Introduction
Globally, tropical forests are being lost at unprecedented rates (Hansen et al. Reference Hansen, Potapov, Moore, Hancher, Turubanova, Tyukavina, Thau, Stehman, Goetz and Loveland2013), leading to fragmentation, degradation and loss of habitat for many organisms (Barlow et al. Reference Barlow, Lennox, Ferreira, Berenguer, Lees, Nally, Thomson, Ferraz, Louzada, Oliveira, Parry, Ribeiro De Castro Solar, Vieira, Aragào, Begotti, Braga, Cardoso, De Oliveira, Souza, Moura, Nunes, Siqueira, Pardini, Silveira, Vaz-De-Mello, Veiga, Venturieri and Gardner2016). Such changes influence the structure and composition of biological communities, often reducing local species richness and diversity, habitat connectivity, and consequently gene flow, the adaptive capacity of species and the integrity of ecosystems (Fahrig Reference Fahrig2003, Laurance et al. Reference Laurance, Camargo, Luizão, Laurance, Pimm, Bruna, Stouffer, Williamson, Benítez-Malvido and Vasconcelos2011). Landscape-level effects of deforestation are manifested in smaller fragment areas, greater isolation and a proliferation of edge effects (Andrén Reference Andrén1994, Ewers & Didham Reference Ewers and Didham2006). Additionally, the remaining forest fragments may be continually degraded by such practices as selective logging (Hill & Hamer Reference Hill and Hamer2004, Morris Reference Morris2010).
Bats (Mammalia: Chiroptera) are taxonomically and ecologically diverse, especially in the tropics. They utilize diverse roosting structures and a wide range of diets (Kunz & Pierson Reference Kunz, Pierson and Nowak1994, Meyer et al. Reference Meyer, Fründ, Pineda and Kalko2008) and their sensitivity to anthropogenic alterations in habitat quality makes bats valuable indicators of habitat disruption (Jones et al. Reference Jones, Jacobs, Kunz, Willig and Racey2009, Sherwin et al. Reference Sherwin, Montgomery and Lundy2013). Forest fragmentation, selective logging and other forms of human disturbance can induce major changes in the distribution and abundance of bat species (Meyer et al. Reference Meyer, Struebig, Willig, Voigt and Kingston2016). At the landscape scale, remaining patches of forest may become too small, too isolated, and too influenced by edge effects to maintain viable populations of some bat species (Meyer et al. Reference Meyer, Fründ, Pineda and Kalko2008). However, not all bat species or functional groups are disadvantaged by human influences, and some species remain unaffected (Presley et al. Reference Presley, Willig, Wunderle and Saldanha2008) or may even benefit from human disturbance (Farneda et al. Reference Farneda, Rocha, López-Baucells, Groenenberg, Silva, Palmeirim, Bobrowiec and Meyer2015, García-García et al. Reference García-García, Santos-Moreno and Kraker-Castañeda2014, Gorresen & Willig Reference Gorresen and Willig2004). Most studies of fragmentation effects on tropical bats have been concentrated in South and Central America and on the New World family Phyllostomidae (Meyer et al. Reference Meyer, Struebig, Willig, Voigt and Kingston2016). Effects of habitat fragmentation on the diverse bat assemblages in tropical Africa, where deforestation rates are high (Hansen et al. Reference Hansen, Potapov, Moore, Hancher, Turubanova, Tyukavina, Thau, Stehman, Goetz and Loveland2013), remain poorly understood (Meyer et al. Reference Meyer, Struebig, Willig, Voigt and Kingston2016).
Equatorial African bat faunas include four functional groups or foraging ensembles: frugivores, forest-interior insectivores, forest-edge insectivores and open-space insectivores. The forest-interior insectivores show specializations for narrow-space foraging and appear dependent on forest interiors (Kingston et al. Reference Kingston, Francis, Zubaid and Kunz2003). We compared bat abundance and species richness in forest habitats varying in fragment size and/or level of degradation. We expected that different bat species and ensembles would respond differentially (Barlow et al. Reference Barlow, Gardner, Araujo, Ávila-Pires, Bonaldo, Costa, Esposito, Ferreira, Hawes, Hernandez, Hoogmoed, Leite, Lo-Man-Hung, Malcolm, Martins, Mestre, Miranda-Santos, Nunes-Gutjahr, Overal, Parry, Peters, Ribeiro-Junior, Da Silva, Da Silva Motta and Peres2007, Pardini et al. Reference Pardini, Faria, Accacio, Laps, Mariano, Paciencia, Dixo and Baumgarten2009), according to their ecomorphology, foraging strategy and echolocation call attributes (Denzinger & Schnitzler Reference Denzinger and Schnitzler2013, Kingston et al. Reference Kingston, Francis, Zubaid and Kunz2003, Schnitzler et al. Reference Schnitzler, Moss and Denzinger2003). Specifically, we predicted that (1) forest-interior insectivores are strongly associated with the interior of the larger, more intact forests; and (2) frugivores and forest-edge or open-space insectivores are more strongly associated with edges and smaller, more degraded forest fragments.
