Ecology and Conservation

Amphibian communities of the dry forest of Western Madagascar:

Taxonomy, Ecology and Conservation

Dissertation zur Erlangung des naturwissenschaftlichen Doktorgrades der Bayerischen Julius-Maximilians-Universität Würzburg

vorgelegt von Julian Glos

aus Werneck

Würzburg, 2006

Eingereicht am: ______

Mitglieder der Promotionskommission:

Vorsitzender: Prof. Dr. ______Gutachter: Prof. Dr. K. Eduard Linsenmair Gutachter: Prof. Dr. Jürgen Tautz

Tag des Promotionskolloquiums: ………………………………………..

Doktorurkunde ausgehändigt am: …………………………………………… List of Contents

Page

PREFACE 1

TAXONOMY, LIFE HISTORY AND BREEDING ECOLOGY Chapter 1 The amphibian fauna of the Kirindy dry forest in western Madagascar. 2

GLOS, J. 2003. Salamandra. 39:75-90. Chapter 2 A new species of Scaphiophryne from Western Madagascar. 19

GLOS, J., F. GLAW, and M. VENCES. 2005. Copeia 2005:252-261. Chapter 3 Description of the tadpoles of Aglyptodactylus laticeps and A. securifer 30 from western Madagascar, with notes on life history and ecology.

GLOS, J., and K. E. LINSENMAIR. 2004. Journal of Herpetology. 38:131- 136. Chapter 4 Description of the tadpoles of Boophis xerophilus and B. doulioti from 37 Western Madagascar with notes on larval life history and breeding ecology.

GLOS, J., and K. E. LINSENMAIR. 2005. Amphibia-Reptilia. 26:459-466.

COMMUNITY ECOLOGY Chapter 5 Living apart together – patterns of tadpole communities in a western 46 Madagascan dry forest.

CONSERVATION Chapter 6 Modeling the habitat use of an endangered dry-forest frog from Western 66 Madagascar Chapter 7 Oviposition site selection in a Madagascan frog – Experimental evaluation 86 of a habitat model

PREDATOR-PREY INTERACTIONS Chapter 8 Mixed-species social aggregations in Madagascan tadpoles – determinants 102 and species composition Chapter 9 Causes and costs of social aggregation: a comparative study on two tadpole 118 species from western Madagascar

Appendix Aquatic Zebras? - The tadpoles of the Madagascan treefrog Boophis 135 schuboeae, compared to those of B. ankaratra

GLOS, J., M. THOMAS, and M. VENCES. Tropical Zoology. In press.

ZUSAMMENFASSUNG (SUMMARY IN GERMAN) 146

LIST OF ABBREVIATIONS 158

ACKNOWLEDGEMENTS 159

CURRICULUM VITAE 160

EHRENWÖRTLICHE ERKLÄRUNG 162

Amphibian communities in Madagascar Julian Glos 1

PREFACE

Horakoraka ny an’ny sahona fa ny tsiboboka no tompon’ny rano.

Frogs make a lot of noise, but it is the tadpoles that are the true kings of the water.

Madagascan proverb

This thesis is the result of four years of fieldwork on larval amphibian communities in the dry forest of Western Madagascar. I present therein various taxonomic, ecological and conservation aspects on this community. The thesis is subdivided in four major topic blocks: Taxonomy, life history and breeding ecology (Chapter 1 to 4), community ecology (Chapter 5), conservation (Chapter 6 and 7), and predator-prey interactions (Chapter 8 and 9). Further taxonomic work on larval amphibians is presented at the end of the thesis (Appendix). The latter work was conducted in the context of this dissertation; however, the species dealt with in this part do not belong to the community that was in my primary focus.

Each topic is presented as a self-contained chapter with its own introduction, methods description, results section, discussion and list of the cited literature. Each chapter can be read without referring to the other chapters. English summaries are given at the beginning of each chapter. A summary in German is given at the end of the thesis.

At the time of submission of this thesis, Chapter 1 to 4 are published in peer reviewed journals, and the Appendix is accepted for publication. These chapters are formatted in the style of the respective journal. Chapter 8 is submitted for publication. All other chapters will soon be submitted to peer reviewed journals.

Amphibian communities in Madagascar Julian Glos 2

Chapter 1

The amphibian fauna of the Kirindy dry forest in western Madagascar

Amphibian communities in Madagascar Julian Glos 3

The amphibian fauna of the Kirindy dry forest in western Madagascar

JULIAN GLOS

Zusammenfassung Die Amphibienfauna des Kirindy-Trockenwaldes in Westmadagaskar. Die Amphibienfauna des Kirindy-Waldes in Westmadagaskar wurde über vier Jahre hinweg im Rahmen einer ökologischen Studie untersucht. Sie besteht aus 15 Arten aus vier verschiedenen Anurenfamilien (Mantellidae, Ranidae, Hyperoliidae, Microhylidae). Unter diesen Arten gibt es ein weites Spektrum von Explosionslaichern und solchen, die sich über die ganze Regenzeit hinweg fortpflanzen. In bezug auf die Wahl des Laichgewässers gibt es ähnlich viele Generalisten und Spezialisten. Vermutlich als Anpassung an die hohe Ephemeralität der Laichgewässer ist die larvale Entwicklungszeit einiger Arten sehr kurz. Zu den Arten des Kirindy-Waldes werden weitere Verhaltensbeobachtungen und ökologische Merkmale vorgestellt. Schlagwörter: Madagaskar; Kirindy-Wald; saisonaler Trockenwald; Amphibien; Habitatwahl; Laichgewässerwahl.

Abstract The amphibian fauna of the Kirindy forest in western Madagascar was analysed over the course of four years within the context of an ecological study. It consists of 15 species out of four anuran families (Mantellidae, Ranidae, Hyperoliidae, Microhylidae). There is a wide spectrum of explosive and prolonged breeders. About the same number of habitat specialists and generalists is present. Larval developmental time of several species is very short presumably as an adaptation to the high ephemerality of the breeding ponds. Further behavioural observations and ecological data are provided. Keywords: Madagascar; Kirindy; dry forest; amphibians; habitat choice; breeding site choice.

1 Introduction The diversity and uniqueness of its fauna and flora makes Madagascar one of the major hot spots of biodiversity (MYERS et al. 2000). Within Madagascar, amphibians are of special interest. Since the break-off of Madagascar from the African continent some 100 million years ago, the Malagasy amphibians evolved strictly separate from their continental counterparts. As a consequence, almost all frogs of Madagascar are endemics to this island (GLAW & VENCES 1994). Within Madagascar, the dry deciduous forest of the western part of the country stands out as it ranks among the most endangered ecosystems of the world (JANZEN 1992). However, very little is known about the fauna and flora present in that region as well as their biology and ecological interactions. In this paper I introduce the amphibian fauna of the Kirindy forest in western Madagascar, give an overview of its breeding sites and present further behavioural and ecological data.

2 Study site The study site was the ‘Forêt de Kirindy’, a deciduous dry forest at the west coast of Madagascar, 60 km north of Morondava and about 20 km inland (44°39’ E, 20°03 S; 18 - 40 m above sea level; SORG & ROHNER 1996; Fig. 1). The area of the Kirindy forest covers about 12.000 ha and thus may be among the largest remaining continuous forests in western Madagascar (NELSON & HORNING 1993, RAKOTONIRINA 1996). The forest

Salamandra, Rheinbach, 30.06.2003, 39(2): 75-90. SALAMANDRA,© 2003 Deutsche Rheinbach,Gesellschaft 39für(2), Herpetologie 2003 und Terrarienkunde e.V. (DGHT) 75 Amphibian communities in Madagascar Julian Glos 4

JULIAN GLOS

is intersected by the Kirindy river and is surrounded by a tree-shrub savanna of anthropogenic origin. This savanna is burned in regular intervals and grazed on by Zebu cows and goats. The Kirindy forest is managed by the ‘Centre de Formation Professionelle Forestière’ (C.F.P.F.) and is exploited by selective logging and eco- tourism. The Kirindy field station, supported by the German Primate Centre (DPZ, Göttingen), is located in the centre of the forest. The climate is characterized by a marked seasonality. Almost all rain falls in the austral summer from November to March, followed by eight months of virtually no rain. Following SORG & ROHNER (1996), annual mean rainfall is 800 mm. Rainfall during the study was 803 mm in the rainy season 1998/99, 1265 mm in 1999/2000 and 874 mm in 2000/01 (own data). Temperatures during the rainy season range between 22 °C at night and > 40 °C during the day (SORG & ROHNER 1996, own data). There are breeding sites for amphibians in three different habitat types: the closed forest (Fig. 2), the bed of the Kirindy river (before the river is running; Fig. 3), and the surrounding savanna (Fig. 4). All breeding ponds but a few pools in the river bed dry out completely during the dry season. They are successively filled with the onset of rain in November and December. The first waters that are used by amphibians for spawning usually arise in the rocky parts of the Kirindy river bed. As a rule, these ponds offer a low risk of desiccation and a low density of invertebrate predators. During the course of the rainy season, these waters connect with each other and eventually fish emigrate from the few permanent

Menabe region

Madagaskar

Fig. 1. Map of Madagascar (left) with the Menabe region (Box), precise location of the Kirindy forest (right). Lage der Region Menabe innerhalb Madagascars (links), genaue Lage des Kirindy-Waldes in der Region Menabe (rechts).

76 SALAMANDRA, Rheinbach, 39(2), 2003 Amphibian communities in Madagascar Julian Glos 5

The amphibian fauna of the Kirindy dry forest in western Madagascar

Fig. 2. Larger breeding pond within the closed forest. Größerer Laichtümpel innerhalb des geschlossenen Waldes.

Fig. 3. Rock pool in the Kirindy river bed with breeding Aglyptodactylus securifer. Felstümpel im Bett des Kirindy-Flusses mit sich paarenden Aglyptodactylus securifer.

Fig. 4. Savanna habitat of human origin surrounding the Kirindy forest. Savannenhabitat anthropogenen Ursprungs außerhalb des geschlossenen Waldes.

SALAMANDRA, Rheinbach, 39(2), 2003 77 Amphibian communities in Madagascar Julian Glos 6

JULIAN GLOS

pools in the river bed, representing then a very high predation risk for amphibian larvae. Usually in December, the first breeding ponds arise in the closed forest. Depending on amount and distribution of precipitation and habitat parameters of the ponds themselves, the persistence of these waters varies from three days to five months. In the savanna, breeding ponds for amphibians arise only after heavy rainfalls. They are fully sun-exposed and often dry out very quickly.

3 Methods The study was conducted in the respective rainy seasons of four years (Dec 1998 to April 1999, Dec 1999 to April 2000, Nov 2000 to June 2001, Nov 2001 to Feb 2002). The activity data are based upon vocal and sight control walks. For that purpose, the study ponds were visited in regular intervals (day and night) and checked for species and number of frogs as well as their calling activity. Additionally, the use of breeding ponds was checked by analysing tadpole communities. This was done by standardized dip-netting and the box-method (RÖDEL 1998). In total, calling activity and tadpoles communities were analysed at 191 breeding sites (65 forest ponds, 111 river bed ponds, 15 savanna ponds). Snout-vent length of frogs was measured to the nearest 0.1 mm using callipers. Systematics follow VENCES & GLAW (2001).

4 Results and discussion

4.1 General aspects In total, the amphibian fauna consists of 15 species representing four anuran families. Snout-vent length data to the most abundant species in Kirindy gives Table 1. The

Males / Männchen Females / Weibchen Species / Art ME ± SD N Min Max ME ± SD N Min Max Aglyptodactylus securifer 3.58 ± 0.25 78 3.00 4.45 4.29 ± 0.56 44 3.24 4.85 Aglyptodactylus laticeps 4.35 ± 0.28 74 3.25 4.91 5.90 ± 0.30 43 5.44 7.20 Laliostoma labrosum 4.89 ± 0.32 9 4.37 5.65 6.18 ± 0.47 9 5.59 6.83 Ptychadena mascareniensis 3.24 ± 0.15 4 3.14 3.49 3.63 ± 0.09 4 3.55 3.78 Boophis doulioti 3.85 ± 0.26 8 3.40 4.24 4.83 ± 0.41 7 3.93 5.24 Boophis xerophilus 3.60 ± 0.44 32 2.93 4.36 4.13 ± 0.56 6 3.39 4.88 Heterixalus luteostriatus 2.86 ± 0.19 11 2.56 3.16 3.09 ± 0.22 5 2.85 3.35 Heterixalus tricolor 2.78 ± 0.17 110 2.02 3.32 3.18 ± 0.19 32 2.88 3.59 Heterixalus carbonei 2.60 ± 1.09 9 2.41 2.80 2.59 ± 0.11 6 2.45 2.81 Dyscophus insularis 3.78 ± 0.39 26 3.44 5.21 4.19 ± 0.25 14 3.78 4.65 Scaphiophryne calcarata 2.84 ± 0.30 15 2.39 3.45 2.98 ± 0.22 7 2.55 3.30 Scaphiophryne brevis 3.81 ± 0.23 3 3.49 4.01 4.03 ± 0.43 3 3.55 4.60

Tab. 1. Adult snout-vent lengths (cm) of most abundant amphibian species at the Kirindy forest. ME = mean, Max = maximum, Min = minimum, N = number of measured specimens, SD = standard deviation. Included are only animals that actively participated in mating. Kopf-Rumpf Längen (cm) der häufigsten Amphibienarten des Kirindy-Waldes. ME = Mittelwert, Max = Maximum, Min = Minimum, N = Anzahl gemessener Tiere, SD = Standardabweichung. Aufgeführt sind nur Tiere, die aktiv am Fortpflanzungsgeschehen teilgenommen haben.

78 SALAMANDRA, Rheinbach, 39(2), 2003 Amphibian communities in Madagascar Julian Glos 7

The amphibian fauna of the Kirindy dry forest in western Madagascar

anuran community of the Kirindy forest covers a wide spectrum of explosive breeders, reproducing only after heavy rainfalls, and prolonged breeders that are found reproducing over large parts of the rainy season (Tab. 2). In this study, explosive breeding was defined as breeding exclusively in the first two nights after rainfalls of > 30 mm. Apart from Mantella betsileo, which is calling regularily both at day and night time, all species are primarily nocturnal. After heavy rains that are creating new breeding sites, however, breeding activity in Aglyptodactylus spp., Boophis spp. and Dyscophus insularis can extend for several hours after sunrise. While Heterixalus spp. lay their

Species / Art Embryonic Breeding Diurnal/ Spawn Pond use Habitat use development strategy nocturnal characteristics Aglyptodactylus < 24 h EB D / N SUR S RB securifer A. laticeps < 24 h EB N SUR S FO Laliostoma < 24 h EB N SUR G FO / RB / SAV labrosum Ptychadena no data PB N SUR G FO / SAV mascareniensis Mantella betsileo no data PB D / N TER G FO / RB Boophis doulioti < 24 h PB N SUR G FO / RB / SAV B. xerophilus < 24 h PB N SUR (S) FO / RB / SAV Heterixalus ~ 3 days PB N SEP (G) FO / RB / SAV luteostriatus H. tricolor ~ 3 days PB N SEP S FO H. carbonei ~ 3 days PB N SEP S FO Dyscophus < 24 h EB N SUR G FO / RB / SAV insularis Scaphiophryne < 24 h EB N SUR G FO / RB / SAV calcarata S. brevis < 24 h EB N no data G FO

Tab. 2. Summarized ecological characteristics of the most abundant amphibians of the Kirindy forest. D = diurnal, EB = explosive breeder, FO = closed forest, G = generalist, N = nocturnal, PB = prolonged breeder, RB = ponds in the river bed, S = specialist, SAV = savanna, SEP = submersed egg packets, SUR = eggs as surface film, TER = terrestrial nest. Data in parentheses indicate that there are rare exceptions from the rule. Zusammenfassung ökologischer Charakteristika der häufigsten Amphibien des Kirindy-Waldes. D = tagaktiv, EB = Explosionslaicher, FO = geschlossener Wald, G = Generalist, N = nachtaktiv, PB = Brutaktivität über lange Zeit hinweg, RB = Tümpel im Flußbett, S = Spezialist, SAV = Savanne, SEP = Eipakete unter der Wasseroberfläche, SUR = Eier als Oberflächenfilm, TER = terrestrisches Nest. Angaben in Klammern beziehen sich auf seltene Beobachtungen, die von dem Angegebenen abweichen.

SALAMANDRA, Rheinbach, 39(2), 2003 79 Amphibian communities in Madagascar Julian Glos 8

JULIAN GLOS

Fig. 5. Amplectant pair of Aglyptodactylus securifer. Paar von Agylptodactylus securifer im Amplexus.

Fig. 6. Aglyptodactylus laticeps female/Weibchen.

eggs in small packets under water and Mantella betsileo has terrestrial clutches, most of the species deposit their eggs as surface layers in stagnant waters. The embryonic and larval development is generally very fast, presumably as an adaptation to the unpredictability of rainfall and subsequent high ephemerality of the breeding ponds.

80 SALAMANDRA, Rheinbach, 39(2), 2003 Amphibian communities in Madagascar Julian Glos 9

The amphibian fauna of the Kirindy dry forest in western Madagascar

Fig. 7. Predation of Boophis doulioti by the iguanid Oplurus cuvieri (GRAY, 1831). Der Dornschwanz Oplurus cuvieri (GRAY, 1831) beim Fressen eines Boophis doulioti.

Fig. 8. Boophis xerophilus female/Weibchen.

SALAMANDRA, Rheinbach, 39(2), 2003 81 Amphibian communities in Madagascar Julian Glos 10

JULIAN GLOS

Accordingly, embryonic development in the majority of species is < 24 h, in Heterixalus about three days (Tab. 2). Moreover, larval development in some species (Aglyptodactylus laticeps, Scaphiophryne calcarata, S. brevis) can be as short as 10 days and is therefore among the fastest known for amphibians (compare RÖDEL 1998 for Bufo pentoni ANDERSSON, 1893). Seven species are specialists in respect to breeding site choice, eight species are generalists in this matter (Tab. 2). Specialists are defined as species using breeding ponds with biotic and abiotic parameters of only a small range, e.g. only small, ephemeral ponds without any water vegetation and low canopy cover. In contrast, generalists use breeding ponds with pond characteristics of a wide range.

4.2 Species accounts Mantellidae Laliostominae Aglyptodactylus securifer GLAW, VENCES & BÖHME, 1998 This species was described first from the Kirindy forest (GLAW et al. 1998). It is known only from the type locality and another area further north (Berara; ANDREONE et al. 2001). Aglyptodactylus securifer is an extreme explosive breeder, reproducing almost exclusively in the rock pools of the Kirindy river while the river is not yet running. These are the only open waters at that time of the season. This narrow spatial niche is accompanied by a very narrow temporal niche. Breeding takes only place in the first days after the very first heavy rains, generally in mid November to early December (1.12.1998, 8.12.1999, 13.11.2000, 16.11.2001). At this time up to several hundreds of calling males and amplectant pairs can be observed at one single pond of < 20 m2. Calling starts early at night, but the breeding activity can extend to the next mid-day. During immediate breeding activity there is a clear sexual dichromatism. While females and non breeding males are uniformly brown coloured, calling males and those in amplectant pairs have yellow lateral bands if breeding occurs at night (Fig. 5), or are bright yellow if breeding during daytime. Amplectant males detached from their females change from yellow to brown within a few minutes. However, it is not known whether the conspicuous yellow colour serves in the context of sexual selection or is simply a non-adaptive by-product of hormonal changes during breeding. When threatened, some individuals of A. securifer reacted with death-feigning behaviour. They turn actively on their backs, laying motionless with spread out legs. This behaviour was observed in both adult and freshly metamorphosed individuals. After breeding took place, adults of A. securifer were found feeding in the forest throughout the rainy season more than 1 km away from the river bed. The developmental time of the tadpoles is fast. Tadpoles raised in the field camp (density: 10 tadpoles / 15 l) reached GOSNER (1960) stage 42 in 21.5 ± 1.7 days (mean ± SD; range 19 - 32, n = 208). Tadpoles tend to form large aggregations as a reaction to an immediate predation risk.

Aglyptodactylus laticeps GLAW, VENCES & BÖHME, 1998 This species is only known from the Kirindy forest so far (GLAW et al. 1998). It is found almost exclusively in the closed forest. Preferred spawning sites are freshly arisen small and medium sized forest ponds, more rarely small puddles on forest roads. Aglypto- dactylus laticeps (Fig. 6) is an explosive breeder that reproduces usually only after

82 SALAMANDRA, Rheinbach, 39(2), 2003 Amphibian communities in Madagascar Julian Glos 11

The amphibian fauna of the Kirindy dry forest in western Madagascar

rainfalls that exceed 30 mm of precipitation. In contrast to A. securifer, the breeding season starts later and extends over the whole rainy season. First calling activity in the study period was recorded on 23.12.1998, 13.12.1999, 25.12.2000 and 18.12.2001. Calling males form choruses of up to 80 individuals that are initiated by single males starting to call. The choruses fade out together and set in again after a few minutes. The members of these choruses are evenly spaced around the breeding pond, 0.3 to 3 m from the water edge. When captured, A. laticeps utters distress calls that are fairly different from their mating calls. Larval development is very short (minimum of 10 days in the field). Tadpoles raised in the field camp (density: 10 tadpoles / 15 l) reached GOSNER (1960) stage 42 in 17.1 ± 3.8 days (mean ± SD; range 10 - 29, n = 391).

Laliostoma labrosum (COPE, 1868) This species (formerly Tomopterna labrosa) together with Aglyptodactylus was recently assigned to the new subfamily Laliostominae (VENCES & GLAW 2001). Lalio- stoma labrosum is widely distributed throughout western Madagascar (GLAW & VENCES 1994). In Kirindy, it is an explosive breeder, reproducing only after heavy rainfalls. It uses ponds of a wide variety of sizes and structures for breeding. Males generally call from the water edge and do not form choruses but are distributed over the breeding ponds. It is found in the closed forest and the river bed as well as in the savanna and in surrounding villages and towns. Tadpoles of L. labrosum were observed to be facultative carnivorous and preyed actively (and successfully) on smaller tadpoles (10-15 mm total length) of Heterixalus sp., Boophis sp. and Dyscophus insularis GRANDIDIER, 1872. However, their main food source was dead plant material. Larval developmental time is about one month.

Boophinae Boophis doulioti (ANGEL, 1934) This species was resurrected from the synonymy of Boophis tephraeomystax (DUMÉRIL, 1853) by VENCES & GLAW (2002). It is clearly adapted to anthropogenic altered habitats. It is found in the villages around the forest and even in the town of Morondava. In Kirindy, it uses all pond types for reproduction and is found in all major habitat types. Its breeding activity is relatively independent of rainfall and time within the rainy season. It calls from the ground, evenly spaced around the breeding pond, as well as from perches at the water edge (> 200 observations). When individuals that were originally calling on the ground were disturbed, they flew into the bushes and continued calling from higher perches within a few minutes. On one occasion, a successful predation event on B. doulioti by Oplurus cuvieri (GRAY, 1831) (Iguanidae) was observed (Fig. 7).

Boophis xerophilus GLAW & VENCES, 1997 This frog (Fig. 8) was described by GLAW & VENCES (1997) from the Kirindy forest. It is known from the type locality and Berenty in southern Madagascar. It resembles superficially the syntopic Boophis doulioti. However, it is different ecologically in respect to a more specialized use of breeding sites. Boophis xerophilus generally uses larger and more permanent waters as breeding ponds. On rare occasions, however, B. xerophilus was observed to breed in small ephemeral puddles. It is found primarily in the closed forest, only rarely in the river bed and the savanna. Breeding activity

SALAMANDRA, Rheinbach, 39(2), 2003 83 Amphibian communities in Madagascar Julian Glos 12

JULIAN GLOS

Fig. 9. Male Heterixalus tricolor in typical calling position during night time. Heterixalus tricolor-Männchen in typischer Rufposition während der Nacht.

Fig. 10. Male Heterixalus tricolor in resting position during daytime. Heterixalus tricolor-Männchen in Ruhestellung während des Tages.

84 SALAMANDRA, Rheinbach, 39(2), 2003 Amphibian communities in Madagascar Julian Glos 13

The amphibian fauna of the Kirindy dry forest in western Madagascar

Fig. 11. Calling male Heterixalus carbonei. Rufendes Heterixalus carbonei-Männchen.

Fig. 12. Green colour morph of Scaphiophryne calcarata. Grüne Farbvariation von Scaphiophryne calacarata.

Fig. 13. Scaphiophryne cf. marmorata male/ Männchen.

SALAMANDRA, Rheinbach, 39(2), 2003 85 Amphibian communities in Madagascar Julian Glos 14

JULIAN GLOS

stretches over the whole rainy season. However, activity peaks after heavy rainfalls. Boophis xerophilus males are about evenly spread around the breeding pond and mostly call from the ground not more than 2 m from the water edge (> 200 observations). Often two males form calling duets alternating their calls.

Mantellinae Mantella betsileo (GRANDIDIER, 1872) This species was recorded for this region by KUCHLING (1993). Mantella expectata BUSSE & BÖHME, 1992 was not recorded for the Kirindy forest as it could be expected following BUSSE & BÖHME (1992). Colouration of M. betsileo of the Kirindy forest ranges from bright red to ochre-yellow. However, the characteristic diamond shaped markings were present in all recorded individuals (n > 100). It stands out ecologically from all other species by its cathemerality and terrestrial breeding site choice. Individuals were found calling and foraging both during day and night (n > 300). Highest levels of calling activity were recorded at night near relatively permanent pools, which is unusual for Mantella species (STANISZEWSKI 2001). Breeding activity was independent of rainfall. On 9 January 1999, one clutch of M. betsileo containing 35 yellow eggs was found under a fallen log about two meters next to a forest pond (500 m2).

Mantidactylus cf. wittei GUIBÉ, 1974 This species is the only member of the very specious genus Mantidactylus in the Kirindy forest. A final species affiliation of two voucher specimens and call is in progress. It is morphologically similar to M. wittei. Mantidactylus cf. wittei is very rare in the Kirindy forest. In four years and almost 200 ponds studied, it was found only in two nights at one pond. This pond was one of the largest (about 8000 m2, depth up to 165 cm) and one of the most permanent (up to five months) ponds in the Kirindy forest. There, males called at night from a bush at the water edge about 1 to 4 m above ground. At Andrakata, GLAW & VENCES (1994) found egg clutches of M. wittei attached to leaves 1 to 1.5 m above water surface. No egg clutches or tadpoles of this species were found in the Kirindy forest.

Ranidae Ptychadena mascareniensis (DUMÉRIL & BIBRON, 1841) This species is one of the few non-endemics to Madagascar. It is distributed throughout Madagascar and is one of the most abundant frogs (GLAW & VENCES 1994). While it is found only sporadically in the closed Kirindy forest, it is the dominant species in the savanna. This is consistent with the view that P. mascareniensis profits by anthropogenic disturbances and is rarely found in primary forests (GLAW & VENCES 1994).

Hyperoliidae Heterixalus tricolor (BOETTGER, 1881) This frog has a scattered distribution in northern and western Madagascar (GLAW & VENCES 1994). It is a prolonged breeder that starts its breeding activities not earlier than one month after the beginning of the rains. It uses only the most permanent ponds as breeding sites (> 500 m2; e.g. Fig. 2) that are grown with submergent and emergent vegetation. Males are calling consistently throughout the rainy season from within the pond, either from the floating leaves of water-lilies or from reed grass stems (> 1000

86 SALAMANDRA, Rheinbach, 39(2), 2003 Amphibian communities in Madagascar Julian Glos 15

The amphibian fauna of the Kirindy dry forest in western Madagascar

observations; Fig. 9). The calling activity is relatively independent of rainfall. Individuals of this species spend the day sitting motionless on leaves of bushes or grasses not more than 1.5 m above ground level, mainly inside the pond and fully sun- exposed (> 30 observations). There, they face a considerable predation risk by spiders. On six occasions, large spiders (Pisauridae) were observed to suck out H. tricolor individuals. These spiders are quite abundant on the reed grass stems within the pond. The colouration of H. tricolor changes from yellow or brownish at night to bright white during the day (Fig. 10). A similar behaviour is found in West-African Hyperolius nitidulus (RÖDEL 2000). There, juveniles spend several months of the dry season on grass stems fully exposed to the sun (SPIELER 1997; RÖDEL 2000). Several behavioural and morphological adaptations help to reduce the desiccation risk of these frogs (SCHMUCK et al. 1988, 1994; KOBELT & LINSENMAIR 1992, 1995; LINSENMAIR 1998). Parallel to Hyperolius (GRAFE et al. 2002), H. tricolor individuals remain vigilant during their daily sunbath as they flee into the water or into bushes when disturbed.

Heterixalus carbonei VENCES, GLAW, JESU & SCHIMMENTI, 2000 This frog was recently described by Vences et al. (2000) from the “Tsingy de Bemaraha”. The Kirindy forest represents a further record for this species (Fig. 11). Males and females show the same colouration, with females being considerable larger (Tab. 1). In Kirindy, this species shares almost all habitats with H. tricolor (e.g. Fig. 2) and shows the same habitat choice and seasonal activity pattern. However, there is almost no overlap in calling activity between Heterixalus tricolor and H. carbonei during one night. While H. tricolor calls from sunset to about midnight, H. carbonei starts calling later at night, being rarely active before 11:00 p.m. (> 50 controls). In contrast to H. tricolor, males of H. carbonei call mainly from high perches at the pond edge (often > 3 m; n > 300). As H. tricolor, they spend the day fully sun-exposed mainly on leaves of bushes that surround the breeding pond (> 30 observations). In full sun, their colour is a shiny silver, and shows characteristic dark grey dorsolateral bands.

Heterixalus luteostriatus (ANDERSSON, 1910) The locally very abundant frog is found throughout all Kirindy habitat types as well as anthropogenic habitats (villages, rice paddies, towns). It is less pretentious compared to the two other Heterixalus species in respect to breeding sites, accepting also smaller (> 20 m2) and less permanent ponds. It starts being active usually in mid December, later than most of the Kirindy amphibians but earlier than H. tricolor and H. carbonei. The first calling activity in the study period was observed on 19.12.1998, 15.12.1999, 5.12.2000 and 18.12.2001. Highest calling activity is after larger rainfalls. Preferred calling sites are usually high perches (often > 2 m) at the water edge (< 100 observations), more rarely grass stems within the pond. On two occassions, two males of this species were observed wrestling for about 15 seconds, presumably fighting for preferred calling sites.

Microhylidae Dyscophinae Dyscophus insularis GRANDIDIER, 1872 This species is widespread mainly over the western part of Madagascar (GLAW & VENCES 1994) and is generalistic in respect to breeding site choice and habitat use. It is found

SALAMANDRA, Rheinbach, 39(2), 2003 87 Amphibian communities in Madagascar Julian Glos 16

JULIAN GLOS

in all pond types and in all habitats, including villages and towns. These frogs are breeding only after heavy rains, calling in choruses aggregated within the pond while they are floating on the water surface. As in Aglyptodactylus laticeps, these choruses set in and fade out together. When threatened, D. insularis digs itself into the mud. When captured, it utters distress calls similar to its mating call. Tadpoles of this species form large aggregations as a reaction to the presence of predatory fish and birds.

Scaphiophryninae Scaphiophryne calcarata (MOCQUARD, 1895) This small, usually brown frog is found mainly in western Madagascar (GLAW & VENCES 1994). In Kirindy, it is an explosive breeder that colonizes freshly arisen ephemeral pools. Waters of all sizes in the closed forest, the river bed and the savanna as well as inside human settlements are used for breeding. After heavy rains, males call in great numbers from crevices and troughs at the edge of these ponds. As described by GLAW & VENCES (1994), a green colour morph exists (Fig. 12). In Kirindy, this colour morph was found in 5.8 % of freshly metamorphosed individuals (n = 104). One population consists of both morphs that were observed interbreeding. When threatened, this frog shows death-feigning behaviour by actively turning on the back, laying motionless and spreading apart its four legs. During the dry season, individuals of S. calcarata are found regularly buried in the ground (depth about 30 cm) by villagers digging their peanut fields. Larval developmental time is very short (10 – 11 days).

Scaphiophryne brevis (BOULENGER, 1896) This species is found in south-western Madagascar (GLAW & VENCES 1994). As the considerably more abundant S. calcarata, it breeds primarily in freshly filled pools after heavy rainfalls, using all major habitat types including human settlements. Often, two males were found at a breeding pond forming alternating calling pairs. At the beginning of the dry season of 1999 (22.4.1999), one individual was dug out in the research camp in hard soil at a depth of 20 cm. This individual appeared to be very globular presumably because it stored water. Larval developmental time is very short (10 – 11 days).

Scaphiophryne cf. marmorata BOULENGER, 1882 A definitive determination of this species is in progress. This frog is a medium sized, brown marbled microhylid frog (Fig. 13) with enlarged toe tips. It was found in Kirindy only at three different ponds, all of them medium sized (~100 m2), ephemeral forest ponds. It is an explosive breeder that breeds after heavy rainfalls (> 40 mm) and only a few times during the rainy season. Males are calling floating on the water exclusively at night.

