http://orcid.org/0000-0001-6635-7027
Figures
Abstract
Shade coffee has shown great promise in providing crucial habitats for biodiversity outside formal protected areas. Insectivorous bats have been understudied in coffee, although they may provide pest control services. We investigated the influence of local and landscape-level features of coffee farms on aerial insectivorous bats in Chikmagalur district in the Western Ghats biodiversity hotspot, India. Bats were monitored in 20 farm sites using ultrasound detectors, and the response of bat species richness and activity to changes in tree density, proportion of built-up area in the neighborhood, and distance of farm from forest areas quantified. We examined if models built to explain the species richness and activity could also predict them in nine additional sites. We detected nine phonic types/species in the study area. The quantified predictors had no effect on assemblage-level species richness and activity of bats. Responses of edge-space and cluttered-space forager guilds mirrored those of the overall assemblage, but some species vulnerable to forest conversion like Rhinolophus beddomei were detected rarely. Best models explained up to 20% and 15% variation in assemblage-level species richness and activity respectively, and were poor predictors of both response variables. We conclude that coffee farms in our study area offer an important commuting space for insectivorous bats across a gradient of shade management. Further research should include species-specific responses to management decisions for at-risk species and quantification of ecosystem services like natural pest control to inform biodiversity conservation initiatives in the Western Ghats coffee landscapes.
Citation: Ongole S, Sankaran M, Karanth KK (2018) Responses of aerial insectivorous bats to local and landscape-level features of coffee agroforestry systems in Western Ghats, India. PLoS ONE 13(8): e0201648. https://doi.org/10.1371/journal.pone.0201648
Editor: Shijo Joseph, Kerala Forest Research Institute, INDIA
Received: February 12, 2018; Accepted: July 19, 2018; Published: August 16, 2018
Copyright: © 2018 Ongole et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper, its Supporting Information files and the Dryad Digital Repository at DOI: 10.5061/dryad.tp14201.
Funding: In 2015-16, SO was funded by Tata Trust and core funding to the National Centre for Biological Sciences - TIFR from the Department of Atomic Energy, Government of India. In 2016-17, SO was supported by funding from Oracle Corporation. MS was supported by core funds from the National Centre for Biological Sciences during the study. KKK was funded by Oracle Corporation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors declare that this research was conducted in the absence of any relationship (including but not restricted to paid employment/consultancy, patents, products in development, marketed products) with Oracle Corporation that could be construed as a potential conflict of interest. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
Introduction
Agricultural landscapes, which occupy almost 40% of the planet’s ice-free land area [1] are much more geographically extensive than protected areas. Consequently, they hold immense potential for biodiversity conservation [2–6]. Agroforestry systems—the intentional management of shade trees with agricultural crops—have been shown to be particularly important in this regard due to their structural similarity to forests and significant vegetation diversity [5]. Indeed, studies have shown that in some cases, species richness and composition of many taxa in agroforestry systems are comparable to neighboring forests [7]. Understanding the characteristics of agroforestry systems that allow species to use them, including specific management practices within farms and the landscape context surrounding the farm can therefore inform management strategies that aim to conserve biodiversity outside of protected areas [8]. This understanding can also enhance the provisioning of ecosystem services like biological pest control and pollination [9,10] in these landscapes.
Among agroforestry systems, coffee is the most widespread (grown in over 50 tropical countries) and economically important crop, with a retail value of ~$US90 billion per annum [11,12]. Several studies show that shade-grown coffee holds significant ecological value for various taxa including birds, insects and mammals [3,13–18]. Local-level features of coffee farms, such as type and density of shade trees, and canopy cover have been shown to influence biodiversity in them [8,19,20]. In addition, landscape characteristics surrounding the farms (such as presence of natural forests) have also been shown to be important for biodiversity in the farms [21,22].
The response of insectivorous bats to the local and landscape characteristics in coffee farms remains unclear, even though they hold immense potential for limiting pests in them [10,23]. Research on the impact of land conversion to agriculture or agricultural intensification on bats has largely emerged from the Neotropics [22,24,25] with few Paleotropical studies. [17,26–28]. However, in the last few decades, coffee production has increased in many regions of the Paleotropics with an associated decline in shade tree diversity [29]. At present, the impacts of such changes on Paleotropical bat communities remains poorly understood.
Here, we examine the relative roles of local and landscape-level features in influencing Paleotropical aerial insectivorous bats in coffee agroforests in India’s Western Ghats biodiversity hotspot. Specifically, we asked how local features, such as shade tree density, and landscape features of coffee farms, such as distance to natural forest and proportion of built-up area (buildings) surrounding the farms influence the overall species richness and activity of bats (assemblage-level). We also quantified the responses of different foraging guilds of insectivorous bats—cluttered-space foragers, edge space foragers and open space foragers [30] in these farms (ensemble-level). These ensembles are likely to show different responses because (i) their echolocation call design results in varying ability to forage in cluttered vegetation (ii) their wing morphologies, which influences their maneuverability while capturing insect prey, also result in differing levels of flight efficiencies when commuting over the landscape matrix [31]. In addition, we also examined how species composition of the assemblage was related to the local and landscape-level features. We expected cluttered space forager richness and activity levels to decrease with increasing distance from forest due to their low flight efficiencies and show no response to increasing vegetation clutter at the local level. The response to distance from forest may be weak if the coffee matrix offers low contrast to the forest. We expected species richness and activity levels of edge and open space foragers to increase or show no response to increasing distance from forest. We also expected increasing vegetation clutter at the local level to negatively influence the edge and open space foragers.
