São Tomé Island Endemic Treefrogs ( Hyperolius spp.) and Land-Use Intensification: A Tale of Hope and Caution

Abstract

Forest degradation is a major driver of the global biodiversity declines. However, responses to forest degradation vary greatly between taxa and are predominantly understudied. This study investigates the effects of land-use change on the endemic amphibian fauna of São Tomé Island (Central Africa), where a fast-growing human population increases pressure on forest resources. We sampled acoustic data on reed frog (Hyperolius spp.) abundance at eight transects in each of four land-use categories, representing different levels of forest degradation: old-growth forest, secondary forest, agroforest, and horticulture. While Hyperolius molleri was most abundant in secondary forest and horticulture, Hyperolius thomensis was almost exclusively found in agroforest. We interpret these differences based on reproductive traits, since both species deposit terrestrial eggs and have aquatic larvae, but reproduce in different microhabitats. H. molleri utilizes open water bodies bordered by vegetation for calling and oviposition, which human disturbance has made available in secondary forest and horticulture. In contrast, H. thomensis breeds in water-filled tree holes (phytotelmata), which predominantly occurred in agroforest, often within coral trees (Erythrina spp.). This first study on Santomean amphibian abundances revealed that H. molleri and H. thomensis males exhibit distinct nonlinear responses to forest degradation and may reap some benefit from human habitat alteration. However, despite their apparent tolerance to some anthropogenic disturbance, further land-use intensification may reduce overall habitat suitability for these and other endemic species.

Keywords

tropical montane forest human-modified landscapes coffee shade plantations arboreal breeders model averaging Gulf of Guinea

Introduction

The current global rate of biodiversity loss is strongly connected to human activity (Ceballos et al., 2015; Newbold et al., 2015), as indicated by the geographic overlap of large human populations and high numbers of threatened species (Cincotta, Wisnewski, & Engelman, 2000; Luck, 2007). Many biodiversity hotspots coincide with populous and impoverished regions (Cincotta et al., 2000; Fisher & Christopher, 2007), mostly located in the tropics (Myers, Mittermeier, Mittermeier, da Fonseca, & Kent, 2000). Here, land-use change is the key driver of terrestrial biodiversity decline (Sala, 2000). As a result of increasing human exploitation, tropical forests are being degraded and lost at accelerated rates (Kim, Sexton, & Townshend, 2015; Lewis, Edwards, & Galbraith, 2015). This poses a serious threat to ecosystem functioning (Foley et al., 2005) and the survival of roughly half of the world's terrestrial biodiversity (Groombridge & Jenkins, 2002).

Many studies on the effects of forest degradation have been dedicated to birds and mammals; far less to the Earth's most threatened group of terrestrial vertebrates: amphibians (Gardner, Barlow, & Peres, 2007; Stuart et al., 2004; Wake & Vredenburg, 2008). The global amphibian extinction crisis has been attributed to multiple factors (Beebee & Griffiths, 2005; Collins & Storfer, 2003; Hof, Araújo, Jetz, & Rahbek, 2011; Navas & Otani, 2007; Wake & Vredenburg, 2008). Land-use conversion is one of the greatest threats and has been known to impact amphibian populations (Beebee & Griffiths, 2005; Collins & Storfer, 2003; Hof et al., 2011). However, it remains unclear how individual species respond to human-modified habitats in the tropics (Barlow et al., 2007; Herrera-Montes & Brokaw, 2010; Wanger et al., 2009). Data on amphibian responses to land-use changes in tropical ecosystems are particularly scarce for most regions and species (Collen et al., 2009; Gardner, Fitzherbert, Drewes, Howell, & Caro, 2007). This is worrying as a growing body of evidence suggests that species-specific approaches are necessary to judge the effects of land use (Cushman, 2006; Nowakowski, Thompson, Donnelly, & Todd, 2017). Complex life histories, strong site fidelity, limited dispersal ability, ectothermy, semi-permeable skin, and egg capsules make amphibians particularly sensitive to human-induced environmental changes (Blaustein & Bancroft, 2007; Navas & Otani, 2007). In particular, species with small geographical ranges and with either unstable or unknown population trends are highly susceptible to habitat modification (Nowakowski et al., 2017).

São Tomé and Príncipe is an African small island nation, facing severe rates of forest degradation and deforestation. Forest resources, including timber, medicinal plants, and bushmeat, constitute an essential part of the inhabitants' livelihoods (Carvalho, Rego, Palmeirim, & Fa, 2015; de Madureira, 2012; Ministry of Infrastructure, Natural Resources and Environment, 2016; Oyono et al., 2014; Salgueiro & Carvalho, 2002). São Tomé, one of the two main islands, is densely populated and the conservation of its endemic-rich natural ecosystems is in strong competition with crop production. However, protecting the biodiversity should be an urgent task (Wilkie, Eckelmann, Laverdiere, & Mathias, 2002), as island species are particularly prone to extinction due to their small population sizes (R. G. Gillespie, 2007). Declining biodiversity may reduce ecological function on islands, potentially impairing ecosystem processes (Cox & Elmqvist, 2000).

The amphibian fauna of São Tomé is composed by five species, all of which are endemic to the island (International Union for Conservation of Nature, 2016), that is, four frogs: Hyperolius molleri (Bedriaga, 1892), Hyperolius thomensis Bocage, 1886 (both Hyperoliidae), Phrynobatrachus leveleve Uyeda, Drewes, and Zimkus, 2007 (Phrynobatrachidae), Ptychadena newtoni (Bocage, 1886) (Ptychadenidae); and one caecilian: Schistometopum thomense (Bocage, 1873) (Dermophiidae). The two Hyperolius are arboreal, Phrynobatrachus and Ptychadena are ground dwellers, and Schistometopum is fossorial. With most studies having focused on the taxonomy of these species (Drewes, 2008), there is little scientific evidence on how they might be affected by land-use change. We focused here on Hyperolius spp. since arboreal species have been recently suggested to be more sensitive indicators to changes in forest structure than terrestrial species (Whitworth, Villacampa, Serrano Rojas, Downie, & MacLeod, 2017; but compare Ernst & Rödel, 2005, 2008). In addition, all other amphibian species were either extremely rare or absent on our study sites. H. molleri reproduces on leaves over stagnant or slow-moving water, whereas H. thomensis breeds in water-filled tree holes (Drewes & Stoelting, 2004). São Tomé offers a diverse array of land-use types (Dallimer, Parnell, Bicknell, & Melo, 2012; de Lima, Dallimer, Atkinson, & Barlow, 2013), including well-preserved native forests that can be used as a valuable baseline for the original state of the island (Jones, 1994). Old-growth forest (OF), secondary forest (SF), and agroforest (AF) each comprise around 30% of the island's surface (Salgueiro & Carvalho, 2002). Due to its limited accessibility, the OF of the higher elevations is among the best preserved on São Tomé (Oyono et al., 2014; Salgueiro & Carvalho, 2002). SFs, being mostly abandoned plantations, form a buffer zone around this OF (de Lima et al., 2013). Both forest types are currently under pressure from agricultural expansion and forest extraction activities, such as logging (Ministry of Infrastructure, Natural Resources and Environment, 2016), despite partly being within the São Tomé Obô Natural Park (Salgueiro & Carvalho, 2002). The region's special climatic conditions favor agricultural production. Most of São Tomé's coffee, one of its key export commodities (Central Intelligence Agency, 2014), is cultivated in the study region. Scattered in-between are horticultural fields (de Lima et al., 2013), used by small-scale farmers to provide vegetables for the internal market (Jones, Burlison, & Tye, 1991). Despite the lack of recent forest inventories, land-use practices are currently assumed to have intensified (de Lima et al., 2013; Oyono et al., 2014).