Acknowledgements I would like to thank K.E. LINSENMAIR for supporting my work, K. DAUSMANN for assisting in the field work, M.-O. RÖDEL for useful comments on the manuscript, F. GLAW for his help in species determination and geographic distribution, P. KAPPELER and the German Primate Center (DPZ) for logistic support and the Biology Departement, Antananarivo University, for their cooperation. Research permits were provided by the Ministère des Eaux et Forêts, Antananarivo. Financial aid was provided by the DAAD (German Academic Exchange Service).

88 SALAMANDRA, Rheinbach, 39(2), 2003 Amphibian communities in Madagascar Julian Glos 17

The amphibian fauna of the Kirindy dry forest in western Madagascar

Résumé La faune d’amphibiens de la forêt sèche de Kirindy à Madagascar. La faune d’amphibiens de la forêt de Kirindy à Madagascar d’Ouest était examinée pendant une analyse écologique. Elle se compose de 15 espèces de quatre familles différentes (Mantellidae, Ranidae, Hyperoliidae, Microhylidae). Parmi ces espèces il y’a un grand spectre de couvers explosives et couvers prolongés. Concernant la choix des eaux pour la reproduction, il y’a à peu prés la même nombre de generalists et specialists. Le développement larvaire de certaines espèces est très court, probablement une adaptation sur l’ephemeralité des eaux de reproduction. Autres observations de la comportement et caractères écologiques des amphibiens de Kirindy sont présenté. Mot-Clés: Madagascar; forêt de Kirindy; forêt sèche saisonnière; amphibiens; choix d’habitat; choix des eaux pour la reproduction.

References ANDREONE, F., M. VENCES, & J.E. RANDRIANIRINA (2001): Patterns of amphibian and reptile diversity at Berara Forest (Sahalamaza Peninsula), NW Madagascar. – Italian Journal of Zoology, 68: 235-241. BUSSE, K. & W. BÖHME (1992): Two remarkable discoveries of the genera Mantella (Ranidae: Mantellinae) and Scaphiophryne (Microhylidae: Scaphiophryninae) from the west coast of Madagascar. – Rev. franc. Aquariol., 19(1/2): 57-64. GLAW, F. & M. VENCES (1994): A fieldguide to the amphibians and reptiles of Madagascar. – Köln. (M. Vences & F. Glaw Verlags GbR). — & — (1997): New species of the Boophis tephraeomystax group (Anura: Ranidae: Rhaco- phorinae) from arid Western Madagascar. – Copeia, 1997(3): 572-578. —, — & W. BÖHME (1998): Systematic revision of the genus Aglyptodactylus BOULENGER, 1919 (Amphibia: Ranidae), and analysis of its phylogenetic relationships to other Madagascan ranid genera (Tomopterna, Boophis, Mantidactylus and Mantella). – Journal of Zoological Systema- tics and Evolutionary Research, 36: 17-37. GOSNER, K. L. (1960): A simplified table for staging anuran embryos and larvae with notes on identification. – Herpetologica, 16: 183-190. GRAFE, T.U., S. DÖBLER & K.E. LINSENMAIR (2002): Frogs flee from the sound of fire. – Proceedings of the Royal Society, London, 269: 999-1003. JANZEN, D. (1992): Tropische Trockenwälder: Die am stärksten bedrohten Ökosysteme der Tropen. – pp.152-161 in: WILSON, E.O. (ed.): Ende der biologischen Vielfalt? – Heidelberg (Spektrum- Verlag). KOBELT, R.F. & K.E. LINSENMAIR (1992): Adaptations of the reed frog Hyperolius viridiflavus (Amphibia, Anura, Hyperoliidae) to its arid environment: VI. The iridophores in the skin of Hyperolius viridiflavus taeniatus as radiation reflectors. – Journal of Comparative Physiology B, 162: 314-326. — & — (1995): Adaptations of the reed frog Hyperolius viridiflavus (Amphibia, Anura, Hyperoliidae) to its arid environment: VII. The heat budget of Hyperolius viridiflavus nitidulus and the evolution of an optimized body shape. – Journal of Comparative Physiology B, 165: 110-124. KUCHLING, G. (1993): Zur Verbreitung und Fortpflanzung von Mantella betsileo in West- madagaskar. – Salamandra, Frankfurt/M., 29(3/4): 273-276. LINSENMAIR, K.E. (1998). Risk-spreading and risk reducing tactics of West African anurans in an unpredictably changing and stressful environment. – pp. 221-242 in: NEWBERRY, D.M., H.H.T. PRINS & N.D. BROWN (eds.): Dynamics of Tropical Communities. – London (Blackwell Science).

SALAMANDRA, Rheinbach, 39(2), 2003 89 Amphibian communities in Madagascar Julian Glos 18

JULIAN GLOS

MYERS, N., R.A. MITTERMEIER, C.G. MITTERMAIER, G.A.B. DA FONSECA & J. KENT (2000): Biodiversity hotspots for conservation priorities. – Nature, 403: 853-858. NELSON, R. & N. HORNING (1993): AVHRR-LAC estimates of forest area in Madagascar, 1990. – International Journal of Remote Sensing, 14: 1463-1475. RAKOTONIRININA (1996): Composition and structure of a dry forest on sandy soils near Morondava. – pp. 81-88 in: GANZHORN, J.U. & J.-P. SORG (eds.): Ecology and economy of a tropical dry forest in Madagascar. Primate Report 46-1. – Göttingen (DPZ). RÖDEL, M.-O. (1998): Kaulquappengesellschaften ephemerer Savannengewässer in Westafrika. – Frankfurt am Main (Edition Chimaira). — (2000): Herpetofauna of West Africa. Vol. I. Amphibians of the West African Savanna. – Frankfurt am Main (Edition Chimaira). SCHMUCK, R., F. KOBELT, & K.E. LINSENMAIR (1988): Adaptations of the reed frog Hyperolius viridiflavus (Amphibia, Anura, Hyperoliidae) to its arid environment: V. Iridophores and nitrogen metabolism. – Journal of Comparative Physiology, 158: 537-546. —, W. GEISE & K. E. LINSENMAIR (1994): Life cycle strategies and physiological adjustments of reedfrog tadpoles (Amphibia, Anura, Hyperoliidae) in relation to the environmental conditions. – Copeia, 1994(4): 996-1007. SORG, J.-P. & U. ROHNER (1996). Climate and tree phenology of the dry deciduous forest of the Kirindy forest. – pp. 57-80 in: GANZHORN, J.U. & J.-P. SORG: Ecology and economy of a tropical dry forest in Madagascar. Primate Report 46-1. – Göttingen (DPZ). SPIELER, M. (1997): Anpassungen westafrikanischer Anuren an Austrocknungsrisiko und Räuber- druck in einem saisonalen Lebensraum. – Berlin (W & T Verlag). STANISZEWSKI, M. (2001): Mantellas. – Frankfurt am Main (Edition Chimaira). VENCES, M. & F. GLAW (2001): When molecules claim for taxonomic changes: New proposals on the classification of Old World treefrogs. – Spixiana, 24(1): 85-91. — & — (2002): Molecular phylogeography of Boophis tephraeomystax: a test case for east-west vicariance in Malagasy anurans. – Spixiana, 25(1): 79-84. —, —, R. JESU & G. SCHIMMENTI (2000): A new species of Heterixalus (Amphibia: Hyperoliidae) from western Madagascar. – African Zoology, 35(2): 269-276.

Author: JULIAN GLOS, Lehrstuhl für Tierökolgie und Tropenbiologie, Biozentrum, Am Hubland, D-97074 Würzburg, Germany, E-Mail: [email protected].

90 SALAMANDRA, Rheinbach, 39(2), 2003 Amphibian communities in Madagascar Julian Glos 19

Chapter 2

A new species of Scaphiophryne from Western Madagascar

Amphibian communities in Madagascar Julian Glos 20

Copeia, 2005(2), pp. 252–261

A New Species of Scaphiophry e from Western Madagascar

JULIAN GLOS,FRANK GLAW, AND MIGUEL VENCES

We describe the adult and larval morphology, advertisement call, ecology, and life history of a new species of Marbled Toad from the dry deciduous forests of western Madagascar on the basis of eight specimens from Kirindy Forest C. F. P. F. in the central Menabe area. Scaphiophry e me abe sis n. sp. is larger, but morphologically similar to S. marmorata from the eastern rainforests. However, DNA sequence analysis of the mitochondrial 16S rRNA gene resulted in a clear differentiation from this species. The strongest mitochondrial afﬁnities are with S. madagascarie sis,a morphologically highly divergent species occurring in montane savanna and forest areas on the high plateau of Madagascar.

AR LED Toads, genus Scaphiophry e, are green with dark markings. In a revision of the M medium-sized, mainly terrestrial anurans Scaphiophry e marmorata complex, however, endemic to Madagascar. Together with the Vences et al. (2003) stated that specimens from monotypic Paradoxophyla they have been classi- Kirindy Forest and from the Tsingy de emar- fied in their own family Scaphiophrynidae (Du- aha in the west could not be reliably distin- bois, 1992), but recent molecular work (van der guished morphologically from eastern S. mar- Meijden et al., 2004) corroborated their inclu- morata. However, in that study only a limited sion in the Microhylidae as subfamily Scaphio- number of specimens were available for com- phryninae ( lommers-Schlo¨sser and lanc, parison, and it was not possible to assess the de- 1991). Most Scaphiophry e species are explosive gree of genetic differentiation between eastern breeders, often using shallow, temporary waters and western populations. as breeding sites (Glaw and Vences, 1994). To clarify this unresolved issue, we tested for Recently, the number of Scaphiophry e species differentiation among eastern and western pop- was raised to seven including two previously un- ulations of S. marmorata. We here report on recognized species (Vences et al., 2003). In con- morphological differences between specimens trast to the mantellid and cophyline radiations, of the two populations and provide a mitochon- which clearly show their maximum diversity in drial DNA sequence analysis. The surprising re- the humid forests of eastern Madagascar, Sca- sult was that the western species is genetically phiophry e appears to have diversified in the not related to eastern S. marmorata but rather eastern, central, and western biogeographic re- to S. madagascarie sis, a morphologically highly gions of the island. divergent species. We consider the observed dif- Within the genus, Scaphiophry e marmorata is ferences as indicative of separation on the spe- unique as it is currently considered to occur in cies level, and describe the morphology, adver- eastern rainforests as well as in remnants of tisement call, ecology, and life history of Sca- western dry forests (Kirindy, emaraha, Namo- phiophry e me abe sis n. sp. roka, Isalo region; Vences et al., 2003). This is surprising, as these two types of forest are very different in respect to climate and seasonality, MATERIALS AND METHODS forest structure, and breeding site characteristics. Along the same line, recent work on other Three amplectant pairs and one single calling amphibians (Vences and Glaw, 2002) and rep- male were collected on 5 February 2002 be- tiles (Nussbaum and Raxworthy, 1998; Nuss- tween 0100–0300 h in the Kirindy Forest. These baum et al., 1998) has shown that eastern and specimens were found floating on the water sur- western populations of several forms are distinct face together with about 20 more individuals in taxa. These examples of sister species distribut- a forest pond in the area locally known as CS5 ed allopatrically in eastern and western Mada- after a heavy rainfall. Additionally, one calling gascar might be a result of vicariant speciation. male was collected on 13 January 2001 in a for- In agreement with these findings, Scaphiophry- est pond (CS7). e marmorata from the west often differ from Photographs of representative frogs were tak- those of the east in their coloration. Specimens en soon after capture to record natural colora- from the west are mostly brown with dark mark- tion (Fig. 1). The specimens were euthanized ings in contrast to those from the east which are with MS222 and fixed and preserved in 70%

᭧ 2005 by the American Society of Ichthyologists and Herpetologists Amphibian communities in Madagascar Julian Glos 21

GLOS ET AL.—A NEW SCAPHIOPHRYNE FROM MADAGASCAR 253

ethanol. Institutional abbreviations follow Levi- el that best fits our data through hierarchical ton et al. (1985). likelihood ratio tests as implemented in Model- ody measurements were taken with calipers test (Posada and Crandall, 1998). Robustness of to the nearest 0.1 mm: SVL (snout-vent length), nodes was tested by full heuristic bootstrapping, HW (maximum head width), HL (head length, with 2000 pseudoreplicates (and 10 random ad- from the maxillary commissure to the snout dition sequence replicates) under maximum tip), ED (horizontal eye diameter), END (eye- parsimony and 500 pseudoreplicates under nostril distance), NSD (nostril-snout tip dis- maximum likelihood. A species of Microhyla tance), NND (distance between nostrils), HAL (Microhylinae), and species of the genus Para- (hand length, from the carpal-metacarpal artic- doxophyla that following lommers-Schlo¨sser ulations to the tip of the longest, third, finger), and lanc (1991) is the closest relative of Sca- FORL (forelimb length, from the axil to the tip phiophry e, were used as outgroups. of the longest finger), HIL (hindlimb length, Clutch size was determined by counting the from the cloaca to the tip of the longest, fourth, eggs of one amplectant pair that we collected toe), FOL (foot length, from the tarsal-metatar- in the field and kept over night in an aquarium sal articulations to the tip of the longest toe), where it spawned. We measured ovum diameter FOTL (foot length including tarsus, from the in a sample of 20 eggs from each clutch of two tibiotarsal articulation to the tip of the longest distinct clutches. toe), IMTL and IMTH (maximum length and Tadpoles of Scaphiophry e were collected in height of inner metatarsal tubercle), FD4 (max- the field from January to March 2001 and 2002 imum width of terminal disk of fourth finger), in the Kirindy Forest. Additionally, fertilized RHL (relative hindlimb length, point reached eggs from one amplectant pair were reared in by tibiotarsal articulation when hindlimb is plastic aquaria filled with rainwater. Tadpoles pressed along the body), coded as follows: 0 ϭ were fed ad libitum with commercial fish food the tibiotarsal articulation does not reach the (TetraMinTabs௡). We preserved 12 tadpoles in forelimb insertion; 1 ϭ it reaches forelimb in- different developmental stages in 5% formalin sertion; 2 ϭ it reaches between forelimb and (ZSM 358/2004). Staging is according to Gos- tympanic region; 3 ϭ it reaches tympanic re- ner (1960) and nomenclature of morphological gion. These measurements were compared to features follows McDiarmid and Altig (1999). those taken on S. marmorata specimens from the Measurements of morphometric variables were east by Vences et al. (2003). taken from preserved specimens using a stereo All specimens were assumed to be adults as microscope (Zeiss௡ Stemi SV 6) with a measur- they were observed to be engaged in breeding ing ocular. Drawings were done with a camera activities (amplexus or calling). Calls were re- lucida. Froglet size at metamorphosis was recorded using a Sony WM TCD–100–S04 DAT– corded on specimens that were caught in late Recorder and Sennheiser microphone. Call pa- developmental stages (Ն Gosner 40) in the field rameters were analysed using Raven 1.0 software and raised to complete metamorphosis (Stage package (᭧ Cornell Lab of Ornithology). 45–46; n ϭ 10, ZSM 628/2003). These speci- Muscle tissue samples were taken from freshly mens were euthanized with chlorobutanol sub- euthanized specimens in the field and pre- sequent to measurements and preserved in 5% served in pure ethanol. DNA was extracted us- formalin. Measurements were taken to the near- ing standard protocols and a fragment of the est 0.05 mm with calipers. mitochondrial 16S rRNA gene amplified using Water chemistry parameters of tadpole habi- the primers 16Sa-L and 16Sb-H (S. R. Palumbi, tats were measured using WA 300 conductivity A. Martin, S. Romano, W. McMillan, L. Stice, analyser (Conrad Electronics), DO–5509 Oxy- and G. Grabowski, 1991, The Simple Fool’s gen analyser (Conrad Electronics) and pHep௡ Guide to PCR, ver. 2, unpubl.), and protocols pH analyser (Hanna Instruments). Sympatric described by Vences et al. (2002b). tadpole species were identified in the natural Phylogenetic analysis was carried out using ponds and breeding habitat choice was assessed PAUP* (vers. 4.0b10, D. L. Swofford, PAUP*: by repeatedly sampling 200 natural ponds with phylogenetic analysis using parsimony [*and the box-method and by standardized dip-net- other methods], Sinauer, Sunderland, MA, ting (Heyer et al., 1994). 2002). We performed unweighted maximum parsimony heuristic searches, with tree-bisection Scaphiophry e me abe sis, new species reconnection branch swapping, and random se- Figure 1, Table 1 quence addition with 100 replicates. In addition, a maximum likelihood analysis was per- Holotype.—ZSM 186/2003, adult male, western formed after determining the substitution mod- Madagascar, province of Toliara, district of Mo- Amphibian communities in Madagascar Julian Glos 22

254 COPEIA, 2005, NO. 2

Fig. 1. (A) Dorsolateral and ( ) Ventral views of living adult specimen of Scaphiophry e me abe sis n. sp. For color photograph see Glos (2003). (C) Lateral view and (D) Dorsal view of a Scaphiophry e me abe sis n. sp. tadpole at stage 39. Scale bars represent 5 mm.

rondava, in the Kirindy Forest C. F. P. F. (forest 219487, one adult male, Kirindy Forest, C. J. pond CS5), 44Њ39ЈE, 20Њ03ЈS, 18–40 m eleva- Raxworthy, J. . Ramanamanjato, A. Raseliman- tion, J. Glos, 5 February 2002. ana, A. Razafimanantsoa, and A. Razafimanant- soa, 31 January 1996, specimen examined by Paratypes.—ZSM 187/2003, ZSM 188/2003 and Vences et al. (2003). ZSM193/2003, three adult females, and ZSM 189/2003, ZSM 190/2003, and ZSM 192/2003, Diag osis.—A medium sized Scaphiophry e; SVL three adult males, western Madagascar, prov- of adult males 40.3–42.0 mm (n ϭ 5), of adult ince of Toliara, district of Morondava, in the females 42.6–45.2 mm (n ϭ 3). Scaphiophry e Kirindy Forest C. F. P. F. (forest pond CS5), J. me abe sis differs from S. brevis, S. calcarata, and Glos, 5 February 2002. ZSM 191/2003, one S. madagascarie sis by highly expanded terminal adult male, Kirindy Forest C. F. P. F. (forest disks on the fingers and toes (vs. not or only pond CS7), J. Glos, 13 January 2001. UMMZ slightly enlarged); from S. gottlebei by a very dif- Amphibian communities in Madagascar Julian Glos 23

GLOS ET AL.—A NEW SCAPHIOPHRYNE FROM MADAGASCAR 255

ferent coloration (without distinct black, red, green, and white areas as visible in living and ethanol-preserved S. gottlebei) and granular dorsal skin (vs. smooth); from S. boribory by smaller snout-vent length (47–60 mm vs. 40–45 mm), granular dorsal skin, and coloration; from S. spi- osa by coloration and by the absence of large dermal spines above forelimbs, at maxilla commissure, and in the tympanic region. Scaphio- phry e me abe sis most resembles S. marmorata, but differs by a larger snout-vent length (males of S. marmorata 32.4–35.9 mm, females 34.9– 43.5 mm), a narrower head, shorter relative

. For abbreviations of measured variables hindlimb length, shorter relative hand length, a longer inner metatarsal tubercle, and a brownish dorsal coloration (Tables 1, 2). FEMALE Scaphiophry e me abe sis furthermore differs ϭ from all other Scaphiophry e species except S. ,F madagascarie sis by a signiﬁcant genetic differ-

MALE entiation in the 16S mitochondrial gene and

ϭ probably from S. spi osa, S. boribory, and S. madagascarie sis by shorter advertisement calls. .M SP .

N Descriptio .—Adult male in breeding state; specimen in excellent state of preservation. For morphometric measurements see Table 1. Me- dium sized, stout frog; snout rounded in dorsal and lateral profile, nostrils directed laterally, closer to snout than to eye; horizontal pupil; tympanum not visible; single, subgular vocal sac; tongue ovoid; maxillary teeth and vomerine teeth absent. Arms slender; fingers without web- Scaphiophry e me abe sis bing; relative length of fingers 1 Ͻ 2 ϭ 4 Ͻ 3; small subdigital tubercles; no metacarpal tuber-

see Materials and Methods section. cles. First toe short, fourth toe much longer than third and ﬁfth toe, third toe as long as ﬁfth

PECIMENS OF toe; relatively large inner metatarsal tubercles; S minute webbing between toes. Tibiotarsal articulation reaches tympanic region. Skin on dor- DULT

A sum with two rows of larger tubercles and many

OF small granules, without dorsolateral folds; ab- ) sence of large tubercles above forelimb inser- MM

( tion and in tympanic region; prominent tubercle before hindlimb insertion. Ventral skin slightly granular; legs dorsally very slightly granular, ventrally not granular. Dorsal coloration in life is brown with sym- EASUREMENTS metrical darker markings, including feet and M arms (Fig. 1A). Ventral pattern with contrasted dark brown–light cream marbling, extending on legs and arms (Fig. 1 ); in the region of inner thighs smaller pattern of marbling with a higher proportion of brown. Color of throat is ORPHOMETRIC dark brown, with many small granules. 1. M

Specimen Sex SVL HW HL EDVariatio END NSD .— NNDZSM HAL 187–193/2003 FORL HIL FOL were FOTL examined IMTL IMTH FD4 RHL

ALE (Table 1) and directly compared with the ho- T ZSM 186/2003ZSM 187/2003ZSM 188/2003 MZSM 189/2003 FZSM 190/2003 40.5 FZSM 191/2003 M 45.2ZSM 192/2003 12.7 M 43.1ZSM 193/2003 M 40.4 13.7 11.0 M 40.3 13.2 12.4 42.0 F 12.9 3.5 11.9 41.3 11.4 10.8 12.5 3.6 42.6lotype. 3.1 11.9 13.2 3.4 4.2 11.2 3.4 14.5 1.5 3.0 11.9 2.9All 3.3 3.0 1.9 11.7 3.0 3.4 3.8 1.9 frogs 1.4 3.0 3.0 3.6 13.1 1.9 3.5 3.2 are 2.4 1.9 13.7 3.2 27.9 2.6 2.5 13.2 largely 12.4 2.8 30.0 1.8 52.1 13.0 3.4 30.8 27.3 12.1 55.4 3.1 concordant 19.4 29.4 11.9 53.3 50.1 28.1 19.5 14.0 25.2 51.1 29.6 19.6 18.4 49.1 26.7 29.8 4.1 18.8 51.8 25.5 in 25.2 18.3 3.8 53.8 2.7mor- 25.6 19.4 4.1 3.0 23.3 2.8 20.8 2.6 4.0 26.4 2.8 2.3 4.1 2.7 27.2 2.8 4.0 3 2.3 2.1 2.4 4.3 2 2.3 2.5 3 2.2 2 2.1 2.4 3 2 2.6 2 3 Amphibian communities in Madagascar Julian Glos 24

256 COPEIA, 2005, NO. 2

phology and coloration, however, the female specimens (ZSM 187, 188, 192/2003) are larger than the males. In life some specimens had small green markings behind the forelimb insertion. 1.811.290.77 0.09 0.52 0.26 0.55 0.52 0.71 1.810.26 0.71 0.09 1.81 0.90 1.29 0.09 0.26 2.32 0.02 2.32 0.02 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Distributio .—The type locality is the Kirindy Forest C. F. P. F. (Centre de Formation Profes- sionelle Forestie`re), a deciduous dry forest near 2 2 2 2 Ϫ Ϫ Ϫ Ϫ the west coast of Madagascar, 60 km northeast 6ZP 10 10 10 10

ϭ of Morondava and about 20 km inland ϫ ϫ ϫ ϫ N Њ Ј Њ Ј 2.8 0.02 0.02 0.82 0.02 0.03 0.0 (44 39 E, 20 03 S; 18–40 m elevation; Sorg and Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Rohner, 1996). The area of the Kirindy forest 0.99 0.43 0.61 0.61 0.05). Data for males and females were covers about 12,000 ha and thus may be among Ϯ Ϯ Ϯ Ϯ S. marmorata

␣ϭ the largest remaining continuous forests in west- ( 8.73 8.29 5.53 3.41 ern Madagascar (Nelson and Horning, 1993). In this forest, the new species was found in the Females 2 2 2 2 areas locally known as CS5, CS6, and CS7. Sur- Ϫ Ϫ Ϫ Ϫ

3 veys in the Menabe region in January and Feb- 10 10 10 10 ϭ

S. marmorata ruary 2004 at ﬁve sites west of the Kirindy Forest ϫ ϫ ϫ ϫ N 1.1 38.8 0.020.060.41 0.34 0.01 0.84 0.020.02 7.64 0.34 0.69 and 1.3 between the Kirindy Forest and emaraha Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ . For abbreviations see Materials and Methods section. Data for and 0.30 0.09 0.73 0.73 ( J. Glos, unpubl. data) and in a distance of 5 to Ϯ Ϯ Ϯ Ϯ

43.6 30 km from Kirindy were unsuccessful in ﬁnd- 0.87 0.32 0.31 S. me abe sis ing this species. We attribute three other local- 8.13 4.32 ities to this species: (1) the ‘‘Tsingy de emar-

S. marmorata aha’’ Reserve (18Њ42.481ЈS, 44Њ42.981ЈE), S. me abe sis UMMZ 219488–219489 and 219491–219495, 8– AND .

0.01 0.01 0.01 0.01 1.24 11 March 1996, C. J. Raxworthy, J. . Ramana- SP Ͻ Ͻ Ͻ Ͻ

. manjato, A. Raselimanana, A. Razafimanantsoa, N and A. Razafimanantsoa; (2) the Namoroka Re- serve about 200 km further north (16Њ28.189ЈS, 45Њ20.906ЈE), UMMZ 227499, 6 December 0.731.280.91 0.54 0.251.462.74 0.43 7.09 0.18 1.00 0.69 9.31 5.81 2. 4 2. 4 2.192.19 0.03 2. 0.03 4 2. 4 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ 1996, C. J. Raxworthy, J. . Ramanamanjato, A. Raselimanana, A. Razafimanantsoa, and A. Ra- zafimanantsoa; and (3) the Isalo region, UMMZ 2 2 2 2 Ϫ Ϫ Ϫ Ϫ

analyzed separately; Mann-Whitney U-Test. 227489, local collector. Specimens from these 6ZP 10 10 10 10 localities agreed morphologically with those ϭ ϫ ϫ ϫ ϫ N Scaphiophry e me abe sis 1.5 0.02 0.04 0.47 0.01 0.02 0.04 from Kirindy. However, since no genetic data OF Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ are available for the populations from emara- 0.62 0.39 0.88 0.37 Ϯ Ϯ Ϯ Ϯ ha, Namoroka, and Isalo, and only single spec- SD) imens were collected from the latter two local- S. marmorata Ϯ

3.93 8.76 5.65 ities, their attribution to S. me abe sis requires 10.34 further conﬁrmation. MEAN Males ( 2 2 2 2 Ϫ Ϫ Ϫ Ϫ Molecular phyloge etic relatio ships.—As stressed 5 10 10 10 10

ϭ by Vences et al. (2002b, 2003), the mitochon- ϫ ϫ ϫ ϫ N 0. 34.8 0.010.050.78 0.35 0.02 0.82 0.020.04 7.52 0.35 0.72 drial 1.36 differentiation among species of Scaphio- Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ

ROPORTIONS phry e is surprisingly low. Of a total of 528 char- 1.24 0.88 0.95 0.38 P

Ϯ Ϯ Ϯ Ϯ acters included in the analysis, 413 were con- 40.9 0.31 0.91 0.31 1.24

S. me abe sis stant and only 77 were parsimony-informative.

8.59 Figure 2 shows a strict consensus of the four most parsimonious trees found; length of these

are from Vences et al. (2003). old indicates signiﬁcant differences between was 217 steps, consistency index was 0.714 and ELATIVE ODY retention index was 0.791. Most clades received very low bootstrap support, probably caused by 2. R the limited number of informative sites. Scaphio-

ALE phry e me abe sis was placed with moderate NND/SVLIMTL/SVL 6.90 9.37 S. marmorata T SVL (mm) HW/SVL HL/HW ED/SVL NSD/SVLHAL/SVL FORL/SVL 4.44 HIL/SVL FD4/HAL 0.70 bootstrap 5.66 support (74% in the maximum par- Amphibian communities in Madagascar Julian Glos 25

GLOS ET AL.—A NEW SCAPHIOPHRYNE FROM MADAGASCAR 257

three different pools, all of them medium sized (range 123–235 m2), ephemeral waters in the closed forest. Scaphiophry e me abe sis has a small spatial niche; it was neither found in rock pools of the river bed (before the river is running), nor in savanna ponds, although these types of breeding waters are abundant around the Kirindy Forest. All breeding ponds had similar characteristics. They were shallow with a water depth of less than 10 cm over 50–95% (range, n ϭ 3) of their surface. Maximum water depth was 21–45 cm when the ponds were completely water-filled. Water was clear to slightly muddy. There was very low coverage of aquatic Fig. 2. Strict consensus phylogram of four most vegetation (floating water plants, grasses, un- parsimonious trees obtained by maximum parsimony derwater vegetation). Water chemistry was com- searches based on 528 nucleotides of the mitochon- parable in these three ponds (oxygen concen- drial 16S rRNA gene. First numbers are bootstrap sup- tration: 1.28–1.42 mg/l, conductivity 100–161 port values from a maximum parsimony bootstrap ␮S, pH 6.55–6.66). Compared to the other 200 analysis (2000 replicates). Second numbers are boot- breeding waters known in Kirindy, invertebrate strap support values in percent from a maximum like- ϩ ϩ predator load was small. lihood search, based on a RTrN I G nucleotide The following syntopic tadpole species were substitution model selected by Modeltest, with empir- ical base frequencies and sustitution rates, a propor- found: Boophis doulioti, B. xerophilus, Aglyptodac- tion of invariable sites of 0.5662, and a gamma distri- tylus laticeps, Ma tella betsileo, Dyscophus i sularis, bution shape parameter of 0.4066. Where only single Scaphiophry e calcarata. numbers are given, they refer to the parsimony boot- We never found adult S. me abe sis during the strap analysis. Values below 50% are not shown. Col- day. Captured specimens were observed to be lecting localities are shown for specimens with reliable fairly good climbers. data; others were obtained from commercial collec- tors. Abbreviations: S. ϭ Scaphiophry e, P. ϭ Paradoxo- ϭ Life history.—Eggs were deposited as a single phyla, M. Microhyla. layered surface film. Clutch size in one female was 450 eggs, ovum diameter was 2.21 Ϯ 0.21 mm (mean Ϯ SD; range 1.68–2.72 mm; n ϭ 40 simony analysis) as sister clade of S. madagascar- from two amplectant pairs that spawned in the ie sis. Two individuals of the latter species, from field camp). One dissected female (ZSM 193/ the Ankaratra and Andringitra Massifs, respec- 2003) contained 670 oocytes (diameter 1.51 Ϯ tively, had fully identical sequences despite their 0.07; range 1.41–1.63). Snout-vent length at different morphology reported by Vences et al. metamorphosis was 10.3 Ϯ 0.7 mm (range 9.4– (2002a). The differentiation of S. me abe sis to 11.4 mm, n ϭ 10). S. madagascarie sis was only 0.4% pairwise sequence divergence (two different nucleotides Advertiseme t call.—The call consists of a series over the whole gene fragment), whereas the dif- of short notes (Fig. 3). At 25 C, call duration ference of S. me abe sis to S. marmorata was of was 0.51–0.76 sec (n ϭ 3). Calls are repeated 2.9% (15 nucleotide differences). with an interval of 0.61–0.77 sec. Calls contained 13–19 notes (n ϭ 5). Note duration was Natural history.—Scaphiophry e me abe sis is an 16–17 ms (n ϭ 6). Intervals between notes were explosive breeder that breeds only after heavy 17–24 ms (n ϭ 6). The intensity of notes was rainfalls (Ͼ 40 mm), and only a few times dur- constant throughout the call. Frequency was ing the rainy season (December to March; data 500–1400 Hz; the dominant frequency was 850 from the rainy seasons 2000/2001, 2001/2002, Hz. The advertisement call of S. me abe sis is 2002/2003, 2004). Males call while floating on structurally similar to that of S. spi osa, S. bori- the water exclusively at night. Several pairs were bory, and S. madagascarie sis as described by observed in axillary amplexus. Within the Kir- Vences et al. (2002a, 2003). However, call and indy forest, S. me abe sis is among the rarest an- single notes appear to be shorter in S. me abe - uran species. Of more than 200 potential breed- sis as compared to the other species. In S. mad- ing waters that we investigated in the Kirindy agascarie sis, note duration was 28–32 ms Forest ( J. Glos, unpubl. data), calling S. me a- (Vences et al., 2002a), in S. spi osa, call duration be sis and/or its tadpoles were found only at was 3539–9117 ms and note duration was 25–34 Amphibian communities in Madagascar Julian Glos 26

258 COPEIA, 2005, NO. 2

Fig. 3. The advertisement call of Scaphiophry e me abe sis n. sp. Audiospectrogram (sonogram: frequency in kHz vs. time) and oscillogram (relative amplitude vs. time).