Methods
Study area and sampled farms
We surveyed bats in Chikmagalur district (13°18–13°21´ N and 75°31´-75°48´ E), a significant coffee growing region in the state of Karnataka, India (Fig 1). This region is located within the Western Ghats and Sri Lanka Biodiversity Hotspot [32]. Besides shade coffee cultivation, the study area is comprised of Bhadra Wildlife Sanctuary, a 493 km2 tropical moist deciduous forest [33], several reserved and revenue forests [19], agriculture fields and human habitations. Grown at altitudes between 800m and 1500m, coffee cultivation (Arabica variety, Coffea Arabica, Robusta variety, C.canephora) covers an area of over 870 km2 in the district and account for ~25% of all coffee produced in India [34]. More than 90% of these comprise small farms with areas <0.1 km2. Canopy is largely closed overall with canopy density ranging between 50 and 99 percent [8]. Shade management in farms can lead to changes in tree density and foliage cover. The shade layer can also vary from being dominated by native tree species (e.g. Ficus glomerata, Artocarpus heterophyllus and Erythrina indica) to one dominated by an introduced species from Australia called silver oak (Grevillea robusta). At the landscape level, large expanses of monoculture shade are not found as coffee grown under exotic shade is interspersed with that grown under native shade. Even at the local level, exotic shade tree stands are in close proximity to native species. In a pilot study, the percentage of native shade tree species in the study area farms ranged from 24% to 97% (n = 20, mean 50%, sd 21%). Following a scheme based on estimates of shade tree species richness, tree density and shade cover used to classify coffee farms in the Neotropics along an intensification gradient [35], most farms in this region can be classified as low-management coffee.
Locations of sampled coffee farms, surrounding forests, built-up and open areas in Chikmagalur district of Western Ghats, India.
Bat surveys
We surveyed bats in 20 coffee farm sites (mean size 5.1 ha, size range from 3.6 to 6.5 ha) in the dry season between November 2015 and April 2016, representing a gradient of shade tree density and located at varying distances from natural forest patches (Fig 1). No wildlife permits were required to monitor bats on these privately owned farms. We obtained verbal permissions from all land owners/managers to sample on their lands. Since we used completely non-invasive acoustic methods to monitor free-flying bats, we did not require approval from an animal ethics committee. The elevations of the sampled sites ranged from 748m to 1267m, with only four sites falling below 1000m. At each site, we acoustically sampled for bats using two Pettersson D240X time expansion bat detectors recording on to Edirol R-09 digital recorders. To account for horizontal spatial variability in bat activity, we sampled at multiple points separated by 100m [36] along narrow walking trails between coffee bushes with three visits to each point across the study duration. However, we recognize that estimates of activity can vary depending on the spatial positioning of detectors even at close distances [37]. Each point in a site was sampled for 15 min in one night between sunset and 22:00 as activity reduced considerably after that. The two detectors were used simultaneously at different points and rotated to the next set of points until all points within the site were exhausted. While sampling for the entire night is a requirement to estimate activity accurately [38], we could not do so due to personnel limitation and note that as a limitation of our study. Detectors were pointed into canopy gaps to maximize detections. Each point in each site was re-surveyed between November 2016 and April 2017, to assess temporal turnover at the assemblage-level. We also sampled an additional 9 sites in 2016–17 (ranging in size from 4 ha to 40 ha) each of which was visited five times during the study period to evaluate models fit with earlier data.
Characterization of local level vegetation
At each site, we estimated shade tree density using the variable-area transect (VAT) method [39](S1 Fig). We first established multiple 40m baseline transects at each site. Each transect was then divided into four 10m long sub-transects. On either side of these sub-transects, we searched for five shade trees (> = 30 cm GBH) in a rectangular cell of length 10m and width up to 20m. If five trees were found within a distance of 20m, the width was calculated as the distance to the fifth farthest tree. If there were fewer than five trees within 20m, the width was taken to be 20m. Tree density estimates were obtained by dividing the total number of trees counted by the total area of the cells, averaged across all transects. We also estimated a measure of vertical foliage cover by counting the number of height classes from among 2 height classes (between the coffee layer and 10m, 10-20m) that contained leafy vegetation within an imaginary cylinder of 0.5 m radius above the observer, at 20 random points on either side of a transect [19].
Characterization of landscape level features
Distance of coffee sites from natural forest was estimated using the ISRO Bhuvan Land Use/Land Cover classification layer (1:50000, 2011–2012) of Chikmagalur district (http://bhuvan5.nrsc.gov.in/bhuvan/wms). As the LULC layer was from an earlier time period, we first visually compared the Bhuvan layer with current satellite imagery from Google maps (2017) to determine if forest cover had significantly changed in the last few years. We restricted our analysis to forest patches larger than 1.8 km2 as they matched up between the two time periods.
Proportion of builtup area was measured using circular buffers at four radii (100m, 200m, 500m and 1000m) from the centre of each site, to account for the possibility that different bats respond to the landscape at different scales. Radii larger than 1000m were not investigated to avoid overlap between buffers of adjacent sites [40–42]. Overlapping buffers leads to statistical non-independence of observations as the same predictor values are used multiple times in the data set [41]). We used QGIS 2.18.10 [43] to estimate distance to forest areas and Microsoft Bing maps to digitize built-up areas.