Tropical amphibian studies rarely cover extensive transitions from OF to deforested land (Gardner, Barlow, et al., 2007) or are based on species richness data only. However, abundance may be a more sensitive measure to compare differences between land-use types in the short term (Collen et al., 2009; Luck, Daily, & Ehrlich, 2003; Schneider-Maunoury et al., 2016). In this study, we aimed to examine the effects of different levels of forest degradation on São Tomé's amphibian abundance. Specifically, the following research questions were explored: (a) How do species abundances change along four different levels of forest degradation? and (b) Can environmental parameters explain this change? We hypothesized that OF, as the original cover of the island, is the main and most important habitat for its amphibian populations.

Methods

Study Area

The 857 km² oceanic island of São Tomé is part of the Cameroon Volcanic Line (Déruelle et al., 1991; Henriques & Neto, 2015) located approximately 250 km off the Gabonese coast, slightly north of the equator (Figure 1). São Tomé was discovered in 1470 by Portuguese navigators. Then, the island had no permanent settlement (Figueiredo, Paiva, Stevart, Oliveira, & Smith, 2011) and was almost entirely covered with lowland (<800 m asl), montane (800–1,400 m asl), and mist (>1,400 m asl) forest (Monod, 1960). Large clearings were subsequently made, especially during the 19th century, when almost all forests up to 1,500 m asl were cut to plant export and subsistence crops (Figueiredo et al., 2011; Juste & Fa, 1994).

Figure 1.

Different land-use types and location of transects within the study area. The inset in the top left shows the geographic location of São Tomé Island (black square) in the Gulf of Guinea. The inset in the bottom left shows the location of the study area (black star) on São Tomé Island. The gray dot indicates Monte Café, the nearest large settlement, and the gray square the capital of São Tomé.

São Tomé has an oceanic equatorial climate, but owing to the moist southwest winds and high mountains, it covers distinct climatic zones (Jones et al., 1991; Juste & Fa, 1994). Our study area was located in the montane forest zone (between 800 and 1,300 m asl, Figure 1). The annual mean temperature at Monte Café (690 m asl), the nearest large settlement, is 20.5℃, with a mean annual maximum temperature of 24.7℃ and a mean annual minimum temperature of 16.5℃. The local absolute minimum temperature is 8.7℃ (Brühl, 1993). The annual precipitation exceeds 2,000 mm, and the dry seasons are not very pronounced at this elevation, with rain falling throughout the year, in varying amounts (Jones et al., 1991).

Sampling Design

Land-use type classification

We considered four main land-use types ranging from OF to horticulture (HC; Salgueiro & Carvalho, 2002). OF includes areas with little human influence, mostly covered by dense, tall vegetation, and large trees, where native species predominate (Figueiredo et al., 2011; Salgueiro & Carvalho, 2002). SF results from agricultural abandonment or intensive logging and includes regenerating areas with high proportions of introduced species and few large trees (de Lima et al., 2013). AF is a multistrata management system in which high-canopy trees shade understorey agricultural plants (Salgueiro & Carvalho, 2002; Tscharntke et al., 2011). In the study area, these were perennial coffee plants, typically Coffea arabica, growing under the canopy of introduced, fast-growing species, such as coral trees (Erythrina spp.). The plantations are often intercropped with banana (Musa spp.) and taro (Colocasia esculenta; de Lima et al., 2013). HC was the only land-use type without a continuous tree cover but comprised solitary trees. This land-use type included annual and perennial crops such as cabbage (Brassica oleracea), bean (Phaseolus vulgaris), tomato (Lycopersicon esculentum), banana, cassava (Manihot esculenta), and taro (Xanthosoma sagittifolium; Jones et al., 1991; Ministry of Infrastructure, Natural Resources and Environment, 2016; Salgueiro & Carvalho, 2002).

Transect selection

Since there are no reliable land-use maps for São Tomé (de Lima et al., 2013), we used our knowledge of the region and Google Earth images (Google Inc., 2016) to classify land-use types in the study area. We then generated eight random points within each land-use category using QGIS (QGIS Development Team, 2016). The random points were identified in the field with the help of a GPS device (Garmin GPSmap 62s) and served as the localities for linear transects (cf. e.g., Doan, 2003; Pearman, Velasco, & López, 1995; Rödel & Ernst, 2004). Transect length was set to 100 m to ensure that transects did not cover more than one land-use type. The direction of each transect was chosen to assure that the terrain was accessible (i.e., not following a steep slope). The resulting 32 transects, eight in each land-use type (Figure 1), assured easy comparability of amphibian abundances across the land-use categories (Hirzel & Guisan, 2002). Open water bodies of any kind were absent from one SF and five AF transects. The average linear distance between transects was 315 m (range: 765 m). All 32 transects were located inside a base area of approximately 4 × 4 km. However, the surface of the study area was considerably larger due to the rugged terrain with altitudinal differences of over 500 m.

Data Collection

Habitat parameters

Prior to amphibian sampling, relevant environmental variables were recorded for all transects (Appendix A). To evaluate canopy openness, we took five hemispherical photographs on each transect; one every 25 m using a camera equipped with a fish-eye lens (Soligor DHG 0.19 × Fish Eye for Digital Ø 37 mm Japan), held vertically at eye level. We applied ImageJ (Schneider, Rasband, & Eliceiri, 2012) to split the red, green, and blue components of each photograph. The blue channel images were then converted into binary images. This conversion was automatic, except for a few cases, when suboptimal light exposure demanded manual thresholding. The percentage of open canopy for each image was obtained employing Gap Fraction Analysis (Gonsamo, Walter, & Pellikka, 2011). We then calculated the mean of the five pictures for each transect: 0% represented a completely closed canopy and at 100% no vegetation obscured the sky. We further recorded vegetation density for each transect as a whole in three different strata: canopy (height: > 20 m), lower tree (height: 3–10 m), and bush and shrub stratum (height: 0.5–1.5 m; Rödel & Ernst, 2004). For each stratum, we applied the following categories by eye: stratum absent, stratum reduced to single trees, stratum present but gaps predominating, stratum present with closed areas predominating, and stratum closed. We further counted all small trees (diameter at breast height, dbh, between 21 and 50 cm) and large trees (dbh larger than 50 cm) within 10 m to each side of the transect. In the same area, we documented presence of plant species or genera, including trees, shrubs, and herbs. To assess the availability of potential amphibian oviposition sites, we counted all tree holes visible in the same area up to 10 m aboveground. We also registered the presence of both stagnant and flowing open water bodies (irrespective of water-filled tree holes), including rain barrels, ponds, puddles, ditches, streams, and rivers, within a distance of 150 m from the transect. In one case in which two transects were only 130 m apart, the water bodies were assigned to the nearest transect. Following Rödel & Ernst (2004), we classified soil into four rough substrate types: forest soil, arenaceous humus soil, loamy soil, and arenaceous soil. We deployed one automatic thermologger per transect (Thermochron® iButtons© DS 1921G; accuracy ± 0.5℃, Maxim Integrated, San Jose, USA) to measure air temperature every hour for 30 days (December 30, 2014, to January 28, 2015). These devices were placed at ground level in the shade. From the data obtained, we extracted average, minimum, and maximum air temperature for each day (24 values), which were then averaged per transect over the 30-day period. The elevation was recorded every 25 m for each transect using the GPS from which the mean elevation of the respective transect was calculated.