ms, and in S. boribory, call duration was up to file, eyes medium sized and directed dorsolat- 12,492 ms and note duration 35–38 ms (Vences erally. Nostrils dorsal, in the same distance to et al., 2003). Although these variables may be snout tip and to eyes (ratio naris-eye distance/ influenced by temperature (recorded at a tem- snout-naris distance 1.02 Ϯ 0.21, mean Ϯ SD), perature range as wide as from 16 C in S. mad- spiracle sinistral, at the rear of the body, low, agascarie sis to 25 C in S. spi osa), this is unlikely inner lateral wall absent, directed posteriorly. to be the only explanation for the differences Medial vent tube short, aperture dextral. of S. me abe sis to the other species. ecause the Tail sturdy; tail fins convex, higher from the calls of S. marmorata are unknown, the existing posterior end of vent tube to the middle of the bioacoustic data cannot be used to assess pos- tail, dorsal fin as high as ventral fin. Tail tip sible differences between S. me abe sis and this rounded. Origin of dorsal fin slightly before the species. Among sympatric microhylids at Kirin- base of tail. dy, the call of Scaphiophry e me abe sis is very Oral disc directed anteriorly, no tooth rows. different compared to those of S. brevis and S. Marginal papillae conical, with rounded tips, calcarata but can be mistaken with that of Dys- slightly pigmented; in stage 39 length of mar- cophus i sularis. ginal papillae 0.18 mm, density 7.6 papillae/ mm. Upper jaw sheath concave, lower jaw Descriptio of tadpole.— ecause the tadpoles sheath V-shaped, both sheaths not serrated and showed no considerable ontogenetic changes in not pigmented. oral disc structure and body proportions (see In vivo, dorsal color is lightly brownish, ven- Table 3), we give data as mean and standard tral color is white. Caudal musculature is entire- deviation for tadpoles of all developmental stag- ly pigmented, dorsal tail fin heavily pigmented es pooled together. dorsally and only slightly pigmented ventrally, ody of the tadpoles depressed ovoid in lat- ventral tail fin only slightly pigmented dorsally eral and wide ovoid in dorsal view (Fig. 1C, D). and heavily pigmented ventrally. Laterally, there is a considerable gap between the outer integument and the body with no vis- Etymology.—The specific name refers to the ible tissue, giving the tadpole a broad, disc-like Menabe region in which the type locality Kir- appearance. Snout flat in dorsal and lateral pro- indy Forest is situated (Malagasy for Menabe ϭ Amphibian communities in Madagascar Julian Glos 27

GLOS ET AL.—A NEW SCAPHIOPHRYNE FROM MADAGASCAR 259

TALE3. MEASUREMENTS (MM) OF TADPOLES OF Sca- the analysis of 16S rDNA sequences leaves little phiophry e me abe sis N. SP. Staging after Gosner doubt that the mitochondrial (maternal) affin- (1960), measurements follow McDiarmid and Altig ities of S. me abe sis are with S. madagascarie sis, (1999). H ϭ body height, L ϭ body length, W ϭ a species typically occurring in montane savan- ϭ ϭ body width, ED eye diameter, IND internarial na and forest areas on the high plateau of Mad- ϭ ϭ distance, IOD interorbital distance, MTH maxi- agascar. Although the genetic divergence ϭ ϭ mum tail height, NED naris–eye distance, ODW among S. madagascarie sis and S. me abe sis is oral disc width, SED ϭ snout–eye distance, SND ϭ ϭ extremely low, these two species are morpholog- snout–naris distance, SSD snout–spiracle distance, ically very distinct. They also appear to slightly TL ϭ total length, TMH ϭ tail musculature height, differ in advertisement calls, which is relevant TMHM ϭ tail musculature height at midlength of tail, because Scaphiophry e appear to be character- TMW ϭ tail muscle width. ized by a highly conserved evolution of call fea- Mean Ϯ SD Range tures (Vences et al., 2003). Vences et al. (2002b) demonstrated that Sca- ϭ Stage 25–28 (N 7) phiophry e gottlebei, the second Scaphiophry e with TL 15.21 Ϯ 2.77 10.15–18.36 enlarged finger disks occurring in western Mad- Ϯ L 6.08 1.05 4.00–7.24 agascar, is an allotetraploid species that origi- Ϯ ED 0.54 0.16 0.29–0.83 nated by hybridization. Hybridization is likely in SSD 5.15 Ϯ 0.97 3.30–6.48 Ϯ explosively breeding species with similar adver- ODW 1.62 0.34 0.97–2.04 tisement calls, such as Scaphiophry e, and phe- Stage 36–39 (N ϭ 5) nomena of recent mitochondrial introgression TL 27.00 Ϯ 2.67 22.90–31.00 could possibly account for the fact that S. me - L 11.87 Ϯ 0.79 11.00–12.96 abe sis is mitochondrially very closely related to ED 0.92 Ϯ 0.15 0.71–1.11 S. madagascarie sis whereas morphologically it SSD 9.96 Ϯ 0.22 9.72–10.26 seems closer to other taxa. Analysis of nuclear ODW 3.34 Ϯ 0.25 3.08–3.80 markers and karyotypes will be necessary to All stages pooled (N ϭ 12) clarify this enigma, and any phylogeographic L/TL 0.42 Ϯ 0.03 0.38–0.48 discussion on the origin of S. me abe sis will L/ W 1.29 Ϯ 0.18 1.14–1.84 have to await these data. L/ H 1.72 Ϯ 0.22 1.41–2.02 W/ H 1.35 Ϯ 0.22 0.83–1.64 Life history.—Scaphiophry e me abe sis can be TMHM/MTH 0.28 Ϯ 0.04 0.21–0.35 considered a typical explosive breeding species IOD/IND 3.48 Ϯ 0.81 2.18–4.71 reproducing only one or a few times per season SND/SED 0.59 Ϯ 0.07 0.43–0.64 in lentic, temporary waters. This is also true for Dorsal/Ventral fin 0.94 Ϯ 0.12 0.79–1.20 most Scaphiophry e species. In contrast to other NED/ L 0.21 Ϯ 0.02 0.17–0.25 Scaphiophry e and in particular to S. marmorata TMH/ H 0.31 Ϯ 0.04 0.24–0.38 that lay a large number of small eggs (S. mar- TMW/ W 0.22 Ϯ 0.05 0.14–0.35 morata 1.2 mm egg diameter, S. boribory 1.7 mm, S. spi osa 1.4–1.7 mm; lommers-Schlo¨sser, 1975; Vences et al., 2002a, 2003), S. me abe sis ‘‘big red’’). This name was chosen to highlight lays fewer and relatively large eggs. Large the unique status of the Menabe region as local clutches and small egg sizes are regularly found center of biodiversity and as one of the largest throughout seasonal areas of the tropics, and remnants of the Malagasy dry forest. are seen as an r-strategy in this often unpredictable environment. However, S. me abe sis clutches with a relatively low number of large DISCUSSION eggs can be regarded as an alternative strategy Relatio ships.—Among Scaphiophry e, and based to cope with rapidly dessicating breeding waters. on morphological features, S. me abe sis clearly A larger size of oocytes and probably also of shows the strongest affinities to S. marmorata. hatchlings may give S. me abe sis tadpoles a These two species share their moderately gran- head start in larval growth and development, ular skin, enlarged disks of fingers and toes, and and may increase the chance to complete meta- size. Although they have different color pat- morphosis before the breeding site dries up. terns, with a basically brown color in S. me abe - In general, the tadpoles of S. me abe sis are sis and a mainly green pattern in S. marmorata, morphologically similar to tadpoles of other Sca- the chromatic differences among them are phiophry e species (e.g., S. brevis, S. calcarata; smaller than those distinguishing other species lommers-Schlo¨sser, 1975; J. Glos pers. obs.). (e.g., the conspicuous S. gottlebei). Nevertheless, These tadpoles are unique in their morphology Amphibian communities in Madagascar Julian Glos 28

260 COPEIA, 2005, NO. 2

as they are intermediate between the ranoid the case for other anuran and reptile species. and the microhylid type (Wassersug, 1984). However, the prevailing climatic conditions in They are easily distinguishable in the field from these two areas are very different. The central other tadpoles as they have a very broad, almost mountain chain, spanning from the north to disc-like appearance caused by a gap with no the south of Madagascar, most likely acts as a apparent tissue between the main body and the dispersal barrier for frogs with low altitude dis- outer integument. tribution and inhibits gene flow between the east and the west. Even if climatic conditions in Distributio a d co servatio .—Scaphiophry e the past might have been similar in the two ar- marmorata is mainly restricted to humid forests eas, the isolation of western populations follow- of the central east of Madagascar (Glaw and ing climatic shifts may have led to vicariant spe- Vences, 1994). In contrast, current knowledge ciation. Along these lines, recent work on am- of geographic distribution patterns in the genus phibians and reptiles detected several cases Scaphiophry e suggests that S. me abe sis is re- where vicariant sister taxa occur in eastern and stricted to the dry deciduous forests of the cen- western Madagascar (Nussbaum et al., 1998; tral west. The type locality, Kirindy C. F. P. F., as Nussbaum and Raxworthy, 1998; Vences and well a second site, the natural reserve of the Glaw, 2002), and a similar hypothesis could ex- ‘‘Tsingy de emaraha,’’ are privately and offi- plain the origin of S. marmorata and S. me abe - cially protected areas, respectively. These sites sis. However, the results obtained so far reject represent some of the largest and most pristine this simple hypothesis, at least from a mitochon- areas of the highly endangered and fragmented drial perspective. Speciation in Scaphiophry e dry forest ecosystem of western Madagascar. Na- might have been characterized by complex pat- moroka Reserve, north of emaraha, might be terns, including recurrent hybridization, intro- a third locality. The identity and precise loca- gression, and admixture among forms occur- tion of a fourth population in the Isalo region ring in the arid west, montane savannas, and in the south remains unknown, but the large eastern rainforests. size of a female specimen (SVL 48.5 mm; Vences et al., 2003) might justify a preliminary MATERIAL EXAMINED determination as S. me abe sis rather than S. marmorata. In a herpetological survey in 2004 at Genbank accession numbers of the sequences five sites between the Kirindy area and emar- used are as follows: Scaphiophry e boribory: aha and around Kirindy C. F. P. F., no S. me a- AJ314810 and AY594126; S. brevis: Kirindy, be sis were found ( J. Glos, unpubl. data). The AF215384 and AY834194; erenty, AY834195; sites of this survey, however, are clearly more Ifaty, AJ314808; S. calcarata: Ampijoroa, disturbed than the Kirindy Forest C. F. P. F., in- AJ314811; erenty, AY834192; Tolagnaro, dicating that S. me abe sis prefers relatively un- AY834193; Isalo region, AY594127; S. gottlebei: disturbed dry forest habitat. AF215385; S. madagascarie sis: Ankaratra, The dry forest of western Madagascar ranks AJ314809; Andringitra, AY834190; S. marmorata: among the world’s most endangered ecosys- Andasibe, AJ417567 and AY834191; S. me abe - tems. In 1990, only 2.8% of its original area re- sis: Kirindy, AY834189; S. spi osa: AF215383. Mi- mained (Laurance and ierregaard, 1997). crohyla pulchra, Vietnam, AF215374; Paradoxophy- Within this ecosystem, S. me abe sis seems to be la palmata: Andranomena, AY834187; Ranoma- restricted to the relatively largest and least dis- fana, AY834188; P. sp.: Masoala, AY834186; Dys- turbed remnants of dry forest, Kirindy and e- cophus a to gili: Maroantsetra, AY834196; D. maraha. There are promising efforts to estab- gui eti: Fierenana, AY834197; D. i sularis: Antsi- lish an integrated conservation concept for the rasira, AY834198. Menabe region including the Kirindy Forest. emaraha is already under governmental pro- ACKNOWLEDGMENTS tection and is comparatively well protected because of its remoteness. In addition to its ap- We are grateful to the Commission Tripartite parently restricted distribution, S. me abe sis was of the Malagasy Government, the Laboratoire never abundant within its known sites of occur- de Primatologie et des Verte´bre´s de l’Universite´ rence, and therefore might be a rare and threat- d’Antananarivo, the Parc otanique et Zoolo- ened species. gique de Tsimbazaza, the Ministe`re pour la Pro- duction Animale, and the De´partement des Co clusio .—Up to now, Scaphiophry e marmorata Eaux et Foreˆts for permits to work in Madagas- was considered to be distributed both in the hu- car (Research permit # 83/MEF/SG/DGEF/ mid east and the arid west of Madagascar, as was DGDRF/SC , Collection and export permit # Amphibian communities in Madagascar Julian Glos 29

GLOS ET AL.—A NEW SCAPHIOPHRYNE FROM MADAGASCAR 261

0105N/EA02/MG02). We are also indebted to of the genus Ebe avia oettger (Reptilia: Squamata: P. Kappeler, L. Razamafimanatsoa, R. Rasoloar- Gekkonidae). Herpetologica 54:18–34. ison, and the German Primate Center (DPZ, ———, ———, AND O. PRONK. 1998. The ghost geck- Go¨ttingen, Germany) for help and logistic sup- os of Madagascar: A further revision of the Mala- gasy leaf-toed geckos (Reptilia, Squamata, Gekkon- port. We thank K. Dausmann for her help and idae). Misc. Pub. Mus. Zool. Univ. Michigan 186:1– useful comments on our study and on earlier 26. versions of this manuscript. JG was supported by POSADA, D., AND K. A. CRANDALL. 1998. Modeltest: a PhD scholarship of the German Academic Ex- testing the model of DNA substitution. ioinfor- change Service (DAAD). In this research all an- matics 14:817–818. imal care protocols as outlined in the ASIH SORG, J.-P., AND U. ROHNER. 1996. Climate and tree ‘Guide for Use of Live Amphibians and Reptiles phenology of the dry deciduous forest of the Kir- in Field Research’ (http://www.asih.org/pubs/ indy forest, p. 57–80. I:Ecology and Economy of herpcoll.html, as of October 2004) have been a Tropical Dry Forest in Madagascar, Primate Re- adhered to. DNA sequences were deposited in port 46–1. J. U. Ganzhorn and J.-P. Sorg (eds.). DPZ, Go¨ttingen, Germany. Genbank (accession numbers AY834186– VAN DER MEIJDEN, A., M. VENCES, AND A. MEYER. 2004. AY834198). Novel phylogenetic relationships of the enigmatic brevicipitine and scaphiophrynine toads as re- LITERATURE CITED vealed by sequences from the nuclear Rag-1 gene. Proc. R. Soc. Lond., (Suppl.) 271:S378–S381. ¨ LOMMERS-SCHLOSSER, R. M. A. 1975. Observations on VENCES, M., F. ANDREONE,F.GLAW,N.RAMINOSOA,J. the larval development of some Malagasy frogs, E. RANDRIANIRINA, AND D. R. VIEITES. 2002a. Am- with notes on their ecology and biology (Anura: phibians and reptiles of the Ankaratra Massif: re- Dyscophinae, Scaphiophryninae and Cophylinae). productive diversity, biogeography and conserva- eaufortia 24:7–26. tion of a montane fauna in Madagascar. Ital. J. Zool. ———, AND C. P. LANC. 1991. Amphibiens (premie`re 69:263–284. partie). Faune de Madagascar 75:1–379. ———, G. APREA,T.CAPRIGLIONE,F.ANDREONE, AND DU OIS, A. 1992. Notes sur la classification des Rani- G. ODIERNA. 2002b. Ancient tetraploidy and slow dae (Amphibiens Anoures). ull. Mens. Soc. Linn. molecular evolution in Scaphiophry e: ecological Lyon 61:305–352. correlates of speciation mode in Malagasy relict am- GLAW,F.,AND M. VENCES. 1994. A Fieldguide to the phibians. Chromosome Res. 10:127–136. Amphibians and Reptiles of Madagascar. Verlags ———, AND F. GLAW. 2002. Molecular phylogeogra- GbR, Ko¨ln. phy of Boophis tephraeomystax: A test case for east- GLOS, J. 2003. The amphibian fauna of the Kirindy west vicariance in Malagasy anurans (Amphibia, An- dry forest in western Madagascar. Salamandra 39: ura, Mantellidae). Spixiana 25:79–84. 75–90. ———, C. J. RAXWORTHY,R.A.NUSS AUM, AND F. GOSNER, K. L. 1960. A simplified table for staging an- GLAW. 2003. A revision of the Scaphiophry e marmor- uran embryos and larvae with notes on identifica- ata complex of marbled toads from Madagascar, in- tion. Herpetologica 16:183–190. cluding the description of a new species. Herpetol. HEYER, W. R., M. A. DONNELLY,R.W.MCDIARMID, L.- J. 13:69–79. A. HAYEK, AND M. S. FOSTER. 1994. Measuring and WASSERSUG, R. 1984. The Pseudohemisus tadpole: a Monitoring iological Diversity: Standard Methods morphological link between microhylid Orton type for Amphibians. Smithonian Institution Press, 2 and ranoid Orton type 4 larvae. Herpetologica Washington. 40:138–149. LAURANCE,W.,AND R. O. IERREGAARD. 1997. Tropical Forest Remnants–Ecology, Management, and Con- servation of Fragmented Communities. The Uni- ( JG) DEPARTMENT OF ANIMAL ECOLOGY AND versity of Chicago Press, Chicago. TROPICAL IOLOGY, IOCENTER,WU¨ RZ URG LEVITON, A. E., R. H. GIS,JR., E. HEAL, AND C. E. UNIVERSITY,AM HULAND, 97074 WU¨ RZ URG, DAWSON. 1985. Standards in herpetology and ich- GERMANY; (FG) ZOOLOGISCHE STAATSSAMM- thyology: part I. Standard symbolic codes for insti- LUNG,MU¨ NCHHAUSENSTR. 21, 81247 MU¨ NCHEN, tutional resource collections in herpetology and GERMANY; (MV) INSTITUTE FOR IODIVERSITY ichthyology. Copeia 1985:802–832. AND ECOSYSTEM DYNAMICS,ZOOLOGICAL MUSE- MCDIARMID,R.W.,AND R. ALTIG. 1999. Tadpoles: The UM,UNIVERSITY OF AMSTERDAM,MAURITSKADE iology of Anuran Larvae. The University of Chi- cago Press, Chicago. 61, 1092 AD AMSTERDAM,THE NETHERLANDS. NELSON, R., AND N. HORNING. 1993. AVHRR-LAC es- E-mail: ( JG) [email protected]. timates of forest area in Madagascar, 1990. Int. J. de. Send reprint requests to JG. Submitted: 24 Remote Sens. 14:1463–1475. Aug. 2004. Accepted: 10 Dec. 2004. Section NUSS AUM,R.A.,AND C. J. RAXWORTHY. 1998. Revision editor: M. J. Lannoo. Amphibian communities in Madagascar Julian Glos 30

Chapter 3

Description of the tadpoles of Aglyptodactylus laticeps and A. securifer from Western Madagascar, with notes on life history and ecology

Amphibian communities in Madagascar Julian Glos 31

SHORTER COMMUNICATIONS 131

Descriptions of the Tadpoles of Aglyptodactylus laticeps and Aglyptodactylus securifer from Western Madagascar, with Notes on Life History and Ecology

1 JULIAN GLOS AND K. EDUARD LINSENMAIR

Department of Animal Ecology and Tropical Biology, Biocenter, Wu¨rzburg University, Am Hubland, 97074 Wu¨rzburg, Germany

ABSTRACT.—The three members of the Malagasy genus Aglyptodactylus (Aglyptodactylus madagascariensis, Aglyptodactylus laticeps, Aglyptodactylus securifer) form, together with Laliostoma labrosum, the subfamily Laliostominae within the Mantellidae (Ranoidea: Anura). In this paper, the morphology, life history, and ecology of sympatric tadpoles of A. laticeps and A. securifer are described and compared to those of A. madagascariensis and L. labrosum. Tadpoles of all Aglyptodactylus species are morphologically more similar to each other than to L. labrosum. However, all species differ with respect to habitat, feeding habits, and life history.

A new classification of the endemic Malagasy genus of running and standing waters dry up every year Aglyptodactylus by Vences and Glaw (2001) placed the except for a few pools in the bed of the Kirindy River. three members of the genus Aglyptodactylus (Aglypto- Mean annual precipitation is about 800 mm (range 390– dactylus madagascariensis, Aglyptodactylus laticeps, Aglyp- 1511 mm; Sorg and Rohner, 1996). Experiments were todactylus securifer) and Laliostoma labrosum (formerly done in the field camp of the German Primate Center Tomopterna labrosum; Glaw et al., 1998) in the mantellid (DPZ). (Ranoidea: Anura) subfamily Laliostominae (for alter- Collected Specimens.—Tadpoles of A. laticeps, A. native classifications see Blommers-Schlo¨sser, 1979a,b; securifer, and L. labrosum were collected in the field Channing, 1989; Dubois, 1992; Glaw et al., 1998). during November to March 1998/1999, 1999/2000 Aglyptodactylus madagascariensis is geographically and 2000/2001 in different ephemeral pools within widely distributed in the humid eastern parts of the Kirindy Forest. Fertilized eggs from amplectant Madagascar. Laliostoma labrosum is also widely distrib- pairs were reared in plastic aquaria filled with uted primarily in the drier regions of the south and rainwater and tadpoles were fed ad libitum with west (Glaw and Vences, 1994; Vences and Glaw, 2000). commercial fish food (TetraMinTabsÒ). Individual In contrast, A. laticeps and A. securifer seem to be tadpoles collected in the field, as well as those reared restricted to western Madagascar. Aglyptodactylus lat- under controlled conditions, were preserved in 4% iceps has been found exclusively in the Kirindy Forest, formalin. The nomenclature of morphological features whereas A. securifer has been found at two localities follows McDiarmid and Altig (1999), and staging is very distant from each other (Kirindy Forest and according to Gosner (1960). Measurements of pre- Berara; Andreone et al. 2001). As for many Malagasy served specimens were taken with a stereo micro- anurans, the tadpoles of A. laticeps and A. securifer have scope (ZeissÒ Stemi SV 6) with a measuring ocular, not been described, and information about the life and drawings were done with a camera lucida. history, ecology, and behavior of both adults and larvae Coloration was noted and measurements of length, are scarce. We describe the tadpoles of A. laticeps and mass, and developmental stage were taken from A. securifer, present morphological and ecological data, living tadpoles. Voucher specimens were deposited point out distinctions of A. madagascariensis and L. in the collection of the Zoologische Staatssammlung labrosum, and provide experimental and observational Mu¨ nchen (A. laticeps: ZSM 1157–2001, A. securifer: notes on larval life history and breeding habitat choice ZSM 1160–2001, L. labrosum: ZSM 1161–2001). The of A. laticeps and A. securifer. only preserved tadpole of A. madagascariensis that we are aware of is from the collection of the Zoo¨logisch MATERIALS AND METHODS Museum Amsterdam (ZMA 7179). This specimen Study Area.—This study was conducted in the served for interspecific morphological comparisons. Kirindy Forest, one of the largest remnants of the dry, Life History, Observations, and Natural History.—Eggs deciduous forest of western Madagascar (120 km2; are laid by A. laticeps, A. securifer, and L. labrosum as Nelson and Horning, 1993). It is located about 50 km a single layered surface film. Clutch size was de- northeast of Morondava and 20 km inland (448399E, termined in the field by taking photographs of distinct 208039S; 18–40 m above sea level; Sorg and Rohner, spawns and counting the eggs afterward. Additionally, 1996). The rainy season of 3–5 months from Novem- eggs were counted from amplectant pairs that de- ber/December to February/March is followed by a dry posited eggs at the field camp. Ovum size was season of 7–9 months of virtually no rain. All stretches determined by measuring each ovum diameter in a sample of 20 from each clutch. Duration of the embryonic and larval period were assessed for tadpoles 1 Corresponding Author. E-mail: glos@biozentrum. in the field as well as on tadpoles reared under uni-wuerzburg.de controlled conditions (plastic aquaria, fed ad libitum Amphibian communities in Madagascar Julian Glos 32

132 SHORTER COMMUNICATIONS

on leaf litter and commercial fish food). Froglet size at and under controlled conditions. Hatchlings attached metamorphosis was recorded for specimens in the field; to the jelly mass for about 12 h before reaching the free- metamorphic length was measured to the nearest 0.05 swimming stage. Larval development was very rapid; mm with calipers and metamorphic mass was assessed under experimental conditions tadpoles reached Stage to the nearest 2 mg with a fine-scale balance (specimens 42 in a minimum of 10 days (17.163.8, range 10.3–29.0, were released). Duration of the larval period was N 5 391). In a natural, rapidly desiccating pond, defined as the time from hatching until the eruption of a minimum developmental time of 12 days was the forelimbs (Stage 42); length and mass at meta- observed. Temperatures in these ponds was very high; morphosis were measured after the complete loss of the temperature (3 cm below water surface) in two natural tail (Stage 46). breeding ponds ranged from 20.08C at night to 41.58C Tadpoles of sympatric species were identified by during the day (data from iButton-TMEX data loggers). repeatedly sampling 145 natural ponds with the box- In the field, mass at metamorphosis was 194 6 101 mg method (Heyer et al., 1994) or by standardized dip- (90–458 mg, N 5 35), and body length at meta- netting. Feeding habits and microhabitat choice were morphosis was 11.3 6 1.7 mm (9.0–15.0 mm, N 5 52; observed in the field as well as in plastic aquaria in the Table 2). field camp. Natural History.—Breeding took place only on nights after heavy rains (.30 l/m2). The breeding season RESULTS starts about one month after the first heavy rain of the Aglyptodactylus laticeps Glaw, Vences, and season and extended over the whole rainy season (Glos, Bo¨hme, 1998 2003). Aglyptodactylus laticeps spawned only in relatively small ephemeral forest ponds (0.5–200 m2) with Description of Tadpole.—Because the tadpoles showed a high risk of desiccation (Glos, 2003). The substrate ontogenetic changes in oral disc structure and apparent of the breeding ponds was usually sand or clay. The changes in body proportions, we give data for lentic, benthic tadpoles were rarely found in midwater. a younger (Stage 28–35; N 5 9) and an older stage 5 Experiments in aquaria suggest that they mainly fed on (Stage 37–39; N 6; Table 1). Within each stage class decaying leaves. However, it was not clear whether there is only little variation in size and body propor- they ate the plant material itself or the microorganisms tions. Body is ovoid in lateral view, elliptical in dorsal 6 on the leaves. Because they usually rasped the leaves view (Fig. 1A–B). Body length 0.35 0.02 (Stage 28–35) and only occasionally bit off larger parts of the leaves, and 0.33 6 0.02 (Stage 37–39) of total length, body 1.6 6 6 we suppose the latter. They never fed on living plant 0.2 (Stage 28–35) and 1.5 0.1 (Stage 37–39) longer than material nor did they actively prey upon smaller wide, 2.0 6 0.3 (Stage 28–35) and 2.0 6 0.2 times longer 6 6 tadpoles; however, they fed on dead tadpoles. In than high, 1.3 0.1 (Stage 28–35) and 1.3 0.1 (Stage natural breeding ponds of A. laticeps, the following 37–39) times wider than high. Snout rounded in dorsal sympatric tadpole species were found: Laliostoma and lateral profiles. Eyes rather small (eye diameter labrosum, Boophis doulioti, Mantella betsileo, Dyscophus 0.860.1 mm for Stage 28–35, 1.460.2 mm for Stage 37– 6 insularis, Scaphiophryne calcarata, Scaphiophryne brevis, 39), and separated by distance 1.60 0.14 (Stage 28–35) and Scaphiophryne cf. marmorata and 1.74 6 0.15 (Stage 37–39) times the internarial distance, and directed dorsolaterally. Nostrils dorsal, apertures directed dorsally, nostrils closer to eyes (1.2 6 Aglyptodactylus securifer 0.2 mm and 1.9 6 0.2 mm) than to snout tip (1.6 6 0.4 Glaw, Vences, and Bo¨hme, 1998 mm and 2.0 6 0.5 mm). Spiracle sinistral, at midbody, Description of Tadpole.—We give data for an earlier below midline, lateral wall not free but present as slight (Stage 25–34; N 5 5) and a later stage class (Stage 37–39; ridge, directed dorsal-posteriorly. Medial vent tube with N 5 3; Table 1). Body ovoid in lateral view, elliptical in lateral displacement, rather short, aperture dextral. dorsal view (Fig. 1D–E). Body length 0.43 6 0.03 (Stage Tail fins convex, higher from the posterior end of 25–34) and 0.33 6 0.07 (Stage 37–39) of total length, vent tube to the middle of the tail, dorsal fin higher than body 1.7 6 0.1 (Stage 25–34) and 1.3 6 0.1 (Stage 37–39) ventral fin. Tail tip rounded. Origin of dorsal fin slightly longer than wide, 2.4 6 0.2 (Stage 25–34), and 1.8 6 0.2 anterior to base of tail. times longer than high, 1.460.1 (Stage 25–34) and 1.46 Oral disc ventral, LTRF 4(2–4)/3(2–3) or 5(2–5)/3(2– 0.1 (Stage 37–39) times wider than high. Snout rounded 3) (Stage 28–35) and 6(2–6)/3(2–3) (Stage 37–39; Fig. in dorsal and lateral profiles. Eyes rather small (eye 1C). Labial teeth small, teeth larger in tooth row 1 and 2 diameter 1.0 60.3 mm for Stage 25–34, 1.3 60.1 mm for of upper labium. Marginal papillae small, conical, with Stage 37–39), and separated by distance 1.796 0.27 mm rounded tips, not pigmented; length of marginal (Stage 25–34) and 1.68 6 0.27 mm (Stage 37–39) times papillae 0.07 mm, density 19.5 papillae/mm. Two rows internarial distance, directed dorsolaterally. Nostrils of marginal papillae interrupted anteriorly; submar- dorsal, slightly closer to snout tip (1.360.4 mm) than to ginal papillae not present. Upper jaw sheath concave, eyes (1.4 6 0.2 mm) in Stage 25–34 and closer to eyes lower jaw sheath V-shaped, both sheath finely serrate (1.6 6 0.1 mm) than to snout tip (1.8 6 0.0 mm) in Stage and fully black pigmented. 37–39; nostril apertures directed dorsally. Spiracle In life, dorsal color is brownish, ventral color is a dull sinistral, at midbody, below midline, lateral wall not gray-brown. Caudal musculature is entirely and free but present as slight ridge, directed posteriorly. heavily pigmented, dorsal tail fin is slightly pigmented, Medial vent tube with lateral displacement, rather ventral tail fin not pigmented. short, aperture dextral. Life History.—Clutch size was 3636 6 470 eggs Tail fins convex, higher from the posterior end of (mean 6 SD; range 2686–4231, N 5 7), ovum diameter vent tube to the middle of the tail, dorsal fin higher than was 1.76 6 0.20 mm (1.26–2.36; N 5 722). Duration of ventral fin. Tail tip rounded. Origin of dorsal fin is embryonic development was 24 h both in the field slightly anterior to base of tail. Amphibian communities in Madagascar Julian Glos 33

SHORTER COMMUNICATIONS 133

TABLE 1. Measurements of tadpoles of Aglyptodactylus laticeps, Aglyptodactylus securifer, and Laliostoma labrosum in an earlier and later stage class; mean 6 SD (range). Staging after Gosner (1960), measurements follow McDiarmid and Altig (1999). BL 5 body length, ED 5 eye diameter, IND 5 internarial distance, IOD 5 interorbital distance, MTH 5 maximum tail height, NED 5 naris–eye distance, ODW 5 oral disc width, SED 5 snout–eye distance, SND 5 snout–naris distance, SSD 5 snout–spiracle distance, TL 5 total length, TMH 5 tail musculature height, TMHM 5 tail musculature height at midlength of tail.

