Life after logging: post-logging recovery of a neotropical bat community

Authors


F. M. Clarke, School of Biological Sciences, University of Aberdeen, Aberdeen AB24 2TZ, UK (e-mail f.clarke@abdn.ac.uk).

Summary

  • 1There is considerable debate about whether tropical forests can be managed for timber production and to conserve biodiversity. Few ‘sustainable forestry’ systems have been evaluated adequately in this respect. Microchiropteran bats may be model taxa for this purpose. They are an important component of mammalian diversity, play key roles in forest dynamics and are indicators of disturbance in neotropical forests.
  • 2We investigated the effect of Trinidad's periodic block system (PBS) on bat species diversity and community organization. PBS is a polycyclic system of selective logging with a 30-year harvesting rotation. We captured bats in primary forest and PBS-managed forest logged 33, 31, 21, 20 and 10 years previously.
  • 3Selective logging did not affect species diversity but did affect community structure. Frugivorous bats were significantly more abundant in logged forest whereas gleaning animalivores were more abundant in primary forest, suggesting that frugivores benefited and gleaning animalivores were adversely affected by logging.
  • 4The bat community showed evidence of recovery. The number of years since forest disturbance was positively correlated with the abundance and number of species of gleaning animalivores and negatively correlated with the proportional abundance of the most common species. Gleaning animalivores increased in abundance with forest regeneration, whereas the community became less dominated by a single generalist frugivore.
  • 5Synthesis and applications. PBS selective logging appears to be compatible with the conservation of bat diversity. This provides evidence that neotropical forests can be managed for timber production in an ecologically sustainable way and that significant biodiversity conservation efforts can occur outside national parks and nature reserves in areas set aside for sustainable development. PBS could serve as a basic blueprint for sustainable forestry in the Guianan Shield where there are forests similar in species composition to those of Trinidad. Key attributes that could be adopted by many tropical countries to manage their forests in an ecologically sustainable way are a low intensity harvest, a long rotation and multiple controls on harvesting.

Introduction

Tropical forests cover < 7% of the Earth's land surface but contain a much larger fraction of the animal, plant, fungal and microbial species of the world. Human activities are destroying or modifying tropical forests, leading to a loss of biodiversity (Myers 1992). There is considerable debate about how best to conserve forest biodiversity. Some urge that more forest should be included in reserves and protected areas (Bowles et al. 1998a,b), whereas others believe that forests can be managed in a sustainable way (Chazdon 1998; Gascon, Mesquita & Higuchi 1998; Hartshorn 1998). The reality is that only a small proportion of tropical forest is adequately protected in national parks and nature reserves, and logging of most of the remaining unprotected forest appears inevitable. Sustainable exploitation of tropical forests may be the most realistic way of conserving biodiversity outside protected forest reserves (Greiser Johns 1997).

Polycyclic systems of selective logging may represent an ecologically sustainable and economically viable use of tropical forests, and there is a growing trend towards such systems of timber extraction (Dawkins & Philip 1998). Only the most valuable timber is harvested and the forest is then allowed to recover through natural regeneration before the next crop of timber is harvested. Although there is an increase in ‘sustainable forestry’ systems throughout the tropics, few have been adequately evaluated with respect to biodiversity conservation.

There is growing evidence that bats are good indicators of forest disturbance in neotropical forests (Fenton et al. 1992; Medellin, Equihua & Amin 2000) and model taxa for evaluating which silvicultural systems are most compatible with biodiversity conservation (F. M. Clarke, unpublished data). Bats comprise 20% of mammalian diversity globally (Patterson 1994). Bat species richness increases with decreasing latitude in the New World and generally decreases with increasing elevation (Findley 1993; Andelman & Willig 2003; Willig, Patterson & Stevens 2004). In lowland forests of the Guianan Shield, bats may comprise more than half the mammal fauna (Simmons & Voss 1998; Lim & Engstrom 2001).

Neotropical bats have diverse diets, which include fruit, pollen, nectar, arthropods, small vertebrates and vertebrate blood (Gardner 1977; Patterson, Willig & Stevens 2004). Fruit- and nectar-feeding species play key roles in tropical forests as seed dispersers and pollinators, maintaining plant diversity and promoting forest regeneration in degraded land; other bat species consume large quantities of insects (Charles-Dominique 1986; Foster, Arce & Wachter 1986; Kalko 1998; Patterson, Willig & Stevens 2004). The enormous variety of foods that bats exploit, coupled with the various foraging methods and roosting structures they use, have led to high levels of abundance and diversity (Patterson, Willig & Stevens 2004). This, coupled with the roles of bats as dispersers and pollinators, suggests they have a significant effect on forest dynamics and, if forests are to be managed in a sustainable way, it is important that the responses of bats to forest disturbance are understood. In the neotropical region 20% of bat species are threatened but to date only one study has investigated the responses of bats to selective logging (Ochoa 2000; Hutson, Mickleburgh & Racey 2001).

Trinidad, the larger of the two islands forming the Republic of Trinidad and Tobago, is an ideal locality for examining the impacts of different systems of forest management on biodiversity conservation. Trinidad is the most southerly of the West Indian islands and lies 11·3 km from the Venezuelan coast of South America. Trinidad was connected to the mainland via a land bridge during the Pleistocene and, despite its relatively small size, supports 66 species of bat. The island has a long history of active forest protection and management since the establishment of the Forestry Division in 1901 (Chalmers 1981; Synnott 1989). Most of Trinidad's forests are state owned and legally proclaimed as forest reserves managed primarily for timber production.