Acoustic data analysis
We used BatSound Pro 3.32 (www.batsound.com) to create spectrograms of the calls (setting for spectrogram: sampling rate 44.1 kHz for time-expanded calls, FFT size = Automatic, FFT Overlap = -1, FFT window = Hanning). Using the power spectrum option, only those with intensity greater than or equal to 20dB above the background noise were used for extracting call structure information (setting for power spectrum: FFT size = Automatic, FFT window = Hanning, Min level = -120 dB). For each call, we extracted frequency of maximum energy (FMAXE), end frequency (EF), start frequency (SF) and call duration (duration). Published echolocation call libraries from the Western Ghats region [44,45] were used to identify bats. Where identification was not possible, we assigned species to phonic types based on similarity of call structure and FMAXE. We classified the detected species/phonic types into guilds based on Denzinger and Schnitzler [46] (2013) by visual inspection of echolocation calls. We labeled species with the following characteristics as cluttered space foragers (i) CF-FM (Constant Frequency-Frequency Modulated) echolocation calls (ii) Broadband calls with low intensity typical of two species expected from the region, Megaderma spasma and M.lyra [47]. We classified species with calls containing a steep, downward frequency modulated component (FM) followed by a shallow, narrowband component (QCF) as edge space foragers. We did not expect to detect edge space trawling foragers as there were no streams and water bodies in the sampling sites. Long, shallow and narrowband echolocation calls were classified as open space foragers.
Statistical analyses
Bat species richness estimation.
We estimated species richness at each site using the software SPECRICH2 (https://www.mbr-pwrc.usgs.gov/software/specrich2.shtml). The software accounts for heterogeneity of detection probabilities among species [48] using the jack knife estimator first developed for capture recapture analysis of closed populations [49].
Bat activity estimation.
We indexed bat abundance based on bat activity as the proportion of the total number of minutes sampled in an estate that yielded any bat call [50]. In order to determine whether detectability of the bat assemblage was similar across sites, we compared average detectability across sites (S1 Dataset). This is a key requirement for comparing activity levels across sites.
Model construction.
At each scale, we used multiple linear regressions to model the response of assemblage-level bat species richness and activity from 2015–16. We did not expect elevation to have an effect on the bats as the maximum difference in elevation between sites was only 250m. We made pair-wise comparisons between the predictor variables, and only those not highly correlated (r<0.7) were used for analysis. The final set of predictor variables contained tree density, distance from natural forest and proportion of built-up area. To assess whether there is spatial autocorrelation in estimated species richness and activity of bats, we plotted Moran’s I correlograms. For each spatial scale, the candidate model set consisted of all possible combinations of predictors included in a particular model and a model also containing an interaction term between proportion of built-up area and distance from forest for a total of eight models. We included this interaction term as we hypothesized that artificial roosting sites may become increasingly important in agricultural landscapes as distance from natural forests increase. In each model, we standardized each input variable by centring it and then standardizing the result by 2*standard deviations [51]. Centring improves the interpretability of input variables even when interaction terms are present, while standardization makes comparison of effect sizes of multiple input variables possible by placing them on a common scale [52]. Exploratory analysis indicated that the proportion of built-up area for one particular site was an extreme value (0.10) due to its proximity to a large town (median value without the outlier 0.0062). We excluded this observation from our analyses, because this is not representative of a typical coffee farm in the study area. At the ensemble-level, we did not conduct multiple regression analyses for species richness as the data were insufficient. To assess if activity differed at the ensemble-level, we plotted guild-wise activity levels with the predictor variables.
For each response variable at each scale, we followed a model selection approach based on small sample-corrected Akaike’s Information Criterion (AICc) to select the best model (ΔAICc<2) [53]. When there were multiple best models, we estimated model averaged regression coefficients of predictors using the zero method, i.e. if a given predictor is absent from a model, a parameter estimate and error of zero is substituted in the model and the model-averaged regression coefficient for that predictor is estimated by averaging over all best models[54]. We only report the results for the spatial scale of 1000m as results are similar at all investigated scales.
We used data from the nine additional sites sampled in 2016–17 to assess if the best models (or model-averaged results) developed for the 2015–16 data predicted species richness and activity for the 2016–17 data well. We obtained the predicted values of species richness and activity based on standardized and centered values of the predictor variables for these nine additional coffee sites [51], and calculated confidence interval bounds for the predicted species richness and activity by substituting the lower and upper confidence limits for the parameter estimates in the model [55].
Species-environment associations
Finally, we used canonical correspondence analysis (CCA) to examine how bat species composition at the assemblage-level changed across sites in relation to changes in predictor variables [56]. We also tested for the significance of these associations using the mock ANOVA function in the vegan Package [57]. In addition, we assessed the degree of species turnover in the 20 sites that were sampled in multiple years. We used the software COMDYN [58] to estimate the species turnover parameter GAMMA, defined as the proportion of sample 2 species present in sample 1. COMDYN uses capture-recapture models for closed animal populations that consider heterogeneous detection probabilities of species in the community [58].
All analyses except species richness estimation and turnover were carried out using R 3.4.2 [59].
Results
Our sites were located at a wide range of distances from natural forests (93–8258 m, mean 1667 m, sd 2246 m). Tree density in sites ranged from 163 trees/ha to 969 trees/ha (mean 482.9, sd 250.9), while the proportion of built up area around sites ranged from 0.006 to 0.1, with the largest value being at a site at very close proximity to a town (mean 0.009, sd 0.018).
