Characteristics of Central American brocket deer resting sites in a tropical mountain cloud forest in eastern Mexico

Study area

The study was carried out in the mountain cloud forest of San Bartolo Tutotepec, which spans over 6070.1 ha in eastern Hidalgo, Mexico (UTM 572661 2261171, 582857 2261149, 572639 2255120, 582880 2255098; Fig. 1). The forest forms part of the priority region “Mountain cloud forests of the Sierra Madre Oriental” and the Ecoregion “Montane Forests of Veracruz” (CONABIO, 2010). The climate is temperate and humid with two seasonal periods throughout the year: a dry season from October to May and a rainy season from June to September (Peters, 1997). The annual rainfall is 1,200 to 2,000 mm, and the average temperature is approximately 12 °C to 18 °C. It ranges in elevation from 200 to 1,944 m and has a rugged topography, with steep slopes where rivers and streams run between the pronounced ravines, emptying into the river Chiflón (CONAGUA, 2012). The hilly countryside is covered by tropical mountain cloud forest fragmented by agricultural lands and pastures. It contains relict-endemic and endangered tree species such as Magnolia schiedeana, Fagus grandifolia subsp. mexicana, Quercus delgadoana, Q. trinitatis, Q. meavei, Symplocos coccinea, Styrax glabrescens, Turpinia insignis, Persea spp. Several tree fern species, including Cyathea fulva, Dicksonia sellowiana var. arachneosa and Alsophila firma, inhabit steep slopes. The understory is mainly composed of Miconia glaberrima. Central American brocket deer share habitat with other mammals such as coyote (Canis latrans), gray fox (Urocyon cinereoargenteus), jaguarundi (Puma yagouaroundi), ocelot (Leopardus pardalis), margay (Leopardus wiedii), American hog-nosed skunk (Conepatus leuconotus), hooded skunk (Mephitis macroura), long-tailed weasel (Mustela frenata), ringtails (Bassariscus astutus), white-nosed coati (Nasua narica), kinkajou (Potos flavus), raccoon (Procyon lotor), nine-banded armadillo (Dasypus novemcinctus), Eastern cottontails (Sylvilagus floridanus), lowland paca (Cuniculus paca) and Mexican red-bellied squirrel (Sciurus aureogaster) among others (Huerta-Valdez, 2017). It is important to mention that until recently it was thought that its natural predator, the cougar (Puma concolor), had been extirpated from the area, but recent observations confirm its presence (A Cruz-Oropeza, 2021, unpublished data).

The study area has been occupied by humans since the 10th century, when the Toltec civilization inhabited the area. Currently there are 12 Otomí communities in the region, ranging from 40 to 258 inhabitants, whose main economic activities are agriculture and extensive livestock production (Muñoz Vazquez, 2013).

Resting site attributes

We conducted eight bimonthly field surveys between November 2017 and March 2019 with four transects per field season (32 in total). Each transect was 500 m long and were distributed uniformly within the part of the study area that is suitable for brocket deer, according to Muñoz Vazquez & Gallina Tessaro (2016); i.e., avoiding human settlements, pastures and crops. The starting points of the transects were randomly selected a priori and the direction was always north-south. The minimum distance between transects was 200 m in order to cover as much habitat as possible. Each every transect was visited once by four people who directly searched for resting sites (Fig. 2).

We identified resting sites based on the presence of fecal pellets following Aranda (2012), which sometimes formed latrines. We also searched for footprints, scrapes (soil disturbed by deer pawing at the ground), and browsed plants (Fig. 3). Whenever we observed a resting site, we recorded its coordinates using a GPS.

To describe the resting site, we assessed habitat characteristics at two different spatial scales: the microhabitat (specific surface that was used for resting and the 20 m² surrounding the resting site) and the landscape (the 30 m² surrounding the resting site).