One of the most important areas for timber production is the Victoria-Mayaro Forest Reserve (VMFR). Part of the reserve is managed using the periodic block system (PBS), a polycyclic system of selective logging that incorporates a set of ecologically sensitive extraction procedures thought to benefit wildlife (Synnott 1989; Fairhead & Leach 2002). Typically each year one block of forest is logged before being closed to allow forest regeneration. A second harvest is taken around 30 years later. PBS has been recognized as an example of a ‘best practice’ management system (World Bank 1991), but is it ecologically sustainable?

To develop sustainable logging systems it is critical that we understand how the intensity (volume of timber removed), frequency (length of interval between harvests) and method of logging affect wildlife. This information can be used to develop and implement measures to maintain biodiversity. We tested whether the PBS is ecologically sustainable by determining the responses of bats to this harvesting system and investigating whether the bat community recovers with post-logging forest regeneration. We used a variety of standardized sampling methods to capture bats in primary forest and sites in PBS-managed forest logged 33, 31, 21, 20 and 10 years previously.

Methods

study area

The VMFR is c. 52 000 ha of tropical moist seasonal forest of Carapa–Eschweilera association located in the south-east of Trinidad at 10°04′−10°18′N, 61°01′−61°18′W. Around 37 000 ha are of Mora type, an evergreen formation with a continuous canopy characterized by the extreme dominance of Mora excelsa Benth. (Beard 1946). Mora forests have been described as single-dominant forests (Connell & Lowman 1989). The next most dominant canopy species are Carapa guianensis Aublet, Terminalia dichotoma G. Mey., Pterocarpus rohrii Vahl and Spondias mombin L. (Beard 1946). Rainfall is seasonally distributed, with the wet season lasting from June to December and the dry season from January to May. The average annual rainfall in south-east Trinidad is 2012 mm. Forestry is the main land use in south-east Trinidad. The Trinity Hills Wildlife Sanctuary borders the VMFR to the south and the Southern Watershed and Rochard Douglas Forest Reserves to the west.

There are no national parks or strict nature reserves primarily dedicated to biodiversity conservation in Trinidad (IUCN categories I and II protected areas; http://www.unep-wcmc.org/protected_areas/data). The protected area network is composed primarily of forest reserves the main function of which is timber production, although priority usage may vary from watershed protection to recreation to conservation (Chalmers 1981). There are a limited number of wildlife reserves but timber production still takes priority. In the VMFR the Forestry Division has experimented with different systems of silviculture, including sustainable forestry and plantation forestry (Synnott 1989). However, other than a small area converted to teak Tectona grandis L., timber production in the VMFR is under PBS management. Mora is the main timber harvested.

In the PBS, forestry officers of the Forest Resource Inventory and Management (FRIM) section of the Forestry Division, rather than logging concessionaires, select which trees are to be harvested in 150–300-ha blocks. These are logged over 1–2 years by concessionaires and then closed for around 30 years (Synnott 1989; Clubbe & Jhilmit 1992; Fairhead & Leach 2002). One or two blocks are opened each year to satisfy local demand from woodworkers and saw millers and logging is confined to the dry season. Before blocks are opened FRIM generally conducts an inventory of all species > 20 cm d.b.h. and commercial species > 10 cm, and estimates the volume of timber. Vines and lianas are then cut. There are no girth limits for trees but there are many controls on harvesting. FRIM uses ecological criteria to select trees to be harvested. These include canopy closure, maturity, value as a seed tree, frequency of occurrence (rarity), value to wildlife and stream bank protection (Fairhead & Leach 2002). Woodworkers and saw millers can only harvest trees that FRIM has selected for felling. Not all valuable trees or useful sizes are selected for felling in order to maintain or improve the condition of the forest (Synnott 1989).

Harvesting is low to medium in intensity, with fewer than eight trees per hectare typically being harvested (Table 1). A harvest of two to four trees per hectare is most common. Natural disturbance rates are only available from PBS forest regenerating for several decades. The overall mortality rate is approximately 1·8 stems ha−1 year−1 (M. Oatham, unpublished data), within the range of 0·6–2·8% quoted for natural forest in the tropics (Phillips & Gentry 1994).

Table 1.  Summary of management history and coordinates of sites sampled for bats
HabitatCoordinatesManagementYear loggedAge (years)Area* (ha)Volume (m3 ha−1)Canopy closure (%)
  • *

    Interval between first cut and start of study.

  • †Calculated from the Forestry Division's estimates of merchantable timber to be harvested (m3) divided by the area of the block (ha).

  • Densiometer recordings made in the wet season.

  • NA, not applicable; UN, data unavailable.