Methods
Study site
Kakamega Forest (0°07’ 0°27’ N, 34°46’ 34°57’ E; Figure 1) is a mid-elevation rain forest (1400–1700 m asl). Rainfall averages 2000 mm y−1 and daily temperatures range from 11°C to 26°C (Glenday Reference Glenday2006). Kakamega is the easternmost outlier of the Guineo-Congolean rain forest (Wagner et al. Reference Wagner, Kohler, Schmitz and Bohme2008). ‘Kakamega Forest’ is used to refer to both the main forest block (8600 ha) and its two satellite fragments, Kisere (400 ha) and Malava (100 ha). Kakamega Forest is managed by two different parastatals, Kenya Forest Service (KFS) and Kenya Wildlife Service (KWS). This study was carried in the northern section of the main Kakamega Forest block (also called Buyangu, ~3950 ha) and Kisere fragment, both managed by KWS as Kakamega National Reserve; KFS manages the southern part of the main Kakamega Forest block (4695 ha) and Malava forest. Malava and Kisere fragments are separated from Buyangu Forest by 9.2 and 1.6 km, respectively, while Malava and Kisere forests are 6.3 km apart. Kisere and Malava forest fragments have been disconnected from the main block for at least 50 y (Mitchell et al. Reference Mitchell, Schaab and Wägele2009), separated by a high-contrast matrix of dense human settlements (> 578 people km−2), subsistence agriculture, exotic forest plantations of Pinus and Eucalyptus, and regenerating forest and pastures (Kokwaro Reference Kokwaro1988, Müller & Mburu Reference Müller and Mburu2009). The dense human population and widespread poverty place unsustainable demands on the forest for timber, charcoal and fuel wood for domestic uses, livestock grazing and conversion for croplands (Guthiga et al. Reference Guthiga, Mburu and Holm-Mueller2008).
Figure 1. Map of Kakamega Forest showing the spatial arrangement of sampling locations (stars) and studied forest fragments: Buyangu, Kisere and Malava forests. Inset is the map of Kenya showing the location of the study area.
Experimental design
We assessed bat species richness and relative abundance at each forest fragment using captures. A two-factor orthogonal experimental design was employed that included three forest fragments (Buyangu, Kisere and Malava) and captures made at edge and interior. We defined forest edge as an area within 100 m of any disturbance (i.e. agricultural crop or pasture, roads) whereas forest interior was an area in the forest interior that showed no detectable edge influence (Harper et al. Reference Harper, MacDonald, Burton, Chen, Brosofke, Saunders, Euskirchen, Roberts, Jaiteh and Esseen2005). Forest interior sites were at least 100 m (Buyangu: 125 ± 5.62 m; Malava: 103 ± 1.62 m; Kisere: 111 ± 2.51 m) from a forest/non-forest boundary (mean ± SE).
We sampled six edge and six interior locations per fragment, and sampled from two sites (≥ 500 m apart) at each location. Locations were > 2 km apart to minimize pseudoreplication and interspersed with respect to the three fragments (Figure 1). Accordingly, captures were conducted at 12 sites per fragment, making a total of 36 sampling sites.
Vegetation and forest-use characteristics
At all 36 capture sites, vegetation and human-use parameters were measured in four 0.04-ha (20 × 20-m) plots. Inside each plot, all trees (≥ 10 cm diameter at breast height - dbh) were counted, their dbh measured and used to calculate basal area, BA (the space covered by tree stems) (Mueller-Dombois & Ellenberg Reference Mueller-Dombois and Ellenberg1974). BA values were combined for all species per location by summing the basal areas of individual trees. Mean canopy cover was estimated from four measurements taken with a concave spherical densiometer, 1.5 m from the centre of the plot in each of the four cardinal directions. Finally, as indicators of human forest use, forest degradation values were indexed in each plot as simple counts of cut tree stumps, charcoal kilns and footpaths.