A. laticeps A. securifer A. madagascariensis L. labrosum Stage 28–35 (N 5 6) 25–34 (N 5 5) 35 (N 5 1) 28–32 (N 5 4) TL 19.3 6 2.5 (15.3–22.5) 18.9 6 2.5 (16.3–22.6) 21.8 21.4 6 1.4 (19.3–23.0) BL 6.8 6 0.7 (5.8–8.2) 8.1 6 1.1 (6.3–9.3) 8.3 8.8 6 0.5 (8.1–9.4) BL/BW 1.6 6 0.2 (1.4–1.9) 1.7 6 0.1 (1.7–1.9) no data 1.7 6 0.1 (1.6–1.8) BL/BH 2.0 6 0.3 (1.7–2.6) 2.4 6 0.2 (2.2–2.6) no data 2.1 6 0.1 (1.9–2.2) BW/BH 1.3 6 0.1 (1.1–1.3) 1.4 6 0.1 (1.3–1.5) no data 1.2 6 0.0 (1.2–1.3) TMHM/MTH 0.4 6 0.1 (0.2–0.5) 0.3 6 0.0 (0.3–0.4) 0.4 0.4 6 0.0 (0.3–0.4) ED 0.8 6 0.1 (0.6–0.9) 1.0 6 0.3 (0.7–1.5) 0.9 1.0 6 0.1 (0.9–1.2) IOD/IND 1.6 6 0.1 (1.5–1.9) 1.8 6 0.3 (1.5–2.3) 1.5 1.4 6 0.5 (0.6–1.9) SND/SED 0.6 6 0.1 (0.6–0.8) 0.5 6 0.1 (0.4–0.6) 0.5 0.6 6 0.0 (0.6–0.7) NED 1.2 6 0.2 (0.9–1.4) 1.4 6 0.2 (1.2–1.6) 1.7 1.5 6 0.1 (1.3–1.7) ODW 1.7 6 0.2 (1.4–1.9) 1.8 6 0.2 (1.6–2.3) 2.1 1.8 6 0.1 (1.6–2.0) Stage 37–39 (N 5 6) 37–39 (N 5 3) 37–39 (N 5 3) TL 30.6 6 4.2 (26.0–36.6) 23.9 6 1.2 (22.3–25.0) 41.6 6 1.7 (40.0–44.0) BL 10.1 6 0.8 (9.1–11.0) 7.9 6 1.3 (6.2–9.2) 15.0 6 0.7 (14.0–15.7) BL/BW 1.5 6 0.1 (1.3–1.6) 1.3 6 0.1 (1.1–1.5) 1.5 6 0.0 (1.4–1.5) BL/BH 2.0 6 0.2 (1.8–2.3) 1.8 6 0.2 (1.6–2.0) 2.0 6 0.0 (2.0–2.0) BW/BH 1.3 6 0.1 (1.2–1.4) 1.4 6 0.1 (1.3–1.4) 1.4 6 0.0 (1.3–1.4) TMHM/MTH 0.4 6 0.1 (1.2–1.4) 0.2 6 0.0 (0.3–0.4) 0.4 6 0.1 (0.3–0.5) ED 1.4 6 0.2 (1.0–1.8) 1.3 6 0.1 (1.2–1.4) 2.2 6 0.1 (2.0–2.3) IOD/IND 1.7 6 0.2 (1.5–2.0) 1.7 6 0.3 (1.3–2.0) 5.4 6 0.4 (4.8–5.6) SND/SED 0.6 6 0.1 (0.4–0.6) 0.6 6 0.0 (0.6–0.7) 0.5 6 0.0 (0.5–0.6) NED 1.9 6 0.2 (1.6–2.1) 1.6 6 0.1 (1.5–1.7) 2.7 6 0.3 (2.3–3.1) ODW 2.4 6 0.3 (1.8–2.8) 2.1 6 0.1 (2.0–2.2) 3.4 6 0.0 (3.4–3.5)

Oral disc ventral, LTRF 4(2–4)/3(2–3), 5(2–5)/3(2–3) one week after the first heavy rains at the beginning of or 6(2–6)/3(2–3) (Stage 25–34) and 6(2–6)/3(2–3) (Stage the rainy season. Only freshly filled pools in the bed of 37–39; Fig. 1F). Labial teeth rather small. Marginal pap- the Kirindy River were used as breeding sites and were illae small, conical, with rounded tip, not pigmented; the only available bodies of waters at this time of the length of marginal papillae 0.06 mm, density 15.6 year. The size of the breeding pools ranged from 1 to papillae/mm. Two rows of marginal papillae interrup- .50 m2, and their substrate was usually rocky. ted anteriorly; submarginal papillae not present. Upper The tadpoles were benthic and were rarely found in jaw sheath concave, lower jaw sheath V-shaped, both midwater. They showed the same feeding habits as A. sheaths finely serrate and fully black pigmented. laticeps. In natural breeding ponds of A. securifer, the In life, dorsal color is brownish, ventral color is a dull following sympatric tadpole species were found: gray-brown. Caudal musculature is entirely but faintly Boophis doulioti, Mantella betsileo, Dyscophus insularis, pigmented, dorsal and ventral tail fins are not Scaphiophryne calcarata, S. brevis and S. cf. marmorata. pigmented. Life History.—Clutch size was 1736 6 416 eggs (mean 6 SD; range 825–2363, N 5 17), ovum diameter DISCUSSION was 1.62 6 0.19 mm (1.14–2.14; N 5 769). Duration of Adults of A. laticeps, A. securifer, and of sympatric L. embryonic development was 24 h both in the field labrosum clearly differ in size and morphological and under controlled condition. Hatchlings attached to characters (Glaw et al., 1998). However, the tadpoles the jelly mass for about 12 h before reaching the free- of these species and of A. madagascariensis are all of the swimming stage. Larval development was rapid; under exotrophic, lentic, and benthic type and are morpho- experimental conditions tadpoles reached Stage 42 in logically very similar. This similarity could be inter- a minimum of 19 days (21.5 6 1.73, range 19–32, N 5 preted as supporting the view of a close phylogenetic 208; Table 2). In natural breeding ponds developmental relationship of these species (Vences and Glaw, 2001) or time was about three weeks. There, temperature during the result of a high ecological convergence in respect to the day reached .358C just below the water surface. In feeding and microhabitat use. These views are not the field, mass at metamorphosis was 343666 mg (178– mutually exclusive. There are differences in coloration 488 mg, N 5 206), and body length at metamorphosis and intensity of pigmentation as well as in the number was 11.2 6 0.7 mm (9.0–12.7, N 5 205). of tooth rows, but these are not fixed characters within Natural History.—Aglyptodactylus securifer was an each species. Coloration can vary intraspecifically extreme explosive breeder that spawns only for about between tadpoles living in different habitats (McDiar- Amphibian communities in Madagascar Julian Glos 34

134 SHORTER COMMUNICATIONS

FIG. 1. Lateral view (A), dorsal view (B), and oral disc (C) of Aglyptodactylus laticeps tadpole at stage 35 and of Aglyptodactylus securifer (D, E, F) tadpole at stage 34. Scale bar represents 5 mm (A, B, D, E) and 1 mm (C, F).

mid and Altig, 1999). To a certain degree, tooth row similarity. Larvae of A. laticeps develop very rapidly, morphology is also as variable a character, as indicated presumably reflecting an adaptation to the high by different LTRFs both within the Kirindy population desiccation risk for their natural habitat. In addition, of A. laticeps and between populations of L. labrosum the larger ovum size of A. laticeps can be seen as from Kirindy Forest and from Toliara (Glaw and advantageous in the very ephemeral and unpredict- Vences, 1994). able habitats (Resetarits, 1996). Apart from the high Aglyptodactylus has five to seven upper teeth rows, desiccation risk, these ponds often offer good con- and a positive imbalance of þ2toþ4 between the ditions for tadpoles (e.g., high food resources, low number of tooth rows on upper and lower labium interspecific competition and low predation pressure; (Altig and Johnston, 1989). This character is rather Roth and Jackson, 1987; Pearman, 1995). In their unusual for lentic tadpoles; it is usually found in stream natural breeding pools, A. laticeps is larger at meta- tadpoles (McDiarmid and Altig, 1999). One possible morphosis than A. securifer, despite its shorter larval explanation is that ancestors of Aglyptodactylus might development. The risk of spawning in ephemeral have had a stream tadpole, and the extra rows are waters and hence a high mortality risk by desiccation vestiges of that ancestry. However, the presence of five might be rewarded by a fitness gain for A. laticeps in or more upper teeth rows is found in several lentic having a head start in the terrestrial life phase (Berven species throughout all three subfamilies of the Man- and Gill, 1983; Smith, 1987; Berven, 1990; Scott, 1994). tellidae; within the subfamily Laliostominae (L. labro- However, growth rate and mortality risk in the larval sum: 5(2–5)/3[1]; pers. obs., or 6(2–6)/3; Glaw and phase have to be balanced against these variables in Vences, 1994) as well as in lentic members of the the terrestrial phase (Werner, 1986, 1988), and there Boophinae (e.g., Boophis xerophilus 5(2–5)/3(1), B. are no data about growth rate and mortality risk of doulioti 5(2–5)/3(1); pers. obs.) and Mantellinae (e.g., postmetamorphic Aglyptodactylus. Development of L. Mantidactylus wittei 5(2–5)/3 or 6(2–6)/3; Glaw and labrosum and A. madagascariensis is clearly longer, and Vences, 1994). Rarely, similar LTRFs are found in lentic size at metamorphosis is greater. tadpoles from other parts of the world (e.g., Hemisus: In contrast to A. laticeps and A. securifer, tadpoles of L. Hemisotidae; Heleioporus, Lechriodus: Myobatrachidae; labrosum are facultatively carnivorous. This could either Aubria, Indirina, Pyxicephalus: Ranidae; (McDiarmid represent a real difference in feeding strategy, or and Altig, 1999). Tadpoles of L. labrosum differ from alternatively, merely a size effect. Tadpoles or Aglypto- those of Aglyptodactylus sp. in size; at a particular dactylus may be simply too small or gape limited to developmental stage they are constantly larger than successively prey upon and consume smaller tadpoles those of the other species. Furthermore, the ratio of (Brodie and Formanowicz, 1983; Formanowicz, 1986; interorbital distance and internarial distance is consid- Gascon, 1989; Crossland, 1998). In an artificial environ- erably greater in Stages 37–39. ment (plastic aquarium) under low food conditions, The different life histories of A. laticeps, A. securifer tadpoles of Laliostoma that were more than twofold and L. labrosum contrast with their morphological larger than their prey were observed to prey upon Amphibian communities in Madagascar Julian Glos 35

SHORTER COMMUNICATIONS 135

TABLE 2. Life-history parameters in natural habitats and overview of morphological differences of larvae of Aglyptodactylus laticeps, Aglyptodactylus securifer. Aglyptodactylus madagascariensis (data from Blommers-Schlo¨sser, 1979a,b) and Laliostoma labrosum (mean 6 SD; range).

Life-history parameter A. laticeps A. securifer A. madagascariensis L. labrosum Clutch size 3636 6 470 1736 6 416 no data 3628 6 2685 2686–4231 825–2363 1717–4420 N 5 7 N 5 17 N 5 3 Egg size (mm) 1.76 6 0.20 1.62 6 0.19 1.60 6 0.12 1.26–2.36 1.14–2.14 1.24–1.90 N 5 722 N 5 769 N 5 20 Larval duration (days) min 12 min 20–25 28–42 in aquarium min . 30 Metamorphic 11.3 6 1.7 9.5 6 1.0 12 16.7 6 2.1 size (mm) (9.0–15.0) (7.1–12.8) (10–15) (12.9–20.1) N 5 52 N 5 206 N 5 14 N 5 24 Metamorphic 194 6 101 102 6 100 no data 529 6 209 weight (mg) (90–458) (38–146) (138–1000) N 5 35 N 5 205 N 5 24 Labial tooth row 5(2–5)/3(1) or 6(2–6)/3(1) 6(2–6)/3(1) or 5(2–5)/3[1] or formula 6(2–6)/3(1) 7(2–7)/3(1) 6(2–6)/3 Dorsal coloration brownish brownish brownish gray Ventral gray-brown gray-brown gray-brown silvery coloration Caudal heavy faint reticulated faint pigmentation Feeding habits detritivorous detritivorous probably detrtivorous, detritivorous facultatively carnivorous Breeding habitat small forest river bed pools stagnant water generalistic pools

tadpoles of Aglyptodactylus. This behavior was not L. labrosum occupy different niches with respect to observed in smaller tadpoles of Laliostoma (pers. obs.). breeding site choice, life history, and feeding habits. Tadpoles of Aglyptodactylus never reached the size at which tadpoles of Laliostoma were observed to attack Acknowledgments.—We thank K. Dausmann, M.-O. other tadpoles. However, they were observed feeding Ro¨del, and F. Glaw for their help and useful comments on dead tadpoles. on our study and on earlier versions of this manu- Aglyptodactylus laticeps and A. securifer are local script. We also thank P. Kappeler, L. Razamaﬁmanat- endemics whose geographic distribution is limited to soa, R. Rasoloarison, and the German Primate Center only a few sites in western Madagascar. Although (DPZ, Go¨ttingen, Germany) for their help and for they are sympatric in the Kirindy Forest and are logistic support. Research and export permits were locally abundant, there is neither spatial nor temporal provided by the DEF/Madagascar (Ministe`re des Eaux overlap of their tadpoles. There was only one et Foreˆts, Madagascar; export permit 0250-EAL/ observation of spatial and temporal overlap of calling MG01/CWN). The German Academic Exchange Ser- adult males of these species within the four-year vice (DAAD) supported this study. study. Aglyptodactylus securifer breeds exclusively in pools within the river bed at the very beginning of LITERATURE CITED the rainy season, and A. laticeps breed in highly ephemeral pools in the dense forest throughout the ALTIG, R., AND G. F. JOHNSTON. 1989. Guilds of anuran whole rainy season. Laliostoma labrosum, in contrast, is larvae: relationships among developmental modes, distributed over a much wider geographic range and morphologies, and habits. Herpetological Mono- locally uses a wide range of pond types for spawning. graphs 3:81–109. Aglyptodactylus madagascariensis occurs mainly in the ANDREONE, F., M. VENCES, AND J. E. RANDRIANIRINA. 2001. humid eastern parts of Madagascar and is not Patterns of amphibian and reptile diversity at sympatric with the other species. Ecological data for Berara Forest (Sahalamaza Peninsula), NW Mada- this species are scarce, but it presumably prefers more gascar. Italian Journal of Zoology 68:235–241. permanent pools as breeding sites (Blommers- BERVEN, K. A. 1990. Factors affecting population Schlo¨sser, 1979a,b). ﬂuctuations in larval and adult stages of the wood The high morphological similarities of A. laticeps and frog (Rana sylvatica). Ecology 71:1599–1608. A. securifer and to a lesser extent of L. labrosum BERVEN, K. A., AND D. E. GILL. 1983. Interpreting combined with the sympatry of these species suggest geographic variation in life-history traits. American niche differentiation. Indeed A. laticeps, A. securifer, and Zoologist 23:85–97. Amphibian communities in Madagascar Julian Glos 36

136 SHORTER COMMUNICATIONS

BLOMMERS-SCHLO¨ SSER, R. M. A. 1979a. Biosystematics of for Amphibians. Smithonian Institution Press, the Malagasy frogs. I. Mantellinae (Ranidae). Washington, DC. Beaufortia 352:1–76. MCDIARMID,R.W.,AND R. ALTIG. 1999. Tadpoles: The ———. 1979b. Biosystematics of the Malagasy frogs. II. Biology of Anuran Larvae. Univ. of Chicago Press, The genus Boophis (Rhacophoridae). Bijdragen Chicago. Dierkunde 49:261–312. NELSON,R.,AND N. HORNING. 1993. AVHRR-LAC BRODIE JR. E. D., AND D. R. FORMANOWICZ JR. 1983. Prey estimates of forest area in Madagascar, 1990. size preference of predators: differential vulnerabil- International Journal of Remote Sensing 14:1463– ity of larval anurans. Herpetologica 39:67–75. 1475. CHANNING, A. 1989. A re-evaluation of the phylogeny PEARMAN, P. B. 1995. Effects of pond size and consequent of Old World treefrogs. South African Journal of predator density on two species of tadpoles. Zoology 24:116–131. Oecologia 102:1–8. CROSSLAND, M. R. 1998. Predation by tadpoles on toxic RESETARITS JR. W. J. 1996. Oviposition site and life toad eggs: the effect of tadpole size on predation history evolution. American Zoologist 36:205–215. success and tadpole survival. Journal of Herpetol- ROTH, A. H., AND J. F. JACKSON. 1987. The effect of pool ogy 32:443–446. size on recruitment of predatory insects and on DUBOIS, A. 1992. Notes sur la classification des Ranidae mortality in a larval anuran. Herpetologica 34: (Amphibiens Anoures). Bulletin Mensuel de la 224–232. Societe Lineenne de Lyon 61:305–352. SCOTT, D. E. 1994. The effect of larval density on adult FORMANOWICZ JR. D. R. 1986. Anuran tadpole/aquatic demographic traits in Ambystoma opacum. Ecology insect predator-prey interactions: tadpole size 75:1383–1396. and predator capture success. Herpetologica 42: SMITH, D. C. 1987. Adult recruitment in chorus frogs: 367–373. effects of size and date at metamorphosis. Ecology GASCON, C. 1989. Predator-prey size interactions in 68:344–350. tropical ponds. Revista Brasileira de Zoologia 6: SORG, J.-P., AND U. ROHNER. 1996. Climate and tree 701–706. phenology of the dry deciduous forest of the GLAW,F.,AND M. VENCES. 1994. A Fieldguide to the Kirindy forest. In J. U. Ganzhorn and J.-P. Sorg Amphibians and Reptiles of Madagascar. M. Vences (eds.), Ecology and Economy of a Tropical Dry and F. Glaw Verlags GbR, Ko¨ln, Germany. Forest in Madagascar, pp. 57–80. Primate Report GLAW, F., M. VENCES, AND W. BO¨ HME. 1998. Systematic 46-1, Erich Goltz GmbH, Go¨ttingen, Germany. revision of the genus Aglyptodactylus Boulenger, VENCES, M., AND F. GLAW. 2000. Laliostoma labrosum. 1919 (Amphibia: Ranidae), and analysis of its Herpetological Review 31:182. phylogenetic relationships to other Madagascan ———. 2001. When molecules claim for taxonomic ranid genera (Tomopterna, Boophis, Mantidactylus changes: new proposals on the classification of Old and Mantella). Journal of Zoological Systematics World treefrogs. Spixiana 24:85–91. and Evolutionary Research 36:17–37. WERNER,E.E.1986.Amphibianmetamorphosis: GLOS, J. 2003. The amphibian fauna of the Kirindy dry growth rate, predation risk, and the optimal size forest in western Madagascar. Salamandra 39: at transformation. American Naturalist 128: 75–90. 319–341. GOSNER, K. L. 1960. A simplified table for staging ———. 1988. Size, scaling and the evolution of anuran embryos and larvae with notes on identi- complex life cycles. In B. Ebenman and L. Persson fication. Herpetologica 16:183–190. (eds.), Size-Structured Populations: Ecology and HEYER, W. R., M. A. DONNELLY,R.W.MCDIARMID, L.-A. Evolution, pp. 60–81. Springer, Berlin, Germany. HAYEK, AND M. S. FOSTER. 1994. Measuring and Monitoring Biological Diversity: Standard Methods Accepted: 31 October 2003. Amphibian communities in Madagascar Julian Glos 37

Chapter 4

Description of the tadpoles of Boophis doulioti and B. xerophilus from Western Madagascar with notes on larval life history and breeding ecology

Amphibian communities in Madagascar Julian Glos 38

Amphibia-Reptilia 26 (2005): 459-466

Description of the tadpoles of Boophis doulioti and B. xerophilus from Western Madagascar with notes on larval life history and breeding ecology

Julian Glos, K. Eduard Linsenmair

Abstract. The genus Boophis is very diverse in species number and ecology within the endemic Malagasy family Mantellidae. Two species, Boophis doulioti and B. xerophilus, stand out as they are breeding in stagnant waters in the dry west and south of Madagascar, while the large majority are specialized to breeding in brooks of the rainforests in the east. In this paper the morphology, larval life history, ecology of the sympatric tadpoles and some aspects of the breeding ecology of B. doulioti and B. xerophilus are described. Tadpoles of the two species differ in many morphological characters and proportions, with differing positions of the eyes and body coloration being the most prominent. There is a considerable temporal and spatial overlap in the choice of breeding ponds by both species in the study area. However, B. xerophilus tadpoles are restricted to larger and more permanent breeding ponds while the generalistic B. doulioti uses also smaller and more ephemeral water bodies. Variability in larval life history traits of B. doulioti individuals growing up in ephemeral ponds compared to those from more permanent ponds indicates considerable developmental plasticity in tadpole life history of this species.

Introduction in all major habitat types (Glos, 2003). Boophis xerophilus Glaw and Vences, 1997 is known Treefrogs of the genus Boophis (Boophinae: from one locality in the west (Kirindy Forest) Mantellidae: Anura; Vences and Glaw, 2001) and one in the south (Berenty Park; Glaw and exhibit some of the most spectacular radia- Vences, 1997). It is more specialized in its habi- tions of the anuran fauna of Madagascar (Glaw tat use as B. doulioti, as it generally uses only and Vences, 1994). While most Boophis species the larger (surface area > 100 m2) and more breed in brooks (48 species, versus 10 species permanent waters as breeding ponds. It is found of pond breeders; Vences et al., 2002) mainly primarily in the closed forest, only rarely in sa- restricted to the humid east, two closely related vannah or anthropogenic habitats (Glos, 2003). species stand out as they are breeding in stag- In Kirindy Forest, both species are sympatric, nant waters in the dry west and south of Mada- and their larvae regularly occur syntopically. gascar. Boophis doulioti (Angel, 1934) was As for many Malagasy anurans, the tadpoles just recently resurrected from the synonymy of and their ecology have not been described so far. B. tephraeomystax (Vences and Glaw, 2002). We describe the tadpoles of both species, and It is a widespread generalist species occur- provide data on larval life history and breeding ring in low- to mid-altitudes throughout west- habitat choice of B. doulioti and B. xerophilus. ern, south-western and southern Madagascar (Blommers-Schlösser and Blanc, 1991; Glaw and Vences, 1994; Vences and Glaw, 2002). At Material and methods Kirindy, one of the largest remnants of Western Malagasy dry forest, a wide spectrum of pond Study Area

types is used for reproduction, and it is found This study was conducted in the Kirindy Forest, one of the largest remnants of the dry, deciduous forest of western Madagascar (120 km2; Nelson and Horning, 1993). It is Department of Animal Ecology and Tropical Biology, located about 50 km northeast of Morondava and 20 km Biocenter, Würzburg University, Am Hubland, 97074 inland (44◦39E, 20◦03S; 18-40 m a.s.l.; Sorg and Rohner, Würzburg, Germany 1996). The rainy season lasts 3 to 5 months (November- e-mail: [email protected] December to February-March) and is followed by a dry

460 Julian Glos, K. Eduard Linsenmair

season with virtually no rain for 7 to 9 months. All stretches natural ponds with the box-method (Rödel, 2000) and by of running and standing waters are temporary except for standardized dip-netting. Feeding habits and microhabitat a few pools in the bed of the Kirindy River. Mean annual choice were observed in the ﬁeld as well as in plastic aquaria precipitation is about 800 mm (range 390-1511 mm; Sorg in the ﬁeld camp. and Rohner, 1996).

Collected Specimens Results Tadpoles of Boophis doulioti and B. xerophilus were collected during November to March 1998-99, 1999-2000 and Boophis doulioti (Angel, 1934) 2000-01 in different waters of Kirindy Forest. Fertilized eggs from two (B. doulioti) and four (B. xerophilus)am- Description of tadpole plectant pairs were reared in plastic aquaria filled with rainwater and tadpoles were fed ad libitum with commer- Because the tadpoles showed no considerable ® cial fish food (TetraMinTabs ). We preserved 20 tadpoles ontogenetic changes in oral disc structure and from each species at six different times in 4% formalin (table 1). Nomenclature of morphological features follows body proportions (see table 1), we give data McDiarmid and Altig (1999), and staging is according to as mean and standard deviation for tadpoles of Gosner (1960). Measurements of preserved specimens were all developmental stages; exceptions are listed taken with a stereo microscope (Zeiss® Stemi SV 6) with a measuring ocular, and drawings were done with a cam- below. In table 1, we summarize data for a era lucida. Coloration was noted on living specimens and young (stage 25), medium (stage 27-31) and measurements of morphometric variables were taken from older stage class (stage 37-41). Body ovoid in preserved specimens. Voucher specimens were deposited in the collection of the Zoologische Staatssammlung München lateral and dorsal view (fig. 1a, b), body length (B. doulioti: ZSM 614/2003, B. xerophilus: ZSM 615/2003). 0.35 ± 0.02 of total length, body 1.54 ± 0.14 longer than wide, 1.84 ± 0.20 times longer than Field observations high, 1.20 ± 0.11 times wider than high, snout Eggs were deposited by both species as a single layered surrounded in dorsal and lateral profile, eyes me- face film. Clutch size was determined in the field by tak- dium sized, and directed dorsolaterally. Nos- ing photographs of distinct clutches and counting the eggs trils dorsal, apertures directed laterally, nostrils afterwards. Additionally, eggs were counted from amplectant pairs that we collected in the field and kept over night closer to snout tip than to eyes (ratio naris-eye in aquaria where they always spawned. We determined the distance/snout-naris distance 1.13 ± 0.14), spir- weight of females before and after spawning to determine acle sinistral, at midbody, below midline, lateral the fraction of egg weight on total weight in gravid females. We measured ovum size as ovum diameter in a sample of wall absent, directed dorso-posteriorly. Medial 20 from each clutch. In one clutch from B. doulioti only vent tube with lateral displacement, rather short, 19 eggs, and in one clutch of B. xerophilus only 15 eggs aperture dextral. were measured (B. doulioti: n = 359 eggs, B. xerophilus: n = 295 eggs). Duration of embryonic and larval develop- Tail fins convex, higher from the posterior ment were assessed for tadpoles in the field as well as for end of vent tube to the middle of the tail, dorsal tadpoles reared under controlled conditions. For the latter, fin slightly higher than ventral fin (ratio dorsal tadpoles of each species were reared from eggs of four am- . ± . plectant pairs in two 20 l plastic aquaria (density about 20 fin height/ventral fin height 1 17 0 05 for tadpoles/l; see above). The aquaria were set up on platforms stage 25, 1.41±0.26 for stage 27-31, and 1.47± shielded from direct sunlight and rain. Temperature range in . ◦ 0 24 for stage 37-41). Tail tip slightly rounded. the aquaria was 23-27 C. Origin of dorsal fin at the base of tail. Froglet size at metamorphosis was recorded in the field and was measured to the nearest 0.05 mm with calipers. Oral disc ventral, Labial Tooth Row Formula Metamorphic weight was assessed to the nearest 2 mg with a (LTRF) 3(2-3)/3(1) or 4(2-4)/3(1) for stage 25, fine-scale balance (Ohaus Navigator, 2 mg precision). These 5(2-5)/3(1) for stage 27-31 and 5(2-5)/3(1) or specimens were released subsequently to measurements. Duration of larval period was defined as the time from 6(2-6)/3(1) for stage 37-41 (fig. 1c). Marginal hatching until the appearance of the forelimbs (stage 42); papillae conical, with rounded tips, not pig- length and weight at metamorphosis were measured after mented; in stage 37, length of marginal papillae the complete resorption of the tail (stage 46). Sympatric tadpole species were identified and breeding 0.23 mm, density 11.1 papillae/mm. Two rows habitat choice was assessed by repeatedly sampling 145 of marginal papillae interrupted anteriorly and Amphibian communities in Madagascar Julian Glos 40

Description of two tadpoles from Madagascar 461

Table 1. Measurements (mm) of tadpoles of Boophis doulioti and B. xerophilus in three stage classes; mean ± SD (range). Staging after Gosner (1960), measurements follow McDiarmid and Altig (1999). BH = body height, BL = body length, BW = body width, ED = eye diameter, IND = internarial distance, IOD = interorbital distance, MTH = maximum tail height, NED = naris – eye distance, ODW = oral disc width, SED = snout – eye distance, SND = snout – naris distance, SSD = snout – spiracle distance, TL = total length, TMH = tail musculature height, TMHM = tail musculature height at midlength of tail, TMW = tail muscle width.

Boophis doulioti Boophis xerophilus Stage 25 Stage 27-31 Stage 37-41 Stage 25 Stage 27-31 Stage 37-41 (n = 5) (n = 9) (n = 6) (n = 4) (n = 8) (n = 5)

TL 10.9 ± 1.823.3 ± 3.939.4 ± 3.612.0 ± 3.022.7 ± 3.037.2 ± 0.8 BL 4.0 ± 0.48.0 ± 1.213.7 ± 0.74.5 ± 0.78.1 ± 1.213.2 ± 0.3 BL/BW 1.3 ± 0.11.6 ± 0.11.6 ± 0.11.4 ± 0.11.5 ± 0.11.5 ± 0.1 BL/BH 1.7 ± 0.21.9 ± 0.21.8 ± 0.11.5 ± 0.11.4 ± 0.11.5 ± 0.1 BW/BH 1.2 ± 0.11.2 ± 0.11.2 ± 0.11.1 ± 0.01.0 ± 0.01.0 ± 0.0 TMH 0.8 ± 0.23.0 ± 0.84.6 ± 0.31.3 ± 0.62.8 ± 0.95.0 ± 0.2 TMW 0.8 ± 0.11.9 ± 0.33.5 ± 0.20.9 ± 0.32.0 ± 0.33.4 ± 0.2 TMHM/MTH 0.2 ± 0.00.7 ± 0.10.7 ± 0.10.5 ± 0.20.5 ± 0.10.6 ± 0.1 ED 0.5 ± 0.11.1 ± 0.22.0 ± 0.20.8 ± 0.11.4 ± 0.22.1 ± 0.1 IOD/IND 2.0 ± 0.12.1 ± 0.12.6 ± 0.32.1 ± 0.12.5 ± 0.22.8 ± 0.4 SND/SED 0.6 ± 0.10.6 ± 0.00.5 ± 0.10.6 ± 0.10.5 ± 0.10.5 ± 0.0 NED 0.9 ± 0.11.9 ± 0.32.7 ± 0.41.0 ± 0.32.3 ± 0.53.3 ± 0.4 ODW 2.1 ± 0.33.2 ± 0.31.3 ± 0.22.3 ± 0.33.4 ± 0.3

Figure 1. Lateral view (A), dorsal view (B), and oral disc (C) of Boophis doulioti tadpole at stage 37 (Gosner, 1960) and of Boophis xerophilus (D, E, F) at stage 37. Scale bares represents 10 mm (A, B, D, E) and 1 mm (C, F). Amphibian communities in Madagascar Julian Glos 41

462 Julian Glos, K. Eduard Linsenmair

posteriorly; submarginal papillae not present. Natural History Upper jaw sheath concave, lower jaw sheath Boophis doulioti uses all pond types (surface V-shaped, both sheath finely serrate and fully area 1-10,000 m2, maximum water depth 10- black pigmented. 100 cm, water holding capacity 3-140 days) for In life, dorsal color is brownish, ventral color reproduction and is found in all major habi- is white. Caudal musculature is entirely and tat types in Kirindy: closed forest, savannah, heavily pigmented at the base of the tail, pig- and riverbed (before the river is running). Its mentation diminishing towards the tail tip, dor- breeding activity is continuous throughout the sal tail fin less intensive pigmented than caudal rainy season and is relatively independent of musculature, ventral tail fin only slightly pig- rainfall (Glos, 2003) when there are water- mented. We found several tadpoles of B. douli- filled pools available. The tadpoles are lentic oti where the rear 1/3 of the tail was completely and benthic. Only on very rare occasions, they dark, in particular at night. This was not a fixed are found in rivers or creeks. Experiments in character; during day respective specimens had aquaria and field observations suggest that they translucent tail fins. mainly feed on decaying leaves. It is not clear whether they eat the plant material itself or the Life History microorganisms living on the leaves. Because they usually rasp the leaves and only occasion- Clutch size was 3445 ± 729 eggs (mean ± SD; ally bite off larger parts of the leaves, they range 1918-4784, n = 13), ovum diameter was supposedly do the latter. They were never ob- 1.22 ± 0.10 mm (0.90-1.62; n = 359 from served to actively prey upon other tadpoles nor 18 amplectant pairs). In the mean, females lost did they feed on dead tadpoles. The following 28% of their body weight during reproduction syntopic tadpole species were found: Boophis (n = 4). Duration of embryonic development xerophilus, Aglyptodactylus laticeps, A. secu- was 24 hours both in the field and under con- rifer, Laliostoma labrosum, Mantella betsileo, trolled conditions. Hatchlings attached to the Ptychadena mascareniensis, Heterixalus luteos- jelly mass for about 12 hours before reaching triatus, H. tricolor, H. carbonei, Dyscophus in- the free-swimming stage. Larval development is sularis, Scaphiophryne calcarata, S. brevis and about one month. Of all individuals measured S. cf.marmorata. in the field, snout-vent length at metamorphosis . ± . n = was 16 6 1 5 mm (13.3-19.7 mm, 31) and Boophis xerophilus Glaw & Vences, 1997 weight at metamorphosis (stage 46 was 386 ± 99 mg, 172-586 mg, n = 31). However, there Description of tadpole was a significant difference in these variables Body is ovoid in lateral view, elliptical in dorsal between individuals from ponds with a water- view (fig. 1d, e). Body length 0.36 ± 0.03 of to- holding duration between 2 and 5 months and tal length, body 1.48 ± 0.10 longer than wide, those from more temporary waters with a high 1.46 ± 0.12 times longer than high, 0.98 ± 0.07 risk of early drying up. The latter were waters times wider than high. Snout rounded in dor- that were drying up within 10 days when there sal and lateral profiles. Eyes medium sized, and was no rain. Individuals from more permanent directed laterally. Nostrils dorsal, apertures di- ponds were larger (17.2 ± 1.3 mm, n = 16) and rected dorsally, nostrils closer to eyes than to heavier at metamorphosis (428±80 mg) than in- snout tip (ratio naris-eye distance/snout-naris dividuals from temporary ponds (16.0±1.6 mm, distance 0.66 ± 0.10). Spiracle sinistral, at mid- 346 ± 107 mg, n = 15) (Mann-Whitney U- body, below midline, lateral wall absent, di- Test; SVL: Z =−2.02, P = 0.04; weight: rected dorsal-posteriorly. Medial vent tube with Z =−2.41, P = 0.01). lateral displacement, short, aperture dextral. Amphibian communities in Madagascar Julian Glos 42

Description of two tadpoles from Madagascar 463

Tail fins convex, higher from the posterior freshly metamorphosed individuals of B. xe- end of vent tube to the middle of the tail, dorsal rophilus are significantly larger (Mann-Whitney fin about as high as ventral fin (ratio dorsal fin U-Test: Z =−3.11, P = 0.002) and heavier height/ventral fin height 1.10±0.02 for stage 25, (Z =−3.25, P = 0.001) than the B. douli- 1.09 ± 0.07 for stage 27-31, and 0.97 ± 0.12 for oti metamorphs from permanent and from tem- stage 37-41). Tail tip rounded. Origin of dorsal porary waters combined. However, there is no fin at the base of tail. significant difference to B. doulioti individuals Oral disc ventral, LTRF 3(2-3)/3(1) or 4(2- from permanent ponds only (SVL at metamor- 4)/3(1) for stage 25, 4(2-4)/3(1) or 5(2-5)/3(1) phosis: Z =−1.15, P = 0.25; weight at meta- for stage 27-31, and 5(2-5)/3(1) for stage 37-41; morphosis: Z =−1.27, P = 0.21). fig. 1f). Marginal papillae conical, with rounded tips, not pigmented; in stage 37, length of mar- Natural History ginal papillae 0.16 mm, density 11.5 papillae/mm. Two rows of marginal papillae inter- Boophis xerophilus differs ecologically from B. rupted anteriorly and posteriorly, submarginal doulioti by its choice of breeding sites. It gener- papillae not present. Upper jaw sheath concave, ally uses larger and more permanent waters. On lower jaw sheath V-shaped, both sheath finely rare occasions, however, B. xerophilus was ob- serrate and fully black pigmented. served to breed also in small ephemeral puddles. In life, dorsal color is dark olive green or dark This species is found primarily in the closed for- blue, sometimes with black spots, in smaller est, only rarely in the riverbed and the savannah. stages with golden spots laterally; ventral color Breeding activity stretches over the whole rainy is white. Caudal musculature is heavily pig- season, showing, however, peaks after heavy mented at the base of the tail, slightly pigmented rainfalls (Glos, 2003). otherwise; dorsal and ventral tail fins are only The tadpoles are lentic and benthic. They slightly pigmented. As in B. doulioti, we found show the same general feeding habits as B. several tadpoles of B. xerophilus where the rear doulioti. Additionally, feeding experiments 1/3 of the tail was completely dark, in particular showed that B. xerophilus tadpoles did feed on at night. This pigmentation vanished during the some living water plant species (Lagerosiphon day. sp., Riccia fluitans) but not on others (Najas sp., Utricularia sp.). Life History In natural breeding ponds, we found tadpoles of Boophis doulioti, Laliostoma labrosum, ± ± Clutch size was 1738 941 eggs (mean Ptychadena mascareniensis, Heterixalus luteos- n = SD; range 999-4235, 9), ovum diame- triatus, H. tricolor, H. carbonei, Dyscophus in- . ± . n = ter was 1 18 0 11 mm (0.92-1.48; 295 sularis and Scaphiophryne calcarata syntopi- from 15 amplectant pairs). Females lost 29% of cally with B. xerophilus. their body weight during reproduction (mean, n = 6). Duration of embryonic development Statistical comparison of morphometric and was 24 hours both in the field and under con- life history variables trolled conditions (see above). Hatchlings attached to the jelly mass for about 12 hours be- A statistical comparison of B. doulioti and B. xe- fore reaching the free-swimming stage. Larval rophilus tadpoles demonstrated significant dif- development was about one month. In the field, ferences (P<0.01) in eight out of 12 ra- snout-vent length at metamorphosis was 17.7 ± tios of morphometric variables considered (ta- 1.3 mm (15.0-21.9 mm, n = 62), and weight ble 3). Egg number was higher in B. douli- at metamorphosis (stage 46; Gosner, 1960) was oti (n = 13; table 2) than in B. xerophilus 484 ± 160 mg (352-1156 mg, n = 62). Hence, (n = 9) (Mann-Whitney U-Test: Z =−3.04, Amphibian communities in Madagascar Julian Glos 43

464 Julian Glos, K. Eduard Linsenmair

Table 2. Life history parameters of tadpoles in their natural habitats and overview of morphological differences of Boophis xerophilus, B. doulioti, and of B. tephraeomystax (data of latter from Blommers-Schlösser, 1979) (mean ± SD; range); SVL = snout-vent length.