Primary 110°08′N, 61°08′WReservedNANANANA92·7
Primary 210°10′N, 61°04′WReservedNANANANA90·5
PB 3310°13′N, 61°08′WPeriodic block196733162UN90·2
PB 3110°12′N, 61°03′WPeriodic block1969312435·390·2
PB 2110°12′N, 61°05′WPeriodic block198021100+9·291·5
PB 2010°14′N, 61°07′WPeriodic block1980202433·890·0
PB 1010°06′N, 61°14′WPeriodic block1991102096·489·8

field methods

Fieldwork was conducted in 2000, 2001 and 2002. We sampled bats in Mora forest at two primary forest sites (replicates) and five sites logged using the PBS 33, 30, 21, 20 and 10 years previously (Table 1). Our choice of sites in PBS-managed forest was dictated by the need to control for forest type and disturbance history. The five sites were selected because they were in Mora forest, had been logged only once and were not damaged by fire. Within sites we captured bats at five sampling points 350–750 m apart, each on two occasions: once in the wet season and again, several months later, in the following dry season. Care was taken to ensure that the sampling effort was equal among sites (Table 2) by standardizing sampling methods and recording environmental factors that may affect catching success, such as rain, cloud cover, light intensity, moon phase and the presence of predators and fruiting plants. We avoided sampling in the period 2 days before to 1 day after the full moon, in case lunar phobia affected capture success. Lunar phobia has been reported in several neotropical bats (Morrison 1980; Fleming & Heithaus 1986) although others have found no influence of moonlight on bat activity (Negraeff & Brigham 1995).

Table 2.  Summary of sampling effort, bat captures by method, and capture rate (bats per mist net hour or harp trap hour) in primary forest and sites in PBS-managed forest regenerating for 33, 31, 21, 20 and 10 years
 Ground mist netsSubcanopy mist netsHarp trap
HabitatnMNHBats net−1 h−1nMNHBats net−1 h−1nHTHBats trap−1 h−1
  1. n, number of individuals captured; MNH, mist net hours; HTH, harp trap hours.

Primary 1 1493500·4  2960·02 9600·2
Primary 2 2473600·7 15840·18 5600·1
PB 33 2953570·8 24960·25 6590·1
PB 31 3203500·9 21960·2235590·4
PB 21 4243601·2 47960·49 8600·1
PB 20 1093510·3  8950·08 6570·1
PB 10 2873600·8 13960·14 7600·1
Total1831  130  76  

As a metric of forest disturbance a spherical densiometer was used to estimate the extent of canopy closure at sampling sites (Forest Densiometers, Bartlesville, OK). At each sampling point, canopy closure values were recorded at 10-m intervals along a 50-m transect (Table 1). Canopy closure was significantly correlated with number of years since disturbance (rs = 0·75, P= 0·05) but differences in canopy closure among primary forest and logged sites were small. Wood et al. (1999) reported that canopy gaps in the VMFR had largely closed 5 years after logging.

We erected six 2·6 × 6-m mist nets at ground level (0–3 m) at each sampling point. Nets were positioned to sample all microhabitats present: ridge tops, valley bottoms and streams, flat well-drained ground, swampy areas, under closed canopy or in tree-fall gaps. One 3 × 12-m mist net was set in the forest subcanopy at four of the sampling points within each site. We used a modification of a relatively quick and easy method for setting nets in the subcanopy, as described by Munn (1991), using a Panzer II crossbow (Barnett International Ltd, Wolverhampton, UK) to shoot lines over trees. All nets were 50-denier weight, 2-ply nylon, with a 38-mm mesh size (Avinet Inc., Dryden, NY). A two-frame harp trap with a catching surface of 4·2 m2 (AUSTBAT Research Equipment, Victoria, Australia) was erected each night for inventory completeness, as there is some evidence of interspecific differences in the effectiveness of mist nets vs. harp traps in catching bats (LaVal & Fitch 1977; Francis 1989).

Mist nets and the harp trap were set at the same time relative to sunset each night and were left open for around 6 hours, typically until 23:30 h. Time of capture was recorded in 1-h increments and bats were kept in separate 1-h lots at the processing station. We used individually numbered aluminium wing bands to mark bats before they were released at the site of capture. The number of nets or harp traps multiplied by the number of hours they were deployed equalled the number of net hours or harp trap hours. Harp trap and mist net samples were used to calculate relative abundance and capture rate (bats net−1 h−1 and trap−1 h−1). Recaptures were not included in abundance estimates.

Bat taxonomy follows Simmons (in press). Following Simmons & Voss (1998), we assigned bats to one of the following broad guilds based on foraging behaviour and diet: aerial insectivores (all non-Phyllostomidae), frugivores (Carolliinae, Stenodermatinae), gleaning animalivores (all Phyllostominae except Phylloderma stenops, Phyllostomus discolor, Phyllostomus hastatus and Vampyrum spectrum), nectarivores (Glossophaginae), omnivores (Phylloderma stenops, Phyllostomus discolor and Phyllostomus hastatus), sanguivores (Desmodus rotundus and Diaemus youngyi) and carnivores (Vampyrum spectrum). We defined locally rare species as those species contributing < 0·5% of the total captures. Voucher specimens were deposited at the University of the West Indies (UWI, St Augustine, Trinidad). The skulls of voucher specimens were later examined and craniodental measurements were taken to confirm taxonomic identification. Examination of bats taken as voucher specimens confirmed 97·4% of field identifications.