Bat trapping
Fieldwork was carried out from May 2013 to April 2014, amounting to 144 sampling nights. Sites were visited in randomized order within each forest fragment, but interior and edge locations were paired during sampling to minimize variation in habitat use, as this may vary temporally due to prey availability and weather conditions. At each sampling site, six monofilament mist nets (6 × 2.5 m or 9 × 2.5 m, denier 75/2, mesh 16 × 16 mm, fiveshelves – Ecotone, Inc., Poland) and two harp traps (two-bank 4.2 m2; Austbat Research Equipment, Victoria, Australia) were set across trails and spaced ∼50 m apart. Each site was sampled over three two-night surveys that were separated by at least 2 weeks. Mist nets were open between 19h00 and 23h00 and checked at 15-min intervals. Harp traps were operational between 19h00 and 06h00 at each site and were checked at 30-min intervals after dark until 23h00 and again at 06h00 the next day. Nights with rain and immediately before, during and after the full moon were not sampled to minimize potential bias (Saldana-Vázquez & Munguía-Rosas Reference Saldana-Vázquez and Munguía-Rosas2013). Except for subadults (Anthony Reference Anthony and Kunz1988), captured bats were marked using coloured and numbered plastic bands placed on the forearm (Handley et al. Reference Handley, Wilson and Gardner1991). A few bats were retained as voucher specimens and deposited at the National Museums of Kenya to both facilitate and document identifications; identifications and nomenclature followed Patterson & Webala (Reference Patterson and Webala2012). Insectivorous species were assigned to one of three foraging modes (ensembles) depending on habitat use (Denzinger & Schnitzler Reference Denzinger and Schnitzler2013, Kingston et al. Reference Kingston, Francis, Zubaid and Kunz2003, Schnitzler & Kalko Reference Schnitzler and Kalko2001). Forest-interior species were considered specialists because they forage exclusively in spatially complex environments (Marinello & Bernard Reference Marinello and Bernard2014). Non-echolocating frugivores were categorized in a separate ensemble.
Statistical analyses
Prior to analyses, all data were transformed [log (x + 1)] and tested for normality by the Shapiro–Wilk Test. Non-parametric tests were applied when transformations failed to render a variable both normal and homoscedastic. All means are presented ± SE. Two-way multivariate ANOVAs were used to test for differences in tree density, canopy cover, basal area and tree stumps among forest fragments and between capture locations (edges and interiors), with the four variables as dependent variables and forest fragments and capture locations as fixed factors. This was followed by post hoc pairwise comparisons of means using Tukey tests (Day & Quinn Reference Day and Quinn1989). Similarly, differences in number of charcoal kilns and footpaths between treatments were tested using the Kruskal–Wallis test and subsequent multiple comparison tests.
Bat assemblages across the three forest fragments and capture locations were described individually with Simpson’s index, D (Simpson Reference Simpson1949) and Pielou’s evenness: J’ (Pielou Reference Pielou1975). Simpson’s index was used because it provides a good estimate of diversity even for small sample sizes (Magurran Reference Magurran2004). Bat compositional similarity among forests was gauged by Bray–Curtis similarity using PRIMER software package (Clarke & Gorley Reference Clarke and Gorley2001).
We tested for differences between forest fragments and locations (edge, interior) in bat abundance (captures) of common species (n > 30) using two-way ANOVAs, with abundance as the dependent variable and forest fragment and location as fixed factors. Spearman’s rank correlation was used to relate bat abundance and the distribution measures (Gorresen & Willig Reference Gorresen and Willig2004, Sokal & Rohlf Reference Sokal and Rohlf1995). Both ANOVA and correlation analyses were conducted using Statistica v.7.0 (www.statsoft.com).
Direct interactions between bat abundance and vegetation/forest-use parameters were explored with Canonical Correspondence Analysis (CCA) in CANOCO 4.5 (Lepš & Šmilauer Reference Lepš and Šmilauer2003) and thereafter Monte–Carlo permutation tests (n = 999) performed to determine which vegetation/forest-use parameters significantly influenced bat distribution at P < 0.05, using conditional automatic forwarding options (Lepš & Šmilauer Reference Lepš and Šmilauer2003). We also tested for significance of the first three canonical axes. Highly correlated vegetation/forest-use parameters were not used in the CCA because highly correlated variables tend to cause redundancy in the set of explanatory variables (Lepš & Šmilauer Reference Lepš and Šmilauer2003).