Across all the 29 sites, we distinguished nine species/phonic types during the study period (Table 1, S2 Fig). Of the nine, four belonging to the families Rhinolophidae and Hipposideridae (Rhinolophus lepidus, Rhinolophus indorouxii, Rhinolophus beddomei, Hipposideros sp., an unknown Rhinolophid) were labelled as cluttered space foragers. The remaining five (P22, P30_40, P50 and P60), which were not identified to species level but most likely belonged to the family Vespertilionidae based on the echolocation call shape, were classified as edge space foragers. We did not detect any cluttered space passive gleaning and open space foragers.
Bat species richness
Estimates of assemblage-level bat species richness for the 20 sites sampled in 2015–16 ranged between 3 and 7 (mean 4.8, sd 1.12). Moran’s I correlogram of estimated species richness did not indicate spatial autocorrelation (S3 Fig). Our model selection exercise indicated comparable support for three models explaining bat species richness. The amount of variance explained by the models was up to 20% and there was no statistically significant effect of any of the considered local- and landscape-level variables on species richness (Table 2).
Bat activity
Bat activity levels at the assemblage-level for the 20 sites ranged from 0.02 to 0.65 (mean 0.29, sd 0.18). There was no indication of spatial autocorrelation from the Moran’s I correlogram (S3 Fig). Through model selection, we found comparable support for three models explaining bat activity. The amount of variance explained was up to 15% and none of the predictor variables were statistically significant (Table 2). Ensemble-level activity also showed no apparent relationship with the predictor variables (Figs 2 and 3).
Y axis is proportion.
Y axis is proportion.
Species-environment associations
The correspondence analysis identified four clusters of sites that could be visually discriminated (Fig 4). Cluster 1 comprised two sites that were characterized by high tree densities and the presence of Rhinolopus beddomei (R.bedd). Clusters 2 (5 sites) and 3 (a single site) occurred closer to natural forest patches, and were characterized by the presence of Rhinolophus indorouxii (R.indo) and a Hipposideros sp., respectively, which were detected rarely. The remaining sites (cluster 4) were characterized by a higher proportion of built-up area in the neighborhood. Distance of site to the nearest natural forest edge was the most important environmental variable discriminating sites, followed by tree density and proportion of built-up area. However, the correspondence analysis did not reveal any statistically significant relationship between species and the environmental variables considered (p = 0.31; Fig 4).
Species codes are in Table 1. Length of the arrow denotes the strength of the association between the species and the corresponding predictor variable. Direction of the arrow denotes positive association.
Evaluation of modeled relationships
The models developed poorly predicted assemblage-level bat species richness and activity for the additional 9 sites we sampled in 2016–17. The correlation coefficient between predicted and estimated species richness was 0.23, and for bat activity was 0.19. We obtained similar estimates of species richness across years for the 20 sites that were sampled in both years (range 3 to 7 [mean 4.8, sd 1.12] in 2015–16; range 3 to 8 [mean 5.65, sd 1.26] in 2016–17). Species turnover was typically low, with ~85% of the species sampled in a site in 2015–16 also sampled in 2016–17.
Discussion
Effect of predictors on species richness
At the site level, we found that tree density had no effect on assemblage-level bat species richness. A potential explanation is that the tree density gradient was not a deterrent for the species detected in this study. Since sun coffee is not grown in the study region, it is difficult to obtain a drastic gradient of tree density, especially at lower end. In a Neotropical study, Bader et al. [60] showed that presence of aerial insectivorous bats was best predicted by canopy cover and canopy height at the local level. Canopy height is largely uniform in the study region. Canopy cover could have led to the observed patterns but we lacked the sample sizes to perform analyses.
At the landscape level, the lack of a clear effect of distance from natural forest areas on species richness is possibly a result of the high tree cover offered by shade coffee in the study area (see Fig 1). Vegetation in the form of trees or hedges can offer increased connectivity in agricultural landscapes for several bat species to exploit resources [61]. In Costa Rica, Mendenhall et al. [22] found that the species richness of bats was similar across forest fragments set in a matrix of coffee farms, concluding that a countryside ecosystem could reduce the contrast between forest areas. No effect of built-up area on species richness likely indicates that the extent of man-made structures in the landscape, while being used by bats, may not be enough to negatively affect species at multiple scales.
Effect of predictors on bat activity
At the local-level, we found that both assemblage and ensemble-level activity were unaffected by shade tree density. Despite constraints based on wing morphology and echolocation, insectivorous bats are highly flexible [62] and demonstrate variable responses to structural changes in their habitat [63,64]. Edge space foragers may show behavioral flexibility through the use of trails in areas with high clutter levels. For example, Law & Chidel [65] found that a number of bats species sensitive to vegetation clutter use forest tracks and riparian areas as flyways. As expected, we found that clutter-tolerant species did not respond to increase in tree density in the sites. However, we remain cautious about this pattern because the cluttered-space forager activity is dominated by only two of the four species we detected. Rhinolophus beddomei (R.bedd) and Hipposideros sp. (Hipp) were detected so rarely that it is difficult to ascertain how these species respond to shade management in this region. This supports conclusions of two recent studies from the Western Ghats [17,28] that species with constant frequency echolocation calls like R.beddomei are at high risk from agricultural intensification.