Typically we collected only one or two individuals per species as voucher specimens. We checked mist nets and the harp trap regularly and captures were always placed individually in cloth bags and processed promptly. During processing, pregnant, lactating and females carrying young were given priority over males and non-reproductive females and typically they were not collected as voucher specimens. Bats were released at the capture site within 3 h of capture. Mortality was low: six individuals out of 2037 captures (0·3%).

analyses

Few inventories approach completion, despite intensive sampling effort. To estimate true species richness (Smax) we used the Chao quantitative estimator, a reliable non-parametric method that gives a lower-bound estimate (Chao 1984). We calculated inventory completeness as species observed/Smax × 100. Alpha diversity was determined using the log series index (α). We determined dominance using the Berger–Parker index (d): an index expressing the proportional importance of the most abundant species (Magurran 1988). The reciprocal form of the index (1/d) is given so that an increase in the value of the index represents an increase in diversity and a reduction in dominance. Kolmogorov–Smirnov two-sample tests were used to compare the species abundance distributions between primary and PBS forest. Distributions were examined in relation to four main models: the geometric series, log series, log normal and broken-stick. Mathematical fit of the data to models was made using chi-square goodness-of-fit tests (χ2).

Any similarity in bat species composition between pairs of sites was determined by calculating 1-Bray–Curtis coefficients: a quantitative measure of percentage similarity (Magurran 1988). We used a one-way anova (F) with post-hoc testing using Fisher's LSD test to compare nightly capture rates (mist nets and harp trap captures combined) among habitats. The number of species and individuals in guilds was compared using chi-square tests. We used data from primary forest to generate expected values (adopting the hypothesis that logging had no effect). Because of the small sample sizes we did not include omnivores, sanguivores and carnivores in chi-square tests.

We used Spearman's rank correlation tests (rs) to analyse the association between forest regeneration (years since forest disturbance) and bat species richness, alpha diversity, proportional abundance of the most common species, number of species and individuals in each guild, and number of rare species. Primary forest was included in correlation tests, as it had been subjected to disturbance 67 years previously by high winds. Although Trinidad lies on the edge of the West Indian hurricane zone and is not normally subject to disturbance from hurricanes, in 1931 and 1933 the south of the island was struck by cyclonic storms that caused low levels of forest disturbance (Beard 1946). Analyses were conducted using the computer packages SDR version 2.6 and CAP version 2.0 (Pisces Conservation Ltd, Lymington, Hants, UK) and Minitab version 12.1 (Minitab Inc., State College, PA, USA).

Results

bat captures

A total of 2037 bats of 38 species and five families was captured (Tables 2 and 3). Examination of capture rates showed that mist nets set at ground level were the most successful method of catching bats (90% of captures), followed by nets in the subcanopy (6%) then the harp trap (4%) (Table 2). There was a significant difference in capture rate among sites (F = 3·93, d.f. = 5, P < 0·01; Fisher's LSD test) for captures in mist nets, but not the harp trap (Table 2). Familial representation was highly skewed in our capture data, with the Phyllostomidae accounting for 98% of captures and 79% of species recorded.

Table 3.  List of bat species recorded and number of individuals captured in primary forest and sites in PBS-managed forest regenerating for 33, 31, 21, 20 and 10 years (recaptures not included). Includes bats captured in ground-level and subcanopy mist nets and harp traps. Sampling effort was equal among sites
TaxonPrimaryPBS-managed forest
Site 1Site 2PB 33PB 31PB 21PB 20PB 10
  • obs, species observed but not captured (escaped from nets).

  • *

    Locally rare species.

Emballonuridae
Peropteryx trinitatus*    1  4   
Saccopteryx bilineata 5  1    3   2
Saccopteryx leptura* 2   1    
Mormoopidae
Pteronotus parnellii* 1    1  1 3 
Phyllostomidae
Carolliinae
Carollia perspicillata6713020117929561195
Desmodontinae
Desmodus rotundus* 2    1  1 
Diaemus youngyi* 1      
Glossophaginae
Choeroniscus minor 3  3   2  1 1  2
Glossophaga soricina20 30 22 39 45 3 30
Phyllostominae
Glyphonycteris daviesi*   2     
Glyphonycteris sylvestris* 4  2  2  1   
Lampronycteris brachyotis*   1     
Lophostoma brasiliense*    1   2 
Micronycteris hirsuta 2  2  2  2  2   2
Micronycteris megalotis 4  8   7  3 3  1
Micronycteris minuta*    2   1   1
Mimon crenulatum12  6  7  2  2 4  5
Phylloderma stenops* 1      
Phyllostomus discolor*    2  1  1   1
Phyllostomus hastatus*     1  2  
Tonatia saurophila 4   7  7  6 1  2
Trachops cirrhosus* 3    2   
Trinycteris nicefori 5 16  3  3  2 1  3
Vampyrum spectrum* 1  1obsobs  1   1
Stenodermatinae
Artibeus cinereus 2 21  8 31 36 1 12
Artibeus jamaicensis   3 17  2 1815  5
Artibeus lituratus   1  4  2  8 3  2
Chiroderma trinitatum*     1  2  
Chiroderma villosum   3  2  5  3 1  4
Mesophylla macconnelli* 1  2  3   1   2
Platyrrhinus helleri 4  1 15 17  6  12
Sturnira lilium*    1  6  1  
Sturnira tildae 6 16 15 49  9 9 11
Uroderma bilobatum 8 16  7  6 24 6 13
Vampyrodes caraccioli   1  1  2  6  
Furipteridae
Furipterus horrens*      2 
Vespertilionidae
Eptesicus brasiliensis*     1   
Myotis nigricans 2  1  1  2  6  1
Number of species per site23 22 24 28 2518 21
Species per habitat29 species  34 species    

species diversity

Bat species richness was similar among sites in primary forest and forest managed with the PBS. In total, 29 species were captured in primary forest and 34 in PBS-managed forest, and the estimated true species richness (Smax) was 33·5 and 34·7, respectively (Table 3). Species were accumulated at a slightly higher rate in primary forest than logged sites (Fig. 1). We found no correlation between the number of years since forest disturbance and the number of species captured (rs = 0·31, P = 0·50).