Treefrog sampling

We performed amphibian counts from mid-December 2014 until the end of January 2015, 30 sampling nights in total. We visited each transect three times, which resulted in 96 transect walks (3 × 32 transects). The transects were surveyed between 5.45 and 8.45 pm at slow pace and constant speed. On average, sampling time per transect walk was 12 min. The order of transect visits followed a predefined random sequence. On all transect walks, we identified every individual detected visually and acoustically (breeding males), following AmphibiaWeb (2017); Drewes and Stoelting (2004); Schiøtz (1999); and Uyeda, Drewes, and Zimkus (2007). Calling males were recorded up to 30 m to each side of the transect. We considered a minimum number of individuals by allocating repeated calls of the same direction and distance to a single individual. H. molleri emits a series of “clack” sounds, whereas the advertisement call of H. thomensis consists of a creaking buzz of lower frequency (Schiøtz, 1999). The two Hyperolius species have been reported to use calling positions at different heights, that is, H. molleri at 0.3 to 1 m and H. thomensis at 5 m aboveground (Fahr, 1993). Visual detections included sightings on the ground and in trees along a band of 2 m on each side of the transect. Species identity, developmental stage (i.e., egg, tadpole, juvenile, or adult), and, if possible, sex were recorded for all detected specimens. We counted egg clutches but not egg numbers within clutches. H. molleri and H. thomensis are known to hybridize in the study area (Bell, Drewes, & Zamudio, 2015). Potential hybrids were only seen in the broader study region. Sightings and calls (cf. Gilbert & Bell, 2018) of all individuals on transects could unambiguously be assigned to one of the two species.

Data Analyses

Our aim was to determine how species' abundances change along different types of forest degradation and which environmental variables best explain this change. Hence, we applied two steps of analysis, for which all species-individuals per transect were summed up over the whole collection period. Abundance data were zero-inflated and not normally distributed.

First, we compared the abundances of each amphibian species between land-use types using Kruskal–Wallis rank sum tests followed by individual pairwise Wilcoxon rank sum tests. We calculated indicator values (IndVals) for each species in all land-use categories (function indval, library labdsv), based on their habitat specificity and fidelity, which range between 0% (no indication) and 100% (excellent indication). The index reaches its maximum when a species only occurs in one land-use category (high specificity) and is present at all sample sites of that land-use category (high fidelity; Dufrêne & Legendre, 1997). A species that has a high IndVal (usually more than 70%) is considered a character species. Species that show medium specificity and medium fidelity are much better indicators of environmental change and are known as detector species (da Mata, McGeoch, & Tidon, 2008; McGeoch, van Rensburg, & Botes, 2002).

Second, we applied model averaging to identify the environmental variables that best explained the abundance of each species. To lower the collinearity, we reduced the number of predictor variables based on Kruskal–Wallis rank sum tests, scatterplots, and variance inflation factors (≤3; Zuur, Ieno, & Elphick, 2010). We standardized all continuous predictor variables between 0 and 1 by subtracting the minimum value and dividing by the maximum value to avoid having comparatively large values distort the results. To rule out systematic biases, we checked the abundance data for temporal autocorrelation by plotting autocorrelation functions (ACFs). No dependence was revealed regarding the timing of observation (Appendix B). However, significant spatial autocorrelation, a frequent problem in ecological data (Hawkins, 2012), was detected using Monte Carlo simulation of Moran's I (H. molleri: p = .0002; H. thomensis: p = .0000). Model averaging based on Akaike weights across spatial autoregressive (SAR) models is a robust method to address this issue (Diniz-Filho, Rangel, & Bini, 2008). We thus chose spatial simultaneous autoregressive lag model estimation and employed the function “lagsarlm” (library spdep) to define a global model (Bivand & Piras, 2015). All possible combinations of the independent predictor variables were fitted using the function “dredge” (library MuMIn) on the global model (Bartón, 2016). To identify key predictor variables, model averaging was carried out with the function “model.avg” (library MuMIn) for a 95% confidence model set, resulting in estimates, standard errors, and relative variable importance.

All statistical analyses were performed in R, Version 3.0.0 (R Core Team, 2013). Additional packages used included MASS (Venables & Ripley, 2002) and vegan (Oksanen et al., 2015). Land-use categorization coincided mostly with differences in canopy density or openness, plant species composition, and substrate type as confirmed by Kruskal–Wallis rank sum tests (p values < .0003). We also conducted a one-way analysis of similarities (ANOSIM) on the plant species data (Appendix C) to validate the land-use type categorization (Clarke, 1993).

Results

During the transect walks, we acoustically documented 72 H. molleri and 26 H. thomensis individuals. Visual data were only obtained for one adult and nine clutches of H. molleri and was therefore not further considered. Neither visual nor acoustic evidence was found for Hyperolius hybrids within transects. As expected, we encountered two additional amphibian species in the study area; however, S. thomense was observed only twice along transects and P. leveleve exclusively off transect.

The two Hyperolius species differed in their abundances across the four land-use types according to the Kruskal–Wallis rank sum (H. molleri: p = .0079; H. thomensis: p = .000) and the pairwise Wilcoxon rank sum tests (Table 1). However, H. molleri abundances were only significantly different between AF (1 individual) and SF (33 individuals). The latter was closely followed by HC (32 individuals) and OF (6 individuals). The abundance of H. thomensis in AF (23 individuals) was significantly different from those in all other land-use types (three individuals in SF and none in OF or HC). H. molleri reached the highest IndVals in SF (40%) and HC (28%; Table 2). Overall, it had a low indicator potential, as it was not restricted to one land-use type and only occurred at some transects of each land-use type. Contrarily, H. thomensis was a character species showing an extraordinary specificity and fidelity to AF (IndVal: 88%; Table 2).

Table 1.

Pairwise Wilcoxon Rank Sum Tests on Differences in Hyperolius Abundances Between Land-Use Types.

