Diversity and distribution of smaU terrestrial rodents along a disturbance gradient in montane Costa Rica

A total of 389 rodent captures in five unequally disturbed habitats in a Costa Rican montane cloud forest corresponded to 185 individuals (seven species, two faroilies). Species richness was similar for al l habitats, averaging 4-5 species/habitat. Population density and capture frequency were higher in moderately disturbed habitats; Peromyscus mexicanus and Scotinomys xerampelinus were four to five times more cornmon than other murids. These species represented 79.4 % of all captures and 73.5 % of all captured individuals. Heteromyid species were rarely trapped. The importance of disturbance-mediated within-habitat microenvironmental heterogeneity fOI terrestrial small-sized rodent populations is stressed.

Terrestrial rodents are known to be important seed dispersers and predators in tropical forest ecosystems (e.g.Adler 1995).Often they form the staple food of many avian and mamm al predators and therefore play a prominent role in maintaining the structure and composition of species-rich tropical habitats.However, Httle is known about their response to changes in the environment and particularly to disturbance by man (Adler 1994).Indeed, it has been recorded that some rodent species populations may flourish due to human activitíes in natural and seminatural ecosystems (Johnson and Vaughan 1993).Untill now, knowledge on the effects of deforestation and habitat fragmentation on the presence and abundance of small terrestrial rodents in tropical forests is stiU limited.This is particularly the case in tropical upland regions where we count with just a few studies (Lanzewizki 1991, Johnson andVaughan 1993).Even less is known on the role terrestrial Rodentia play in forest succession, canopy closure and biodiversity recovery following clearing and burning.
In order to gain insight in the intrinsic recovery capacity of the biodiversity inhabiting a shredded Costa Rican montane cloud forest area and to provide knowledge needed for the conservation and sustainable use of this threatened ecosystem, we studied the diversity and distribution of small terrestrial rodent species along a disturbance gradient, ranging from undisturbed mature primary oak forest to graminoid dominated pastureland suffering from intensive cattle grazing.The principal aim of the present study is to identify key indicator rodent species characteristic for certain levels of disturbance and recovery.

MATERIALS AND METHODS
This study was conducted in the Los Santos Forest Reserve near San Gerardo de Dota in the NW part of the Costa Rican Cordillera de Talamanca (9°35'40" N; 83°44'30").Originally, the watershed area (2000-3200 m altitude) was completely covered with mature, primary Quercus dominated montane cloud forest, but since the 1950s forest conversion for agricultural purposes has led to a landscape mosaic with vegetation patches of different size, structure, composition, level of disturbance and successionaI age (Kappelle andJuárez 1995, Kappelle et al. 1995).In Villa Mills, at a distance of 10 km E ofthe selected forest pIot, the average annual temperature is 1O.9°C and the average annual rainfall is 28 12 mm.A detaiIed description of the study area's physicaI setting and botanicaI aspects can be found in Kappelle (1993Kappelle ( , 1996)).
At the end of the dry season (March-April 1996), using B&W aerial photographs (1992, scaIe 1: 15,000) five different habitats were selected aIong a man induced disturbance gradient in the upper montane forest beIt (2200-2800 m a.s.l.).These habitats are, from low to high le veIs of disturbance: (i) 30 to 40 m taH, dense oak dominated 'CIosed Mature Forest' (CMF); (ii) oak dominated 'Open Mature Forest' with a reIatively open, 30 to 35 m high canopy (OMF); (iii) 3 to 7 m high secondary 'SuccessionaI Shrubland' (SSL); (iv) 0.5 to 2 m taIl secondary 'Abandoned Pastureland' (APL); and (v) less than 0.5 m high 'Grazed Pastureland' (GPL); see Table 1.At each habitat site a 0.25 ha pIot (50 x 50 m) was established, with exception of APL where a 0.28 ha plot (37.5 x 75 m) was laid out, since the site's patch dimensions required so.
In all but APL a total of 25 Sherman traps (trap size: 23 x 9 x 8 cm) were pIaced in a 5 x 5 grid with a 12.5 m distance between two neighbouring traps.Similarly, in APL 28 traps were Iocated in a 4 x 7 grid.A standardized capture-recapture method was used to estimate species diversity, distribution and abundance in each habitat type (Leslie et al. 1952).Plots were studied from April to June 1996, covering the transition from the dry to the wet season.Sherman traps were checked ten consecutive days before noon.Only in OMF traps were checked nine days.Bait consisted of a mixture of roUed oats and peanut butter with a touch of vanilla flavour.In the afternoon traps were checked for diurnal catches and bait was renewed for the next trap-night.Traps were covered with litter and/or leaves for camouflage and insuIation.
Data on the presence, sex, weight and length of trapped individuals were recorded.Weight was measured using a Pesola pocket scale (max.weight 300 g).Length measurements included the head-body Iength measured from the tip of the nose to the inflection point of the tail, and the tail length measured from the inflection point with the body to the fleshy tip of the tail.In situ species identification was done with the aid of Emmon & Feer's (1990) fieId guide, after studying rodent specimens in Costa Rica's national collections.Control specimens were collected and stored at the Museo Nacional.Specimens were ídentified with help of B. Rodríguez (pers.com.) and on basis of field guides (Mora & Moreira 1984, Emmons & Feer 1990).Trapped individuals were marked upon first capture with acrylic paint before being reIeased.Additionally, terrestrial vascular plant inventories were undertaken in each habitat and collected plant specimens stored and identifíed at the Museo Nacional.Species data were grouped and analysed for habitat sites along the disturbance gradient.Richness, abundance, density and capture frequency were assessed and preliminar conclusions drawn.Mean values with n = 5 habitat sites .
Vegetation layering is expressed in maximum height and relative aerial crown cover of each stratum. 2 Capture frequency = mean number of captures per trap night.
Table 1 presents some general data on the terrestrial rodent population in the five habitats.Richness was very similar for all habitat sites, with 4 to 5 species per habitat regardless its position along the disturbance axis.Population density and capture frequency (absolute and per area values), however, were particular1y higher in habitats with interrnediate levels of disturbance, being twice as high in OMF, SSL and APL when compared to both undisturbed CMF and intensively-used GPL.Abundance values differed strongly among species (Table 2).