Figure 1.

Species accumulation curves of bats of primary forest and sites in PBS-managed forest regenerating for 33, 31, 21, 20 and 10 years.

Alpha diversity was similar among sites in primary forest and forest managed with the PBS (Table 4). Moreover, the rank abundance distributions of bats of primary and PBS forests were not significantly different (Kolmogorov–Smirnov two-sample test: maximum positive difference = 0·21, P > 0·50; Fig. 2). Both distributions were described by the log series (primary χ2 = 6·3, P= 0·28; PBS χ2 = 7·9, P= 0·24) and truncated log normal models (primary χ2 = 3·9, P= 0·41; PBS χ2 = 2·4, P= 0·88), although they showed a better fit to the latter model. Primary forest was typically less dominated by a single bat species (Carollia perspicillata) than PBS-managed forest (Table 4). Furthermore, we found a negative correlation between the number of years since forest disturbance and the proportional abundance of the dominant species (rs = −0·69, P= 0·08).

Table 4.  Numerical abundance, species richness and diversity of bats of primary forest sites and sites in PBS-managed forest regenerating for 33, 31, 21, 20 and 10 years
 Primary forestPBS-managed forest
Site 1Site 2PB 33PB 31PB 21PB 20PB 10
Diversity
Individuals (n)160267325376479123307
Species richness (Sobs) 23 22 24 28 25 18 21
Log series index (α)  7·36  5·70  5·98  7·00  5·61  5·81  5·11
Berger–Parker index (1/d)  2·39  2·05  1·62  2·10  1·62  2·02  1·56
Figure 2.

Rank abundance plots of bats of primary forest and forest under PBS management forest in the VMFR.

Similarity in bat species composition was greater when comparing primary forest sites than between sites in primary forest and PBS-managed forest (Table 5). There was no evidence of an association between the number of years that sites have been left to regenerate after logging and similarity to primary forest. In general, forest recovering for more than 30 years was not more similar to primary forest than forest logged only 10 years previously (Table 5).

Table 5.  Percentage similarity in bat species composition between pairs of forest sites. Values are 1-Bray–Curtis coefficients × 100
SitePrimary 1Primary 2PB 33PB 31PB 21PB 20
Primary 272     
PB 335369    
PB 31487177   
PB 2140617370  
PB 206449483936 
PB 10536689804573

guild species richness and abundance

Primary forest did not differ from PBS-managed forest in the numbers of species in each guild (all comparisons χ2 < 7·1, d.f. = 3, P > 0·05). The frugivorous guild was by far the most speciose guild in both habitats, followed by gleaning animalivores and aerial insectivores (Table 6). We found no correlation between the number of years since forest disturbance and the number of species of frugivores (rs = −0·07, P= 0·88) and aerial insectivores (rs = 0·33, P= 0·46). In contrast, we found a strong positive correlation between the number of years since forest disturbance and the number of species of gleaning animalivores (rs = 0·84, P= 0·02).

Table 6.  Abundance of bat guilds in primary forest and PBS-managed forest
GuildPrimary forestPBS-managed forest
Site 1Site 2PB 33PB 31PB 21PB 20PB 10
n%n%n%n%n%n%n%
  1. Numerical abundance (n) and relative abundance (%) of guilds: F, frugivore; N, nectarivore; GA, gleaning animalivore; AI, aerial insectivore; OM, omnivore; S, sanguivore; C, carnivore.

F8855·019472·727484·330079·840985·49678·125683·4
N2314·4 3312·4 22 6·8 4110·9 46 9·6 4 3·3 3210·4
GA3421·2 3713·9 24 7·4 24 6·4 16 3·311 8·9 14 4·6
AI10 6·2  2 0·7  3 0·9  8 2·1  4 0·811 8·9  3 1
OM 1 0·6  0 0  2 0·6  2 0·5  3 0·6 0 0  1 0·3
S 3 1·9  0 0  0 0  1 0·3  0 0 1 0·8  0 0
C 1 0·6  1 0·4  0 0  0 0  1 0·2 0 0  1 0·3

Frugivores were the most commonly captured bats, accounting for 55–73% of captures in primary forest and 78–85% of captures in logged sites (Table 6). The next most frequently captured bats were nectarivores, gleaning animalivores and aerial insectivores. Omnivores, sanguivores and carnivorous bats were represented by few captures. Comparisons of the number of individuals in guilds revealed a significant difference between primary forest and all logged sites (all comparisons χ2 > 32·1, d.f. = 3, P < 0·01). Examination of standardized residuals for chi-square tests indicated that this result was attributable to the large difference between primary and logged forest in the abundance of frugivores and gleaning animalivores. Frugivorous bats were far more abundant in logged forest, whereas gleaning animalivores were more abundant in primary forest (Table 6).