Microhabitat

Resting sites were characterized by recording vertical and horizontal thermal cover, protection from predators for fawns and adults, comfort signs, and availability of food and water. First, we performed 20 readings of canopy closure (five in each of the four cardinal directions) between 9:00 am and 12:00 pm using a densiometer to estimate vertical thermal cover (Model C, Lemmon, 1957, Fig. 3C). Second, we followed the Canfield method along four 10 m-long transects (beginning at the resting site and progressing in each cardinal direction), to record all understory vegetation cover ≤2 m in height. These surveys were used to calculate the understory density and coverage height as measures of horizontal thermal cover (see Canfield, 1941).

We measured concealment cover following Griffith & Youtie (1988) at 10 m from the resting site; the only modification was that we considered the cover closest to the ground (“first segment”, 0–50 cm) as protection from predators for fawns and the cover in the second (50–100 cm) and third segments (100–150 cm) as protection from predators for adults (Fig. 3B). We also counted deer footpaths using a manual counter (Base Mount Tally Counter), which we considered the potential escape routes from predators.

As comfort signs, we measured the slope from the ground by using a clinometer (SUUNTO), and we registered the presence/absence of scrapes and the tree and/or tree fern species under which the site was located to determine whether there was a preference for a particular species of tree/tree fern for resting sites.

To evaluate food availability, we identified the plant species along the transect and compared to the to the lists of plant species consumed by deer compiled by Villarreal-Espino-Barros et al. (2008) and Flores-Vazquez (2021) and calculated the cumulative importance value of the edible plant species with the following formulas:

Finally, to record access to water, we measured the linear distance from the resting site to the nearest water source (e.g., permanent and ephemeral ponds, creeks), using a measuring tape (Fig. 3).

Landscape

Since we hypothesized that the habitat preferences of the Central American brocket deer influenced the selection of its resting sites, we used four landscape-scale variables to describe resting sites. Type of biotope was derived from a Landsat 8 OLI image, using an Iso cluster function with unsupervised classification to obtain a biotope type layer that was categorized into the following discrete classes: beech forest, oak forest, secondary vegetation, pine forest, rainforest, ravines, houses and roads, and grazing areas (areas devoid of native vegetation and dedicated to livestock grazing). Elevation was derived directly from a digital elevation model (DEM), and aspect and slope that were calculated with the Surface toolbox in Arc Map 10.3 from the DEM. Erdas Imagine software v.14.0 was used to reproject the Landsat and DEM images and to perform atmospheric and radiometric corrections (Intergraph Corporation). Pixel size for both images was 30 m².

For each resting site surveyed, we established a paired control plot where we measured the same microhabitat and landscape attributes. Control plots were selected by measuring at the point 50 m from the identified resting site in a randomly selected cardinal direction.

Statistical analyses

We first tested whether the explanatory variables were normally distributed. Spearman rank tests were used to evaluate correlations (r_s ≥ —0.7—). Then, multiple logistic regressions were used to differentiate combinations of variables associated with the resting sites. We chose this analysis because multiple logistic regressions are especially useful when the data consist of both discrete and continuous variables (Powell, 2000).

Resting site distribution

We used a point pattern analysis to describe resting site distribution. First, we used the resting site locations to construct a planar point pattern (ppp), then used the Kernel-smoothed intensity to measure the mean number of occurrences per unit at a point (u) defined by λ (u). Finally, we performed first-order characteristics analysis of a spatial process for a general location (s; Eq. (1)):

where K_b() is a Kernel with band b > 0, and C_b() is an edge correction factor (Yang et al., 2007).

We performed a quadrat count test to determine whether there was complete spatial randomness. Once we determined that the pattern was not random, we calculated the inhomogeneous K and L functions (Eq. (2)): (2)${\displaystyle {\hat{K}}_{inhom}\left(r\right)=\frac{1}{\left|W\right|}\sum _i\sum _{j\ne i}\frac{1\left\{∥x_i-x_j∥\le r\right\}}{\hat{\lambda }\left(x_i\right)\hat{\lambda }\left(x_j\right)}e\left(x_i,x_j;r\right)}$

where e (u, v, r) is an edge correction weight and $\hat{\lambda }\left(u\right)$ is an estimate of the intensity function λ(u).