DISCUSSION
The varied abundance values were concordant with classical rank-abundance Figs.(Magurran 1988).High abundances (60 -75 individual s) were recorded for P. mexicanllS sumichrasti, appeared to be four to five times les s common.Together, P. mexicanus and S. xerampelinus represented 79.4 % of all captures (íncluding recaptured individuals) and 73.5 % of all captured individuals.Both Heteromyidae species, in their tum, seemed to be rather rare in the area ($ 6 indiv.)and appeared to avoid both abandoned and grazed pasturelands (Table 2).P. mexicanus and S.
xerampelinus occured at all five sites, although they differed in habitat preference with P. mexicanus appearing with higher numbers in CMF and OMF, and S. xerampelinus showing a higher abundance in SSL and APL.This difference is explained by the fact that the latter, more insectivorous species requires a rather dense vegetation cover at ground level, as its peak in activity is during moming hours (Hooper 1972).range given by these authors would suggest (49-103 g;1O.8-14.8cm).This outcome slresses the difficulties that occurred in the preliminary species identificaríon.The weight distribution for male and female individual s of the most abundant species, P. mexicanus, c1early illustrates within-species weight differences found between sexes in rodents (Fig. 1).Whereas the weight distribution (expressed in J g weight c1asses) for males depicts a normal bell curve, the weight distribution for female indivíduals is asymmetrical, with an abrupt change at the 53 g weight level.This striking discrepancy may be explainedby the presence of a number of pregnant adults in the 50-53 g weight c1ass.If non-pregnant, these individuals would have been recorded in lower weight c1asses such as the 44-47 and 47-50 g c1asses.Indeed, four heavy adults (13 %) out of a total of 32 females were in a certain stage of pregnancy.

Fig, 1 .
Fig, 1. Weight distributíon for maJe and female índividuals of Peromyscus mexicanus in the montane cloud forest zone of the upper Río Savegre watershed area, Cordillera de Talamanca, Costa Rica,

TABLE 1
General data on physiographic aspects, vegetatíon layering and terrestrial radent populations ojfive diff erent habitat sites in the montane cloud jorest zone oj ¡he upper Río Savegre watershed area, Cordillera de Talamanca, Costa Rica •

TABLE 2
Abundance data of seven terrestrial rodent species found in five diff erent habitat sites in the montane cloud forest zone of the upper Río Savegre watershed area, Cordillera de Talamanca, Costa Rica.Abundance data are subdivided per sex abo capto abo capt.abo capt.abo capt.abo capt.ab.capt.sex = abundance measured as the total number of individuals captured; capt.= total number of captures including recaptures.and S. xerampelinus, whereas the other three Muridae, O. albigularis, R. creper and R. cf.