.6° C for the paramo (^{Herrera 2005}). The temperature oscillates drastically during the day, particularly in the paramo (−5° to 35°). Forests dominated by oaks (Quercus costaricensis) with abundant epiphytes, bushes (e. g., ferns, Ericaceae, Asteraceae, Onagraceae), and large patches of bamboo (Chusquea talamancensis) cover most of the CMBS study site (^{Calderón-Sanou et al. 2019}). On the contrary, the paramo is dominated by an herbaceous layer with a large diversity of Asteraceae and Poaceae (mainly Chusquea subtesellata), and some scattered bushy patches in which Ericaceae, Asteraceae, and Hypericaceae are abundant (^{Vargas and Sánchez 2005}).

We selected two microhabitats in each habitat, the montane forest and the paramo, to assess rodent habitat use, species richness, and abundance. In the montane forest, we sampled mice in (1) patches dominated by bamboo (Chusquea talamancensis), and in (2) patches dominated by bushes. In the paramo, we sampled in (1) patches dominated by bushes and an herbaceous layer (e. g., Asteraceae, Cyperaceae) and short bushes (e. g., Pernettya, Vaccinium, and Hypericum); and (2) in homogeneous patches dominated by Chusquea subtessellata. In each microhabitat we established a circular 10 m-diameter plot, so that we had the following combination of microhabitats in each ecosystem: paramo-bush, paramo-Chusquea, montane forest-bush, and montane forest-Chusquea (Figure 2).

Figure 1 Sample sites at the Talamanca mountain range, Costa Rica. A) High montane oak forest at Cerro de la Muerte Biological Station; B) paramo habitat. 

Capture, recapture, and tattooing mice. We used baited Sherman live traps (5 × 6 × 16 cm; H. B. Sherman Traps, Inc., Florida) to capture mice. The bait was made of barley, oatmeal, banana, peanut butter, and vanilla extract. In each habitat, we placed 50 traps (25 in each microhabitat) for two consecutive nights, each 3 to 4 weeks from May 2015 through April 2016. We placed each trap at 5 to 10 m from a transect we established within each plot and at 5 to 10 m from any other trap. The distance from the transect and from other traps varied depending on the topographic characteristics of the terrain. We registered the geographic position of each trap in each sampling using a GPS Garmin 60Csx (Garmin Corp., Olathe, Kansas) and changed the location of the traps in each sampling to maximize the capture in each microhabitat. In addition, we placed 4 to 8 traps on branches at 1.5 to 3 m above the ground to capture arboreal or climbing mice.

We identified each individual captured to species level using the key published by ^{Villalobos-Chaves et al. (2016}). For each individual captured, we recorded species, date, habitat, microhabitat, and trap coordinates. Given that taxonomy within the group is in debate, voucher specimens were collected prior and after the sampling period as a reference (Appendix 1).

We tattooed large adult Peromyscus using a Spaulding Revolution I (Spaulding and Rogers, Inc., New York) machine for animal tattooing. Individuals were tattooed at the base of the tail. For juvenile P. nudipes and small species, we cut a patch (or patches) of hair from the lumbar region to identify each individual; the tattooing machine is not recommended for small animals. In addition, we recorded any distinctive mark of an individual that facilitated their identification (e. g., tip tail coloration, scars, ear cuts), in case they were recaptured. For the recaptured individuals, we recorded the individual mark or tattoo, trap coordinates, habitat, and microhabitat. After each individual was marked and its information recorded, we released it at the site where it was captured. We then waited until each individual had found a retreat or a protected site before leaving the release site to avoid predation.

Data analyses. We captured four mouse species and compared their abundance among the four microhabitats with a chi-squared contingency analysis and a Fisher paired test. We did not analyze richness since we only captured four species. We used the statistical language R, version 3.4.0 (R Core Team 2017) for the analyses.

Results

Species richness and abundance. We captured 108 mice of four species in the family Cricetidae, from a 2,100 night-trap effort distributed equally between habitats. We captured three species at each habitat, two of which were present in both habitats (Table 1). Scotinomys xerampelinus was captured only in montane forest, Reithrodontomys creper only in the paramo, R. sumichrasti was found in the Montane forest-Chusquea and Paramo-bush microhabitats, and P. nudipes was the only species captured in all four microhabitats (Table 1).