Life history B. xerophilus B. doulioti B. tephraeomystax parameter

Clutch size 1738 ± 941 3445 ± 729 no data 999-4235 1918-4784 n = 9 n = 13 Ovum diameter (mm) 1.18 ± 0.11 1.22 ± 0.10 no data 0.92-1.48 0.90-1.62 n = 295 n = 359 Larval duration about 30 days about 30 days no data SVL (mm) at 17.7 ± 1.316.6 ± 1.5 15-20 metamorphosis 15.0-21.9 13.3-19.7 (stage 46) n = 62 n = 31 Metamorphic weight 484 ± 160 386 ± 99 no data (mg) at stage 46 352-1156 172-586 n = 62 n = 31 Labial tooth row 3(2-3)/3(1) to 3(2-3)/3(1) to 4(2-4)/3(1) or formula 5(2-5)/3(1) 6(2-6)/3(1) 5(2-5)/3(1) Dorsal coloration of dark olive green or dark blue, ± uniform brownish yellowish with beige tadpoles sometimes with black spots; or greenish spots, greenish in in smaller stages golden older stages spots laterally Ventral coloration of white white, golden shiny in whitish with small dark spots tadpoles very small tadpoles on its anterior part Feeding habits of detritivorous detritivorous no data tadpoles herbivorous Breeding habitat larger and more permanent waters all pond types sunlit, often temporary primarily in the closed forest, all major habitat types stagnant water, with aquatic only rarely in the riverbed in Kirindy: closed vegetation, fully sun-exposed and the savannah forest, savannah, and small, shallow pools riverbed (before the river is running)

P = 0.001; table 2). There was no signiﬁcant gasy family Mantellidae (Bossuyt and Milin- difference in ovum diameter between B. douli- kovitch, 2000; Vences and Glaw, 2001). Most oti (n = 18; table 2) and B. xerophilus (n = 15) Boophis species occur in the eastern rainfor- (Z = 1.50, P = 0.13). To ensure indepen- est belt. Boophis doulioti and B. xerophilus are dence of data in this analysis, the mean diame- exceptional within the genus, as they inhabit ter of 20 eggs/female was used as sample unit. the drier parts of the west and south of Mada- Egg number and ovum diameter was in the two gascar (Glaw and Vences, 1997; Vences and species not correlated (Spearman; B. doulioti: Glaw, 2002). The two species are closely related r = . P = . n = S 0 42, 0 15, 13, B. xerophilus: (Vences et al., 2002) and are combined within rS =−0.57, P = 0.14, n = 8). the B. tephraeomystax group. Members of this phenetic group are distinguished from other Boophis species by morphological traits and Discussion breeding characters (e.g. shallow, often tempo- The genus Boophis is very diverse in species rary breeding pools, and small eggs with a small number and ecology within the endemic Mala- animal pole; Blommers-Schlösser and Blanc, Amphibian communities in Madagascar Julian Glos 44

Description of two tadpoles from Madagascar 465

Table 3. Statistical comparison (Mann-Whitney U-Test) of fects, or, alternatively, by faster growth caused ratios of morphometric variables between B. doulioti (n = 20) and B. xerophilus (n = 20) tadpoles. Measurements by the choice of a microhabitat that offers bet- follow McDiarmid and Altig (1999). For abbreviations see ter food resources. In contrast to B. xerophilus table 1. that breeds mainly in ponds that exist for three Morphometric Z P to five months, B. doulioti uses a wide vari- variables ety of breeding ponds including more tempo- BL/TL −0.90 0.38 rary waters drying partly already up within a BL/BW −1.53 0.18 few days, depending on their initial depth, sub- BL/BH −4.90 <0.001 BW/BH −5.21 <0.001 soil and the frequency of rainfall. Tadpoles of B. TMH/BH −4.21 <0.001 doulioti are thus often exposed to a high risk of TMW/BW −1.38 0.17 mortality due to desiccation. The lower size of − TMHM/MTH 1.20 0.23 freshly metamorphosed B. doulioti individuals DF/VF −5.08 <0.001 NED/BL −5.41 <0.001 from such temporary, desiccation-prone ponds IOD/IND −2.45 0.01 versus those from longer-lived ponds indicates − < SND/SED 2.64 0.01 that this species possesses developmental plas- NED/SED −5.41 <0.001 ticity allowing for adaptive variation of metamorphic sizes. Theory suggests that tadpoles 1991; Glaw and Vences, 1994; Andreone et al., should — under such circumstances — adjust 2002). both their developmental time and their growth The tadpoles of both species are of the ex- to effectively cope with the variable conditions otrophic, lentic and benthic type. Both species in the pond (Wilbur, 1972; Newman, 1992). are characterized by a prominent muscular base Accordingly, tadpoles like those of B. douli- of the tail. In many morphological characters oti from temporary waters facing a high risk of and proportions, however, the two species differ desiccation may speed up development accept- considerably. The most obvious differences are ing costs on metamorphic size (Newman, 1988, found in the relative position of the eyes that are 1989). Whereas individuals from more perma- situated more lateral, and in color that is more nent ponds may exploit abundant food resources uniform in B. doulioti; this applies particularly in the aquatic stage and utilize the opportunity to younger stages that lack the golden lateral to grow before the terrestrial phase of life begins spots of B. xerophilus. In older stages, the tad- (Wilbur and Collins, 1973). Accordingly, B. xe- poles have a positive imbalance of +3(B. douli- rophilus and B. doulioti from more permanent oti) and +2(B. xerophilus) between the number waters were not different in size at metamor- of tooth rows on the upper and the lower lip. phosis. This is supporting the view of an adap- This character is rather unusual for lentic tad- tive trade off between larval duration and size at poles as it is usually found in stream tadpoles metamorphosis in B. doulioti. (McDiarmid and Altig, 1999). At any rate, there is a high general degree Adults of B. doulioti are considerably larger of sympatry of the two species and syntopy of than those of B. xerophilus (Glos, 2003). How- their larvae. Boophis doulioti is a generalistic ever, this is generally not reflected in their size species that is found throughout all habitat types at metamorphosis. Freshly metamorphosed in- in western and southern Madagascar (Vences dividuals of B. xerophilus are larger and heav- and Glaw, 2002). Boophis xerophilus is known ier than those of B. doulioti. Several expla- from only two very distant localities (Kirindy nations are conceivable for this larger size at Forest, Berenty Park; Glaw and Vences, 1997). metamorphosis. It might be caused by species- In these two areas, both species occur. How- specific (physiological) differences, by a gen- ever, the knowledge on the geographic distrib- erally longer larval period, by competitive ef- ution of B. xerophilus is far from complete, and Amphibian communities in Madagascar Julian Glos 45

466 Julian Glos, K. Eduard Linsenmair

it is likely that more localities between these Glaw, F., Vences, M. (1994): A Fieldguide to the Amphib- sites are inhabited by this species. Within the ians and Reptiles of Madagascar, Second Edition. Köln, M. Vences & F. Glaw Verlags GbR. Kirindy Forest, there is a considerable tempo- Glaw, F., Vences, M. (1997): New species of the Boophis ral and spatial overlap of breeding pond choice tephraeomystax group (Anura: Ranidae: Rhacophori- of both species and hence of the habitat of their nae) from arid Western Madagascar. Copeia 1997: 572- tadpoles (Glos, 2003). Niche theory, however, 578. predicts that the processes of either competitive Glos, J. (2003): The amphibian fauna of the Kirindy dry forest in western Madagascar. Salamandra 39: 75-90. exclusion or niche differentiation act on sym- Gosner, K.L. (1960): A simplified table for staging anuran patric and closely related species with similar embryos and larvae with notes on identification. Her- resource requirements such as B. doulioti and petologica 16: 183-190. McDiarmid, R.W., Altig, R. (1999): Tadpoles: The Biology B. xerophilus (Begon et al., 1990). The mor- of Anuran Larvae. Chicago, The University of Chicago phological differences between the tadpoles of Press. the two species indicate that they indeed oc- Nelson, R., Horning, N. (1993): AVHRR-LAC estimates of cupy different niches in respect to feeding and forest area in Madagascar, 1990. Int. J. Remote Sens. 14: 1463-1475. possibly still other parameters of microhabitat Newman, R.A. (1988): Adaptive plasticity in development choice. of Scaphiopus couchii tadpoles in desert ponds. Evolu- tion 42: 774-783. Newman, R.A. (1989): Developmental plasticity of Scaphiopus couchii tadpoles in an unpredictable Acknowledgements. We thank M.-O. Rödel, K. Daus- environment. Ecology 70: 1775-1787. mann and F. Glaw for their help and/or useful comments Newman, R.A. (1992): Adaptive plasticity in amphibian on our study and on earlier versions of this manuscript. We metamorphosis. Bioscience 12: 671-678. also thank P. Kappeler, L. Razamafimanatsoa, R. Rasoloari- Rödel, M.-O. (2000): Herpetofauna of West Africa. Vol. I. son and the German Primate Center (DPZ, Göttingen, Ger- Amphibians of the West African Savanna. Frankfurt am many) for help and logistic support. Research and export permits were provided by the DEF/Madagascar (Min- Main, Chimaira. istère des Eaux et Forêts, Madagascar; export permit 0250- Sorg, J.-P., Rohner, U. (1996): Climate and tree phenology EAL/MG01/CWN). JG was supported by a PhD scholarship of the dry deciduous forest of the Kirindy forest. In: of the German Academic Exchange Service (DAAD). Ecology and Economy of a Tropical Dry Forest in Madagascar, Primate Report 46-1, p. 57-80. Ganzhorn, J.U., Sorg, J.-P., Eds, Göttingen. Vences, M., Andreone, F., Glaw, F., Kosuch, J., Meyer, A., References Schaefer, H.C., Veith, M. (2002): Exploring the potential of life-history key innovation: Brook breeding in the Andreone, F., Vences, M., Guarino, F.M., Glaw, F., Ran- radiation of the Malagasy treefrog genus Boophis.Mol. drianirina, J.E. (2002): Natural history and larval mor- Ecol. 11: 1453-1463. phology of Boophis occidentalis (Anura: Mantellidae: Vences, M., Glaw, F. (2001): When molecules claim for Boophinae) provide new insights into the phylogeny and taxonomic changes: new proposals on the classification adaptive radiation of endemic Malagasy frogs. J. Zool., of Old World treefrogs. Spixiana 24: 85-91. 257 Lond. : 425-438. Vences, M., Glaw, F. (2002): Molecular phylogeography Begon, M., Harper, J.L., Townsend, C.R. (1990): Ecology: of Boophis tephraeomystax: A test case for east-west Individuals, Populations and Commmunities, Second vicariance in Malagasy anurans (Amphibia, Anura, Edition. Oxford, Blackwell. Mantellidae). Spixiana 25: 79-84. Blommers-Schlösser, R.M.A. (1979): Biosystematics of the Wilbur, H.M. (1972): Competition, predation, and the struc- Malagasy frogs. II. The genus Boophis (Rhacophoridae). ture of the Ambystoma — Rana sylvatica community. Bijdr. Dierk. 49: 261-312. Ecology 75: 1368-1382. Blommers-Schlösser, R.M.A., Blanc, C.P. (1991): Amphibi- Wilbur, H.M., Collins, J.P. (1973): Ecological aspects of ens (première partie). Faune de Madagascar 75: 1-379. Bossuyt, F., Milinkovitch, M.C. (2000): Convergent adap- amphibian metamorphosis. Science 182: 1305-1314. tive radiations in Madagascan and Asian ranid frogs reveal covariation between larval and adult traits. Proc. Natl. Acad. Sci. U.S.A. 97: 6585-6590. Received: May 14, 2004. Accepted: July 7, 2004. Amphibian communities in Madagascar Julian Glos 46

Chapter 5

Living apart together – patterns of tadpole communities in a western Madagascan dry forest

Julian Glos1, Kathrin H. Dausmann2, K. Eduard Linsenmair1

1 Department of Animal Ecology and Tropical Biology, Biocenter, Würzburg University, Am Hubland, 97074 Würzburg, Germany 2 Department of Animal Ecology and Conservation, Biocenter Grindel and Zoological Museum, Martin Luther King-Platz 3, 20146 Hamburg, Germany

ABSTRACT

Whether communities are established in a deterministic or in a stochastic manner depends to a large degree on the spatial scale considered. In this study we use a tadpole community in the dry forest of western Madagascar to show that when within-site habitat diversity is considered, communities may also differ in two community parameters (species composition and species richness) within one geographic scale. Forest ponds and riverbed ponds are two types of breeding habitat that are both used by anurans but that differ generally in their temporal availability, predation pressure, and environmental characteristics. In forest ponds, tadpole communities were very predictable by the physical properties of the ponds and by their vegetation characteristics. In contrast, the riverbed communities were not predictable. We offer two hypotheses to explain this phenomenon. This study clearly demonstrates differing patterns in community organization in two natural habitats within one site, and therefore, highlights the importance of considering local conditions and within-site habitat diversity in community studies. Amphibian communities in Madagascar Julian Glos 47

INTRODUCTION

How species communities are organized is one of the most intriguing questions in community ecology. Factors influencing the distribution and abundance of species, and thus the composition of communities, include moisture and temperature regimes, availability of nutrients, physical structure of the habitat (e.g., Carey and Bryant, 1995; Parris, 2004), as well as processes such as predation, competition, dispersal, disturbance and disease (e.g., McCarthy, 1997; Woodward, 1983). Communities are traditionally classified as either being to a large extent deterministically or mostly stochastically determined (e.g., Huston, 1994). The species composition and other attributes of deterministic communities are predictable by the environmental characteristics of the habitat. Therefore, sites with similar habitats will support similar communities (e.g., Parris and McCarthy, 1999). Stochastically organized communities, in contrast, are not predictable by biotic or abiotic characteristics of the habitat. In these communities, strong stochastic elements determine the recruitment in the component species and unpredictable environmental changes or disturbances result in variations of community composition (e.g., Floren and Linsenmair, 1998). The final composition of stochastic communities is therefore mainly the result of chance. Studies classifying communities as either deterministic or stochastic are influenced not only by the organismic group and taxonomic level, but also by the spatial scale considered (e.g., Wright et al., 1993). Accordingly, patterns observed on a local scale might deviate from those observed at a regional or global scale. In some systems, even more refinement within the spatial scale is needed when the habitat in question is separated into two or more sub-types. Each of these different habitat types can show particular community patterns, e.g., in species composition and species richness. Therefore, when the local situation is not taken into account, the integrated analysis of communities over different habitat types might conceal existing community patterns that occur on a habitat type- based scale. The Kirindy dry forest of western Madagascar represents such a system of subdivided habitat for breeding anurans. Breeding takes place either in ephemeral ponds in the closed forest, or in ponds in the bed of a seasonal river. Both types of breeding habitat are obviously different in their environmental characteristics. However, they are not spatially separated as the riverbed is meandering through the forest and there is no distinct river bank area. Both pond types are easily accessible by all anuran species, thus potentially recruiting anurans from the same species pool. This subdivision of the anuran breeding habitat remains elusive when traditional models of community ecology are applied. In Madagascar, the distribution of anurans on the scale of biogeographic regions is relatively well studied (e.g., Glaw and Vences, 1994; Veith et al., 2004). However, only little is known about the local distribution and habitat choice of single species and about patterns of community composition and species richness. This statement is in particular true for larval anuran communities, although the distribution of tadpoles reflects anuran breeding habitat choice. The selection of breeding ponds, Amphibian communities in Madagascar Julian Glos 48 however, is one of the most important factors influencing local distribution in anurans (e.g., Blaustein et al., 2004). Moreover, the larval stage represents a crucial stage in amphibian life history because larval performance has great effects on individual fitness (e.g., Wilbur and Collins, 1973). Because of their moderate diversity and clear taxonomy, tadpoles are particularly appropriate organisms for community studies. Standardized methods exist for estimating their species richness and abundance (Heyer et al., 1994). Additionally, tadpoles are not moving between ponds and thus constitute well- defined entities.

OBJECTIVES

In this paper, we examine the influence of environmental characteristics on two community parameters, namely species composition and species richness of tadpole communities by studying anuran breeding ponds in the dry forest of western Madagascar. In detail, we ask the following questions: (1) Do the breeding ponds in two habitat types of the dry forest, namely the closed forest and the bed of a seasonal river, differ in their temporal availability, predation pressure, environmental characteristics, and tadpole species composition? (2) Is there a predictable pattern in the composition and species richness of tadpole assemblages in each of the habitat types? Are there differences in the patterns of tadpole assemblages between the two habitat types? Do environmental characteristics and/or geographical distance between the ponds play an important role in determining these patterns? (3) Exemplarily of all analyzed environmental characteristics of the breeding ponds, is there a species-area relationship?

METHODS

Study site The Kirindy Forest is one of the largest remnants of the dry, deciduous forest of western Madagascar (120 km2; Nelson and Horning, 1993). It is located about 50 km northeast of Morondava and 20 km inland (44°39’E, 20°03’S; 18 – 40 m above sea level; Sorg and Rohner, 1996). The rainy season of 3 – 5 months from November-December to February-March is followed by a dry season of 7 – 9 months of virtually no rain. Mean annual precipitation is about 800 mm (range 390 – 1511 mm; Sorg and Rohner, 1996). Precipitation in the rainy season of 1999-2000 was 1265 mm, and 874 mm in 2000- 2001 (Glos, 2003). Within the forested area of the Kirindy concession, the potential breeding sites for amphibians can be subdivided into two different habitat types, namely the closed forest and the bed of Amphibian communities in Madagascar Julian Glos 49 the seasonal Kirindy river before the river is running. All stretches of running and standing waters dry up every year during the dry season except for a few pools in the Kirindy riverbed (Glos, 2003). Amphibians in the Kirindy Forest use only lentic waters as breeding sites. Further information about Kirindy Forest is provided by Sorg and Rohner (1996) and Glos (2003).

Data collection To examine interannual variation, this study was conducted over two entire rainy seasons, 1999-2000 and 2000-2001. Each season was statistically analysed separately. We examined the use of water bodies as breeding sites by recording the presence or absence of anuran tadpoles. In contrast to the analysis of breeding activity of adult frogs, this method is independent of any activity patterns of the adults and considers that not all water bodies with calling male frogs are actually used for reproduction (J. Glos, unpubl. data). Each breeding pond represents one replicate. To include interspecific differences in breeding ecology (Glos, 2003) and to get a full picture of the community, the analyses were done several times per season and pond. In detail, we repeated the analysis about one week after each rainfall that exceeded 30 mm. The period of larval development for all tadpole species in the Kirindy Forest is longer than the sampling intervals (Glos, 2003). Therefore, this method excludes the possibility that some species may have been present but escaped the sampling. Accordingly, species composition was defined as the presence and absence of each species, and species richness as the number of species irrespective of species identity, found in one pond over one entire rainy season, respectively. Species presence was analyzed by standardized dip-netting (Heyer et al., 1994). In each pond, 30 dip net strokes were performed, randomly distributed over the pond. The dip net was triangular shaped with a base of 400 cm2 (30 x 30 x 30 cm). Each dip net stroke was 1 m long, and was touching the ground substrate. All tadpoles were determined in the field or in the field camp to species level, using a stereo microscope (Blommers-Schlösser and Blanc, 1991; Glaw and Vences, 1994; Glos et al., 2005; Glos and Linsenmair, 2004; 2005). Most of the specimens were subsequently released. Only a few voucher specimens were retained for comparison with a reference collection. Nineteen variables were measured on each study pond that characterize its abiotic and biotic properties, for definition and methods see Appendix 1. The habitat variables were recorded parallel to the control checks for species presence. For practical reasons, not all variables could be measured in some ponds, leading to different sample sizes in some analyses. For each sampling date and water body, we calculated a mean for each variable and then used these means to calculate a second order mean for the entire study period. Geographic distances between breeding ponds were calculated using GPS coordinates (Garmin GPS 12) and ArcView GIS software. Maximum distances between breeding ponds are 3.6 km (riverbed) and 5.5 km (forest), respectively. Amphibian communities in Madagascar Julian Glos 50

MRPP – Differences in environmental characteristics and species composition between habitat types Using MRPP (Multi-Response Permutation Procedures; McCune and Grace, 2002), we tested whether the two habitat types (forest ponds, riverbed ponds) differed in their environmental characteristics and in their species composition. MRPP has the advantage (over e.g., Manova) that it does not require distributional assumptions (such as multivariate normality and homogeneity of variances) which are seldom met with ecological community data (Mielke, 1984; Mielke and Berry, 2001). As the distance measure for environmental variables, we chose relative euclidian distance, for species assemblage Sørensen (Bray-Curtis) coefficients were used. Analyses were performed using PC-ORD for Windows version 4.17 MjM Software Design. Additionally, we tested each environmental variable separately for differences between both habitat types using non-parametric statistics (Table 1), to show up in more detail which variables were different between the forest and the riverbed.

Mantel Tests – Predictability of community composition In order to test if species composition of the communities is predictable by habitat variables and/or geographic distance, we compiled the field data into matrices based on (1) species distribution vectors (absence-presence data), (2) environmental characteristic vectors (habitat variables), and (3) geographic distance vectors (true geographic distances). We included 14 habitat variables in the analysis: Desiccation risk, Pond size, Pond depth, Shallowness, Foliage coverage, Water transparency, Surrounding vegetation, Emergent vegetation, Submerse plants, Surface water plants, Leaf litter, Rock substrate, Soil substrate, and Sand substrate (see Appendix 1). Matrices of distance based on these original matrices (3 x 2 + 2 matrices: forest and riverbed habitat, separately for each season) were constructed. For species distribution vector matrices, we used the Sørensen (Bray-Curtis) coefficient (for discussion see Faith et al., 1987). For environmental vector matrices, we chose the relativized Euclidian distance (Legendre and Legendre, 1998). We performed Mantel tests using the program PC- ORD for Windows version 4.17 MjM Software Design. A randomization (Monte Carlo) test with 10,000 permutations was used for the calculation of the p-value.

Linear regression models – Predictability of species richness In order to test whether tadpole species richness is predictable by habitat variables, we constructed habitat models using multiple linear regression. In a first step of the prae hoc analysis, we constructed a correlation matrix for each season and habitat type with the habitat variables, and subsequently eliminated high collinearity within the variables (exclusion of variables in cases of Spearman rho 0.7, see suggestions by Fielding and Haworth, 1997). Additionally, using single linear regressions with species richness on the y-axis and each of the environmental habitat variables on the x-axis, we tested whether non-transformed or log-transformed data would provide the better fit. The results Amphibian communities in Madagascar Julian Glos 51 showed that we should use Pond size as log-transformed variable in future analysis. Subsequently, we assigned each of the remaining 13 habitat variables to one of four variable groups, representing physical pond properties (variables Desiccation risk, Pond size, Pond depth, Isolation), vegetation characteristics (Surrounding vegetation, Emergent vegetation, Submergent vegetation, Floating water plants, Leaf litter), invertebrate predators (Predators), and water chemistry (Oxygen, Conductivity, PH) (Appendix 1). Using inclusive linear regression (SPSS© 12.0), we included the number of tadpole species (Species richness) as the dependent variable, and the environmental variables of the breeding ponds as independent variables. We constructed one model for each variable group, habitat type, and season. For evaluation of the goodness-of-fit of the model, we used R2.

1999 / 2000

2000 / 2001

Nov Dec Jan Feb March April

FIG. 1. Temporal availability of breeding ponds in the Kirindy riverbed (open bars) and closed forest (grey bars). The dashes on the x-axis indicate the beginning of the month, the arrows indicate the time when the river started to flow. Note that in particular the breeding ponds in the closed forest were not necessarily continuously filled but dried out in periods of low rainfall during one season and refilled after rainfall. Reproductive activity was highest at the beginning of the rainy season (Dec – Jan) and was only weak towards the end (March – April). Amphibian communities in Madagascar Julian Glos 52

RESULTS

Differences between the two types of breeding habitat Temporal availability – The breeding ponds in the rocky parts of the Kirindy riverbed differed from those of the closed forest in their temporal availability (Fig. 1). Oviposition was possible earlier in the riverbed ponds, but for a considerable longer period in forest ponds. Nevertheless, temporal overlap existed in the availability of breeding ponds between the forest and the riverbed. The time of this overlap (December to February) was the period when reproductive activity was highest (J. Glos, unpubl. data).

Predation pressure - While the average density of invertebrate predators was not different between the riverbed and the forest ponds (Table 1), larval anurans in riverbed ponds potentially faced a high predation risk by fish. At the time of actual oviposition, only the few permanent riverbed ponds contained fish; and these ponds were among the few waters that were completely avoided for breeding. During the course of the rainy season, however, most waters in the riverbed eventually connected and fish immigrated into the other ponds, creating a very high predation risk for amphibian tadpoles that have not yet completed their larval development. In contrast, all forest ponds were fish- free over their entire duration.

Environmental characteristics – MRPP analyses showed that significant overall differences existed between forest ponds and riverbed ponds in their environmental characteristics. This finding was consistent for both study seasons (1999 / 2000: chance-corrected within-group agreement A = 0.12; 2000 / 2001: A = 0.15; p < 0.0001). When examining each habitat variable separately, breeding ponds of anurans inside the closed forest were larger, had a higher desiccation risk, had a different substrate (mainly soil, in contrast to mainly rock and sand in the riverbed ponds), had less surrounding vegetation, and contained more aquatic vegetation (higher coverage of emergent vegetation, surface water plants, and submerse plants) than those in the riverbed (Table 1).

Differences in species composition Using MRPP analyses, significant differences were found between forest ponds and riverbed ponds in their species compositions. This result was consistent for both study seasons (1999 / 2000: A = 0.11; 2000 / 2001: A = 0.22; p < 0.0001). Fourteen tadpole species were encountered in the Kirindy area (Fig. 2). All 14 species used one or more forest ponds, whereas only eight species used riverbed ponds. Although the total species number was lower, the mean number of species per pond was higher at riverbed sites than at forest sites. This finding was consistent for both study seasons (Table 1). Boophis xerophilus, Ptychadena mascareniensis, and three reed frogs (Heterixalus sp.) were exclusively forest pond breeding species, Amphibian communitiesinMadagascar TABLE 1. Differences between forest and riverbed ponds in respect to environmental variables and species numbers. Mann-Whitney tests for each of the environmental variables, given are mean ± SD, test statistic Z, and probability level P. Season 1999 / 2000: n (Forest) = 42, n (Riverbed) = 17; season 2000 / 2001: n (Forest) = 44, n (Riverbed) = 41.

Season 1999 / 2000 Season 2000 / 2001 Forest ponds Riverbed ponds Z P Forest ponds Riverbed ponds Z P Total number of species 14 7 12 8 Mean number of species / site 3.02 3.29 2.48 2.90 Pond size (m2) 1011.24 ± 3486.37 24.35 ± 21.39 -1.81 0.07 994.04 ± 3405.96 25.44 ± 30.27 -3.61 <0.001 Pond depth (cm) 36.59 ± 35.70 34.94 ± 17.33 -1.23 0.22 38.50 ± 34.65 37.77 ± 26.75 -0.42 0.68 Shallowness (%) 59.38 ± 39.16 42.41 ± 35.68 -1.58 0.11 53.14 ± 38.47 41.66 ± 39.01 -1.10 0.27 Desiccation risk (cat.) 3.49 ± 1.23 2.85 ± 1.14 -2.13 0.03 3.45 ± 1.23 2.96 ± 0.96 -2.26 0.02 Water transparency (cat.) 1.46 ± 0.63 1.67 ± 0.75 -0.61 0.54 1.47 ± 0.55 1.75 ± 0.81 -1.07 0.27 Surrounding vegetation (cat.) 3.98 ± 0.94 2.18 ± 0.68 -5.15 < 0.001 4.28 ± 0.78 2.63 ± 0.55 -7.15 < 0.001 Rock substrate (%) 0.48 ± 2.42 41.57 ± 34.58 -6.79 < 0.001 0.11 ± 0.75 38.48 ± 37.12 -6.50 < 0.001 Soil substrate (%) 73.57 ± 40.03 0.56 ± 1.85 -5.06 < 0.001 82.29 ± 33.97 4.20 ± 13.02 -7.52 < 0.001 Sand substrate (%) 25.95 ± 39.47 57.87 ± 34.64 -3.01 < 0.01 16.59 ± 34.00 57.33 ± 37.74 -5.03 < 0.001 Leaf cover (cat.) 3.10 ± 1.35 3.73 ± 1.10 -1.59 0.11 3.31 ± 1.24 3.72 ± 0.87 -1.39 0.16 Surface water plants (%) 6.32 ± 13.94 0.00 ± 0.00 -2.18 0.03 8.14 ± 18.57 0.00 ± 0.00 -3.57 < 0.001 Submerse plants (cat.) 1.45 ± 1.22 0.22 ± 0.31 -4.57 < 0.001 1.42 ± 1.30 0.09 ± 0.26 -6.34 < 0.001 Vegetation cover (%) 7.75 ± 13.77 0.59 ± 2.43 -3.58 < 0.001 15.42 ± 26.17 0.30 ± 1.60 -5.83 < 0.001 Invertebrate predators (cat.) 2.45 ± 1.30 2.62 ± 1.20 -0.48 0.63 2.60 ± 1.30 2.48 ± 1.39 -0.54 0.59 Julian Glos53 Amphibian communities in Madagascar Julian Glos 54 and Aglyptodactylus laticeps was mainly found in forest ponds. In contrast, Aglyptodactylus securifer was restricted to riverbed ponds, with only one exception. Ptychadena mascareniensis, Scaphiophryne brevis and S. menabensis were rare within Kirindy Forest as they were found in less than 5 % of the study ponds.