Carollia perspicillata was the most abundant (dominant) bat at all sites (Table 3). Of the frugivorous species captured, Carollia perspicillata was less abundant in primary forest compared with sites in PBS-managed forest, accounting for 42–48% of captures in primary forest and 48–63% of captures in logged forest (Table 3). Artibeus jamaicensis was not uncommon in logged forest but was rarely captured in primary forest. The nectar-feeding Glossophaga soricina was frequently captured and appeared to be unaffected by logging (Table 3). Bats in other guilds were represented by too few captures to allow rigorous comparisons.

We found no significant correlation between the number of years since forest disturbance and the abundance of frugivores (rs = −0·29, P= 0·53), nectarivores (rs = 0·04, P= 0·94) and aerial insectivores (rs = −0·20, P= 0·67). In contrast, there was a strong positive correlation between the number of years since forest disturbance and the abundance of gleaning animalivores (r = 0·95, P= 0·001), indicating that the bat community recovers with forest regeneration.

rarity

Twenty bat species, more than half the total number of species captured, were locally rare (Table 3). All carnivorous and sanguivorous (vampire), 71% of aerial insectivore and 55% of gleaning animalivore species captured were rare. In contrast, only 25% of the frugivorous species captured were locally rare. We found no significant correlation between the number of years since forest disturbance and the number of rare species (rs = 0·56, P= 0·19), although there appeared to be an increasing trend.

Discussion

inventory completeness

We took care to equalize sampling effort among sites and adopt a standardized sampling protocol utilizing several capture methods to maximize inventory completeness and allow valid quantitative comparisons to be made. We captured 87% and 98% of the species likely to be sampled using our protocol in primary forest and PBS-managed forest, respectively. Despite setting nets in the subcanopy and using harp traps, we probably undersampled non-phyllostomid bats. Twenty-one per cent of the species recorded were aerial insectivores and most were represented by few captures. Aerial insectivores may constitute 22–39% of known species at neotropical rain forest sites but are undersampled by mist nets as they rely almost exclusively on echolocation for orientation in space and when foraging, and mostly avoid nets (Kalko 1998; Simmons & Voss 1998).

We infrequently captured aerial insectivores belonging to the Emballonuridae and Vespertilionidae and did not capture Molossidae. In Trinidad, all molossids, a few vesper bats and a single emballonurid species (Diclidurus albus) forage in the unobstructed airspace outside or above the forest, explaining why they were not recorded. For non-phyllostomid species (aerial insectivores) that forage in the forest interior, but which may remain undetected or under-sampled using mist nets and harp traps, time-expansion bat detectors may prove to be a useful tool in the future. These species include Thyroptera tricolor, Furipterus horrens, Natalus tumidorostris, most emballonurids and all mormoopids.

bats and logging

We found gleaning animalivores to be adversely affected by logging. These bats primarily glean arthropods and/or small vertebrates from surfaces (Belwood 1988; Patterson, Willig & Stevens 2004) and their sensitivity to forest disturbance is well documented (Fenton et al. 1992; Brosset et al. 1996; Wilson, Ascorra & Solari 1996). The vulnerability of gleaning animalivores to forest disturbance most probably derives from altered resource distributions and we suggest that a specialized diet may be a key factor.

Logging operations fragment forests as roads and skid trails are built to extract trees. Bats may be affected if predation risk or commuting costs increase as they traverse unsuitable matrix habitats. Gleaning animalivores may be especially vulnerable to predation while commuting through deforested areas as their flight morphology indicates they are adapted for slow manoeuvrable flight within forests (Norberg & Rayner 1987). Gleaning bats typically have more restricted geographical ranges than aerial foragers and may be more at risk of extinction as a result of habitat modification and loss (Arita et al. 1997).

The nectarivorous Glossophaga soricina appears to be unaffected by logging. Aerial insectivores, omnivores and sanguivorous bats are represented by too few captures to conclude whether they are affected by PBS management, although other studies have found that omnivorous bats are often more abundant in forest that has been subjected to high levels of disturbance (Wilson, Ascorra & Solari 1996; Simmons & Voss 1998; F. M. Clarke, unpublished data). None of the bat species we captured appears to be restricted to logged forest. A few species were captured only in logged forest, but with the exception of Sturnira lilium all have previously been recorded from undisturbed forest in the VMFR (Clarke & Downie 2001). Sturnira lilium favours more disturbed forest sites than its congener Sturnira tildae (Simmons & Voss 1998), which we captured at all sites.

In general we found frugivorous bats to be far more abundant in logged forest, indicating that some species may benefit from changes to the forest brought about by timber extraction. Our findings are in agreement with other studies that show that disturbed neotropical forests harbour many more individual bats than undisturbed forests, providing that forest composition and structure remain essentially intact. This is principally a result of an increase in the abundance of a few frugivorous species (Brosset et al. 1996; Wilson, Ascorra & Solari 1996; Medellin, Equihua & Amin 2000; Ochoa 2000). Frugivorous bats in the palaeotropics may also benefit from some forest disturbance, as evidenced by an increase in the number of pteropodid species in the forests of the Philippines following low or moderate levels of disturbance (Utzurrum 1998).