Species		OF	SF	AF
Hyperolius molleri	OF
	SF	.104
	AF	.364	.017
	HC	.364	.873	.119
Hyperolius thomensis	OF
	SF	.342
	AF	.002	.005
	HC		.342	.002

Note. n = 8 for each land-use type. Results described through p values; significant p values (<.05) marked in boldface. OF = old-growth forest; SF = secondary forest; AF = agroforest; HC = horticulture.

Table 2.

Indicator Values (IndVals) of Hyperolius spp. According to Land-Use Types.

Species	OF	SF	AF	HC
Hyperolius molleri	4	40	0	28
Hyperolius thomensis	0	3	88	0

Note. n = 8 for each land-use type. OF = old-growth forest; SF = secondary forest; AF = agroforest; HC = horticulture, given as percentages with 0% meaning no indication (species not present in the respective land-use type) and 100% meaning excellent indication (species present at all transects of the respective land-use type and absent from all other land-use types).

For our global SAR models, land-use type, lower tree density, bush and shrub density, number of small trees, number of large trees, number of tree holes, water body availability, and maximum air temperature were kept as independent predictor variables. Model-averaged results for H. molleri and H. thomensis only included those variables that were part of the respective 95% confidence model set. Model averaging for the abundance of H. molleri revealed highest relative importance of the variables land-use type and water body availability (Table 3). The highest H. molleri abundances were reached in SF and HC. No frogs were observed when open water bodies were absent (Figure 2). Model averaging for H. thomensis resulted in highest relative importance of water body availability, number of tree holes, and land-use type (Table 3). Few H. thomensis individuals were found in the presence of open water bodies, but the number of tree holes was positively correlated with H. thomensis abundances. H. thomensis was considerably more abundant in AF than in any of the other three categories (Figure 2). Most tree holes were found in AF (15), while few were recorded for SF (5) and OF (1), and none for HC. Sixteen of the 21 tree holes occurred in trees of the nonnative genus Erythrina.

Figure 2.

Relationship between Hyperolius abundances and variables with the highest relative variable importance values (>30; Table 3): Land-use type (H. molleri and H. thomensis; n = 8 for each land-use type), water body availability (H. molleri and H. thomensis), and number of tree holes (H. thomensis). Box and whisker values refer to median, interquartile range, minimum and maximum, and extreme values. OF = old-growth forest; SF = secondary forest; AF = agroforest; HC = horticulture; A = absence of water; P = presence of water.

Table 3.

Summary Results Obtained After Model Averaging (Natural Average: Non-shrinkage).

Species	Variable	Estimate	Standard error	Relative variable importance
Hyperolius molleri	(Intercept)	0.092	1.376
	Land-use type 2^a	2.852	1.194	0.68
	Land-use type 3	−0.275	1.584	0.68
	Land-use type 4	2.796	1.294	0.68
	Water body presence^b	2.277	1.202	0.50
	Maximum air temperature	−8.941	10.735	0.21
	Number of small trees	−1.643	1.839	0.20
	Number of tree holes	1.198	1.584	0.20
	Number of large trees	−0.876	2.392	0.18
	Bush and shrub density 2^c	0.367	1.541	< 0.01
	Bush and shrub density 3	0.814	1.140	< 0.01
	Bush and shrub density 4	−1.062	1.318	< 0.01
Hyperolius thomensis	(Intercept)	1.594	0.434
	Water body presence^b	−1.552	0.401	1.00
	Number of tree holes	1.295	0.390	0.94
	Land-use type 2^a	0.043	0.276	0.31
	Land-use type 3	1.282	0.506	0.31
	Land-use type 4	−0.020	0.283	0.31
	Number of small trees	−0.507	0.413	0.26
	Number of large trees	−0.356	0.495	0.19
	Maximum air temperature	0.381	2.325	0.15

Note. Effects of each variable on Hyperolius abundances on transects, including estimates, standard errors, and relative variable importance (land-use type 2: secondary forest, land-use type 3: agroforest, land-use type 4: horticulture; bush and shrub density 2: reduced to single trees, bush and shrub density 3: gaps predominating, bush and shrub density 4: closed areas predominating).

^aLand-use type 1 (old-growth forest) was the reference category. ^bWater body absence was the reference category. ^cBush and shrub density 1 (absent) was the reference category.

Discussion

Our amphibian data contribute to understanding the responses of arboreal frog species to tropical land use. Amphibian studies measuring abundances commonly show conflicting outcomes with regard to land-use change. Many studies using species richness reveal declines toward disturbed areas (Gardner, Fitzherbert, et al., 2007; Hillers, Veith, & Rödel, 2008; Pineda, Moreno, Escobar, & Halffter, 2005; Wanger et al., 2010), presumably owing to lower structural complexity (Gardner, Barlow, et al., 2007) or altered microclimate (Hillers et al., 2008). However, inconsistencies among results of any measure hamper the search for cross-species generalities. The underlying cause may be species-specific sensitivities to habitat modification and associated traits (Nowakowski et al., 2017). We try to shed light on the link between landscape variables and reproductive traits in Hyperolius spp., advocating for species-centered approaches (Becker, Loyola, Haddad, & Zamudio, 2010; Betts et al., 2014).

Treefrog Species Response to Land-Use Intensification

São Tomé's endemic treefrogs showed species-specific complex nonlinear responses to the different levels of forest degradation. H. molleri has been considered a generalist that can adapt to modified landscapes (International Union for Conservation of Nature, 2016), occurring in primary forest, SF, and coffee plantations and using water reservoirs within agricultural sites (Bell et al., 2015). We confirmed its presence in all of these land-use types but in distinctly different abundances. The species was, however, most numerous in SF, as indicated by Fahr (1993) and Drewes and Stoelting (2004). H. thomensis, in contrast, was almost exclusively restricted to AF, with a few records in SF. This suggests that this species tolerates intermediate levels of forest degradation and may benefit from human-induced increased availability of tree holes. This finding contradicts previous data, assuming the species to be an OF specialist (Drewes & Stoelting, 2004), as we did not find it in this land-use type. However, the apparent absence of H. thomensis from our OF transects might be due to its overall low abundance.

Species-Specific Habitat Associations

Since H. molleri tadpoles develop in stagnant or slow-flowing water, the species' association with the presence of open water bodies was not surprising. We could not determine any preference with regard to stagnant versus flowing. So far it is not known whether tadpoles of this species develop equally in different water body types. Water bodies were overall more common in SF, HC, and OF than in AF, where the deep-rooted plantation systems do not require artificial watering. Regarding microstructures, SF appeared to offer the ideal combination for the development of H. molleri offspring, namely, puddles on roads and numerous streams framed by overhanging vegetation. H. molleri requires vegetation with large, stable leaves overhanging water to deposit clutches (Drewes & Stoelting, 2004). Likewise, HC frequently provided adequate man-made reproduction sites such as irrigation ponds and rain barrels. Here, the walls of the water bodies functionally replaced overhanging vegetation as clutch oviposition sites. Man-made water sources in tropical farmland have been suggested to be crucial for amphibian biodiversity as natural ecosystems become scarcer (Mendenhall et al., 2014). OF, mostly located on steep slopes, harbored only few streams or rivers, usually with high-velocity flows. These habitats are likely unsuitable for Hyperolius larvae, which usually have no morphological adaptations for fast flowing streams (Channing, Rödel, & Channing, 2012; Drewes & Stoelting, 2004; International Union for Conservation of Nature, 2016). The higher abundances of H. molleri in SF and HC may thus be explained by its breeding requirements, since both land-use types provide a suitable combination of vegetation or man-made structures and open water bodies. We frequently spotted clutches of H. molleri on transects with considerable vegetation gaps, which suggests that it might also require the specific microclimatic conditions best found in SF and HC. In the study area, and elsewhere, increased disturbance correlated with greater canopy openness and increased ambient temperature (Halverson, Skelly, Kiesecker, & Freidenburg, 2003).