Results
Vegetation and forest-use intensity characteristics
Vegetation and forest-use characteristics differed significantly between forest fragments: tree density (F 2,18 = 277, P < 0.001), canopy cover (F 2,18 = 554, P < 0.001), basal area (F 2,18 = 104, P = 0.001), cut tree stumps (F 2,18 = 57.6, P = 0.001), charcoal kilns (H 5,24 = 21.3, P = 0.05) and footpaths (H 5,24 = 19.0, P = 0.05). Tree density, canopy cover, basal area did not differ between Buyangu and Kisere forests (P > 0.05), but all were significantly higher than at Malava forest (P < 0.001). Conversely, cut tree stumps, charcoal kilns and footpaths did not differ between Buyangu and Kisere forests (P > 0.05), but were all significantly lower than at Malava forest (P < 0.001, Figure 2). Tree density (F 1,18 = 48.4, P < 0.001), canopy cover (F 1,18 = 46.5, P < 0.001) and basal area (F 1,18 = 7.42, P = 0.018) differed between edge and interior locations, with higher values of each in the interior. Canopy cover differed more at interior than edge locations, resulting in a significant fragment × location interaction (F 2,18 = 5.34, P = 0.022).
Figure 2. Differences in vegetation canopy cover (a), tree density (b), basal area (c) – and disturbance characteristics – charcoal kilns (d), cut tree stumps (e) and footpaths (f) – (units ha−1; untransformed mean ± SE) in three forest fragments (Buyangu forest = stripes, Malava forest = black, and Kisere forest = blue), and at forest edges and interiors. Different letters denote significant differences at P < 0.05.
Bat abundance responses to fragmentation
A total of 3456 mist-net h and 3168 harp-trap h yielded 4983 unique bat captures representing 26 species, eight families and all four foraging ensembles (Table 1). We recaptured 204 bats but excluded recaptures from analyses.
Table 1. F-values from two-way ANOVAs for effects of forest fragment and capture location on captures of common bat species at Kakamega Forest, western Kenya (n > 30), with captures as the dependent variable and forest fragment and capture location as fixed factors. Significant results are denoted by asterisks. *P < 0.05; **P < 0.01; ***P < 0.001
More captures were made at Buyangu forest, whereas the fewest were made at Malava forest (Appendix 1). Bat captures differed significantly between the forests (F 2, 18 = 352, P < 0.001), as well as between forest edges and interiors (F 1, 18 = 63.3, P < 0.001), with a significant forest × location interaction (F 2, 18 = 23.7, P < 0.001) (Table 1). Mean captures of bats did not differ (P > 0.05) at Buyangu (119 ± 37.2) and Kisere (57.4 ± 21.5) forests, but each differed significantly from Malava forest (14.8 ± 7.91; P < 0.001). At the edges, captures were significantly lower at Malava forest than at Buyangu and Kisere forests, with no difference between the latter two. Similarly, the interiors of the three forests differed significantly in bat captures (F 2, 18 = 3.62, P = 0.036), with fewer captures at Malava forest (14.2 ± 4.20) than at either Buyangu (69.9 ± 28.1) or Kisere forest (40.8 ± 4.69), which did not differ significantly (P > 0.05).
Frugivores made up 52% of all captures, although most (89%) were of two dominant species, Epomophorus wahlbergi (n = 1185) and Epomophorus labiatus (n = 1158). These two and Neoromicia capensis (n = 510) were widely distributed on the edges of all three forests. While captures of E. wahlbergi and N. capensis did not differ at the edges of the three forests (P > 0.05), captures of E. labiatus were significantly lower at the edges of Kisere (P < 0.05) than at the edges of either Buyangu and Malava forests, with no significant difference between the latter two (P = 0.943). Captures at Malava forest were dominated by the two disturbance-adapted frugivores, Epomophorus labiatus and E. wahlbergi. Two forest-interior insectivores, Kerivoula cuprosa (n = 504) and Hipposideros beatus (n = 435), were also fairly common, and were captured mainly in harp traps and exclusively in the interiors of the Buyangu and Kisere forests. Overall, forest-interior insectivores such as Doryrhina camerunensis, Hipposideros beatus, H. ruber, Nycteris arge, Glauconycteris humeralis, Hypsugo crassulus and K. cuprosa were captured more often in harp traps (203 ± 86.2) than in mist nets (11.8 ± 4.70; F 1,13 = 6.11, P < 0.05).
Further analyses of comparison of bat species composition between capture locations showed different degrees of similarities. The degree of similarity was high for the pairwise comparisons of edges (87.5%) and interiors (77.0%) of Buyangu versus Kisere forests as well as edges (70.6%) of Malava and Kisere forests. On the other hand, species similarities were far lower between Buyangu edges with interiors of Buyangu (19.1%) and Kisere (32.4%).