Insect prey availability at the site level, which was not included as a predictor variable, may also have played a role in the observed patterns [42,66]. Studies show that changes in insect abundance in open areas affect bat activity more strongly than in dense areas [67,68]. This is because prey density is inherently low in open areas and bats adapted to foraging in open spaces respond strongly to any changes in prey density [68]. Since coffee in the study area is grown under a fairly dense canopy (see Methods), insect abundance is likely to have a weak effect on bat activity. Insect abundance can also affect bat activity through its interaction with agricultural practices such as agrochemical use [69]. Organic farming has been shown to increase insect abundance and positively affect bat activity [69]. Alternatively, Poocock & Jennings [70] found bats to be more sensitive to field boundary loss than increased agrochemical inputs. During the time period of our study, coffee is harvested in the region and agrochemical use is minimal. However, pesticides are used through growing season and residual effects of agrochemicals might have impacted the activity of bats.
Increasing distance to forest areas did not affect assemblage and ensemble-level activity of bats. In tree-covered or low-management agricultural landscapes such as this, forest/woodland patches may be less influential for insectivorous bats when compared to homogeneous, intensively managed agriculture [71,72].
Species-environment associations
Drawing conclusions from species richness alone could mask important trends in compositional changes in bat communities [73]. Through correspondence analysis to elucidate species-environment relationships, we found that most of the species in the study area were not strongly related to the predictor variables and were distributed similarly across the study area. Another explanation for the observed effect of distance from forests could be a limited range of distance values which bats in our study are able to cover. Dispersal distances of bats in Western Ghats are currently unknown [17] but we nevertheless speculate that the low contrast nature of the coffee farms is an important driver of the observed pattern.
Evaluation of modeled relationships
Our models did not lend to highly accurate predictions of species richness and activity in the new sites. Perhaps this is not very surprising because the models contained non-significant predictors. We also believe that predicting bat activity accurately is a bigger challenge as it may be influenced by other ecological factors such as food availability, presence of roosting sites and high temporal variability of bat activity.
We also found that the complement of species turnover in the sampled sites across the two years was high on an average but we speculate that the low precision was due to the species-poor nature of the assemblage. Wordley [74] also found no annual variation in bat communities in a similar landscape consisting of coffee, tea, forest fragments.
Conclusions and future research
Our study shows no response of species richness and activity of a Paleotropical bats at both assemblage and ensemble-level to current local shade management practices (characterized by shade tree density and foliage cover) and landscape-level factors in a coffee agroforestry system. However, impacts on particular species might be missed in a low contrast landscape when community level metrics are used as response variables [75]. This is hinted at by our finding that species such as Rhinolophus beddomei and Hipposideros sp. appear to be rare in our study region. Further research should investigate species-environment relationships in this landscape; especially for such rare species, which may also be vulnerable. Complementary methods like mist netting must also be employed in the future to gather morphological information which has been shown to predict the vulnerability of insectivorous bats to anthropogenic changes [60].
Given that local features on a farm are more easily manipulated than landscape level factors by farmers, we must caution that these neutral responses should not be a reason for large-scale reduction of tree density, as a heterogenous vegetation structure benefits several bat species, particularly ones that have small home ranges [72]. However, farmer decisions in agroforestry systems are subject to non-biological considerations such as commodity price fluctuations, input and labor costs [2] which may be detrimental to biodiversity conservation. To this end, future investigations should quantify levels of biological pest control services across gradients of shade tree densities and other local-level features through manipulative experiments, so that farmers can balance these ecosystems services against economic factors in making decisions that could impact biological diversity.
Supporting information
S1 Fig. Schematic of the Variable Area Transect (VAT) method.
The line in the centre is the baseline transect, the rectangles on either side are the cells, open circles are trees. The width of the cells (d1-d8) is the distance of the fifth farthest tree in that cell. The length of each cell is 10m.
https://doi.org/10.1371/journal.pone.0201648.s001
(TIF)
S2 Fig. Representative call pulses of bats detected in the study.
A time expansion detector was used for acoustic monitoring. Therefore, for actual values of time and frequency, divide values of time on x-axis by a factor of 10 and multiply values of frequency on y-axis by a factor of 10.
https://doi.org/10.1371/journal.pone.0201648.s002
(TIFF)
S3 Fig. Spatial correlograms of estimated species richness and activity.
Error bars denote standard deviations.
https://doi.org/10.1371/journal.pone.0201648.s003
(TIFF)
S1 Dataset. Acoustic recording data and predictor variables in Chikmagalur district.
Observed and estimated species richness, total and guildwise bat activity, predictor variables at each of the 20 sites sampled in 2015–16. Predicted and estimated species richness, predicted and observed total activity, predictor variables at each of the 9 sites sampled in 2016–17.
https://doi.org/10.1371/journal.pone.0201648.s004
(XLSX)
Acknowledgments
We are grateful to the Wildlife Conservation Society, Centre for Wildlife Studies and National Centre for Biological Sciences (NCBS, TIFR), for logistical support. We thank Yashwanth, Anil, Fayaz, Aamir, Spoorthi, Ankit, Parvathi, Anisha and Sharath who provided field assistance at various stages. MO Anand, Suhel Quader, Kavita Isvaran, Devcharan Jathanna, Ajith Kumar, Jayashree Ratnam and Claire Wordley provided insights during study design, implementation and analysis. We thank Danilo Russo, Shijo Joseph and two anonymous reviewers for critical feedback on the manuscript. We thank D.V Girish and Shreedev Hulikere for logistical support during fieldwork and all the coffee planters—Sharan, Steve, Rajagopal, Philomena, Aveen, Sripal, Asim, Jayakumar, Dattatreya, Harshith, Krishnadev, Pradeep, Wesley, Shashidhar, Aravind, Arun, Madhu, Aranha, Gurunath and Satish who provided their support and cooperation.