The habitat preferences of H. thomensis also reflect a compromise between water availability and vegetation structure, linked to breeding ecology. The species likewise deposits its eggs outside of water and has aquatic larvae. By contrast, it heavily depends on the availability of water-filled tree holes which serve as breeding sites (Drewes & Stoelting, 2004). The species was clearly more abundant in AF, since the fast-growing shade trees, including Erythrina spp., tend to form cavities (Ishiguri et al., 2007). These nitrogen-fixing shade tree species do not naturally occur in OF (Lejoly, 2000). OF and SF mostly host native, slow-growing trees, and in HC, only few trees are present; both factors likely account for lower phytotelmata availability. The absence of aquatic habitats in forested environments may have led to the evolution of the phytotelmata-breeding strategy of H. thomensis and to its divergence from the H. molleri lineage (Bell et al., 2015). Our results even suggest that the absence of open water bodies favors the occurrence of H. thomensis; however, such a causal relationship seems unlikely.

Spatial behavior in amphibians is a trade-off between the need to find suitable microhabitats for reproduction (G. R. Gillespie, Lockie, Scroggie, & Iskandar, 2004; Refsnider & Janzen, 2010), thermoregulatory requirements (Magnuson, Crowder, & Medvick, 1979), and predator avoidance (Vredenburg, 2004). Females selecting oviposition sites search for a trade-off between these pressures (Thurman, Garcia, & Hoffman, 2014). Males may equally take them into account when calling to attract females (G. R. Gillespie et al., 2004; Rudolf & Rödel, 2005). The different responses of H. molleri and H. thomensis males to the shift in anthropogenic land use in our area seem mostly related to their differences in reproductive biology. To varying degrees, both species appear to profit from some anthropogenic disturbance. This may, however, only concern the breeding periods and life stages of these species. Other land-use types may be essential during the nonbreeding periods, with a possible negative effect of habitat degradation on survival rates (Becker, Fonseca, Haddad, & Prado, 2010; Pittman, Osbourn, & Semlitsch, 2014). In addition, connectivity and minimum size of land-use patches are likely to influence population viability (Becker, Fonseca, et al., 2010; Halpern, Gaines, & Warner, 2005). Overall, H. molleri may be better suited to survive major habitat changes than H. thomensis, since it is well adapted to the nonforested habitats found in intensive land-use areas. Hybrids, which are likely to be H. molleri carrying H. thomensis mitochondrial haplotypes (Bell et al., 2015), may deal similarly with land-use change depending on their breeding habitat. This raises the issue of whether Hyperolius hybrids contribute to conservation goals, such as the maintenance of ecological functions, or negatively affect the genetic fitness of the “pure” species (Jackiw, Mandil, & Hager, 2015).

Implications for Conservation

Despite the apparent tolerance of H. molleri and H. thomensis to some forest degradation, the importance of OF to overall biodiversity conservation remains unquestioned (Gibson et al., 2011). The richness and abundance of São Tomé's endemic bird species decrease toward more intensively used ecosystems, which highlights the irreplaceability of São Tomé's natural forests (de Lima et al., 2013). Different taxonomic groups and species within groups do not necessarily exhibit similar responses to forest degradation (Barlow et al., 2007; Wolters, Bengtsson, & Zaitsev, 2006). Thus, optimal conservation strategies will be species dependent. Further research including all amphibian species at any life cycle phase, and other taxonomic groups, as well as additional systems such as lowland forests, cocoa, and oil palm plantations would help clarify the island-wide impacts of land-use intensification.

São Tomé's human-modified landscapes currently show potential for the conservation of endemic amphibians. The ongoing intensification in all land-use types may, however, exacerbate in the future. The current trend of forest conversion to HC, palm oil plantations, and other unshaded cultivations (Oyono et al., 2014) may be further fueled by the explosive population growth and foreseeable increment of coffee, cocoa, palm oil, and coconut exports (United Nations, Department of Economic and Social Affairs, Population Division, 2017). Climate change is likely to contribute to a reduction in forested land by forcing small-scale farmers to move horticultural parcels to higher altitudes, where OF and SF currently predominate. While deforestation could primarily affect H. thomensis through the loss of tree habitats, any impact on H. molleri may be less direct but likewise urges caution. Concerted conservation strategies, taking life-history traits into account, are essential.

Footnotes

Appendices

Appendix B

Appendix C

Acknowledgments

The authors would like to thank Joel dos Santos Oquiongo and Esmael Vunge Rocha da Luz for their assistance during field work; Gabriel dos Santos Oquiongo for providing means of transportation to the sites; Rayna Camille Bell who shared her knowledge with us; Eng. Arlindo de Ceita Carvalho for supporting and authorizing our research in the study area; and two anonymous reviewers, whose comments improved the manuscript. Finally, we want to express our thanks to the members of the Rödel lab who gave consistent advice throughout the study process and to Simone Allen and Sami Asad for language check.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We gathered financial support for research-related costs on São Tomé Island by the BirdLife International São Tomé and Príncipe Initiative, which has been funded by the Royal Society for the Protection of Birds, the Prentice family through the BirdLife International Preventing Extinctions Programme (BirdLife's Species Champion Programme), and the Disney Worldwide Conservation Fund. The Commission for the Advancement of Women (Frauenförderkommission) at the Albrecht Daniel Thaer-Institute of Agricultural and Horticultural Sciences, Humboldt-Universität zu Berlin, financially contributed to cover LS' travel costs between Germany and São Tomé and Príncipe. RFL was funded by a “Fundação para a Ciência e a Tecnologia” postdoc grant (SFRH/BPD/91494/2012).

References

AmphibiaWeb. (2017). Search the database. Retrieved from http://amphibiaweb.org/search/index.html.

Barlow

Gardner

T. A.

Araujo

I. S.

Ávila-Pires

T. C.

Bonaldo

A. B.

Costa

J. E.

Peres

C. A.

(2007) Quantifying the biodiversity value of tropical primary, secondary, and plantation forests. Proceedings of the National Academy of Sciences, USA 104: 18555–18560.

Bartón, K. (2016). MuMIn: Multi-Model Inference. R package version 1.13.4. Retrieved from https://cran.r-project.org/web/packages/MuMIn/.