Assemblage-level responses to fragmentation
Buyangu and Kisere forest samples contained more than thrice (n = 26) and twice (n = 22) as many species as Malava forest (n = 8). The three forests and capture locations differed in both evenness (F 5,12 = 5.20, P < 0.05) and species diversity (F 5,12 = 11.2, P < 0.05). While the edges of the three forests did not differ significantly in species diversity (P > 0.05), post hoc tests showed that the interior of Kisere forest had significantly higher species diversity than the interiors of either Buyangu (P < 0.05) or Malava (P < 0.001); the latter two forests also differed significantly (P < 0.001).
Relationship of bats to vegetation and disturbance
Spearman rank correlation showed bat abundance, species richness and diversity were positively correlated with canopy cover, tree density and basal area, and negatively correlated with cut tree stumps, charcoal kilns and footpaths (Table 2). Permutation analyses of the CCA revealed that bats were significantly influenced by canopy cover (F = 11.7, P = 0.002), basal area (F = 10.3, P = 0.002) and tree density (F = 2.9, P = 0.04). As predicted, the CCA biplot ordered bat species into three major groups related to their forest utilization (Figure 3). The first group is comprised of the forest-interior insectivores Doryrhina camerunensis, Hipposideros beatus, H. ruber, Nycteris arge and Kerivoula cuprosa. These species were strongly associated with the less-disturbed interior sites of Buyangu and Kisere forests. The second group was associated with forest edges and included both frugivores (Eidolon helvum, Myonycteris angolensis, Hypsignathus monstrosus and Micropteropus pusillus) and edge-tolerant insectivores (Nycteris thebaica, Chaerephon major, Miniopterus inflatus, Glauconycteris argentata, Myotis bocagii, M. welwitschii, Neoromicia capensis, N. nana, Scotophilus dinganii and S. nux). The third group included the open-space insectivore Chaerephon pumilus and two frugivores, Epomophorus labiatus and E. wahlbergi, and was associated with heavily disturbed sites.
Table 2. Spearman’s rank correlation of bat abundance and distribution measures with vegetation characteristics and disturbances levels at Kakamega Forest, western Kenya. Significant correlation (r s) values at P = 0.05
Figure 3. Canonical Correspondence Analysis (CCA) triplots of bat species, vegetation and human-use characteristics at Kakamega Forest, western Kenya. The three major bat species groups ordered according to forest use are circled in different colours, with green representing forest-interior insectivores, blue (specialist frugivores and forest-edge insectivores) and red (generalist frugivores and open-space insectivores). Symbols represent forest fragments with open triangles representing Buyangu, filled triangles (Kisere) and open squares (Malava). In addition, the samples labels are in bold and italicized with the first two letters representing forest fragment: BU = Buyangu forest; KI = Kisere; MA = Malava whereas the next two letters represent sampling location, with ed = edge; in = interior. Also note some symbols have overlapped. Bat species names are in roman and italicized fonts and the first three letters represent genus and the next three letters represent species, with full names provided in Appendix 1. Vegetation variables that significantly influenced bat species were canopy cover, basal area and tree density.
Discussion
This is the first study of species-specific and assemblage-wide responses of bats to rain-forest fragmentation in Kenya. Fragmentation negatively affected bat abundance and species richness, in agreement with fragmentation studies elsewhere (Meyer et al. Reference Meyer, Struebig, Willig, Voigt and Kingston2016, Watling & Donnelly Reference Watling and Donnelly2006). The larger, better protected forests had higher bat abundances and species richness and lower human impacts than the smaller, more degraded forest (Malava), which was inhabited by generalists. Additionally, the larger forests supported interior-forest specialists. Interior-forest insectivores appear more vulnerable to forest fragmentation (Farneda et al. Reference Farneda, Rocha, López-Baucells, Groenenberg, Silva, Palmeirim, Bobrowiec and Meyer2015, Struebig et al. Reference Struebig, Kingston, Zubaid, Adnan, Nichols and Rossiter2008).
Impacts of forest fragmentation on bat abundance and species richness
At the assemblage level, bat abundance, species richness and diversity were all higher in the larger, less-disturbed Buyangu and Kisere forests than in Malava forest (Appendix 1). Previous studies have also demonstrated higher bat abundance and species richness in larger, more intact forests (Cosson et al. Reference Cosson, Pons and Masson1999, Struebig et al. Reference Struebig, Kingston, Petit, Le Comber, Zubaid, Mohd-Adnan and Rossiter2011).