Frugivorous phyllostomid bats have broad diets and most can be considered to be dietary generalists (Gardner 1977; Patterson, Willig & Stevens 2004). However, some species specialize on a core plant taxon the fruit of which is available throughout the year (Fleming 1986). Carollia feeds selectively on the fruits of Piper (Piperaceae), which thrives in disturbed forest (Gardner 1977; Opler, Frankie & Baker 1980; Cloutier & Thomas 1992) and in the VMFR is abundant under the gaps in the canopy where trees have been felled (F. M. Clarke, personal observation). The greater abundance of Carollia perspicillata at most sites in logged forest in the VMFR was probably the result of an increased abundance of Piper and other bat-dispersed pioneer plants. Others have also found Carollia to be more abundant in disturbed forest (Brosset et al. 1996; Wilson, Ascorra & Solari 1996; Simmons & Voss 1998; Medellin, Equihua & Amin 2000) although, in Venezuela, at least, Carollia perspicillata appears to be unaffected by selective logging (Ochoa 2000).

Artibeus jamaicensis, Artibeus lituratus and Chiroderma villosum feed principally on Ficus (Moraceae) and Cecropia (Cecropiaceae) in the canopy (Gardner 1977; Fleming 1986; Kalko 1996), plant species which appear to be more abundant in disturbed forest, perhaps accounting for a greater abundance of these bats in PBS-managed forest. Most frugivorous phyllostomids are dietary generalists and it is likely that many species may take advantage of an increase in fruit production by pioneer plants in logged forest, even if these fruits do not form the core of their diet. Indeed many bats are able to shift their diets partially or adapt their behaviour to compensate for an alteration in the availability of their main food types, perhaps explaining why selective logging is not detrimental to many species. For example, most nectarivorous and frugivorous bats also consume some insects and may alter the quantities of nectar, fruit and insects in their diet as the availability of these foods change seasonally (Patterson, Willig & Stevens 2004). The plasticity in diet, foraging behaviour and high mobility of many frugivorous phyllostomids may make them less vulnerable to localized habitat disturbances than gleaning animalivores. Similarly nectarivorous, omnivorous and frugivorous birds in the neotropics are less affected by selective logging than birds that forage on insects terrestrially or in the forest understorey (Thiollay 1992; Mason 1996; Greiser Johns 1997).

neotropical bats as indicators of forest disturbance

Bats are indicators of disturbance in neotropical forests (Fenton et al. 1992; Brosset et al. 1996; Wilson, Ascorra & Solari 1996; Medellin, Equihua & Amin 2000). In Mexico, conversion of forests to cacao plantations Theobroma cacao L., agricultural fields and cornfields reduces bat species richness, diversity and the number of rare species, and increases the relative abundance of the most common bat species (Medellin, Equihua & Amin 2000). We found that selective logging affects the abundance and species richness of gleaning animalivores and the proportional abundance of the most common (Carolliinae) species. Bat species diversity and the number of rare species may be indicative of high levels of disturbance, such as unsustainable logging and conversion of forest to secondary vegetation or agricultural habitats (Estrada, Coates-Estrada & Meritt 1993; Brosset et al. 1996; Wilson, Ascorra & Solari 1996; Medellin, Equihua & Amin 2000), but our findings suggest that these community variables are insensitive (indicators) to low or moderate levels of forest disturbance, as occurs with low intensity selective logging.

species of conservation concern

In the western hemisphere 82% of threatened and restricted range bat species are not adequately represented within national parks and strictly protected nature reserves (Andelman & Willig 2003). No bat species occurring from Trinidad are categorized as threatened, but we captured five species, Choeroniscus minor, Vampyrum spectrum, Glyphonycteris daviesi, Glyphonycteris sylvestris and Lampronycteris brachyotis, identified as being of special conservation concern or categorized by IUCN as near threatened (Arita 1993; Hutson, Mickleburgh & Racey 2001). Given the absence of national parks and protected nature reserves in Trinidad, the continued survival of these species and other bats may largely depend on adequate protection of existing forest reserves. Indeed throughout the neotropics most forest exists outside national parks and nature reserves, and our study illustrates that significant biodiversity conservation efforts can occur in other types of reserves, such as those set aside for sustainable development. However, national parks should be established in Trinidad and protection of forest reserves needs to be strengthened against agricultural encroachment, squatting, illegal hunting and wildfires. Current legislation is inadequate to prevent illegal activities, perpetrators are rarely prosecuted and wardens infrequently patrol forest reserves.

sustainable forestry in trinidad

Forest logged with the PBS provides one of the best examples of tropical forest managed for timber extraction (Poore 1989; World Bank 1991). We found that PBS management does not appear to affect bat species diversity in Trinidad. Our findings are in agreement with Ochoa (2000), who found that low-intensity selective logging in Venezuela does not affect bat species richness. In general, studies on neotropical birds have found that selective logging of low to moderate intensity (2–10 trees ha−1) has a negative effect on bird communities but that most species persist in the short term (Thiollay 1992; Mason 1996; Greiser Johns 1997).