A 100

0 s i s s r i s s s p er m t eo si e tus ta i ri su o il n a ra vi ns la o hilu s e a e e u ouli p et ni c br d e ostri al br ab ns a s r e c ne i lus latice i ca e y y B h s lus carbon r hus ct op yn emen p phis xero antella b r stoma l o rixa alus lut ph ryn sco o Bo M na ma te ix o ptoda100 Bo e Heterixalus tricolo r i ph y Lali He te o Dy ad aph Scaphioph hi ch He p Agl Aglyptodactylus securif ty Sc ca P S 80 Poportion of occupied sites (%) 60 Proportion of occupied of sitesProportion (%) 40

0 r ps er m i s o s i us a s s s sile riat erophilu eost x t alus carbone x ri Boophis douliot Mantella bet Boophis Heterixalus tricolo erixalus lu Dyscophus insulari Laliostoma labrosu Hete t Scaphiophryne brevi He AglyptodactylusAglyptodactylus latice securif Scaphiophryne calcarat Ptychadena mascareniensi Scaphiophryne menabensi Mantellidae Ran Hyperoliidae Microhylidae

FIG. 2. Percentage of study ponds containing the species, for (A) forest ponds and (B) riverbed ponds. Open bars represent data from 1999 / 2000, grey bars from 2000 / 2001. Ran = Ranidae.

Predictable pattern of assemblage composition When looking at forest ponds, Mantel tests revealed a significant correlation between species distribution and environmental characteristics, indicating that sites with similar environmental characteristics had similar species assemblages (environmental response signal). Species distribution and geographic distance were not correlated, indicating that sites in close proximity did not have Amphibian communities in Madagascar Julian Glos 55 similar species assemblages (no spatial response signal). This result was found for both study seasons (Table 2a). In contrast, when considering data from riverbed ponds, neither a correlation between species distribution and environmental characteristics nor between species distribution and geographic distance (neither environmental nor spatial response signal) was found. Again, these results were consistent throughout both study seasons. In both habitat types and in both seasons, geographic distance was significantly correlated (albeit only at a low to moderate level) with environmental characteristics, indicating that breeding ponds were more similar in their environmental characteristics to neighbouring ponds than to those that were more distant (Table 2a). One explanation for this difference is that the whole community’s predictability, in contrast to the unpredictability encountered in the riverbed, may occur if all or most of the species exclusively found in the forest ponds are highly predictable in their breeding pond choice. To test whether this hypothesis might be the case, we performed Mantel tests with a reduced data set. In this analysis, species were only included when they had been present in > 5% of the study ponds in each of both habitat types. This scenario occurred for six species (A. laticeps, L. labrosum, B. doulioti, M. betsileo, S. calcarata, D. insularis). We used the same matrices of environmental characteristics and geographic distance as in the previous analysis. We found that the results did generally not differ to those of the full community (Table 2b).

TABLE 2. Mantel tests on comparison of three different distance matrices. (a) complete community (14 species), (b) reduced community (6 species). Randomization (Monte Carlo) test (10,000 permutations), r = standardized Mantel statistic, distance measure for environmental variables = relative euclidian distance, for species assemblage = Sørensen (Bray-Curtis). Season 1999 / 2000: n (Forest) = 42, n (Riverbed) = 17; season 2000 / 2001: n (Forest) = 44, n (Riverbed) = 41.

(a) Season 1999 / 2000 Season 2000 / 2001 Forest ponds Riverbed ponds Forest ponds Riverbed ponds r p r p r p r p Species/ environmental 0.309 0.0001 0.015 0.406 0.347 0.0001 0.046 0.168 Species/ geographic 0.015 0.397 0.0430 0.319 0.001 0.460 0.034 0.279 Geographic/ environmental 0.163 0.0002 0.179 0.039 0.144 0.010 0.214 0.0003

(b) Season 1999 / 2000 Season 2000 / 2001 Forest ponds Riverbed ponds Forest ponds Riverbed ponds r p r p r p r p Species/ environmental 0.266 0.0002 0.067 0.260 0.196 0.002 0.082 0.073 Species/ geographic -0.042 0.330 -0.042 0.345 -0.023 0.459 0.069 0.164 Geographic/ environmental 0.097 0.043 0.119 0.115 0.129 0.031 0.197 0.001 Amphibian communities in Madagascar Julian Glos 56

Predictable pattern of species richness In forest ponds, the species richness was moderately to highly (R2=0.45 to 0.62; Table 3) and significantly predictable by the ponds’ physical properties and vegetation. Species richness was also significantly related, although less strongly (R2 = 0.22 to 0.29), to the density of invertebrate predators and to water chemistry variables. These results were generally similar for season 1999-2000 and 2000- 2001 (Table 3). A different picture was found in the riverbed ponds. The species richness was to a much lower degree predictable by environmental variables. In neither of the two seasons was species richness significantly related to physical properties, vegetation or water chemistry. We found a significant relationship of species richness and invertebrate predator density only in season 1999-2000. This relationship was not found in season 2000-2001.

TABLE 3. Multiple linear regression analysis of breeding habitat use (dependent variable: species richness) with four groups of independent variables (inclusive models). Shown are regression coefficient (R2), F – statistic, degrees of freedom (DF), and probability level (P). For definition of the independent variables see text and Appendix 1.

Forest ponds Riverbed ponds Season 1999 / 2000 R2 F DF P R2 F DF P Physical properties 0.62 15.68 43 < 0.001 0.18 0.65 16 0.64 Vegetation 0.56 10.90 47 < 0.001 0.12 0.51 19 0.73 Predators 0.23 11.23 38 0.002 0.54 16.43 15 0.001 Water chemistry 0.22 3.00 34 < 0.05 0.17 1.11 19 0.37

Season 2000 / 2001 Physical properties 0.55 22.87 41 < 0.001 0.12 0.91 31 0.47 Vegetation 0.45 6.25 43 < 0.001 0.06 0.63 40 0.64 Predators 0.29 14.84 38 < 0.001 0.05 1.64 34 0.21 Water chemistry 0.22 2.82 33 0.06 0.15 1.94 35 0.14

Pond area – species richness relationship To exemplify different patterns of community structure in the two habitats, we conducted regressions of species richness (dependent variable) on Pond size (independent variable) separately from the multi-variable models presented above, one for each habitat type and season. In the forest ponds, there was a strong positive semi-logarithmic relationship between pond area and the number of anuran species in both years (1999-2000: F = 47.12, p < 0.001, R2 = 0.48, df = 50; 2000-2001: F = 41.64, p < 0.001, R2 = 0.50, df = 42) (Figure 3). In the riverbed ponds, no such relationship was found (1999-2000: F = 0.41, p = 0.53, R2 = 0.03, df = 15; 2000-2001: F = 2.27, p = 0.14, R2 = 0.05, df = 39). Amphibian communities in Madagascar Julian Glos 57

As all ponds in the riverbed are below 100 m2, the differing results between forest and riverbed ponds might be an effect of the limited maximum pond size in the riverbed rather than a true difference in the species richness – patch size pattern (Fig. 3). This phenomenon is called the ‘small island effect’ (Lomolino and Weiser, 2001; Lomolino, 2000) and is known from several taxa. Beyond some minimum area, species richness increases as described by traditional models (e.g., semi-log or log-log models). On smaller islands, however, richness may vary independently of island area. To exclude the possibility that the significant species – area relation in forest ponds was mainly caused by data points of larger (> 100 m2) ponds, we conducted additional regressions considering only ponds below 100 m2. In this analysis, therefore, the study ponds of both habitat types were appropriately comparable in respect to their size. Results from both seasons showed a weaker but still significant relation of pond area on species richness in the forest (1999-2000: F = 6.87, p = 0.01, R2 = 0.19, df = 31; 2000-2001: F = 4.38, p < 0.05, R2 = 0.15, df = 25). This result confirmed our finding that there is no small island effect over the pond size range considered, and accordingly that there is a true difference in the species – area pattern between forest and riverbed ponds.

1999 - 2000 2000 - 2001

A C 9 9 8 8 7 7 6 6 5 5 4 4 3 3 2 2 Species richness Species richness 1 1 0 0 0 1 10 100 1000 10000 100000 0 1 10 100 1000 10000 100000 Pond area (m2) Pond area (m2)

7 B 7 D 6 6

5 5

4 4

3 3

2 2 Species richness 1 Species richness 1

0 0 0 1 10 100 1000 0 1 10 100 1000 Pond area (m2) Pond area (m2)

FIG. 3. Species – area relationships of forest ponds (A, C) and riverbed ponds (B, D). The number of tadpole species increased significantly with increasing pond area in the forest ponds, but not in the riverbed ponds. Species richness gives the number of tadpole species within one season. The dashed line in A and C shows the threshold for the analysis of the small island effect, see text. Amphibian communities in Madagascar Julian Glos 58

DISCUSSION

Community patterns depend on the habitat type Whether tadpole communities are established in a deterministic or in a stochastic manner depends on the respective habitat type in the Kirindy Forest. This study shows that tadpole community patterns are not only dependent on the geographical scale (Fulton and Harcombe, 2002; Ricklefs, 1987), which is the usual level of resolution (e.g., Wiens, 1984), but may also differ within one scale when within-site habitat diversity is considered. In forest ponds, the communities were very predictable on the species level by the physical properties of the ponds that served as breeding sites and by their vegetation characteristics. Additionally, simple geographical distance between the ponds had no influence on community composition. Therefore, breeding sites with similar characteristics tend to have similar tadpole assemblages independent of their geographic position. A different situation was found in the riverbed habitat: The communities in the riverbed ponds were neither predictable by environmental characteristics nor by their geographical distance. The findings that different habitats show different community patterns are in accordance with previous studies investigating communities in habitats with different states of anthropogenic disturbance, corresponding to differing habitat types within one area. For example, community composition of ants in Borneo (Floren et al., 2001) and leaf litter anurans in Western Africa (Rödel and Hillers, unpubl.; Ernst and Rödel, 2005) in primary tropical rainforest habitats were not predictable, while secondary, anthropogenically disturbed habitats showed a deterministic pattern. However, these studies differ from ours in two important ways: (1) they investigated community patterns in spatially separated habitats, as compared to our study, which investigated different habitat types within one area, and (2) they compared natural habitats with disturbed habitats, compared to two natural habitat types in our study. To our knowledge, Rödel’s (1998) study on tadpole assemblages in a Western African savannah is the only other study comparing amphibian community patterns of two habitat types within one spatial scale. He separately analyzed tadpole communities in large, semi- permanent ponds and those communities in smaller, temporary water bodies. The patterns of community structures, based on microhabitat preferences of the tadpoles, also differed between these habitat types with predictable communities in the former and non-predictable communities in the latter. The differences in community patterns between the two habitat types in the Kirindy Forest are not restricted to species composition. The analysis of species richness generated similar patterns. Species richness of forest ponds was highly predictable by physical and vegetation characteristics, as well as by predator density and water chemistry. In the riverbed ponds, in contrast, species richness was usually not predictable by these parameters. This finding is exemplified by the species-area relationships. According to this tenet, species richness should increase with the area of an “island” and is often referred to as the closest thing to a rule in ecology (Schoener, 1976; Lomolino, 2000). The Amphibian communities in Madagascar Julian Glos 59 positive effect of patch area on the number of frog species is well documented (e.g., Tocher et al., 1997) but has so far only sporadically been reported for tadpoles (Peltzer and Lajmanovich, 2004). In the riverbed ponds of the Kirindy Forest, this universal rule seems to have been abandoned, which substantiates the fact that profound differences existed in assemblage patterns between forest and riverbed ponds. This study clearly demonstrates differing patterns in community organization in two natural habitats within one site, and therefore, highlights the importance of considering local conditions and within-site habitat diversity in community studies.

Why do the species and the community patterns differ between the two habitats? The anuran species breeding in the riverbed constitute a subset of the species breeding in the forest ponds. This fact means that all (but one) species breeding in the riverbed ponds also breed in the forest ponds but several species breed exclusively in the forest ponds. One possible explanation for this phenomenon might be that differences in the spatial position of the ponds lead to different species pools in the forest and the riverbed from which breeding anurans are recruited. However, this explanation does not seem plausible for two reasons. First, the riverbed is surrounded by forest with practically no river bank and maximal distances of a few meters between riverbed ponds and forest habitat. Therefore, both habitats are easily accessible even for anurans with low mobility. Second, no significant correlation was found between species presence and geographic distance in both habitats, showing that geographic distance between ponds is not important when tadpole communities are formed. Because of the differences in environmental characteristics, predation pressure, and temporal availability, a more likely explanation is that a site filter is effective in the riverbed habitat type. Hence the conditions at the riverbed site exclude the successful breeding of certain species, making the site suitable for only some species that are able to cope with the prevailing conditions. For example, even though the desiccation risk in the riverbed ponds is lower, the development nevertheless needs to be rather quick in order to avoid the invasion of fish during the tadpole stage when the river starts to flow. Such a site filter was recently assumed to be responsible for differences in community patterns of leaf-litter anurans in western Africa (Ernst and Rödel, 2005). In contrast to our study, however, Ernst and Rödel found predictable communities in the habitat with a reduced subset community (secondary forest) and stochastic communities in the full community (primary forest) (for ants see Floren et al., 2001). They argued among other things that the subset species in the secondary forests are predictable because highly adapted species are able to breed there and because the occurrence of these species is heavily dependent on habitat characteristics. The opposite is the case in our system as the subset species also breeding in the riverbed ponds are actually generalists (Glos, 2003), utilizing all ponds available in this habitat type. Because the riverbed ponds have less species breeding, this subset may consist of those species whose occurrence is not predictable. This stochastic pattern is concealed in the forest habitat by the deterministically distributed forest species. When comparing Amphibian communities in Madagascar Julian Glos 60 only the subset of species found in riverbed ponds with the community in the forest habitat, however, the same differing patterns were found. Hence the same species, or possibly even the same individuals, show a different pattern of pond colonization, depending on the type of habitat. This surprising result emphasizes the importance of carefully choosing the scale and habitat types considered for studies on community ecology. We have taken two hypotheses into consideration to explain this phenomenon. First, differing spans of the habitat characteristics between the two habitat types could be responsible for the different patterns of species composition. The riverbed ponds are generally less heterogeneous and variable, especially concerning size, depth, ground substrate and vegetation parameters, as shown by smaller standard deviations in the data. Explicitly, they are all mainly of medium size and depth, most have a rocky substrate and only little vegetation. They are thus much more similar to each other the forest ponds are. Even though the species show a deterministic pattern of colonization when given all options (as is the case in the forest), within the restricted choice of habitat parameters of the riverbed ponds, the breeding adults possibly do not discriminate between the ponds. Therefore, their occurrence is not predictable. The second hypothesis concerns the temporal aspect of the pond availability. Even though a considerable temporal overlap of both breeding habitat types existed, the riverbed ponds were the first available after the very first rainfalls of each rainy season. Maybe the species breeding in these ponds use all available breeding sites rather indiscriminately in a stochastic manner at the beginning of the season and become later choosier, at a time when the forest habitat is also available. At this point, they show a deterministic pattern.

Conclusions Our study establishes a new approach for future studies on community patterns. It expands results from previous studies investigating community patterns in adjacent sites, e.g. of differing anthropogenic disturbance, by demonstrating that pattern differences are also found between different natural habitats. When community patterns which are integrated over different habitat types are analysed (e.g., Amazonian tadpoles in terra firme and streamside pools; Gascon, 1991), existing community patterns that occur on a habitat type-based scale might be concealed. The factors responsible for the existence of these differing patterns so far remain elusive. Furthermore, additional investigation is needed to determine what governs the direction of the pattern differences in different ecosystems (stochastic vs. deterministic). Regardless of the actual mechanisms that lead to these variations, our results make clear that within site-habitat diversity should always be considered when investigating patterns in communities. Amphibian communities in Madagascar Julian Glos 61

LITERATURE CITED

BLAUSTEIN, L., M. KIFLAWI, A. EITAM, M. MANGEL, and J. E. COHEN. 2004. Oviposition habitat selection in response to risk of predation in temporary pools: mode of detection and consistency across experimental venue. Oecologia. 138:300-305.

BLOMMERS-SCHLÖSSER, R. M. A., and C. P. BLANC. 1991. Amphibiens (première partie). Faune de Madagascar. 75:1-379.

CAREY, C., and C. J. BRYANT. 1995. Possible interactions among environmental toxicants, amphibian development, and decline of amphibian populations. Environmental Health Perspectives. 103:13-17.

ERNST, R., and M.-O. RÖDEL. 2005. Anthropogenically induced changes of predictability in tropical anuran assemblages. Ecology. 86:3111-3118.

FAITH, D., P. R. MINCHIN, and L. BELBIN. 1987. Compositional dissimilarity as a robust measure of ecological distance. Vegetatio. 69:57-68.

FIELDING, A. H., and P. H. HAWORTH. 1997. A review of methods for the assessment of predicting errors in conservation presence/absence models. Environmental Conservation. 24:38-49.

FLOREN, A., A. FREKING, M. BIEHL, and K. E. LINSENMAIR. 2001. Anthropogenic disturbance changes the structure of arboreal tropical ant communities. Ecography. 24:547-554.

FLOREN, A., and K. E. LINSENMAIR. 1998. Diversity and recolonization dynamics of selected arthropod groups on different tree species in a lowland rainforest in Sabah, Malaysia with special reference to Formicidae, p. 344-382. In: Canopy Arthropods. N. E. Stork, J. A. Adis, and R. K. Didham (eds.). Chapman Hall, London.

FULTON, M. R., and P. A. HARCOMBE. 2002. Fine-scale predictability of forest community dynamics. Ecology. 83:1204-1208.

GASCON, C. 1991. Population- and community level analyses of species occurrences of central Amazonian tadpoles. Ecology. 72:1731-1746.

GLAW, F., and M. VENCES. 1994. A Fieldguide to the Amphibians and Reptiles of Madagascar. M. Vences & F. Glaw Verlags GbR, Köln.

GLOS, J. 2003. The amphibian fauna of the Kirindy dry forest in western Madagascar. Salamandra. 39:75-90.

GLOS, J., F. GLAW, and M. VENCES. 2005. A new species of Scaphiophryne from Western Madagascar. Copeia. 2005:252-261.

GLOS, J., and K. E. LINSENMAIR. 2004. Description of the tadpoles of Aglyptodactylus laticeps and A. securifer from Western Madagascar, with notes on life history and ecology. Journal of Herpetology. 38:131-136. —. 2005. Description of the tadpoles of Boophis doulioti and B. xerophilus from Western Madagascar with notes on life history and breeding ecology. Amphibia-Reptilia. 26:459-466. Amphibian communities in Madagascar Julian Glos 62

HEYER, W. R., M. A. DONNELLY, R. W. MCDIARMID, L.-A. HAYEK, and M. S. FOSTER. 1994. Measuring and Monitoring Biological Diversity: Standard Methods for Amphibians. Smithonian Institution Press, Washington.

HUSTON, M. A. 1994. Biological Diversity The coexistence of species on changing landscapes. Cambridge University Press, Cambridge.

LEGENDRE, P., and L. LEGENDRE. 1998. Numerical ecology. Elsevier Science, Amsterdam.

LOMOLINO, M. V. 2000. Ecology's most general, yet protean pattern: the species-area relationship. Journal of Biogeography. 27:17-26.

LOMOLINO, M. V., and M. D. WEISER. 2001. Towards a more general species-area relationship: diversity on all islands, great and small. Journal of Biogeography. 28.

MCCARTHY, M. A. 1997. Competition and dispersal from multiple nests. Ecology. 78:873-883.

MCCUNE, B., and J. B. GRACE. 2002. Analysis of Ecological Communities. MjM Software Design, Gleneden Beach, Oregon.

MIELKE, P. W., JR. 1984. Meteorological applications of permutation techniques based on distance functions, p. 813-830. In: Handbook of statistics. Vol. 4. P. R. Krishnaiah and P. K. Sen (eds.). Elsevier Science Publishers.

MIELKE, P. W., JR., and K. J. BERRY. 2001. Permutation methods: a distance function approach. Springer Series in Statistics. Springer.

NELSON, R., and N. HORNING. 1993. AVHRR-LAC estimates of forest area in Madagascar, 1990. International Journal of Remote Sensing. 14:1463-1475.

PARRIS, K. M. 2004. Environmental and spatial variables influence the composition of frog assemblages in sub-tropical eastern Australia. Ecography. 27:392-400.

PARRIS, K. M., and M. MCCARTHY. 1999. What influences the structure of frog assemblages at forest streams? Australian Journal of Ecology. 24:495-502.

PELTZER, P. M., and R. C. LAJMANOVICH. 2004. Anuran tadpole assemblages in riparian areas of the Paraná River, Argentina. Biodiversity and Conservation. 13:1833-1842.

RICKLEFS, R. E. 1987. Community diversity: relative roles of local and regional processes. Science. 235:167-171.

RÖDEL, M.-O. 1998. Kaulquappengesellschaften ephemerer Savannengewässer in Westafrika. Edition Chimaira, Frankfurt a.M.

SCHOENER, T. W. 1976. The species-area relationship of terrestrial isopods (Isopoda, Onisciea) from the Aegean archipelago (Greece): a comparative study. Global Ecology and Biogeography Letters. 5:149-157.

SORG, J.-P., and U. ROHNER. 1996. Climate and tree phenology of the dry deciduous forest of the Kirindy forest, p. 57-80. In: Ecology and Economy of a Tropical Dry Forest in Madagascar, Primate Report 46-1. J. U. Ganzhorn and J.-P. Sorg (eds.). Amphibian communities in Madagascar Julian Glos 63

TOCHER, M. D., C. GASCON, and E. AL. 1997. Fragmentation effects on a central Amazonian frog community: a ten-year study, p. 124-137. In: Tropical Forest Remnants. W. F. Laurance and J. Bierregaard, R.O. (eds.). The University of Chicago Press, Chicago & London.

VEITH, M., S. LÖTTERS, F. ANDREONE, and M.-O. RÖDEL. 2004. Measuring and monitoring amphibian diversity in tropical forests. II. Estimating species richness from standardised transect census. Ecotropica. 10:85-99.

WIENS, J. A. 1984. On understanding a non-equilibrium world: myth and reality in community patterns and processes, p. 439-457. In: Ecological Communities. Conceptual Issues and the Evidence. D. R. Strong, D. Simberloff, L. G. Abele, and A. B. Thistle (eds.), Princeton, New Jersey.

WILBUR, H. M., and J. P. COLLINS. 1973. Ecological aspects of amphibian metamorphosis. Science. 182:1305-1314.

WOODWARD, B. D. 1983. Predator-prey interactions and breeding-pond use of temporary-pond species in a desert anuran community. Ecology. 64:1549-1555.

WRIGHT, D. H., D. J. CURRIE, and B. A. MAURER. 1993. Energy supply and patterns of species richness on local and regional scales, p. 416 pp. In: Species Diversity in Ecological Communities. Historical and Geographical Perspectives. R. E. Ricklefs and D. Schluter (eds.). The University of Chicago Press, Chicago & London. Amphibian communitiesinMadagascar APPENDIX 1. Independent variables in the initial linear regression model.

Variable name Variable definition Unit Methods

Pond size Absolute surface area of pond when completely m2 Calculated from pond length, width, and shape water-filled Pond depth Maximum pond depth when pond is completely cm Measured at the deepest point in the pond water-filled Shallowness Relative surface area of the pond with < 3 cm water 0 – 100 % Estimated depth Desiccation risk Time until pond dries up after complete water- 5 categories, from I (low) to V Measured after heavy rain that left all study filling (high); categories were defined as ponds completely water-filled and a consecutive days until desiccation of the pond dry phase without rain I: > 150 days, II: 55 – 150 days, III: 20 – 54 days, IV: 5 – 19 days, V: < 5 days Isolation Number of water bodies within 50 m distance Number; note that high values Calculation from GPS coordinates and ArcView indicate low isolation and vice GIS software versa Water Transparency of the pond water influenced by 5 categories The visibility of a metal stick was estimated in a transparency dissolved and suspensed matters I (low) to V (high) depth of 15 cm at randomly chosen points within the pond and averaged Surrounding Density of surrounding vegetation (trees, bushes) > 5 categories Estimated vegetation 1m height I (low) to V (high) Rock substrate Relative area of the pond bottom composed of rocky 0 – 100 % Estimated substrate Soil substrate Relative area of the pond bottom composed of soil 0 – 100 % Estimated substrate Sand substrate Relative area of the pond bottom composed of sandy 0 – 100 % Estimated substrate Shading Relative area of pond shaded by foliage of shrubs 0 – 100 % Estimated as relative area shaded at 12:00 a.m. and trees Julian Glos64 Amphibian communitiesinMadagascar APPENDIX 1. continued

Leaf litter Relative area of pond bottom covered with leaf litter 6 categories Measured at plots of 0.25 m2 within the pond and I: 0 %, II: 1 – 20 %, III: 21 – 40 averaged %, IV: 41 – 60 %, V: 61 – 80 %, VI: 81 – 100 % Surface water Relative area of pond surface covered with floating 0 – 100 % Estimated plants water plants (mainly Nymphea lotus, Salvinia sp.) Submerged Relative volume of pond filled with submersed 6 categories; see Leaf litter Measured at plots of 0.25 m2 within the pond and plants water plants (mainly Lagarosiphon madagascar- averaged iensis, Najas sp.) Emergent Relative area of pond surface covered with 0 – 100 % Estimated vegetation emergent water plants (grasses, herbs) Invertebrate Density of invertebrate predators, including: 5 categories Measured at 12 plots of 0.25 m2 within the pond Predators Adult water beetles, water bugs (Belostomatidae), I (low) to V (high) using dip-netting and the box-method (Heyer et larval dragonflies (Anisoptera) > 0.5 cm al., 1994) and averaged; determined to higher Larval water beetles, water scorpions (Nepidae) > 1 taxonomic levels in the field and subsequently cm released Larval damselflies (Zygoptera) > 2 cm Oxygen Concentration of oxygen dissolved in water mg / l Measured at a depth of 5 cm and 30 cm from the concentration pond edge with a DO-5509 Oxygen analyser (Conrad Electronics) PH Concentration of H+ dissolved in water pH Measured at a depth of 5 cm and 30 cm from the pond edge with a pHep® pH analyzer (Hanna Instruments) Conductivity Concentration of ions dissolved in water μS Measured at a depth of 5 cm and 30 cm from the pond edge with a WA 300 conductivity analyser (Conrad Electronics) Julian Glos65 Amphibian communities in Madagascar Julian Glos 66

Chapter 6

Modeling the habitat use of an endangered dry-forest frog from Western Madagascar

Julian Glos1, Kathrin H. Dausmann2 & K. Eduard Linsenmair1

ABSTRACT

A crucial factor for the successful reproduction and thus conservation of an amphibian species is the availability of suitable waters as breeding sites. In this chapter, we examine the use of breeding sites of an endangered, local endemic frog of Western Madagascar, Aglyptodactylus laticeps, over a three year period. Logistic regression was used to model the relationship between the species’ breeding habitat use and environmental variables. This model was aimed to be predictive, rather than explanatory, and only environmental variables were included that are assessable in a time and cost effective manner, and that can therefore be used as an easy-to-use management tool in applied conservation. On the local scale of the Kirindy concession, A. laticeps is restricted to forest with a relatively low degree of disturbance and closed canopy cover. The model identified three environmental variables that suffice to satisfactorily predict the use of respective breeding sites, namely leaf litter, vegetation coverage and surface water plants. Based on these results, we present recommendations for the conservation management of this frog. Furthermore, the presence or absence of this species within its natural range indicates the relative degree of environmental integrity of its habitat, and we therefore consider this species as a suitable indicator species of temporary aquatic habitats within the dry forest that are characterized by a low water permanency and high leaf litter coverage. This study demonstrates that models constructed from basic ecological knowledge of relevant species may serve as valuable management tools in applied conservation. Amphibian communities in Madagascar Julian Glos 67

INTRODUCTION

Madagascar is one of the world’s hot spots for biodiversity (Myers et al., 2000). After its separation from mainland Africa and India, Madagascar remained evolutionarily highly isolated for about the last 80 Mio years leading to extraordinary degrees of endemism in many taxa (Goodman and Benstead, 2003). Within this exceptional playground of evolution, the dry deciduous forests along the west coast stick out as local centers of biodiversity. Moreover, the animals and plants inhabiting this ecosystem have developed remarkable adaptations to cope with the marked seasonality of their habitat. Nonetheless, the dry deciduous forest ranks among the most endangered ecosystems, within Madagascar as well as worldwide (Janzen, 1988; Smith, 1997). Forest clearance by illegal logging activities, slash-and-burn agriculture, and fragmentation, have led to virtually completely isolated forest blocks (Smith, 1997). The largest remaining tracts of dry deciduous forests in Madagascar (mainly Kirindy Forest and Ambadira Forest; Sorg et al., 2003) are situated in the Central Menabe region, which lies in the center of Western Madagascar and includes the area surrounding Morondava from the sea to the foothills of the central highlands. The Menabe region has been the focus of national and international development and research programs for the past few decades. Its biodiversity was recognized as being very important (Randrianandianina et al., 2003), and the region is considered as a priority area for future protection (FANAMBY, 2003). Even optimistic prognoses do not assume that all of the remaining forest with its extant biodiversity can be conserved in its present status. The rural population of Madagascar largely depends on the forest and its products for everyday life (Favre, 1996). Thus, human pressure from the surrounding villages will lead to further degradation of at least some parts of the forest. In a pragmatic view, research and conservation efforts must therefore focus on selected sites of extraordinary conservation interest. These sites might be chosen either according to high general levels of biodiversity, or based on the distribution and ecology of species of particular interest (“flagship species”). During the last decade, conservation research and its application on flagship species was established in the Menabe region. The main focus of these activities centered on vertebrate species with extremely restricted ranges within the Menabe region (e.g., giant jumping rat, Hypogeomys antimena; Sommer and Hommen, 2000; Sommer et al., 2002, flat-tailed tortoise, Pyxis planicauda; Gibson and Buley, 2004; Bloxam et al., 1996, pygmy mouse lemur, Microcebus berthae). Amphibians were so far not included as target species into the conservational efforts. This is unfortunate for a number of reasons. Amphibians are locally and globally highly threatened, they are seen as good bioindicators of ecological integrity (e.g., Welsh and Ollivier, 1998), and they are important for ecosystem functions (e.g., as prey and predators in the food chains). However, general habitat requirements of amphibians are very different to those of larger vertebrates and therefore the conservation of amphibians is not adequately accommodated for when their special needs are not taken into consideration by the design and management of protected areas. Especially limnic systems, Amphibian communities in Madagascar Julian Glos 68 which are essential habitats in amphibian life histories (e.g., as oviposition sites and habitats for larval stages), are regularly not adequately taken into consideration.

STUDY OBJECTIVES

For the above reasons, we strongly recommend to include amphibians into conservation activities in Central Menabe. To realize this aim, it is necessary to increase the knowledge on ecological requirements of amphibians in this area. A crucial factor for the successful reproduction and thus conservation of an amphibian species is the availability of suitable waters as breeding sites. Therefore, this study examines the use of breeding sites of an endangered, local endemic Madagascan frog, Aglyptodactylus laticeps, over a three year period. Logistic regression was used to model the relationship between the species’ breeding habitat use and environmental variables at a local scale in Kirindy Forest, one major forest block in Central Menabe. In our model, we included environmental variables that are assessable in a time and cost effective manner, and that can therefore be used as an easy-to-use management tool in applied conservation. Our model was aimed to be predictive, rather than explanatory.

MATERIAL AND METHODS

Study site The Kirindy Forest is one of the largest remnants of the dry, deciduous forest of Western Madagascar (120 km2; Nelson and Horning, 1993). It is located about 50 km northeast of Morondava and 20 km inland (44°39’E, 20°03’S; 18 – 40 m above sea level; Sorg and Rohner, 1996). The forest is managed by the “Centre de Formation Professionelle Forestière” (CFPF), and is so far not included in the official network of protected areas (Randrianandianina et al., 2003). Until about 1994, Kirindy/CFPF was the focus of silvicultural research projects and agroforestry experiments including sustainable timber exploitation with subsequent reforestation (Sorg et al., 2003). In surrounding areas with cleared forest, the vegetation consists of secondary-forest formations, scrub, and savanna. Human density is low and concentrated in small villages spread out over the area (Sorg et al., 2003). The rainy season of 3 – 5 months from November-December to February-March is followed by a dry season of 7 – 9 months with virtually no rain. Mean annual precipitation is about 800 mm (range 390 – 1511 mm; Sorg and Rohner, 1996). The potential breeding sites for amphibians in this area can be found in three different habitat types, namely the forest (forest ponds), the bed of a river that is meandering through the forest (before the river is running; riverbed ponds), and the surrounding savanna (savanna ponds). All stretches of running and standing waters dry up every year except for Amphibian communities in Madagascar Julian Glos 69 very few pools in the riverbed. For further information on amphibian breeding sites in Kirindy see chapter 5 and Glos (2003).