Increased canopy cover, tree density and basal area were all correlated with the higher bat abundance at Buyangu and Kisere forests. Malava forest was not only the smallest and most isolated of the three forests, but it was also the most disturbed. This forest had higher numbers of charcoal kilns, cut tree stumps and footpaths, and lower values of tree density, basal area and canopy cover than either Buyangu or Kisere forests (Figure 2). Highly degraded forests often have an open and simplified structure, more clearings, higher solar penetration and fewer food resources (Lagan et al. Reference Lagan, Mannan and Matsubayashi2007), making them suboptimal for bats (Grindal & Brigham Reference Grindal and Brigham1998). The lower tree density, canopy cover and basal area at Malava forest could limit food and/or roosting resources for bats (Kunz & Lumsden Reference Kunz, Lumsden, Kunz and Fenton2003), and these are often the basis for their vulnerability (Fenton Reference Fenton1997). Roost sites, particularly for tree-cavity and foliage-roosting species that depend directly on the forest itself, may be less common at the highly degraded Malava forest.
Although our study concurs with many previous studies finding negative effects of fragmentation on bats (Farneda et al. Reference Farneda, Rocha, López-Baucells, Groenenberg, Silva, Palmeirim, Bobrowiec and Meyer2015, Meyer & Kalko Reference Meyer and Kalko2008, Struebig et al. Reference Struebig, Kingston, Zubaid, Adnan, Nichols and Rossiter2008), it contrasts with others that found little or no evidence of negative effects at the assemblage level (Faria Reference Faria2006, García-García et al. Reference García-García, Santos-Moreno and Kraker-Castañeda2014, Gorresen & Willig Reference Gorresen and Willig2004, Montiel et al. Reference Montiel, Estrada and Leon2006, Pardini et al. Reference Pardini, Faria, Accacio, Laps, Mariano, Paciencia, Dixo and Baumgarten2009). The area around Kakamega Forest is one of the most densely populated rural areas in the world, with over 570 people km−2 (Müller & Mburu Reference Müller and Mburu2009), ten times Kenya’s average (World Bank 2008). Dense human settlements and intense small-scale farming create a matrix in sharp contrast to the forests. Unsuitability of the matrix at Kakamega Forest may restrict movement across fragment boundaries, especially by forest-interior insectivores, and limit bat utilization of Malava forest (Meyer & Kalko Reference Meyer and Kalko2008, Meyer et al. Reference Meyer, Fründ, Pineda and Kalko2008, Struebig et al. Reference Struebig, Kingston, Zubaid, Adnan, Nichols and Rossiter2008). Buyangu and Kisere forests were not only better protected, but were separated by 2 km, compared to the 6–9 km isolation of Malava forest. The short distance likely mitigates against the adverse effects of the matrix, allowing even less vagile species to move between the two forests (Ewers & Didham Reference Ewers and Didham2006).
Impacts of fragmentation on individual species and ensembles
Species and ensembles differentially used the three forests. Usage reflects the area, isolation and degradation of fragments (Figure 2), as well as species-specific ecomorphology, foraging behaviour, and echolocation abilities (Denzinger & Schnitzler Reference Denzinger and Schnitzler2013, Schnitzler & Kalko Reference Schnitzler and Kalko2001, Schnitzler et al. Reference Schnitzler, Moss and Denzinger2003). Frugivores and open-space insectivores were more abundant at the edges of the larger forests, and the ubiquitous and disturbance-tolerant Epomophorus labiatus and E. wahlbergi (Cunto & Bernard Reference Cunto and Bernard2012, Meyer & Kalko Reference Meyer and Kalko2008) were the only frugivores recorded at the smaller, more disturbed Malava forest. The severity of anthropogenic disturbance at Malava forest probably explains the absence there of three other frugivorous bats (Hypsignathus monstrosus, Micropteropus pusillus, Myonycteris angolensis) that are fairly common in other fragments.
Other studies have found more frugivorous species at disturbed than in intact sites (Meyer & Kalko Reference Meyer and Kalko2008, Rocha et al. Reference Rocha, López-Baucells, Farneda, Groenenberg, Bobrowiec, Cabeza, Palmeirim and Meyer2016). However, unlike Kakamega’s frugivores, Neotropical frugivores can echolocate and are able to forage on early-successional plants, including shrubs (Thies et al. Reference Thies, Kalko and Schnitzler1998). Like frugivores, forest-edge and open-space insectivores were present in edges and disturbed sites at Malava forest, but we recorded higher abundance, species richness and diversity of these ensembles at the less-disturbed edges of Buyangu and Kisere forests. Small area and greater isolation might also contribute to the reduced abundance and species richness of frugivores, and forest-edge and open-space insectivores at Malava forest.