References
- 1. Ramankutty N, Evan AT, Monfreda C, Foley JA. Farming the planet: 1. Geographic distribution of global agricultural lands in the year 2000. Global Biogeochem Cycles. 2008;22(1):1–19.
- 2. Robbins P, Chhatre A, Karanth K, Knight A. Political Ecology of Commodity Agroforests and Tropical Biodiversity. Conserv Lett. 2015;8(April):77–85.
- 3. Daily GC, Ceballos G, Pacheco J, Suzan G, Sanchez-Azofeifa A. Countryside Biogeography of Neotropical Mammals : Conservation Opportunities in Agricultural Landscapes of Costa Rica. Conserv Biol. 2003;17(6):1814–26.
- 4. Beukema H, Danielsen F, Vincent G, Hardiwinoto S, Van Andel J. Plant and bird diversity in rubber agroforests in the lowlands of Sumatra, Indonesia. Agrofor Syst. 2007;70(3):217–42.
- 5. Harvey CA, Villalobos JAG. Agroforestry systems conserve species-rich but modified assemblages of tropical birds and bats. Biodivers Conserv. 2007;16:2257–92.
- 6. Schroth G, Harvey CA. Biodiversity conservation in cocoa production landscapes: an overview. Biodivers Conserv. 2007;16:2237–44.
- 7. Bhagwat SA, Willis KJ, Birks HJB, Whittaker RJ. Agroforestry : a refuge for tropical biodiversity ? Trends Ecol Evol. 2008;23:261–7. pmid:18359125
- 8. Chang C, Karanth KK, Robbins P. Distinct differences in biodiversity and socioeconomic values of arabica and robusta coffee plantations in India’s Western Ghats. Nat Sci Reports. 2018;8(1):3143.
- 9. Boreux V, Kushalappa CG, Vaast P, Ghazoul J. Interactive effects among ecosystem services and management practices on crop production : Pollination in coffee agroforestry systems. Proc Natl Acad Sci. 2013;110(21):8387–92. pmid:23671073
- 10. Maas B, Karp DS, Bumrungsri S, Darras K, Gonthier D, Huang JC-C, et al. Bird and bat predation services in tropical forests and agroforestry landscapes. Biol Rev. 2015;43:1–21.
- 11. Jaramillo J, Muchugu E, Vega FE, Davis A, Borgemeister C, Chabi-olaye A. Some Like It Hot : The Influence and Implications of Climate Change on Coffee Berry Borer (Hypothenemus hampei) and Coffee Production in East Africa. PLoS One. 2011;6(9):1–14.
- 12. Karp DS, Daily GC. Cascading effects of insectivorous birds and bats in tropical coffee plantations. Ecology. 2014;95(4):1065–74. pmid:24933824
- 13. Perfecto I, Vandermeer J. Quality of Agroecological Matrix in a Tropical Montane Landscape : Ants in Coffee Plantations in Southern Mexico. Conserv Biol. 2002;16(1):174–82.
- 14. Numa C, Verdú JR, Sánchez-Palomino P. Phyllostomid bat diversity in a variegated coffee landscape. Biol Conserv. 2005;122(1):151–8.
- 15. Bali A, Kumar A, Krishnaswamy J. The mammalian communities in coffee plantations around a protected area in the Western Ghats, India. Biol Conserv. 2007;139(1–2):93–102.
- 16. Dolia J, Devy MS, Aravind NA, Kumar A. Adult butterfly communities in coffee plantations around a protected area in the Western Ghats, India. Anim Conserv. 2008;11:26–34.
- 17. Wordley CFR, Sankaran M, Mudappa D, Altringham JD. Landscape scale habitat suitability modelling of bats in the Western Ghats of India: Bats like something in their tea. Biol Conserv. 2015;191:529–36.
- 18. Karanth KK, Sankararaman V, Dalvi S, Srivathsa A, Parameshwaran R, Sharma S, et al. Producing Diversity : Agroforests Sustain Avian Richness and Abundance in India ‘ s Western Ghats. Front Ecol Evol. 2016;4(September):1–10.
- 19. Anand MO, Krishnaswamy J, Das A. Proximity to forests drives bird conservation value of coffee plantations: Implications for certification. Ecol Appl. 2008;18(7):1754–63. pmid:18839769
- 20. Philpott SM, Arendt WJ, Armbrecht I, Bichier P, Diestch T V, Gordon C, et al. Biodiversity Loss in Latin American Coffee Landscapes : Review of the Evidence on Ants, Birds, and Trees. Conserv Biol. 2008;22(5):1093–105. pmid:18759777
- 21. Anand MO, Krishnaswamy J, Kumar A, Bali A. Sustaining biodiversity conservation in human-modified landscapes in the Western Ghats : Remnant forests matter. Biol Conserv. 2010;143(10):2363–74.
- 22. Mendenhall CD, Karp DS, Meyer CFJ, Hadly EA, Daily GC. Predicting biodiversity change and averting collapse in agricultural landscapes. Nature. 2014;0:1–5.
- 23. Maas B, Clough Y, Tscharntke T. Bats and birds increase crop yield in tropical agroforestry landscapes. Ecol Lett. 2013;16(12):1480–7. pmid:24131776
- 24. Williams-Guillen K, Perfecto I. Ensemble Composition and Activity Levels of Insectivorous Bats in Response to Management Intensification in Coffee Agroforestry Systems. PLoS One. 2011;6(1):1–10.