Study species Agylptodactylus laticeps (Laliostominae: Mantellidae: Anura) was described first from the study site (Glaw et al., 1998). From current knowledge it is endemic to the Central Menabe region, from Kirindy Forest in the South to the Tsiribinha river in the north (Glos and Volahy, 2004). It is classified as an endangered species (EN) on the basis of its restricted extent of occurrence and the observed shrinking of its habitat area (criteria for EN B1a and B1b (iii); IUCN, 2001) (Andreone et al., 2005). However, within its habitat it is locally abundant (Glos, 2003). This frog shows distinct adaptations to the dry forest habitat. It is active only during the rainy season, while spending the cool dry season presumably hibernating under ground. Breeding starts about two to four weeks after the beginning of the rainy season and extends over its whole duration. A. laticeps is an explosive breeding species that usually reproduces only after heavy rainfalls that exceed 30 mm of precipitation (Glos, 2003). Larval development is very short, with a minimum of ten days (Glos and Linsenmair, 2004). The tadpoles have been shown to exhibit adaptive plasticity in metamorphic traits (J. Glos, unpubl. data).

Logistic regression model Statistics In order to extract the key habitat factors that predict the choice of breeding waters by A. laticeps we designed a habitat model using multiple logistic regressions (Hosmer and Lemeshow, 1989). Logistic response functions are appropriate when there is a sigmoid relationship between a species’ probability of occurrence and the independent variables (e.g., in contrast to hump-shaped response curves; McCune, 2004). In A. laticeps, we expected the response variable (probability of occurrence) to fit this assumption over the range of the relevant independent variables. Using a backward stepwise logistic regression (LR – method; SPSS© 12.0) and likelihood ratio statistics, we included binary presence / absence data of A. laticeps tadpoles as the dependent variable, and environmental characteristics of the breeding ponds as independent variables. Each breeding pond represents one replicate. We used a significance level of 0.10 for exclusion of variables as this is considered to provide better discrimination performance than less conservative levels (Adler and Wilson, 1985). For model calibration and evaluation of the goodness-of-fit of the model, we used Nagelkerke’s R2 (Nagelkerke, 1991). In order to express the classification accuracy of the models irrespective of their threshold criteria, ‘Receiver Operating Characteristics’ (ROC) – plots were constructed, and ‘Area Under Curve’ (AUC) – values were calculated providing a single quantitative index of the diagnostic accuracy of the model (Zweig and Campbell, 1993). This method measures the probability that the model will assign a higher probability of occurrence to cases with observed Amphibian communities in Madagascar Julian Glos 70 presence in any data pair, randomly chosen from the presence / absence data. AUC values vary from 0.5 (no apparent accuracy) to 1.0 (perfect accuracy). For model discrimination, we calculated from the classification matrix: prevalence (proportion of presences in the total sample), sensitivity (proportion of the correct predictions for presence), and specificity (proportion of correct predictions for absence).

Sensitivity and specificity were calculated for different classification thresholds: Poptimal (maximizes the proportion of correct classifications), Pbalanced (when sensitivity § specificity), and P0.5 (classification threshold = 0.5). We tested for spatial autocorrelation of the variables using standardised deviation residuals to calculate Moran’s I as an index of covariance between different pond locations (CrimeStat® 2.0 software).

Data acquisition of dependent variable Amphibian breeding habitat choice was analyzed over three consecutive rainy seasons (1999-2000, 2000-2001, 2001-2002), in a total of 157 different potential breeding waters (n = 60 for forest ponds, n = 84 for riverbed ponds, n = 13 for savanna ponds). The sampled forest and riverbed ponds represent all breeding sites within an area of 3 km2 that are locally known as CS5, CS6, and CS7. Additionally, three forest ponds outside this area were included. The ponds analyzed in the savanna represent all ponds available in this habitat. This savanna is surrounded by the Kirindy Forest, and is situated in the area locally known as CS 9-12 within the Kirindy concession. Whether a pond was used as a breeding site or not was evaluated by repeatedly (see below) recording the presence or absence of A. laticeps tadpoles. This procedure has the advantage of directly measuring the actual reproduction, in contrast to the analysis of e.g., calling activity of adult frogs. Moreover, this method is independent of any activity patterns of the adults and considers that not all breeding waters with calling male frogs are actually used for reproduction (J. Glos, unpubl. data). Presence of A. laticeps larvae was recorded by standardized dip-netting (Heyer et al., 1994). In each pond, 30 dip net strokes were performed, randomly distributed over the pond. The dip net was triangular shaped with a base of 400 cm2 (30 x 30 x 30 cm). Each dip net stroke was 1 m long, and was touching the ground substrate. This is the preferential microhabitat of A. laticeps tadpoles (Glos and Linsenmair, 2004). All captured tadpoles were determined in the field or in the field camp to species level, using a stereo microscope, existing literature (Blommers-Schlösser and Blanc, 1991; Glaw and Vences, 1994; Glos and Linsenmair, 2004; 2005; Glos et al., 2005), and a reference collection. Specimens were released subsequently to determination. Depending on pond properties (e.g., ground substrate, exposition, water depth) and the length of periods with low precipitation, some study ponds dried out during the rainy season. When they were refilled after rainfalls, they were repeatedly used for breeding by amphibians, leading each time to a new set of tadpoles. To get a complete picture of pond use by A. laticeps throughout the rainy season, therefore, our analysis was performed several times per season and breeding site. We repeated our analysis within ten days after each rainfall > 30 mm, and after each refilling of a pond when it had Amphibian communities in Madagascar Julian Glos 71 been dried out before. Our method thus was adjusted to the breeding ecology of A. laticeps (Glos, 2003), and therefore reliably detected its presence in the respective ponds. Presence of A. laticeps tadpoles in a pond during at least one control check, irrespective of year and date, was coded as 1, and absence in all control checks at that particular pond was coded as 0. We pooled the data for all years for numerous reasons. First, the fundamental idea of our study is to analyze overall breeding habitat use, i.e. in as many ponds as possible over a time span as long as possible. Pooling the data meets this claim, and furthermore avoids pseudo-replications. Additionally, the differences in the environmental characteristics of individual ponds between seasons are negligible (unpubl. data), thus providing the legitimate basis for their pooling over different years. Moreover, there is an implicit assumption in most presence / absence designs that breeding habitats are saturated (Capen et al., 1986). In A. laticeps, however, particular characteristics of its breeding ecology (e.g., forming of breeding aggregations, colonizing preferably freshly filled ponds) make it likely that not all suitable habitats are occupied when tadpole presence is measured only at one point of time. Therefore, this assumption might then not be met.

Data acquisition of environmental variables Thirteen variables that are characterizing abiotic and biotic properties were measured at each study pond. Detailed nomenclature and definitions of these variables are given in Table 1. Habitat variables were recorded parallel to the control checks for species presence. The variables were chosen with regard to their applicability in practical conservation management. First, these variables are known to constitute important resources or habitat requirements to anurans in other systems. Second, they are measurable with passable effort and in relatively short time, independent of extensive preparatory training and any academic ecological knowledge. In a first step of prae hoc analysis we eliminated two variables (Macrohabitat, Fish) out of statistical reasons from the logistic regression analysis. Macrohabitat was excluded because A. laticeps was mainly restricted to forest ponds, and was found only in two riverbed ponds and never in savannah ponds (see Results section). Hence, the subsequent analysis was run only for forest ponds (n = 60). Fish was excluded as only two riverbed ponds contained fish at all, and all ponds occupied by A. laticeps were fish free. In a second step, we constructed a correlation matrix with the remaining independent variables, and subsequently eliminated high collinearity within the environmental variables (exclusion of variables in cases of Spearman rho 0.7, see suggestions by Fielding and Haworth, 1997). In these cases, one of two or more correlated variables were eliminated according to their practical applicability, i.e. those variables were kept that required less effort to measure or proved to be more reliable to measure, being thus more useful in applied conservation management. As a result, we eliminated the variables Pond origin (rho = 0.87 with Rock substrate), Water permanency (rho = -0.72 Amphibian communities in Madagascar Julian Glos 72 with Pond size), and Shallowness (rho = 0.80 with Water permanency and rho = -0.79 with Pond depth). Finally, we included eight environmental variables in the initial regression model (Table 2). Amphibian communitiesinMadagascar TABLE 1. Environmental variables in the logistic regression model.

Variable name Variable definition Unit Methods Macrohabitat Forest, Riverbed, or Savanna pond 3 categories Pond origin Pond is of natural or anthropogenic origin 2 categories Pond size Absolute surface area of pond when completely m2 Calculated from pond length, water-filled width, and shape Pond depth Maximum pond depth when pond is completely cm Measured at the deepest point in water-filled the pond Shallowness Relative area of the pond < 3 cm water depth 0 – 100 % Estimated Water permanency Time until pond dries up after complete water- 5 categories, from I (high permanency) to Measured after heavy rain that left filling V (low permanency); categories were all study ponds completely water- defined as days until desiccation of the filled and a consecutive dry phase pond without rain I: > 150 days, II: 55 – 150 days, III: 20 – 54 days, IV: 5 – 19 days, V: < 5 days Rock substrate Relative area of the pond bottom composed of 0 – 100 % Estimated rocky substrate Shading Relative area of pond shaded by foliage of shrubs 0 – 100 % Estimated as relative area shaded and trees at 12:00 a.m. Leaf litter Relative area of pond bottom covered with dead 6 categories Measured at plots of 0.25 m2 leaves I: 0 %, II: 1 – 20 %, III: 21 – 40 %, IV: 41 within the pond and averaged – 60 %, V: 61 – 80 %, VI: 81 – 100 % Surface water plants Relative area of pond surface covered with water 6 categories Estimated plants (mainly Nymphea lotus, Salvinia sp.) I: 0 %, II: 1 – 20 %, III: 21 – 40 %, IV: 41 – 60 %, V: 61 – 80 %, VI: 81 – 100 % Submerged plants Relative volume of pond filled with submersed 6 categories Measured at plots of 0.25 m2 water plants (mainly Lagarosiphon madagascar- I: 0 %, II: 1 – 20 %, III: 21 – 40 %, IV: 41 within the pond and averaged iensis, Najas sp.) – 60 %, V: 61 – 80 %, VI: 81 – 100 % Emergent vegetation Relative area of pond surface covered with 6 categories Estimated emergent water plants I: 0 %, II: 1 – 20 %, III: 21 – 40 %, IV: 41 – 60 %, V: 61 – 80 %, VI: 81 – 100 % Julian Glos73 Fish Presence of fish (mainly Oreochromis sp.) 2 categories (presence / absence) Observation, dip-netting Amphibian communities in Madagascar Julian Glos 74

RESULTS

Of all habitats, 153 of 157 ponds (98.1 %) were used by at least one amphibian species. A. laticeps tadpoles were found predominantly in forest ponds, only in two riverbed ponds and never in savanna ponds. Correspondingly we never noticed any calling choruses as are typical of A. laticeps (Glos, 2003) from other riverbed ponds or from savanna ponds during extensive acoustic monitoring of amphibian breeding activity (unpubl. data). In the forest ponds, the breeding sites of A. laticeps were not restricted to primary forest, but in general to forest areas with closed canopy cover. Accordingly, A. laticeps was found both in primary forest and in parts of the forest that had been used for low impact sustainable logging during the last decades (Sorg et al., 2003). In addition, puddles that had been unintentionally created by humans on abandoned dirt roads were frequently used for breeding. A. laticeps was never found in ponds that were inhabited by fish. However, only two rock pools of the river bed contained fish at all. Within the closed forest, the potential breeding waters cover a wide range of all environmental variables (Table 3).

TABLE 2. Logistic regression analysis of the breeding habitat use of A. laticeps (presence = 1, absence = 0), using a model that incorporates eight independent variables (stepwise backwards, LR – method). Initial logistic regression model: n = 60, Ȥ2 = 22.58, Nagelkerke R2 = 0.42, AUC = 0.80, p = 0.004. Final logistic regression model using variables from the initial regression (stepwise backwards analysis): n = 60, Ȥ2 = 18.54, Nagelkerke R2 = 0.35, AUC = 0.78, p < 0.001. Coefficient = regression coefficient B; SE = standard error; P = probability level. In a prae hoc analysis five variables were eliminated: Macrohabitat, Pond origin, Shallowness, Water permanency, Fish. Settings of analysis were: P (exclusion of variable) = 0.10, number of iterations = 20. For definition of the environmental variables see Table 1.

Variables Coefficient SE P Exp(B)

Initial logistic regression model Pond size 0.00 0.00 0.70 1.00 Pond depth - 0.02 0.02 0.32 0.98 Rock substrate 3.15 981.67 1.00 23.29 Shading 0.08 0.01 0.48 1.01 Leaf cover 0.32 0.25 0.21 1.37 Surface water plants - 0.05 0.05 0.28 0.95 Submerse plants 0.05 0.33 0.88 1.05 Vegetation cover - 0.02 0.01 0.17 0.98

Final logistic regression model Leaf cover 0.32 0.10 < 0.01 0.97 Surface water plants - 0.09 0.04 0.04 0.92 Vegetation cover - 0.03 0.01 0.17 0.36

The parameters of the initial and of the final multiple logistic regression model for the occurrence of A. laticeps tadpoles are listed in Table 2. Their importance for the model predictions is reflected by the correlation coefficient (r). Five variables were stepwise eliminated from the model: Amphibian communities in Madagascar Julian Glos 75

Submerged plants (step 2), Pond size (step 3), Shading (step 4), Pond depth (step 5), and Rock substrate (step 6). In the final model, the observed and predicted presence of A. laticeps tadpoles rises with increasing cover of leaf litter at the bottom of the pond (variable Leaf litter), decreases with the proportion of the pond area covered by surface water plants (Surface water plants), as well as decreases with the relative area of the pond surface covered with emergent water plants (Emergent vegetation) (Figure 1).

TABLE 3. Descriptive analysis of the environmental variables. For variable definition see Table 1.

Variable N Mean SD Minimum Median Maximum Pond size (m2) 67 840.3 3099.3 0.1 26.67 20000.0 Pond depth (cm) 67 34.3 30.7 5.0 26.0 165.0 Shallowness (%) 67 60.2 36.9 3.0 75.0 100.0 Water permanency (cat.) 67 3.8 1.2 1.0 4.2 5.0 Rock substrate (%) 67 1.8 8.9 0.0 0.0 60.0 Shading (cat.) 67 4.1 0.9 2.0 4.3 5.0 Leaf litter (cat.) 67 3.3 1.3 1.0 3.1 5.0 Surface water plants (%) 67 5.3 15.5 0.0 0.0 90.0 Submerged plants (cat.) 67 1.1 1.2 0.0 1.0 4.2 Emergent vegetation (%) 67 10.3 22.3 0.0 0.14 100.0

Prevalence of A. laticeps at the ponds was 56.7 %, meaning that this proportion of the ponds was used by A. laticeps for breeding at least once during our study. The overall percentage of correctly predicted presences and absences was 73.3 % (for threshold value Poptimal = 0.40), and 68.3 % (for threshold value Pbalanced = 0.65 and P = 0.50; Table 4). The AUC – value of 0.78 (95% confidence interval 0.66 – 0.90) confirms a good ability of the final model to give correct predictions for all possible classification thresholds. There was no evidence for spatial autocorrelation in our data (Moran’s I coefficient = -0.022).

TABLE 4. Classification accuracy for the logistic regression model; for model settings see Table 2. Popt = classification threshold for highest overall classification accuracy; Pbalanced = classification for sensitivity § specificity.

Correct predictions (%) Sensitivity (%) Specificity (%) Popt = 0.40 73,3 97,1 42,3 P = 0.50 68,3 88,2 42,3 Pbalanced = 0.65 68,3 70,6 65,4 Amphibian communities in Madagascar Julian Glos 76

1,00

0,90

0,80

0,70

0,60

Incidence 0,50

0,40

0,30

0,20 0 to 20 21 to 40 41 to 60 61 to 80 81 to 100 GroundLeaf litter leaf cover (%) (%)

1,00

0,80

0,60

0,40 Incidence

0,20

0,00 0 0.1 to 5 5.1 to 20 21 to 50 > 50 EmergentVegetation vegetation cover (%) (%)

1,00

0,80

0,60

0,40 Incidence

0,20

0,00 0 0.1 to 10 10.1 to 20 20.1 to 50 81 to 100 Surface water water plants plants (%) (%)

FIG. 1. Incidence, given as the probability of occurrence (mean ± SD), of Aglyptodactylus laticeps depending on the variables included in the final logistic regression model; (A) Leaf litter, (B) Emergent vegetation, (C) Surface water plants. For definition of these variables see Table 1. Amphibian communities in Madagascar Julian Glos 77

DISCUSSION

The Madagascan dry forest is an extremely threatened ecosystem The knowledge of a species’ ecological requirements is often a prerequisite for its successful conservation (Araújo et al., 2002). Especially in the tropics, where anthropogenic landscape modifications rapidly reduce many natural habitats, the lack of detailed biological knowledge handicaps effective conservation of many rare species. Therefore, this study examines one important ecological requirement, the choice of breeding habitat, by an endemic and endangered frog species in Madagascar, Aglyptodactylus laticeps. This frog exclusively occurs in the Central Menabe region (Glaw et al., 1998). Since the arrival of humans some 2000 years ago, this eco-system has seen severe pressure from slash and burn agriculture, illicit and licit harvesting, and expanding human populations surrounding the forest corridor (FANAMBY, 2003; Sorg et al., 2003). It has been reduced to about 3 % of its initial surface area and has become extremely fragmented (Smith, 1997). Consequently, conservation of the remaining dry deciduous forests of Western Madagascar and its fauna and flora have become a matter of great concern (Hannah et al., 1998; Ganzhorn et al., 1997). Even though dry deciduous forests generally rank among the most endangered major ecosystems of the world (Lerdau et al., 1991; Janzen, 1988) very little is known about the ecological processes and requirements of most of the species inhabiting these forest ecosystems. Several factors render A. laticeps a suitable focus of conservation priorities: The high level of local endemism, the restriction to relatively little disturbed forests, the high local abundance that is indicating an important functional role in the ecosystem, and its remarkable ecological and life history traits (Glos and Linsenmair, 2004).

Occurrence of A. laticeps can be predicted by environmental variables On the local scale of the Kirindy concession, A. laticeps is restricted to forest with a relatively low degree of disturbance and closed canopy cover. The species was absent from waters in secondary vegetation formations that are surrounding the Kirindy Forest, such as scrub and savanna. Furthermore, A. laticeps did not use any of the rock pools in the bed of the Kirindy River that are usually formed at the beginning of the rainy season as breeding sites. Corresponding with these results, on a regional scale, A. laticeps was found only in the two largest and least disturbed forest blocks (Ambadira Forest, Kirindy Forest) and its connecting corridor forest during a survey within the Central Menabe region in 2004 (Glos and Volahy, 2004). By no means can these sites be classified as pristine primary forest, as they have a long history of timber harvesting, honey collecting, and hunting (Sorg et al., 2003). At the moment, however, these forest blocks are certainly among the largest and best preserved remnants of the dry forest of western Madagascar (Nelson and Horning, 1993), and the only known sites of occurrence for A. laticeps. Amphibian communities in Madagascar Julian Glos 78

Our model highlights the relationship between environmental conditions and species occurrences. The model identified three environmental variables that suffice to satisfactorily predict the use of respective breeding sites by A. laticeps, namely leaf litter, vegetation coverage and surface water plants. The probability of A. laticeps occurrence increases with the proportional area of the pond bottom that is covered by leaf litter, decreases with the proportional surface area that is covered by standing vegetation such as grasses, and also decreases with the area covered by water plants such as water lilies (Nymphea sp.), water fern (Salvinia sp.) and duckweed (Lemna sp.). In particular, a vegetation cover of below 5 % of the pond’s surface area, the complete absence of surface water plants, and the coverage of leaf litter on the pond’s bottom of over 60 % of the bottom area best meet the requirements of A. laticeps. Our model was designed to be predictive rather than explanatory. Consequently, we do not imply that the avoidance of A. laticeps to breed in highly vegetated waters is necessarily based on a negative causal relationship, although there is a negative statistical correlation. However, the growth of this type of vegetation requires a combination of sufficient exposure to direct sunlight and an ample duration of water permanency of at least over a month. Accordingly, these plants can be considered as an indicator of low forest canopy cover, and high water permanency. Two of the variables with the highest explanatory power in our main model, vegetation cover and surface water plants, can therefore be merged into one key factor that is causally (negatively) related to breeding habitat choice of A. laticeps, namely water permanency. This variable, however, is known for not being proximately assessable by naïve frogs, although it is critically important for reproductive success. In contrast to the vegetation parameters, breeding habitat choice is presumably causally determined by the quantity of leaf litter in a pond. Dead and decaying leaves and / or its microfauna represent an important food resource for many tadpole species (McDiarmid and Altig, 1999), including A. laticeps (Glos and Linsenmair, 2004). Furthermore, leaf litter constitutes an important structural component for tadpoles, offering retreat sites and camouflage against predators.

Management implications and recommendations The three variables with the highest explanatory power in the model, two concerning vegetation cover and the 3rd quantity of leaf litter, are easy and quick to assess even by inexperienced persons. Other variables either proved to be not very predictive, although easy to measure (e.g., pond size, pond depth, water chemistry), or might be predictive but not measurable with reasonable effort (e.g., water permanency). Therefore, the use of these three variables for the prediction of A. laticeps occurrence represents an effective management tool that can be easily applied to identify potentially suitable habitat within and outside the Menabe region. By sampling the relevant waters within its range of distribution, the species’ presence in formerly not inspected habitats can be predicted, and the suitability of the habitat for this species can thus be judged. Accordingly, the distribution of A. laticeps could be narrowed down. Furthermore, the effects of habitat alteration on the presence of this species Amphibian communities in Madagascar Julian Glos 79 could be predicted, and forest use options could subsequently be ranked according to the estimated effect on the species’ distribution. When sampling potentially suitable habitat, the date of sampling within the season is a crucial factor for sampling success. Breeding of A. laticeps starts two to four weeks after the beginning of the first heavy rainfall, and extends over the entire rainy season (Glos, 2003). Therefore, sampling will be most effective during the main part of the rainy season, which usually is January and February. A. laticeps is an explosive breeding species, reproducing only after heavy rainfalls (> 30 mm) in freshly filled ponds (Glos, 2003). As larval developmental time can be very short (minimum 10 days; Glos and Linsenmair, 2004), A. laticeps tadpoles might be missed in actual breeding ponds when the timing of sampling is inadequate. Therefore, sampling should preferably be done in the time span between one and two weeks after a heavy rainfall (except the first rains in a season), when A. laticeps tadpoles are easily distinguishable from other species, but have not yet completed metamorphosis. Within its range of distribution, A. laticeps also uses breeding waters that were unintentionally created by humans, mainly puddles on dirt roads. These waters are generally shallow and free of any aquatic vegetation, and very much exposed to sunlight. A. laticeps obviously assesses this habitat as suitable for breeding, as does our model (no vegetation cover and low water permanency). However, in periods of low rainfall that regularly interrupt periods of high precipitation during the rainy season in this region, a high proportion of these waters quickly dries up before even the fast metamorphosis of A. laticeps is completed, leading to a complete breeding failure in those waters (unpubl. data). Therefore, this anthropogenically created habitat may act as a population sink rather than increasing the number of suitable breeding sites for A. laticeps. The availability of suitable breeding habitat is not the only ecological requirement for the successful establishment and persistence of a species at a given site. Suitable habitat for juveniles and adults (e.g., retreat sites during the day, overwintering or aestivating sites), qualitative and quantitative food availability, microclimatic conditions (e.g., temperature, moisture level), the identity and density of predators (e.g., lizards; Glos, 2004) and of competitors, or a combination of any of these factors might influence the presence and density of A. laticeps at a site. When transferring our model to other habitat types, therefore, these factors must be considered. For example, environmental characteristics of the breeding sites in the secondary (e.g., savanna) habitat that is surrounding Kirindy Forest do not in all cases satisfactorily explain the absence of A. laticeps from these waters (unpubl. data). In this case, the absence of A. laticeps tadpoles is more likely caused by the lack of suitable habitat for adults surrounding these ponds. This is certainly not the case for riverbed ponds as the river meanders through the closed forest without pronounced river banks, and therefore suitable adult frog habitat is nearby. However, there is a temporal incongruity between the formation of the riverbed ponds and the breeding phenology of A. laticeps. A. laticeps preferably spawns first in newly filled ponds, usually within the first two nights (see also chapter 7). However, when riverbed ponds are filled, A. laticeps is Amphibian communities in Madagascar Julian Glos 80 not yet active in breeding (Glos, 2003). Later on, the riverbed ponds are already populated by other tadpole species and therefore no longer the preferred breeding habitat for A. laticeps.

Aglyptodactylus laticeps – a suitable indicator species It is not realistic to inventory all species and their requirements when establishing promising sites for protected areas. Therefore, it is often helpful for conservation purposes to appoint representative species for the specialised fauna of a certain habitat as target species (New, 1995). Such indicator species should have narrow ecological amplitudes with respect to one or more environmental factors and its presence can thus serve as an indicator for a particular environmental condition or set of conditions (Allaby, 1994). Due to specific characteristics of their group, e. g., low mobility and permeable skin, amphibians are generally seen as good indicators of environmental integrity (e.g., (Welsh and Ollivier, 1998; Wilson and McCranie, 2003). In the Western Madagascan dry forest, A. laticeps is especially suitable for this function. Adults of A. laticeps occur exclusively in relatively little disturbed forest parts, and the presence of its tadpoles is an indicator of temporary aquatic habitats within this forest that are characterized by a low water permanency and high leaf litter coverage. Although the integrity of the terrestrial components of the dry forest might be equally well indicated by e.g., botanical variables, these are of only limited significance in respect its aquatic components. Therefore, the presence or absence of this species within its natural range indicates the relative degree of environmental integrity of this habitat as a whole (Wilson and McCranie, 2003). However, the appointment and use of indicator species in conservation remains controversial (Simberloff, 1998). They possess an undeniable appeal for practical conservationists, land managers, and governments as they provide a cost- and time-efficient mean to assess the impacts of environmental disturbances on an ecosystem. Their use is particularly advantageous when several species representing different taxa and life histories are included as indicator species in a monitoring program (Carignan and Villard, 2002). Ideally, the habitat requirements of the target species should also reflect the demands of other species in need of protection (“umbrella effect”; Simberloff, 1998; New, 1995). This is the case for A. laticeps and several other anuran species in the Menabe region. The protection of suitable habitat for A. laticeps will also promote the survival of Boophis doulioti, Laliostoma labrosum, Mantella betsileo (Mantellidae), Dyscophus insularis, Scaphiophryne calcarata and S. menabensis (Microhylidae) (Glos and Linsenmair, 2004). Of these, Scaphiophryne menabensis is of particular conservation interest, as it is among the rarest frogs within this area (Glos et al., 2005), due to its extremely restricted distribution.

Conclusions The model presented in this study opens up the possibility to assess suitable habitat of A. laticeps as an indicator species and thus predict presence and absence of this species, by using only few, easy-to- obtain and easy-to measure variables. It demonstrates that models constructed from basic ecological Amphibian communities in Madagascar Julian Glos 81 knowledge of relevant species may serve as valuable management tools in applied conservation. With their aid, not only can recommendations be made as to which areas of an ecosystem conservation should focus on, but also inferences be made as to which direction a habitat should be improved, or advice may be derived on how to effectively restore an already impoverished habitat. Amphibian communities in Madagascar Julian Glos 82

LITERATURE CITED

ADLER, G. H., and M. L. WILSON. 1985. Small mammals on Massachusetts island: the use of probability functions in clarifying biogeographic relations. Oecologia. 66:178-186.

ALLABY, M. 1994. The Concise Oxford Dictionary of Ecology. Oxford University Press, Oxford New York.

ANDREONE, F., J. E. CADLE, N. A. COX, F. GLAW, R. A. NUSSBAUM, C. J. RAXWORTHY, S. N.

STUART, D. VALLAN, and M. VENCES. 2005. Species review of amphibian extinction risks in Madagascar: conclusions from the global amphibian assessment. Conservation Biology. 19:1790-1802.

ARAÚJO, M. B., P. H. WILIAM, and R. J. FULLER. 2002. Dynamics of extinction and the selection of nature reserves. Proceedings of the Royal Society Biological Sciences Series B. 269:1971- 1980.

BLOMMERS-SCHLÖSSER, R. M. A., and C. P. BLANC. 1991. Amphibiens (première partie). Faune de Madagascar. 75:1-379.

BLOXAM, Q. M. C., J. L. BEHLER, E. R. RAKOTOVAO, H. RANDRIAMAHAZO, K. T. HAYES, S. J.

TONGE, and J. U. GANZHORN. 1996. Effects of logging on the reptile fauna of the Kirindy Forest with special emphasis on the flat-tailed tortoise (Pyxis planicauda), p. 189-203. In: Ecology and Economy of a tropical dry forest in Madagascar. J. U. Ganzhorn and J.-P. Sorg (eds.), Göttingen.

CAPEN, D. E., J. W. FENWICK, B. B. INKLEY, and A. C. BOYNTON. 1986. Multivariate models of songbird habitat in New England forests, p. 171-175. In: Wildlife 2000: modelling habitat relationships of terrestrial vertebrates. J. Verner, M. L. Morrison, and C. J. Ralph (eds.). University of Wisconsin Press, Madison & London.

CARIGNAN, V., and M.-A. VILLARD. 2002. Selecting indicator species to monitor ecological integrity: a review. Environmental Monitoring and Assessment. 78:45-61. FANAMBY. 2003. Biodiversity Conservation in the Central Menabe Region through Protected Areas Creation - A proposal presented to the Global Conservation Fund (GCF). Association FANAMBY, Antananarivo, Madagascar.

FAVRE, J.-C. 1996. Traditional utilization of the forest, p. 33-40. In: Economy and Ecology of a Tropical Dry Forest in Madagascar. Vol. 46-1. J. U. Ganzhorn and J.-P. Sorg (eds.), Göttingen.

FIELDING, A. H., and P. H. HAWORTH. 1997. A review of methods for the assessment of predicting errors in conservation presence/absence models. Environmental Conservation. 24:38-49.

GANZHORN, J. U., B. RAKOTOSAMIMANANA, L. HANNAH, J. HOUGH, L. IYER, S. OLIVIERI, S.

RAJAOBELINA, C. RODSTROM, and G. TILKIN. 1997. Priorities for Biodiversity Conservation in Madagascar. Primate Report. 48-1:1-81. Amphibian communities in Madagascar Julian Glos 83

GIBSON, R. C., and K. R. BULEY. 2004. Biology, captive husbandry, and conservation of the malagasy flat-tailed tortoise, Pyxis planicauda Grandidier, 1867. Herpetological Review. 35:111-116.

GLAW, F., and M. VENCES. 1994. A Fieldguide to the Amphibians and Reptiles of Madagascar. M. Vences & F. Glaw Verlags GbR, Köln.

GLAW, F., M. VENCES, and W. BÖHME. 1998. Systematic revision of the genus Aglyptodactylus Boulenger, 1919 (Amphibia: Ranidae), and analysis of its phylogenetic relationships to other Madagascan ranid genera (Tomopterna, Boophis, Mantidactylus and Mantella). Journal of Zoological Systematics and Evolutionary Research. 36:17-37.

GLOS, J. 2003. The amphibian fauna of the Kirindy dry forest in western Madagascar. Salamandra. 39:75-90. —. 2004. ZONOSAURUS LATICAUDATUS (NCN). PREDATION. Herpetological Review. 35:271.

GLOS, J., F. GLAW, and M. VENCES. 2005. A new species of Scaphiophryne from Western Madagascar. Copeia. 2005:252-261.

GLOS, J., and K. E. LINSENMAIR. 2004. Description of the tadpoles of Aglyptodactylus laticeps and A. securifer from Western Madagascar, with notes on life history and ecology. Journal of Herpetology. 38:131-136. —. 2005. Description of the tadpoles of Boophis doulioti and B. xerophilus from Western Madagascar with notes on larval life history and breeding ecology. Amphibia-Reptilia. 26:459-466.

GLOS, J., and A. T. VOLAHY. 2004. An inventory of amphibians and reptiles in the Central Menabe region, p. 17. Report to DWCT (Durrell Wildlife Conservation Trust), DPZ (Deutsches Primatenzentrum), and FANAMBY, Antananarivo.

GOODMAN, S. M., and J. P. BENSTEAD. 2003. The Natural History of Madagascar. Chicago University Press, Chicago and London.

HANNAH, L., B. RAKOTOSAMIMANANA, J. U. GANZHORN, R. A. MITTERMEIER, S. OLIVIERI, L. IYER,

S. RAJAOBELINA, J. HOUGH, F. ANDRIAMIALISOA, I. BOWLES, and G. TILKIN. 1998. Participatory planning, scientific priorities, and landscape conservation in Madagascar. Environmental Conservation. 25:30-36.

HOSMER, D. W., and S. LEMESHOW. 1989. Applied logistic regression. Wiley, New York. IUCN. 2001. IUCN Red List Categories: Version 3.1. Prepared by the IUCN Species Survival Commission. IUCN, Gland and Cambridge.

JANZEN, D. H. 1988. Tropical dry forests: the most endangered major tropical ecosystem, p. 130-137. In: Biodiversity. E. O. Wilson (ed.). National Academy Press, Washington, D.C., USA.

LERDAU, M., J. WHITBECK, and N. M. HOLBROOK. 1991. Tropical deciduous forest: death of a biome. Trends in Ecology and Evolution. 6:201-202. Amphibian communities in Madagascar Julian Glos 84

MCCUNE, B. 2004. Nonparametric multiplicative regression for habitat modelling.