Becker

C. G.

Fonseca

C. R.

Haddad

C. F. B.

Prado

P. I.

(2010) Habitat split as a cause of local population declines of amphibians with aquatic larvae. Conservation Biology 24: 287–294.

Becker

C. G.

Loyola

R. D.

Haddad

C. F. B.

Zamudio

K. R.

(2010) Integrating species life-history traits and patterns of deforestation in amphibian conservation planning. Diversity and Distributions 16: 10–19.

Beebee

T. J. C.

Griffiths

R. A.

(2005) The amphibian decline crisis: A watershed for conservation biology?. Biological Conservation 125: 271–285.

Bell

R. C.

Drewes

R. C.

Zamudio

K. R.

(2015) Reed frog diversification in the Gulf of Guinea: Overseas dispersal, the progression rule, and in situ speciation. Evolution 69: 904–915.

Betts

M. G.

Fahrig

Hadley

A. S.

Halstead

K. E.

Bowman

Robinson

W. D.

Lindenmayer

D. B.

(2014) A species-centered approach for uncovering generalities in organism responses to habitat loss and fragmentation. Ecography 37: 517–527.

Bivand

Piras

(2015) Comparing implementations of estimation methods for spatial econometrics. Journal of Statistical Software 63: 1–36.

10.

Blaustein

A. R.

Bancroft

B. A.

(2007) Amphibian population declines: Evolutionary considerations. BioScience 57: 437–444.

11.

Brühl

(1993) Überblick über Geographie, Geologie, Klima und Vegetationsformen der Insel São Tomé [Golf von Guinea]. Faunistische Abhandlungen: Staatliches Museum für Tierkunde Dresden 19: 5–16.

12.

Carvalho

Rego

Palmeirim

J. M.

J. E.

(2015) Wild meat consumption on São Tomé Island, West Africa: Implications for conservation and local livelihoods. Ecology and Society 20: 27.

13.

Ceballos

Ehrlich

P. R.

Barnosky

A. D.

Garcia

Pringle

R. M.

Palmer

T. M.

(2015) Accelerated modern human-induced species losses: Entering the sixth mass extinction. Science Advances 1: e1400253–e1400253.

14.

Central Intelligence Agency. (2014). The world factbook: Sao Tome and Principe. Retrieved from https://www.cia.gov/library/publications/the-world-factbook/geos/tp.html.

15.

Channing

Rödel

M.-O.

Channing

(2012) Tadpoles of Africa: The biology and identification of all known tadpoles in Sub-Saharan Africa, Frankfurt am Main, Germany: Edition Chimaira, Frankfurt Contributions to Natural History. (Vol. 55).

16.

Cincotta

R. P.

Wisnewski

Engelman

(2000) Human population in the biodiversity hotspots. Nature 404: 990–992.

17.

Clarke

K. R.

(1993) Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18: 117–117.

18.

Collen

Loh

Whitmee

McRAE

Amin

Baillie

J. E. M.

(2009) Monitoring change in vertebrate abundance: The Living Planet Index. Conservation Biology 23: 317–327.

19.

Collins

J. P.

Storfer

(2003) Global amphibian declines: Sorting the hypotheses. Diversity and Distributions 9: 89–98.

20.

Cox

P. A.

Elmqvist

(2000) Pollinator extinction in the Pacific Islands. Conservation Biology 14: 1237–1239.

21.

Cushman

S. A.

(2006) Effects of habitat loss and fragmentation on amphibians: A review and prospectus. Biological Conservation 128: 231–240.

22.

Dallimer

Parnell

Bicknell

J. E.

Melo

(2012) The importance of novel and agricultural habitats for the avifauna of an oceanic island. Journal for Nature Conservation 20: 191–199.

23.

da Mata

R. A.

McGeoch

Tidon

(2008) Drosophilid assemblages as a bioindicator system of human disturbance in the Brazilian Savanna. Biodiversity and Conservation 17: 2899–2916.

24.

de Lima

R. F.

Dallimer

Atkinson

P. W.

Barlow

(2013) Biodiversity and land-use change: Understanding the complex responses of an endemic-rich bird assemblage. Diversity and Distributions 19: 411–422.

25.

de Madureira, M. do C. (2012). Plantas medicinais e medicina tradicional de S. Tomé e Príncipe. Actas Do Colóquio Internacional São Tomé E Príncipe Numa Perspectiva Interdisciplinar, Diacrónica E Sincrónica, 433–453.

26.

Déruelle

Moreau

Nkoumbou

Kambou

Lissom

Njonfang

Nono

(1991) The Cameroon Line: A review. In: Kampunzu

P. D. A. B.

Lubala

P. D. R. T.

(eds) Magmatism in extensional structural settings, Berlin, Germany: Springer, pp. 274–327.

27.

Diniz-Filho

J. A. F.

Rangel

T. F. L. V. B.

Bini

L. M.

(2008) Model selection and information theory in geographical ecology. Global Ecology and Biogeography 17: 479–488.

28.

Doan

T. M.

(2003) Which methods are most effective for surveying rain forest herpetofauna?. Journal of Herpetology 37: 72–81.

29.

Drewes

R. C.

(2008) Biogeographic implications of intriguing relationship patterns of endemic amphibians of the Gulf of Guinea Islands. In: Corti

(ed.) Herpetologia Sardiniae, Belvedere. Latina, Italy: Societas Herpetologica Italica, pp. 231–234.

30.

Drewes

R. C.

Stoelting

R. E.

(2004) The California Academy of Sciences Gulf of Guinea expedition (2001): II. Additions and corrections to our knowledge of the endemic amphibians of São Tomé and Príncipe. Proceedings of the California Academy of Sciences 55: 573–587.

31.

Dufrêne

Legendre

(1997) Species assemblages and indicator species: The need for a flexible asymmetrical approach. Ecological Monographs 67: 345–366.

32.

Ernst

Rödel

M.-O.

(2005) Anthropogenically induced changes of predictability in tropical anuran assemblages. Ecology 86: 3111–3118.

33.

Ernst

Rödel

M.-O.

(2008) Patterns of community composition in two tropical tree frog assemblages: Separating spatial structure and environmental effects in disturbed and undisturbed forests. Journal of Tropical Ecology 24: 111–120.

34.

Fahr

(1993) Ein Beitrag zur Biologie der Amphibien der Insel São Tomé (Golf von Guinea) (Amphibia). Faunistische Abhandlungen: Staatliches Museum für Tierkunde Dresden 19: 75–84.

35.

Figueiredo

Paiva

Stevart

Oliveira

Smith

G. F.

(2011) Annotated catalogue of the flowering plants of São Tomé and Príncipe. Bothalia 41: 41–82.

36.

Fisher

Christopher

(2007) Poverty and biodiversity: Measuring the overlap of human poverty and the biodiversity hotspots. Ecological Economics 62: 93–101.

37.