Figure 2 Location of the sample sites in the Talamanca Mountain range. Light green background represents paramo; dark green represents montane forest. Black dots show bush microhabitat; black triangles, Chusquea microhabitat. 

Peromyscus nudipes was the most abundant species captured with 85% of all captures (Table 1). The number of mice captured (all species combined) differed between habitats (X² = 43.75, d. f. = 1, P < 0.001) and among microhabitats (X² = 36.83, d. f. = 9, P < 0.001). We captured more mice in the montane forest than in the paramo and more in the montane forest/bush microhabitat (n = 58) than in any other microhabitat (Table 1). On the contrary, the microhabitat with the fewest mice captured was paramo-Chusquea with only 6 individuals (Table 1).

Discussion

We found four mouse species in the montane forest and the paramo, the two dominant ecosystems at the Costa Rican highland. At this elevation, richness and abundance of mice vary according to the characteristics of the habitat and microhabitat. For instance, the vegetation structure directly influences density (e. g., ^{Pardini et al. 2005}; ^{Blaum et al. 2006}), diversity (e. g., ^{Johnson and Vaughan 1993}; ^{Muñoz et al. 2009}), and richness of small rodents (e. g., ^{Brehme et al. 2011}; ^{Thompson and Gese 2013}) in Costa Rica and other ecosystems.

Our results showed that some species are present only in a particular habitat and some microhabitats, but not in others. Specifically, we captured S. xerampelinus only in the montane forest and R. creper only in the paramo; although both species have sporadically previously been recorded in both habitats (^{Reid 2009}; JDRF unpublished data). Therefore, our findings suggests that even when both species are present in both habitats, each of them preferred one particular habitat over the other, likely related to the structural differences in the vegetation (^{McCloskey 1976}; ^{Johnson and Vaughan 1993}).

The abundance of mice varied largely between habitats and among microhabitats, as reported in other studies (^{Mohammadi 2010}). In the montane forest, a more complex vegetation structure and greater diversity of food resources and hiding places likely permit the coexistence of more individuals than in the paramo. The abundance of mice among microhabitats showed a similar pattern. Structurally complex microhabitats, such as the montane forest / bush microhabitat, allow more species to coexist (^{Torres-Pulliza et al. 2020}). On the contrary, the microhabitat dominated by Chusquea subtessellata in the paramo, which is structurally simple with large exposed open areas, had the lowest mice abundance.

Differences in abundance among microhabitats of P. nudipes, further support that habitat structure influences in the abundance of mice in Costa Rican highlands. This species is present in all microhabitats and its abundance increased with the microhabitat complexity. More complex structure in the vegetation elicits a series of synergetic effects that affect the abundance of different species. For example, a structurally complex vegetation provides a larger number of microhabitats, food resources, and hiding places that can be used for more rodents than those provided by more simple microhabitats (^{Tews et al. 2004}).

The paramo-Chusquea is the simplest microhabitat, composed by homogeneous patches of Chusquea subtessellata with open spaces in between. The vegetation composition and the simple structure likely provide fewer food resources and retreats than any other microhabitat (^{Johnson and Vaughan 1993}; ^{Mohammadi 2010}; ^{García et al. 2011}). In addition, rodents tend to avoid foraging in areas deprived of vegetation, since they are more likely detected by predators (^{Kotler et al. 1988}; ^{Morris and Davidson 2000}).

Summarizing, the species richness of mice at the highest vegetation ecosystems, the montane forest and paramo, in the Talamanca mountain range is low. However, the abundance of some species is very high, and this abundance varies greatly among microhabitats. The variation in mice abundance is presumably determined by the characteristics of the habitat, primarily related to the structure of vegetation. A more complex structure is expected to provide rodents with more diverse and abundant food resources (^{Johnson and Vaughan 1993}; ^{García et al. 2011}), more retreats, and better protection from predators (^{Kotler et al. 1988}; ^{Morris and Davidson 2000}).