Although PBS management affects bat community structure, we found evidence of community recovery. As the forest regenerates there is an increase in the abundance and number of species of gleaning animalivores, which are the bats most sensitive to forest disturbance and characteristic of mature and undisturbed forest in the neotropics. Furthermore, the community appears to become less dominated by Carollia perspicillata, a ubiquitous and generalist frugivore that feeds primarily on the fruits of pioneer plants. Aerial insectivores were probably undersampled and we cannot predict the effect of PBS management on these species. Although the bat community shows signs of recovery, at the end of the harvesting rotation the abundance of gleaning animalivores is lower than in primary forest.

In PBS-managed forest the response of plants to different periods of forest regeneration has been investigated. The epiphyte community recovered after logging, and a study of tree dynamics revealed evidence of forest recovery but also reported that succession was still ongoing 30 years after logging (Clubbe & Jhilmit 1992; Wood et al. 1999). Together with our findings, studies on epiphytes and forest dynamics suggest that plant and animal communities recover but may not recover fully before a second harvest. It is unrealistic to think that most floral and faunal communities can recover completely over a few decades. Furthermore, some species may be lost as a result of the increasing forest homogeneity that occurs over successive harvesting rotations. However, this potential loss of biodiversity must be put in perspective. Pressure on Trinidad's forestry base is high and selective logging is competing against more destructive land-use interests that have a much larger impact on biodiversity. In some forest reserves large areas have already been converted to plantations of teak and Caribbean pine Pinus caribaea Morelet.

The extent to which logging affects biodiversity depends principally on the frequency and intensity of timber extraction (Bawa & Seidler 1998). PBS is a polycyclic system with a low intensity harvest, and a relatively long harvesting rotation, and these attributes may largely account for the maintenance of bat species diversity. In the VMFR the PBS results in a mosaic of lightly disturbed forest blocks at different stages of succession and unexploited forest. The multiple controls on harvesting with ecological criteria used to select trees to be harvested may further account for the compatibility of the PBS with biodiversity conservation. Although the length of the harvesting rotation could be extended by several decades to allow greater recovery of floral and faunal communities, this would be unwise if it resulted in logging more primary forest to satisfy the annual demand for timber. A reduction in the size of the remaining undisturbed forest may result in a greater loss of biodiversity than would be conserved by lengthening the harvesting rotation. Although the high mobility of bats and the plasticity in foraging behaviour and diet of many species may allow them to adapt to forest disturbance, species with small home ranges and specialized diets and/or roosting requirements, such as gleaning animalivores, may be especially threatened by disturbance to forests. The survival of some species may require that some undisturbed areas in the VMFR are protected from exploitation. Currently the Forestry Division is resisting pressure from woodworkers and saw millers, and aims to keep primary forest unexploited (Fairhead & Leach 2002). Unfortunately, the Forestry Division regards undisturbed forest primarily as a strategic timber reserve. We recommend that it should be not be logged.

Improvements to the PBS could be made. Since 1998 the forestry division has begun post-liberation thinning in logged forest, felling over-mature and dead standing trees (snags) to liberate smaller trees of commercial value from competition. Post-liberation thinning should be halted until its effects on wildlife can be ascertained, as snags and over-mature trees often contain cavities that form important roost sites for bats and critical habitat for wildlife (Greiser Johns 1997; McComb & Lindenmayer 1999).

The PBS is unique to Trinidad and it is unclear whether other countries can easily adopt the system to sustainably manage their forests. Timber produced by the PBS is not exported to generate foreign revenue but instead ensures a continuous supply of wood to Trinidad's local market while maintaining forest integrity. Moreover, in contrast to most other countries in tropical America, forestry has not needed to be financially autonomous because of heavy state subsidies from Trinidad's gas- and oil-rich revenue base. This enables Trinidad to maintain a culture of scientific and sustainable forestry as opposed to economic forestry (Fairhead & Leach 2002). Finally, although the PBS is a good example of sustainable timber production, this may partly be because Mora is a single-dominant forest that is floristically simple and regenerates well after logging with little management intervention.

Single-dominant forests are not uncommon, occurring in each of the three major world regions of rain forest in Africa, the Americas and Asia (Richards 1996). For countries with single-dominant forests similar in species composition to those of Trinidad, the PBS could serve as a basic blueprint for sustainable timber production, for example Guyana's Mora and Morabukea forests (ter Steege et al. 1996). Key attributes of the PBS that other countries could adopt to manage their forests sustainably are a low intensity harvest, a long rotation and comprehensive controls on harvesting overseen by experienced forestry officers. However, it should be noted that silvicultural systems should not be regarded as packages to be applied to forests about which little is known, and it will be a challenge for individual countries to develop or refine existing systems of sustainable forestry to suit local biotic and socio-economic conditions. This task is vital, as sustainable management holds the key to conserving the world's biodiversity.

Acknowledgements

This research was funded by a research grant from The Leverhulme Trust, reference no. F/00 152/B. We thank Ruari Allan, Daveka Boodram, Katherine Fisher, Anita Hogan, Dorothea Pio, Donna Piccini, Gaya Sriskanthan and Ana Peña Tuñas for assistance in fieldwork. We are indebted to Mike Oatham and Sheeba Sreenisivan, University of the West Indies, for PBS data. We thank Nadra Gyan and Krishnna John, Forestry Division, and Petrotrin Ltd for providing accommodation.

Ancillary