Foley

J. A.

DeFries

Asner

G. P.

Barford

Bonan

Carpenter

S. R.

Snyder

P. K.

(2005) Global consequences of land use. Science 309: 570–574.

38.

Gardner

T. A.

Barlow

Peres

C. A.

(2007) Paradox, presumption and pitfalls in conservation biology: The importance of habitat change for amphibians and reptiles. Biological Conservation 138: 166–179.

39.

Gardner

T. A.

Fitzherbert

E. B.

Drewes

R. C.

Howell

K. M.

Caro

(2007) Spatial and temporal patterns of abundance and diversity of an East African leaf litter amphibian fauna. Biotropica 39: 105–113.

40.

Gibson

Lee

T. M.

Koh

L. P.

Brook

B. W.

Gardner

T. A.

Barlow

Sodhi

N. S.

(2011) Primary forests are irreplaceable for sustaining tropical biodiversity. Nature 478: 378–381.

41.

Gilbert

C. M.

Bell

R. C.

(2018) Evolution of advertisement calls in an island radiation of African reed frogs. Biological Journal of the Linnean Society 123: 1–11.

42.

Gillespie

G. R.

Lockie

Scroggie

M. P.

Iskandar

D. T.

(2004) Habitat use by stream-breeding frogs in South-East Sulawesi, with some preliminary observations on community organization. Journal of Tropical Ecology 20: 439–448.

43.

Gillespie

R. G.

(2007) Oceanic islands: Models of diversity. In: Levin

S. A.

(ed.) Encyclopedia of biodiversity, Oxford, England: Elsevier Ltd, pp. 1–13.

44.

Gonsamo

Walter

J.-M. N.

Pellikka

(2011) CIMES: A package of programs for determining canopy geometry and solar radiation regimes through hemispherical photographs. Computers and Electronics in Agriculture 79: 207–215.

45.

Google Inc. (2016). Google Earth (version 7.1.2.2041). Retrieved from https://www.google.de/intl/de/earth/.

46.

Groombridge

Jenkins

(2002) World atlas of biodiversity: Earth's living resources in the 21st century, Berkeley, CA: University of California Press.

47.

Halpern

B. S.

Gaines

S. D.

Warner

R. R.

(2005) Habitat size, recruitment, and longevity as factors limiting population size in stage-structured species. The American Naturalist 165: 82–94.

48.

Halverson

M. A.

Skelly

D. K.

Kiesecker

J. M.

Freidenburg

L. K.

(2003) Forest mediated light regime linked to amphibian distribution and performance. Oecologia 134: 360–364.

49.

Hawkins

B. A.

(2012) Eight (and a half) deadly sins of spatial analysis. Journal of Biogeography 39: 1–9.

50.

Henriques

Neto

(2015) Geoheritage at the equator: Selected geosites of São Tomé Island (Cameron Line, Central Africa). Sustainability 7: 648–667.

51.

Herrera-Montes

Brokaw

(2010) Conservation value of tropical secondary forest: A herpetofaunal perspective. Biological Conservation 143: 1414–1422.

52.

Hillers

Veith

Rödel

M.-O.

(2008) Effects of forest fragmentation and habitat degradation on West African leaf-litter frogs. Conservation Biology 22: 762–772.

53.

Hirzel

Guisan

(2002) Which is the optimal sampling strategy for habitat suitability modelling. Ecological Modelling 157: 331–341.

54.

Hof

Araújo

M. B.

Jetz

Rahbek

(2011) Additive threats from pathogens, climate and land-use change for global amphibian diversity. Nature 480: 516–519.

55.

International Union for Conservation of Nature. (2016). The IUCN Red List of threatened species (version 2017-1). Retrieved from http://www.iucnredlist.org.

56.

Ishiguri

Eizawa

Saito

Iizuka

Yokota

Priadi

Yoshizawa

(2007) Variation in the wood properties of Paraserianthes falcataria planted in Indonesia. IAWA Journal 28: 339–348.

57.

Jackiw

R. N.

Mandil

Hager

H. A.

(2015) A framework to guide the conservation of species hybrids based on ethical and ecological considerations. Conservation Biology 29: 1040–1051.

58.

Jones

P. J.

(1994) Biodiversity in the Gulf of Guinea: An overview. Biodiversity & Conservation 3: 772–784.

59.

Jones

P. J.

Burlison

J. P.

Tye

(1991) Conservação dos ecossistemas florestais na república democrática de São Tomé e Príncipe, Gland, Switzerland: International Union for Conservation of Nature.

60.

Juste

B. J.

J. E.

(1994) Biodiversity conservation in the Gulf of Guinea Islands: Taking stock and preparing action. Biodiversity & Conservation 3: 759–771.

61.

Kim

D.-H.

Sexton

J. O.

Townshend

J. R.

(2015) Accelerated deforestation in the humid tropics from the 1990s to the 2000s. Geophysical Research Letters 42: 3495–3501.

62.

Lejoly

(2000) Taxonomy and biodiversity conservation in Central Africa. Portugaliae Acta Biologica 19: 61–70.

63.

Lewis

S. L.

Edwards

D. P.

Galbraith

(2015) Increasing human dominance of tropical forests. Science 349: 827–832.

64.

Luck

G. W.

(2007) A review of the relationships between human population density and biodiversity. Biological Reviews 82: 607–645.

65.

Luck

G. W.

Daily

G. C.

Ehrlich

P. R.

(2003) Population diversity and ecosystem services. Trends in Ecology & Evolution 18: 331–336.

66.

Magnuson

J. J.

Crowder

L. B.

Medvick

P. A.

(1979) Temperature as an ecological resource. American Zoologist 19: 331–343.

67.

McGeoch

M. A.

van Rensburg

B. J.

Botes

(2002) The verification and application of bioindicators: A case study of dung beetles in a savanna ecosystem. Journal of Applied Ecology 39: 661–672.

68.

Mendenhall

C. D.

Frishkoff

L. O.

Santos-Barrera

Pacheco

Mesfun

Quijano

F. M.

Pringle

R. M.

(2014) Countryside biogeography of Neotropical reptiles and amphibians. Ecology 95: 856–870.

69.

Ministry of Infrastructure, Natural Resources and Environment. (2016). National Biodiversity Strategy and Action Plan 2015-2020 (NBSAP II). Democratic Republic of São Tomé and Príncipe. Retrieved from https://www.cbd.int/doc/world/st/st-nbsap-v2-en.pdf.

70.

Monod

(1960) Notes botaniques sur les îles de São Tomé et de Príncipe. Bulletin de l'Institut Français de l'Afrique Noire 22: 19–83.

71.

Myers

Mittermeier

R. A.

Mittermeier

C. G.

da Fonseca

G. A.

Kent

(2000) Biodiversity hotspots for conservation priorities. Nature 403: 853–858.

72.

Navas

C. A.

Otani

(2007) Physiology, environmental change, and anuran conservation. Phyllomedusa 6: 83–103.

73.

Newbold

Hudson

L. N.

Hill

S. L. L.

Contu

Lysenko

Senior

R. A.

Purvis

(2015) Global effects of land use on local terrestrial biodiversity. Nature 520: 45–50.