Five species of forest-interior insectivores (Doryrhina camerunensis, Hipposideros beatus, H. ruber, Nycteris arge and Kerivoula cuprosa) exhibit morphology and echolocation calls that identify them as narrow-space foragers (sensu Schnitzler & Kalko Reference Schnitzler and Kalko2001). Unlike frugivores and forest-edge and open-space insectivores, forest-interior species not only avoided the edges and open habitats, but were virtually absent at Malava forest.
Narrow-space-foraging insectivores are poorly adapted to long-distance flight by their wing morphology, echolocation call design and foraging behaviour (Aldridge & Rautenbach Reference Aldridge and Rautenbach1987, Neuweiler Reference Neuweiler1984, Norberg & Rayner Reference Norberg and Rayner1987). Some, like the hipposiderids (Doryrhina camerunensis, Hipposideros beatus, H. ruber), are clutter specialists (Norberg & Rayner Reference Norberg and Rayner1987), and are adapted for foraging close to or within dense vegetation (Kingston et al. Reference Kingston, Francis, Zubaid and Kunz2003). Their avoidance of Malava forest, intolerance of forest-edge conditions and dependence on the less-perturbed interiors of Kisere and Buyangu forests all suggest the vulnerability of these forest-interior specialists (Jones et al. Reference Jones, Purvis and Gittleman2003, Lane et al. Reference Lane, Kingston and Lee2006, Safi & Kerth Reference Safi and Kerth2004). Other studies have also demonstrated the greater response of forest-interior species to habitat fragmentation and degradation (Estrada-Villegas et al. Reference Estrada-Villegas, Meyer and Kalko2010, Farneda et al. Reference Farneda, Rocha, López-Baucells, Groenenberg, Silva, Palmeirim, Bobrowiec and Meyer2015, Meyer & Kalko Reference Meyer and Kalko2008, Struebig et al. Reference Struebig, Kingston, Zubaid, Adnan, Nichols and Rossiter2008, Reference Struebig, Kingston, Zubaid, Lecomber, Adnan, Turner, Kelly, Bozek and Rossiter2009).
Edge effects
Edge effects were most pronounced at the small, highly disturbed Malava forest, where the edges and interiors differed modestly in canopy cover and tree density and had similar levels of disturbance. Only two species of frugivores (E. labiatus, E. wahlbergi) were present, and these generalists used forest edges and interiors indiscriminately. Generalist frugivores and forest- edge and open-space insectivores were typically absent from the interiors of less-disturbed forests, where forest-interior insectivores predominated. Like cavity-roosting species in Peninsular Malaysia (Struebig et al. Reference Struebig, Kingston, Zubaid, Adnan, Nichols and Rossiter2008) and gleaning animalivorous bats in the Neotropics (Farneda et al. Reference Farneda, Rocha, López-Baucells, Groenenberg, Silva, Palmeirim, Bobrowiec and Meyer2015, Gorresen & Willig Reference Gorresen and Willig2004, Henry et al. Reference Henry, Cosson and Pons2010, Klingbeil & Willig Reference Klingbeil and Willig2009, Reference Klingbeil and Willig2010; Meyer & Kalko Reference Meyer and Kalko2008, Pardini et al. Reference Pardini, Faria, Accacio, Laps, Mariano, Paciencia, Dixo and Baumgarten2009, Rocha et al. Reference Rocha, López-Baucells, Farneda, Groenenberg, Bobrowiec, Cabeza, Palmeirim and Meyer2016), forest-interior insectivores appear highly susceptible to edge effects. Besides eliminating suitable roosting structures, edges cause tree mortality and open the canopy, increasing desiccation stress, windshear and wind turbulence (Laurance et al. Reference Laurance, Laurance, Ferreira, Rankin-De Merona, Gascon and Lovejoy1997), ultimately modifying forest composition (Laurance et al. Reference Laurance, Perez-Salicrup, Delamonica, Fearnside, D’Angelo, Jerozolinski, Pohl and Lovejoy2001, Reference Laurance, Nascimento, Laurance, Andrade, Ribeiro, Giraldo, Lovejoy, Condit, Chave and D’Angelo2006).