- 25. Park KJ. Mitigating the impacts of agriculture on biodiversity : bats and their potential role as bioindicators. Mamm Biol. 2015;80(3):191–204.
- 26.
Graf S. Diversity and Habitat Use of Understorey Bats in Forest and Agroforestry Systems at the Margin of Lore National Park (Central Sulawesi, Indonesia). University of Vienna; 2010.
- 27.
Kingston T. Response of Bat Diversity to Forest Disturbance in Southeast Asia: Insights from Long-Term Research in Malaysia. In: Adams RA, Pedersen SC, editors. Bat Evolution, Ecology, and Conservation. First. Springer; 2013. p. 169–85.
- 28. Wordley CFR, Sankaran M, Mudappa D, Altringham JD. Bats in the Ghats : Agricultural intensification reduces functional diversity and increases trait filtering in a biodiversity hotspot. Biol Conserv. 2017;210(April):48–55.
- 29. Jha S, Bacon CM, Philpott SM, MÉndez VE, LÄderach P, Rice R a. Shade coffee: Update on a disappearing refuge for biodiversity. Bioscience. 2014;64(5):416–28.
- 30. Schnitzler H-U, Moss CF, Denzinger A. From spatial orientation to food acquisition in echolocating bats. Trends Ecol Evol. 2003;18(8):386–94.
- 31. Schnitzler H-U, Kalko EK V. Echolocation by Insect-Eating Bats. Bioscience. 2001;51(7):557–69.
- 32. Myers N, Mittermeier RA, Mittermeier CG, Fonseca GAB, Kent J. Biodiversity hotspots for conservation priorities. Nature. 2000;403(February):853–8.
- 33. Karanth KU. Bhadra Wildlife Sanctuary and its endangered ecosystem. J Bombay Nat Hist Soc. 1981;79:79–86.
- 34.
Coffee Board of India [Internet]. Ministry of Commerce and Industry, Government of India. http://www.indiacoffee.org/
- 35. Williams-Guillen K, Perfecto I. Effects of Agricultural Intensification on the Assemblage of Leaf-Nosed Bats (Phyllostomidae) in a Coffee Landscape in Chiapas, Mexico. Biotropica. 2010;42(5):605–13.
- 36. Wordley CFR, Sankaran M, Mudappa D, Altringham JD. Heard but not seen: Comparing bat assemblages and study methods in a mosaic landscape in the Western Ghats of India. Ecol Evol. 2018;(October 2017):1–12.
- 37.
Adams AM. Assessing and Analyzing Bat Activity With Acoustic Monitoring: Challenges and Interpretations. 2013;(July):176.
- 38. Skalak SL, Sherwin RE, Brigham RM. Sampling period, size and duration influence measures of bat species richness from acoustic surveys. Methods Ecol Evol. 2012;3:490–502.
- 39. Sheil D, Ducey MJ, Sidiyasa K, Samsoedin I. A NEW TYPE OF SAMPLE UNIT FOR THE EFFICIENT ASSESSMENT OF DIVERSE TREE COMMUNITIES IN COMPLEX FOREST LANDSCAPES. J Trop For Sci. 2003;15(1):117–35.
- 40. Chambers CL, Cushman SA, Medina-Fitoria A, Martínez-Fonseca J, Chávez-Velásquez M. Influences of scale on bat habitat relationships in a forested landscape in Nicaragua. Landsc Ecol. 2016;31(6):1299–318.
- 41. Eigenbrod F, Hecnar SJ, Fahrig L. Sub-optimal study design has major impacts on landscape-scale inference. Biol Conserv. 2011;144(1):298–305.
- 42. Mendes ES, Fonseca C, Marques SF, Maia D, Ramos Pereira MJ. Bat richness and activity in heterogeneous landscapes: guild-specific and scale-dependent? Landsc Ecol. 2017;32(2):295–311.
- 43.
Quantum GIS Development Team. Quantum GIS Geographic Information System. Open Source Geospatial Foundation Project; 2017.
- 44. Raghuram H, Jain M, Balakrishnan R. Species and acoustic diversity of bats in a palaeotropical wet evergreen forest in southern India. Curr Sci. 2014;107(4):631–41.
- 45. Wordley CFR, Foui EK, Mudappa D, Sankaran M, Altringham JD. Acoustic Identification of Bats in the Southern Western Ghats, India. Acta Chiropterologica. 2014;16(1):213–22.
- 46. Denzinger A, Schnitzler HU. Bat guilds, a concept to classify the highly diverse foraging and echolocation behaviors of microchiropteran bats. Front Physiol. 2013;4(July):1–15.
- 47. Korad V, Yardi K, Raut R. Diversity and distribution of bats in the Western Ghats of India. Zoos’ Print J. 2007;22(7):2752–8.
- 48. Boulinier T, Nichols JD, Sauer JR, Hines JE, Pollock KH. ESTIMATING SPECIES RICHNESS : THE IMPORTANCE OF HETEROGENEITY IN SPECIES DETECTABILITY. Ecology. 1998;79(3):1018–28.
- 49. Burnham KP, Overton WS. Robust Estimation of population size when capture probabilities vary among animals. Ecology. 1979;60(5):927–36.
- 50.