74.

Nowakowski

A. J.

Thompson

M. E.

Donnelly

M. A.

Todd

B. D.

(2017) Amphibian sensitivity to habitat modification is associated with population trends and species traits. Global Ecology and Biogeography 26: 700–712.

75.

Oksanen, J., Blanchet, F. G., Kindt, R., Legendre, P., Minchin, P. R., O'Hara, R. B., . . . Wagner, H. (2015). Vegan: Community ecology package (R package version 2.2-1). Retrieved from http://www.pelagicos.net/MARS6910_spring2015/manuals/R_vegan.pdf.

76.

Oyono

P. R.

Morelli

T. L.

Sayer

Makon

Djeukam

Hatcher

Bigombé

(2014) Allocation and use of forest land: Current trends, issues and perspectives. In: de Wasseige

Flynn

Louppe

Hiol Hiol

Mayaux

(eds) The forests of the Congo Basin: State of the forest 2013, Neufchâteau, Belgium: Weyrich, pp. 215–240.

77.

Pearman

P. B.

Velasco

A. M.

López

(1995) Tropical amphibian monitoring: A comparison of methods for detecting inter-site variation in species' composition. Herpetologica 51: 325–337.

78.

Pineda

Moreno

Escobar

Halffter

(2005) Frog, bat, and dung beetle diversity in the cloud forest and coffee agroecosystems of Veracruz, Mexico. Conservation Biology 19: 400–410.

79.

Pittman

S. E.

Osbourn

M. S.

Semlitsch

R. D.

(2014) Movement ecology of amphibians: A missing component for understanding population declines. Biological Conservation 169: 44–53.

80.

QGIS Development Team. (2016). QGIS Geographic Information System (version 1.8.0) [Computer software]. Open Source Geospatial Foundation Project. Retrieved from http://www.qgis.org/.

81.

R Core Team. (2013). R: A language and environment for statistical computing (version 3.0.0) [Computer software]. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from https://www.r-project.org/.

82.

Refsnider

J. M.

Janzen

F. J.

(2010) Putting eggs in one basket: Ecological and evolutionary hypotheses for variation in oviposition-site choice. Annual Review of Ecology, Evolution, and Systematics 41: 39–57.

83.

Rödel

M.-O.

Ernst

(2004) Measuring and monitoring amphibian diversity in tropical forests: I. An evaluation of methods with recommendations for standardization. Ecotropica 10: 1–14.

84.

Rudolf

V. H. W.

Rödel

M.-O.

(2005) Oviposition site selection in a complex and variable environment: The role of habitat quality and conspecific cues. Oecologia 142: 316–325.

85.

Sala

O. E.

(2000) Global biodiversity scenarios for the year 2100. Science 287: 1770–1774.

86.

Salgueiro

Carvalho

(2002) Proposta de plano nacional de desenvolvimento florestal. Direcção das florestas da República democrática de São Tomé e Príncipe & ECOFAC. 98.

87.

Schiøtz

(1999) Treefrogs of Africa, Frankfurt am Main, Germany: Edition Chimaira.

88.

Schneider

C. A.

Rasband

W. S.

Eliceiri

K. W.

(2012) NIH Image to ImageJ: 25 years of image analysis. Nature Methods 9: 671–675.

89.

Schneider-Maunoury

Lefebvre

Ewers

R. M.

Medina-Rangel

G. F.

Peres

C. A.

Somarriba

Pfeifer

(2016) Abundance signals of amphibians and reptiles indicate strong edge effects in Neotropical fragmented forest landscapes. Biological Conservation 200: 207–215.

90.

Stuart, S. N., Chanson, J. S., Cox, N. A., Young, B. E., Rodrigues, A. S. L., Fischman, D. L., & Waller, R. W. (2004). Status and trends of amphibian declines and extinctions worldwide. Science, 306, 1783–1786.

91.

Thurman

L. L.

Garcia

T. S.

Hoffman

P. D.

(2014) Elevational differences in trait response to UV-B radiation by long-toed salamander populations. Oecologia 175: 835–845.

92.

Tscharntke

Clough

Bhagwat

S. A.

Buchori

Faust

Hertel

Wanger

T. C.

(2011) Multifunctional shade-tree management in tropical agroforestry landscapes: A review. Journal of Applied Ecology 48: 619–629.

93.

United Nations, Department of Economic and Social Affairs, Population Division. (2017). World population prospects: The 2017 revision, custom data acquired via website. Retrieved August 10, 2017, from https://esa.un.org/unpd/wpp/DataQuery/.

94.

Uyeda

J. C.

Drewes

R. C.

Zimkus

B. M.

(2007) The California Academy of Sciences Gulf of Guinea expeditions (2001, 2006): VI. A new species of Phrynobatrachus from the Gulf of Guinea Islands and a reanalysis of Phrynobatrachus dispar and P. feae (Anura: Phrynobatrachidae). Proceedings of the California Academy of Sciences 58: 367–385.

95.

Venables

W. N.

Ripley

B. D.

(2002) Modern applied statistics with S, (4th ed.). New York, NY: Springer.

96.

Vredenburg

V. T.

(2004) Reversing introduced species effects: Experimental removal of introduced fish leads to rapid recovery of a declining frog. Proceedings of the National Academy of Sciences, USA 101: 7646–7650.

97.

Wake

D. B.

Vredenburg

V. T.

(2008) Are we in the midst of the sixth mass extinction? A view from the world of amphibians. Proceedings of the National Academy of Sciences, USA 105: 11466–11473.

98.

Wanger, T. C., Iskandar, D. T., Motzke, I., Brook, B. W., Sodhi, N. S., Clough, Y., & Tscharntke, T. (2010). Effects of land-use change on community composition of tropical amphibians and reptiles in Sulawesi, Indonesia. Conservation Biology, 24, 795–802.

99.

Wanger, T. C., Saro, A., Iskandar, D. T., Brook, B. W., Sodhi, N. S., Clough, Y., & Tscharntke, T. (2009). Conservation value of cacao agroforestry for amphibians and reptiles in South-East Asia: Combining correlative models with follow-up field experiments. Journal of Applied Ecology, 46, 823–832.

100.

Whitworth

Villacampa

Serrano Rojas

S. J.

Downie

MacLeod

(2017) Methods matter: Different biodiversity survey methodologies identify contrasting biodiversity patterns in a human modified rainforest – A case study with amphibians. Ecological Indicators 72: 821–832.

101.

Wilkie

M. L.

Eckelmann

C. M.

Laverdiere

Mathias

(2002) Forests and forestry in Small Island Developing States. International Forestry Review 4: 257–267.

102.

Wolters

Bengtsson

Zaitsev

A. S.

(2006) Relationship among the species richness of different taxa. Ecology 87: 1886–1895.

103.

Zuur

A. F.

Ieno

E. N.

Elphick

C. S.

(2010) A protocol for data exploration to avoid common statistical problems: Data exploration. Methods in Ecology and Evolution 1: 3–14.