Caveats
We used both ground-level mist nets and harp traps to sample Kakamega’s diverse bat fauna. The absence of canopy-level mist nets may have caused us to overlook some high-flying species also found there (Meyer et al. Reference Meyer, Aguiar, Aguirre, Baumgarten, Clarke, Cosson, Villegas, Fahr, Faria, Furey, Henry, Hodgkison, Jenkins, Jung, Kingston, Kunz, Gonzalez, Moya, Patterson, Pons, Racey, Rex, Sampaio, Solari, Stoner, Voigt, Staden, Weise and Kalko2011). In addition, many of the insectivorous bats inhabiting Kakamega forest commonly use their echolocation calls to evade capture and might be better sampled via acoustic monitoring (MacSwiney et al. Reference MacSwiney, Clarke and Racey2008, O’Farrell & Gannon Reference O’Farrell and Gannon1999). However, acoustic monitoring requires an exhaustive and corroborated call library. During the course of this fieldwork, we assembled a partial call library that can be used in future surveys (Webala et al. Reference Webala, Rydell, Dick, Musila and Patterson2019). Even though we might have failed to capture some of Kakamega’s bats, the method used is repeatable and would have introduced no systematic bias (Meyer et al. Reference Meyer, Aguiar, Aguirre, Baumgarten, Clarke, Cosson, Estrada-Villegas, Fahr, Faria, Furey, Henry, Jenkins, Kunz, MacSwiney Gonzalez, Moya, Pons, Racey, Rex, Sampaio, Stoner, Voigt, Von Staden, Weise and Kalko2015).
We also focused sampling on the fragments themselves rather than on the bats’ use of the surrounding matrix. There are virtually no data on the home ranges or foraging behaviour of the bat species documented by our work. Radio-tracking studies will be needed to elucidate matrix effects on the bats’ space use and whether individual bats are able to treat the Kakamega fragments as foraging patches.
Conclusions
Bat species and ensembles at Kakamega forest respond to fragmentation in predictable ways, likely determined by differences in their foraging ecology, wing morphology and movement behaviour (Klingbeil & Willig Reference Klingbeil and Willig2009). As expected, generalist frugivores and forest-edge and open-space insectivores predominated at degraded sites because of their tolerance of a range of habitats. Remarkably, these disturbance-tolerant species nevertheless preferred the edges of less-disturbed forests. Forest-interior insectivores, on the other hand, appear vulnerable to habitat fragmentation and degradation because they were confined to habitats within the interiors of less-disturbed forests. Continued logging, charcoal production, livestock grazing or altered disturbance regimes that modify the quality of habitats within fragments at Kakamega Forest are likely to adversely impact forest-interior specialists. Clearly, the protection and maintenance of undisturbed forest interiors is crucial for these species (Andrén Reference Andrén1994, Metzger & Décamps Reference Metzger and Décamps1997, Meyer & Kalko Reference Meyer and Kalko2008).
Acknowledgements
This project was conducted with Kenya Wildlife Service (KWS) and Kenya Forest Service (KFS) Permits KWS/4001 and RESEA/1/KFS/75, respectively. We thank Geoffrey M. Wambugu of Karatina University for digitizing the map of Kakamega Forest. Beryl Makori, Dedan Ngatia and Simon Masika (Karatina University), and Michael Bartonjo and Sospeter Kibiwot (National Museums of Kenya) provided invaluable help in the field. This study would not have been possible without the collaboration of staff of KWS and KFS. We especially thank Dr Samuel Kasiki and Mr James Mwang’ombe, from KWS and KFS, respectively, for facilitating access and research permits. The support and confidence of Bud and Onnolee Trapp and Walt and Ellen Newsom were instrumental to the effective execution of this work. PWW thanks Dr David Jacobs for funding his 3-month stay at the University of Cape Town for writing up the manuscript. We thank Ricardo Rocha and Adrià López Baucells for highly beneficial comments on an earlier version of this manuscript.
Financial support
Financial support was provided by grants from the British Ecological Society (BES 3571/4375), International Foundation of Science (IFS D/5278-1), Kenya’s NACOSTI, and Field Museum’s Council on Africa to PWW, the Barbara E. Brown Fund for Mammal Research to BDP, and the JRS Biodiversity Foundation to both lead and senior authors.
Appendix 1
Bat captures, species richness, evenness and diversity at Kakamega Forest, western Kenya. For ensembles: F, frugivore; FI, forest interior; FE, forest edge; OS, open space