Walsh AL, Barclay RMR, Mccracken GF. Designing Bat Activity Surveys for Inventory and Monitoring Studies at Local and Regional Scales. In: Brigham RM, Kalko EK V., Jones G, Parsons S, Limpens HJGA, editors. Bat Echoloction Research: tools, techniques and analysis. Austin, Texas: Bat Conservation International; 2004. p. 157–65.
- 51. Gelman A. Scaling regression inputs by dividing by two standard deviations. Stat Med. 2008;27:2865–73. pmid:17960576
- 52. Schielzeth H. Simple means to improve the interpretability of regression coefficients. Methods Ecol Evol. 2010;1:103–13.
- 53.
Burnham KP, Anderson DR. Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach. Second Edi. Springer; 2002.
- 54. Nakagawa S, Freckleton RP. Model averaging, missing data and multiple imputation : a case study for behavioural ecology. Behav Ecol Sociobiol. 2011;65(1):103–16.
- 55. Grueber CE, Nakagawa S, Laws RJ, Jamieson IG. Multimodel inference in ecology and evolution : challenges and solutions. J Evol Biol. 2011;24:699–711. pmid:21272107
- 56. Ter Braak CJF. Canonical Correspondence Analysis : A New Eigenvector Technique for Multivariate Direct Gradient Analysis. Ecology. 1986;67(5):1167–79.
- 57.
Oksanen J, Blanchet GF, Friendly M, Kindt R, Legendre P, McGlinn D, et al. vegan: Community Ecology Package. 2017. https://cran.r-project.org/package=vegan
- 58. Nichols JD, Boulinier T, Hines JE, Pollock KH, Sauer JR. Inference Methods for Spatial Variation in Species Richness and Community Composition When Not All Species Are Detected. Conserv Biol. 1998;12(6):1390–8.
- 59.
R Core Team. R: A language and environment for statistical computing [Internet]. Vienna: R Foundation for Statistical Computing; 2017. https://www.r-project.org/
- 60. Bader E, Jung K, Kalko EK V, Page RA, Rodriguez R, Sattler T. Mobility explains the response of aerial insectivorous bats to anthropogenic habitat change in the Neotropics. Biol Conserv. 2015;186:97–106.
- 61. Frey-Ehrenbold A, Bontadina F, Arlettaz R, Obrist MK. Landscape connectivity, habitat structure and activity of bat guilds in farmland-dominated matrices. J Appl Ecol. 2013;50(1):252–61.
- 62. Denzinger A, Schnitzler H-U. Bat guilds, a concept to classify the highly diverse foraging and echolocation behaviors of microchiropteran bats. Front Physiol. 2013;4(July):164.
- 63. Hanspach J, Fischer J, Ikin K, Stott J, Law BS. Using trait-based filtering as a predictive framework for conservation : a case study of bats on farms in southeastern Australia. J Appl Ecol. 2012;49:842–50.
- 64. Blakey R V, Law BS, Kingsford RT, Stoklosa J, Tap P, Williamson K. Bat communities respond positively to large-scale thinning of forest regrowth. J Appl Ecol. 2016;53:1694–703.
- 65. Law B, Chidel M. Tracks and riparian zones facilitate the use of Australian regrowth forest by insectivorous bats. J Appl Ecol. 2002;39:605–17.
- 66. Fuentes-Montemayor E, Goulson D, Cavin L, Wallace JM, Park KJ. Fragmented woodlands in agricultural landscapes: The influence of woodland character and landscape context on bats and their insect prey. Agric Ecosyst Environ. 2013;172:6–15.
- 67. Adams MD, Law BS, French KO. Vegetation structure influences the vertical stratification of open- and edge-space aerial-foraging bats in harvested forests. For Ecol Manage. 2009;258:2090–100.
- 68. Müller J, Mehr M, Bässler C, Fenton MB, Hothorn T, Pretzsch H, et al. Aggregative response in bats: Prey abundance versus habitat. Oecologia. 2012;169(3):673–84. pmid:22218944
- 69. Wickramasinghe LP, Harris S, Jones G, Jennings NV. Abundance and Species Richness of Nocturnal Insects on Organic and Conventional Farms : Effects of Agricultural Intensification on Bat Foraging. Conserv Biol. 2004;18(5):1283–92.
- 70. Pocock MJO, Jennings N. Testing biotic indicator taxa: The sensitivity of insectivorous mammals and their prey to the intensification of lowland agriculture. J Appl Ecol. 2008;45(1):151–60.
- 71. Rodriguez-San Pedro A, Simonetti PJA. The relative influence of forest loss and fragmentation on insectivorous bats : does the type of matrix matter ? Landsc Ecol. 2015;30:1561–72.
- 72. Fuentes-Montemayor E, Watts K, Macgregor NA, Lopez-Gallego Z, Park JK. Species mobility and landscape context determine the importance of local and landscape-level attributes: Ecol Appl. 2017;27(5):1541–54. pmid:28370641
- 73. Struebig MJ, Kingston T, Zubaid A, Mohd-Adnan A, Rossiter SJ. Conservation value of forest fragments to Palaeotropical bats. Biol Conserv. 2008;141:2112–26.
- 74.
Wordley CFR. Ecology and conservation of bat species in the Western Ghats of India. 2014.
- 75.
Meyer CFJ, Struebig MJ, Willig MR. Responses of Tropical Bats to Habitat Fragmentation, Logging, and Deforestration. In: Kingston T, Voigt CC, editors. Bats in theAnthropocene: Conservation of Bats in a Changing World. Springer; 2016. p. 63–103.