MCDIARMID, R. W., and R. ALTIG. 1999. Tadpoles: The Biology of Anuran Larvae. The University of Chicago Press, Chicago.

MYERS, N., R. A. MITTERMEIER, C. G. MITTERMAIER, G. A. B. DA FONSECA, and J. KENT. 2000. Biodiversity hotspots for conservation priorities. Nature. 403:853-858.

NAGELKERKE, N. J. D. 1991. A note on general definition of the coefficient of determination. Biometrica. 78:691-692.

NELSON, R., and N. HORNING. 1993. AVHRR-LAC estimates of forest area in Madagascar, 1990. International Journal of Remote Sensing. 14:1463-1475.

NEW, T. R. 1995. The species: Elements of a management plan. Victorian Naturalist. 112:40-42.

RANDRIANANDIANINA, B. N., L. R. ANDRIAMAHALY, F. M. HARISOA, and M. E. NICOLL. 2003. The role of the protected areas in the management of the island's biodiversity. In: The Natural History of Madagascar. S. M. Goodman and J. P. Benstead (eds.). The University of Chicago Press, Chicago and London.

SIMBERLOFF, D. 1998. Flagships, umbrellas, and keystones: Is single-species management passe in the landscape era? Biological Conservation. 83:247-257.

SMITH, A. P. 1997. Deforestation, Fragmentation, and Reserve Design in Western Madagascar, p. 415- 441. In: Tropical Forest Remnants: Ecology, Management, and Conservation of Fragmented Communities. W. F. Laurance and R. O. Bierregard, Jr. (eds.). The University of Chicago Press, Chicago & London.

SOMMER, S., and U. HOMMEN. 2000. Modelling the effects of life-history traits and changing ecological conditions on the population dynamics and persistence of the endangered Malagasy giant jumping rat (Hypogeomys antimena). Animal Conservation. 4:333-343.

SOMMER, S., A. T. VOLAHY, and U. S. SEAL. 2002. A population and habitat viability assessment for the highly endangered giant jumping rat (Hypogeomys antimena), the largest extant endemic rodent of Madagascar. Animal Conservation. 5:263-273.

SORG, J.-P., J. U. GANZHORN, and P. M. KAPPELER. 2003. Forestry and research in the Kirindy Forest/Centre de Formation Professionelle Forestrière. In: The Natural History of Madagascar. S. M. Goodman and J. P. Benstead (eds.). The University of Chicage Press, Chicago and London.

WELSH, H. H., JR., and L. M. OLLIVIER. 1998. Stream amphibians as indicators of ecosystem stress: a case study from California's Redwoods. Ecological Applications. 8:1118-1132. Amphibian communities in Madagascar Julian Glos 85

WILSON, L. D., and J. R. MCCRANIE. 2003. Herpetofaunal indicator species as measures of environmental stability in Honduras. Carribean Journal of Science. 39:50-67.

ZWEIG, M. H., and G. CAMPBELL. 1993. Receiver-operating characteristic (ROC) plots: A fundamental evaluation tool in clinical medicine. Clinical Chemistry. 39:561-577. Amphibian communities in Madagascar Julian Glos 86

Chapter 7

Oviposition site selection in a Madagascan frog - Experimental evaluation of a habitat model

Julian Glos, Frank Wegner & K. Eduard Linsenmair

Department of Animal Ecology and Tropical Biology, Biocenter, Würzburg University, Am Hubland, 97074 Würzburg, Germany

ABSTRACT

The selection of the oviposition site is crucial for the fitness in many organisms, including the endangered Madagascan frog Aglyptodactylus laticeps. In chapter 6, we presented a conservation oriented model of habitat use for this species that extracted key habitat factors that reliably predict the utilization of ponds as oviposition sites. However, habitat use models are descriptive and thus have only limited explanatory power for true causal relationships between habitat variables and species distributions. Furthermore, habitat use models are generally affected by low temporal resolution and are mostly insufficient in regard to incorporating dynamic components of habitat choice, such as the colonization history. Therefore, we experimentally investigated the explanatory power of the habitat use model within the context of dynamic components, by creating artificial breeding ponds in the natural habitat of this frog species and by manipulating two environmental factors within these breeding ponds (presence of leaf litter at the bottom of the ponds, presence of conspecific tadpoles). A. laticeps proved to be an exceptionally quick colonizer of freshly filled temporary forest ponds. Contrary to conclusions derived from the habitat use model, leaf litter on the ponds’ bottom did not prove to be a decisive factor for oviposition site selection. In contrast, the presence of conspecific tadpoles proved to be much more decisive. These results highlight the need to cautiously treat conclusions derived from descriptive habitat models in respect to their significance for explaining true causal relationships. Furthermore, they show that the influence of behavioural variables on observed distribution patterns can often only inadequately be addressed in descriptive studies, and experimental studies are needed to elucidate the true relationships. Through this, experimental studies expand the explanatory and predictive power of habitat models, and their conclusions must be incorporated in the recommendations for species management and conservation derived from the respective habitat models.

Amphibian communities in Madagascar Julian Glos 87

INTRODUCTION

The site of oviposition considerably influences hatching success, larval performance, and, thereby, parental fitness in many organisms. Depending on the species’ requirements, the survival, growth and development of juveniles or larvae can be optimized in respect to abiotic (e.g., temperature) and biotic factors (e.g., predators, competitors) by choosing a suitable breeding habitat (Reich and Downes, 2003; Blaustein et al., 2004). Therefore, selection for the ability to discriminate between potential oviposition sites on the basis of expected larval performance should be strong. This is particularly true when the respective life stages (e.g., larvae) are not very mobile, i.e. cannot migrate to better habitat patches, or when respective movements involve high fitness costs (e.g., high mortality or high energy expenditure). The knowledge about the ecological requirements of species, with the characteristics of the oviposition sites being an essential part of it, is a prerequisite for successful conservation of endangered species, including amphibians (Araújo et al., 2002). Especially in the tropics, where anthropogenic landscape modifications rapidly reduce the size of many natural habitats and change their character, the lack of detailed biological knowledge handicaps effective conservation of many rare species. In this study we analyze the oviposition site choice of a tropical anuran species that is of particular conservation interest. Aglyptodactylus laticeps, a mantellid frog endemic to the Menabe region in western Madagascar, is facing a very high risk of extinction in the wild due to the destruction of its natural habitat, the dry deciduous forest (EN based on IUCN criteria; Andreone et al., 2005). On the basis of the responsiveness of this species to disturbances, A. laticeps is considered to be an adequate indicator species for the relative degree of environmental integrity of ephemeral lentic systems within the dry forest of the Menabe region, and it was proposed to expand the species oriented conservation concepts of this highly threatened region. In a preceding study, Glos et al. (chapter 6) presented a conservation oriented model of habitat use for this species that extracted the key habitat factors that reliably predict the utilization of ponds as oviposition sites. In this model, the most influential factors proved to be the existence of pond vegetation and the amount of leaf litter at the bottom of the breeding pond. Based on the well-established significance of leaf litter as food resource and structural component for tadpoles (McDiarmid and Altig, 1999), this factor was considered to be directly and causally related to oviposition site choice of A. laticeps. However, habitat use models are descriptive and thus have only limited explanatory power for true causal relationships between habitat variables and species distributions. In other words, statistical models provide a description of the realized niche of a species, but can say little about its fundamental niche. Furthermore, habitat use models are generally affected by low temporal resolution and mostly insufficient in regard to incorporating dynamic components of habitat choice, such as colonization history, and effects of intraspecific competition. Amphibian communities in Madagascar Julian Glos 88

The actual influence of these dynamic components on oviposition site selection can only be assessed experimentally. In particular, the colonization history, i.e. the question whether a respective site has already been used as a breeding habitat before by a conspecific, is considered to be an important factor either increasing (e.g., Rudolf and Rödel, 2005) or decreasing (e.g., Spieler and Linsenmair, 1997) the probability of oviposition by an individual. In order to e.g., avoid intraspecific competition individuals might choose oviposition sites that are suboptimal in respect to other biotic and abiotic habitat characteristics. Consequently, the true habitat requirements of a species, i.e. its fundamental niche, might be only imprecisely identified in a descriptive analysis that is based on the actually prevailing situation, and subsequent conservation actions based on these analyses might therefore be less effective.

STUDY OBJECTIVES

We experimentally investigated the explanatory power of the previously established habitat use model of the endangered Madagascan frog Aglyptodactylus laticeps within the context of dynamic components, by creating artificial breeding ponds in the natural habitat of this frog species and by manipulating two environmental factors within these breeding ponds. First, we examined if these anurans are able to assess the suitability of potential breeding sites. Second, given that they do differentiate between ponds, we tried to identify the habitat variables that are most decisive for oviposition site selection. Specifically, we evaluated whether the presence of leaf litter at the bottom of the ponds is a causal factor in the choice of breeding habitat, as predicted by the descriptive habitat model. Furthermore, we tested whether the presence of conspecific tadpoles in a potential breeding pond is influencing oviposition site selection. Finally, general information about the colonization ability of A. laticeps was acquired, and is discussed in respect to conservation issues.

MATERIAL AND METHODS

Study site The Kirindy Forest/CFPF (Centre de Formation Professionnelle Forestière) is one of the largest remnants of the dry, deciduous forest of western Madagascar (120 km2; Nelson and Horning, 1993). It is located about 50 km northeast of Morondava and 20 km inland (44°39’E, 20°03’S; 18 – 40 m above sea level; Sorg and Rohner, 1996). Until about 1994, Kirindy/CFPF was the focus of silvicultural research projects and agroforestry experiments including sustainable timber exploitation with subsequent reforestation (Sorg et al., 2003). Amphibian communities in Madagascar Julian Glos 89

The climate of this area is characterized by a marked seasonality. The rainy season of 3 – 5 months from November / December to February / March is followed by a pronounced dry season of 7 – 9 months with virtually no rain. Mean annual precipitation is about 800 mm (range 390 – 1511 mm; Sorg and Rohner, 1996). All stretches of running and standing waters dry up every year during the dry season except for very few pools in the bed of the Kirindy river. Depending on amount and distribution of precipitation and habitat parameters of the pools themselves, the persistence of these waters varies from several days to five months. Further information on the study site is provided by Sorg and Rohner (1996) and Glos (2003).

Study animal Agylptodactylus laticeps (Laliostominae: Mantellidae: Anura; Vences and Glaw, 2001) was described first from the study site (Glaw et al., 1998). It is endemic to the Central Menabe region, from Kirindy Forest in the south to the Tsiribinha river in the north (Glos and Volahy, 2004). It is classified as endangered species (EN) on the basis of its restricted occurrence and the rapid decline of its habitat area (Andreone et al., 2005). However, it is locally abundant within its habitat (Glos, 2003). A. laticeps shows distinct adaptations to its dry forest habitat. It is active only during the rainy season, while spending the dry season presumably hibernating underground. Breeding starts about one month after the beginning of the rainy season and extends over its whole duration (Glos, 2003). It is a typical explosive breeding species, usually reproducing only after heavy rainfalls that exceed 30 mm of precipitation. Larval development is very short (minimum 10 days; Glos and Linsenmair, 2004), which is presumably an adaptation to the potentially high desiccation risk of its breeding ponds. In accordance with this assumption, the tadpoles of A. laticeps seem to exhibit phenotypic plasticity in metamorphic traits (J. Glos, unpubl. data).

Experimental setup Creation of artificial pools We created artificial breeding pools (on January 24, 2002) to experimentally test habitat use in A. laticeps. A total of 18 pools with a surface area of about 1.5 m2 each and a maximum water depth of about 20 cm were dug (Table 1). The pools were lined with gray plastic to prevent quick desiccation. They were about 100m apart from each other and were laid out along two lines in an area that was exposed to only low impact logging within the last decades (locally known as CS5 within the Kirindy forest). We recorded seven habitat variables characterizing these artificial breeding pools to test whether there were any differences between the treatment groups other than those that were manipulated (Table 1): Depth (maximum water depth), Area (surface area when pool was maximum water-filled, calculated from dimensions and shape), Exposition (percentage of the pool sun-exposed at 12:00 hours, estimated during on-site checks), Conductivity (measured several times at a depth of 5 Amphibian communities in Madagascar Julian Glos 90 cm, 30 cm from the pond’s edge with a WA 300 conductivity analyzer, Conrad Electronics, and averaged), PH (measured several times at a depth of 5 cm, 30 cm from the pond’s edge with a pHep® pH analyzer, Hanna Instruments, and averaged), Nearest plant (distance from the edge of the pool to the nearest plant > 10 cm in dbh, Predators (number of larvae of Anisoptera and Zygoptera, and of larvae and adults of Dytiscidae, Belostomatidae and Nepidae), Distance to next source pond (distance to next natural breeding pond of A. laticeps, calculated from GPS coordinates and ArcView© Software). We tested if there were statistical differences between the ponds of the three treatment groups (see below) with non-parametric statistics. The relevant data were collected when all pools were water filled due to a heavy rain the night before.

Experimental manipulation We tested for oviposition site selection of A. laticeps in respect to two habitat variables, namely the presence of leaf litter on the bottom of the pool, and the presence of conspecific tadpoles in the pool. Therefore, we randomly assigned the ponds to one of three groups (six pools per group): (1) – leaves – tadpoles: The bottom of the pool was not covered with leaf litter; there were no tadpoles introduced. (2) + leaves – tadpoles: The bottom of the pool was covered with a 3 cm layer of leaf litter from the adjacent forest floor; no tadpoles were introduced. (3) + leaves + tadpoles: The bottom of the pool was covered with a 3 cm layer of leaf litter from the adjacent forest floor ; when the pools were filled by rain, we immediately introduced 100 tadpoles of A. laticeps. These tadpoles were raised from 5 amplectant pairs that we collected in the field, and were kept in mixed-sibling groups in stock tanks in the field camp until the start of the experiment. At the beginning of the experiment, the initial body length of a random sample of 30 tadpoles from the stock tanks was 5.0 – 6.0 mm, and developmental stage of these tadpoles was 25 - 28 (staging after Gosner, 1960).

Data acquisition - Response variables The experiment started when all pools were simultaneously filled by rain at the end of January 2002. From then on, we recorded six response variables over 10 consecutive nights during five minute visits per night at each pool: (a, b) The (cumulative) number of males and females present at the pond (number of males, number of females). When actively breeding, A. laticeps males are evenly spaced around breeding pools in a distance of 0.3 to 3 m from the water’s edge (Glos, 2003). Five minutes are therefore a sufficient time span to thoroughly check the surrounding area of a pool (≤ 3 m) for frogs. (c) The number of days that passed since the beginning of the experiment before the first males were observed that uttered advertisement calls (time to first calling males). (d) The number of days that passed before the first egg clutch was deposited, i.e. the first breeding occurred (time to first spawning). (e) The (cumulative) number of egg clutches in the pool (number of egg clutches). This variable was recorded by checking the pools each night and each morning for egg clutches. After heavy rainfall, the egg clutches of A. laticeps were dissolved by raindrops, and they were thus no Amphibian communities in Madagascar Julian Glos 91 longer identifiable as units. In these cases, these data were excluded from the analysis. Finally, when we stopped the experiment after 10 days, (f) the number of tadpoles in the artificial pools was counted (number of tadpoles). This was done by completely scooping the pools and transferring all tadpoles of each pond into a 10 l tub. All tadpoles in a subsample of 405 ml (three randomly taken cups à 135 ml per tub) were counted, and this number extrapolated to the total volume of the tub. Subsequently, we analyzed species identity, and measured body length, total length, and developmental stage in another subsample of these tadpoles (n = 15). To distinguish between the tadpoles originating from clutches spawned during the experiment and those that we initially introduced into the pools as part of the experiment in the + leaves + tadpoles treatment, we separated the tadpoles in two size classes in these ponds, and included only the smaller size class in this analysis. The two size classes were clearly distinguishable (body length 4.9 ± 0.5 versus 8.4 ± 0.7 mm, total length 14.0 ± 1.8 versus 22.5 ± 1.2 mm, developmental stage 27.1 ± 0.7 versus 35.6 ± 2.8, and were significantly different (Mann- Whitney U-Test, Z = -2.88 for body length, Z = 2.88 for total length, Z = 2.89 for Gosner stage; p < 0.01; n = 6).

Exclusion of spatial effects This study was designed to be performed within the natural breeding habitat of A. laticeps. Therefore, the artificially created pools were located within a network of natural breeding sites of this frog. However, the distance to the nearest natural breeding pond might possibly influence the utilization (e.g., time and probability of colonization) of the experimental pools. To test for these effects, we first analyzed differences in the variable distance to next source pond between the treatment groups. Thereafter, we performed correlations of all five response variables with distance to next source pond, and accordingly, significant correlations were interpreted as unintented effects of distance to next source pond on time and probability of breeding.

RESULTS

Characteristics of artificial breeding pools The artificial breeding pools, although smaller in surface area than the average natural breeding pond of A. laticeps, were within their range in all variables (Glos and Linsenmair, 2004, chapter 6) (Table 1). We tested whether there were any differences in pond characteristics apart from the variables that we had manipulated experimentally (presence of leaf litter and of conspecific tadpoles) between the ponds of the different treatment groups. There was no such difference in any of the variables (Table 1).

Amphibian communities in Madagascar Julian Glos 92

TABLE 1. Variables characterizing the artificial breeding pools, explanations see text. Kruskal-Wallis tests did not detect any differences between the three treatment groups. All values are given as mean ± SD.

Pond parameter - Leaves + Leaves + Leaves H P - Tadpoles - Tadpoles + Tadpoles (N = 6) (N = 6) (N = 6) Depth (cm) 18.8 ± 1.2 17.8 ± 2.7 19.8 ± 1.7 2.29 0.32 Area (m2) 1.6 ± 0.2 1.6 ± 0.4 1.4 ± 0.3 2.45 0.29 Exposition (%) 26.7 ± 22.7 39.2 ± 22.9 30.8 ± 18.8 1.37 0.50 Conductivity (μS) 34.8 ± 18.8 44.7 ± 14.3 37.5 ± 5.7 2.75 0.25 PH 6.6 ± 0.4 6.4 ± 0.3 6.3 ± 0.4 1.55 0.46 Nearest plant (m) 0.8 ± 0.6 0.9 ± 0.6 0.7 ± 0.5 0.21 0.90 Predators (number) 0.0 ± 0.0 0.0 ± 0.0 0.0 ± 0.0 0.00 1.00 Distance to next 262.6 ± 145.0 290.6 ± 86.6 267.1 ± 115.4 0.19 0.91 source pond (m)

Spatial effects The three treatment groups did not differ in their distance to the next natural breeding pond (“source pond”) of A. laticeps (Table 1). Furthermore, there were no significant correlations of the distance to the next natural breeding pond with number of males (Spearman: rho = -0.03, n = 18, p = 0.92), number of egg clutches (rho = -0.23, n = 18, p = 0.36), time to first spawning (rho = 0.28, n = 18, p = 0.25), and number of tadpoles (rho = 0.16, n = 18, p = 0.53), confirming the equality of these ponds. Solely the number of females at the artificial pools was negatively correlated to the distance to the next source pond (rho = -0.51, n = 18, p = 0.03), but only when no correction of α was applied (e.g., Bonferroni-correction; Sachs, 1984). Therefore, the distance to the next natural breeding pond did not bias our experiment.

Treatment effects The experimental treatment had a distinct influence on the utilization pattern of the artificial ponds as breeding sites by A. laticeps. The presence of conspecific tadpoles had a significant effect in two out of six response variables (Figure 1). In ponds without an introduced set of tadpoles, first reproduction took place notably earlier, and there were on average ten times more egg clutches deposited than in ponds that were initially equipped with A. laticeps tadpoles. In both variables, this was independent of the presence or absence of leaf litter in the pond, i.e. none of the response variables differed significantly between ponds with leaf litter and those without. In accordance with these results, there was a strikingly higher, although not significant, number of tadpoles in the + leaves – tadpoles treatment ponds on the one side, and the two other groups on the other side (Kruskal-Wallis test and multiple post hoc comparisons, DUNN-tests with Holm’s alpha-correction; Dunn, 1964; Holm, 1979: Number of males: H = 0.18, p = 0.91; Number of females: H = 2.37, p = 0.30; Time to first calling: H = 1.07, p = 0.59; Time to first spawning: H = 6.70, p = 0.03; Number of egg clutches: H = 9.92, p < 0.01; Number of tadpoles: H = 4.68, p = 0.09). When we stopped the experiment after 10 days, A. laticeps male frogs were calling at all pools at some point, and all but one pool (in the – leaves – tadpoles Amphibian communities in Madagascar Julian Glos 93

a a b

a a

FIG. 1. Response variables of A. laticeps tadpoles at the end of the experiment in the ponds of the different treatments. -/- = no leaf litter, no introduced tadpoles, +/- = leaf litter, no introduced tadpoles; +/+ = leaf litter and introduced tadpoles. Box = 25%- and 75%-percentiles, whiskers = 5%- and 95%- percentiles, asterisks = maximum and minimum. Significant differences (Kruskal–Wallis and multiple post hoc comparisons, DUNN-tests with Holm’s alpha-correction; Holm, 1979; Dunn, 1964) were found in C and E and are indicated by different letters.

Amphibian communities in Madagascar Julian Glos 94 treatment) contained tadpoles of A. laticeps in the smaller size class (i.e. non-introduced tadpoles). Then, tadpoles between treatments did not differ significantly in size, and did marginally not significantly differ in developmental stage (Table 2).

TABLE 2. Size and developmental stage of A. laticeps tadpoles at the end of the experiment in the ponds of the different treatments. Kruskal-Wallis test (α = 0.05); all values are given as mean ± SD. One of the pools of the – leaves – tadpoles treatment was not utilized for reproduction during the study period by A. laticeps, hence N = 5.

- Leaves + Leaves + Leaves H P - Tadpoles - Tadpoles + Tadpoles (N = 5) (N = 6) (N = 6) Body length (mm) 4.43 ± 0.64 4.75 ± 0.51 4.90 ± 0.50 1.18 0.55 Total length (mm) 11.79 ± 1.35 12.74 ± 1.18 13.97 ± 1.63 3.90 0.14 Gosner stage 25.79 ± 0.67 26.83 ± 0.98 27.15 ± 0.60 5.86 0.053

In general, A. laticeps accepted the artificial ponds exceptionally well and quickly as breeding sites. Within only a few days after the ponds had filled, the first male frogs were found calling, and reproduction took place. Apart from A. laticeps we found adults of the sympatric frog species Laliostoma labrosum (Mantellidae), and Dyscophus insularis and Scaphiophryne calcarata (Microhylidae) in the immediate vicinity of the pools, but only sporadically so. Indeed, the analysis of tadpole species identity revealed that none of these species had used the pools as breeding sites during the duration of the study.

DISCUSSION

Aglyptodactylus laticeps is an excellent colonizer of freshly arisen forest ponds The dry forests of Western Madagascar represent a variable and unpredictable environment for temporary pond breeding organisms, such as Aglyptodactylus. The results of this study reveal mechanisms, which make the successful reproduction under these harsh seasonal conditions possible. A. laticeps proved to be an exceptionally quick colonizer of freshly arisen temporary forest ponds. Within only a few days, the first male frogs were calling in ultimate proximity of the artificially created ponds, and egg clutches were deposited in the ponds. This is especially remarkable in consideration of the fact, that the distance of these artificial ponds to the next natural breeding pond was up to 400 m, indicating that home ranges of A. laticeps are not restricted to the direct vicinity of existing breeding waters, but, in contrast, that these frogs regularly go on extensive forays in the forest. The coverage of comparable distances is known from migrations to and from their hibernation sites in anurans, and from migrations between different breeding habitat types (> 600 m in Hoplobatrachus occipitalis (Spieler and Linsenmair, 1998), but is exceptional as part of routine Amphibian communities in Madagascar Julian Glos 95 movements within breeding seasons. In Hyla versicolor, for example, a study on colonization rates of artificial pools in varying distances found that 95 % of all eggs were deposited within 15 m of the natural breeding pond (Johnson and Semlitsch, 2003). As an exception, Lampert (2001) reports comparable migration distances between breeding ponds in the reed frog Hyperolius nitidulus (regular movements of 5 to 200 m, and rarely of up to several hundred meters). The long-distance travelling of A. laticeps might be an adaptation to the comparative unpredictability of the emergence sites of temporary ponds in the dry forest, making a regular check of a wide area necessary to quickly find and utilize newly arisen ponds, before they dry up again. Accordingly, A. laticeps was not only the first anuran species to colonize the artificial ponds, but it was the only species to do so within the time span of our study and beyond it (J. Glos, unpubl. data). This behaviour is seen as beneficial for individual fitness. Early arrival at breeding sites can decrease the strength of competition (Morin et al., 1990; Alford and Wilbur, 1985; Wilbur and Alford, 1985), the risk of predation (Blaustein and Margalit, 1996), and tadpole mortality due to pond desiccation. Therefore, together with their very fast larval development (minimum of 10 days; Glos and Linsenmair, 2004) and an adaptive plasticity in the timing of metamorphosis (J. Glos, unpubl. data), this colonization ability gives them the possibility to exclusively open up breeding possibilities, not available for other species, and is a key to successfully use temporary ponds as breeding sites in the dry forest.

Oviposition site selection is not based on the amount of leaf litter in the pond The selection of an adequate oviposition site plays a crucial role for an animal’s reproductive success. Where an animal places its annual, or possibly lifetime, reproductive investment can be as important as how that investment is packaged or what quality its mate possesses. Factors such as the presence of other species at a pond, and the pond’s age, temperature, vegetation structure, ground consistence, and degree of permanence form an interactive complex of characters that influences its suitability as an oviposition site. Contrary to one conclusion derived from the descriptive model of habitat use of A. laticeps (chapter 6), leaf litter on the ponds’ bottom did not prove to be a decisive factor for oviposition site selection in this experimental approach. This highlights the need to cautiously treat conclusions derived from descriptive habitat models in respect to their significance for explaining true causal relationships. Nevertheless, we consider it unlikely that the presence of leaf litter in a pond does not positively influence growth, development and survival, and therefore fitness of A. laticeps tadpoles. Leaf litter, either the plant material itself or microorganisms living on it, is known to be an important food resource for tadpoles of many anurans, including A. laticeps (Glos and Linsenmair, 2004). Furthermore, it is known to constitute an important structural component of the habitat, e.g., as hide- out against predators (e.g., Semlitsch and Reyer, 1992). The significance of leaf litter on tadpole Amphibian communities in Madagascar Julian Glos 96 survival is indeed indicated by our study. Although the presence of leaf litter did not influence the number of egg clutches deposited in the pools, a considerably higher (albeit statistically not significant) number of tadpoles was found at the end of the experiment in those pools that had initially been equipped with leaf litter. If the fitness of tadpoles is increased in pools containing leaf litter, it would be expected that adult A. laticeps prefer these ponds for spawning to increase their individual fitness. That this is not the case, gives reason to presume that adult A. laticeps are not able to (proximately) assess the presence of leaf litter in a breeding pond. One reason might be that the presence of leaf litter normally is a common property of all breeding ponds in the closed forest (i.e., our experiment represents an unnatural situation). Therefore, selection pressure in the evolutionary history did not act on A. laticeps to avoid leaf litter free pools within the forest. Another possible explanation is that it is not the leaf litter itself that is proximately assessed by A. laticeps to evaluate habitat quality, but a variable which is highly correlated with leaf litter under natural conditions (e.g., canopy cover at the site). This mechanism might have evolved when the respective variable is easier and more efficient for frogs to assess. In respect to species conservation, the crucial consequence is as follows: if leaf litter itself is not a proximate factor in oviposition site selection, A. laticeps might not be able to avoid less suitable, viz. leaf litter free, ponds for oviposition under altered environmental conditions (e.g., in logged parts of the forest), despite the significance of leaf litter for tadpole performance.

Presence of conspecifics is decisive for oviposition site selection In our experiments, a different factor proved to be much more decisive for the choice of the oviposition site than the presence of leaf litter, namely the presence of conspecific tadpoles. For many species, the presence of conspecifics has particular influence on oviposition site selection (Anderson and Löfqvist, 1996; Crump, 1991; Edgerly et al., 1998; Huth and Pellmyr, 1999; Resetarits and Wilbur, 1989). Individuals either avoid (e.g., Spieler and Linsenmair, 1997) or prefer (e.g., Rudolf and Rödel, 2005) sites already utilized by conspecifics. This is seen as a consequence of the fact that many (especially temporary pond breeding) species are sensitive to the density of conspecifics in their immediate environment. Accordingly, the density dependence on survival, growth, duration of larval development, and size at metamorphosis has been demonstrated for insects (e.g., Anderson and Löfqvist, 1996) and larval anurans (e.g., Newman, 1987; 1998; Semlitsch, 1987; Wilbur, 1977), and therefore potentially affects parental fitness (Smith, 1987; Berven, 1988; 1990). In temporary ponds, density effects may be particularly critical when density slows growth to the point that few or none of the larvae reach minimum size for metamorphosis before the pond dries up (Wilbur, 1987). This threat is indeed imminent in the Western Madagascan dry forests, as the rainy season is regularly interrupted by periods of low rainfall of up to a week, and desiccation risk in ponds is high (J. Glos, unpubl. data). Additionally, retarded growth increases the time span where larvae are vulnerable to gape-limited Amphibian communities in Madagascar Julian Glos 97 predators (Cronin and Travis, 1986) and predators that are more effective when preying on smaller individuals (e.g., Semlitsch, 1990). The ability to detect the presence of (interacting) conspecifics in potential oviposition sites and to adjust the breeding activities accordingly should therefore be selected for. In A. laticeps, the presence of conspecifics was indeed a significant factor in oviposition site selection. We found less egg clutches in ponds with an introduced set of conspecific tadpoles and reproduction took place later. Strikingly, these ponds were only selected after most of the other, initially tadpole-free ponds, had already been used for reproduction during the course of the experiment, and thus there was no longer the choice of ponds without conspecific tadpoles. As tadpoles of A. laticeps are not known to be cannibalistic (Glos and Linsenmair, 2004), there is no reason to assume that the observed behavior is a consequence of predator avoidance. It is more likely that A. laticeps tadpoles therewith do avoid competition by larger conspecific tadpoles. The influence of behavioural variables on observed distribution patterns can often only inadequately be addressed in descriptive studies, and experimental studies such as the one presented here are needed to elucidate the true relationships. Through this, experimental studies expand the explanatory and predictive power of habitat models, and their conclusions must be incorporated in the recommendations for species management and conservation derived from the respective habitat models. In this respect, one main objective resulting from this study is to emphasise the fact that selected oviposition sites found in the field need not necessarily represent the optimal conditions. In our case, the avoidance of ponds already colonized by conspecifics might lead to an evasion of these pools, irrespective of whether they offer the best biotic and abiotic conditions in other respects (their first choice) or not, and subsequently to the selection of oviposition sites that are suboptimal in respect to at least some of these conditions (their second choice). Therefore, the second choice might suddenly become first choice, which is not readily seizable by the observer. Ultimately, this conclusion becomes significant in conservation practice, e.g., when judging habitat quality or creating artificial breeding habitat.

The females select the oviposition site Male frogs did not behave differentially between the different pond types. However, more clutches were deposited in tadpole-free ponds. This clearly indicates that the females select the oviposition site in A. laticeps, as it is the case in most anuran species (e.g., Rudolf and Rödel, 2005). Due to the poor light conditions during the exclusively nocturnal activities of A. laticeps and the tadpoles’ cryptic appearance in respect to coloration and microhabitat use (Glos and Linsenmair, 2004), we consider it unlikely that tadpole presence is visually or mechanically detected by the females. More likely, chemical cues are used for the detection of conspecific tadpoles in the water. Several frog species are known to proximately assess the colonization history of potential oviposition sites via chemical cues (e.g., Hoplobatrachus occipitalis, Spieler and Linsenmair, 1997). Furthermore, behavioural reactions Amphibian communities in Madagascar Julian Glos 98 to water-borne chemical cues are known from a closely related species (Aglyptodactylus securifer; chapter 9), arguing for the utilization of chemical cues also in A. laticeps.

CONCLUSIONS

This study highlights the necessity to combine descriptive habitat models with experimental studies for full explanatory and predictive power. Furthermore, it emphasizes the importance of the colonization history for oviposition site selection in temporary waters. In our system, the combination of the descriptive habitat model and the experimental approach proved indispensable for sensible conservation efforts of the endangered A. laticeps. Having investigated the general dependencies responsible for the observed pattern of occurrences, predictions can be made on the probability of occurrence of the species of interest in further habitats without extensive zoological investigations by measuring these variables (Schröder, 2000; Morrison et al., 1998; Fielding and Haworth, 1995). Accordingly, our results can be used to identify suitable habitat for A. laticeps within and outside the Menabe region to predict species presence or absence in formerly unknown habitat, and to predict the effects of habitat alteration on the survival prospects of this species. Lastly, the extraordinary colonization ability of A. laticeps provides the opportunity to use artificial breeding sites as a promising management tool in the conservation of this frog.

Amphibian communities in Madagascar Julian Glos 99

LITERATURE CITED

ALFORD, R. A., and H. M. WILBUR. 1985. Priority effects in experimental pond communities: competition between Bufo and Rana. Ecology. 66:1097-1105.

ANDERSON, P., and J. LÖFQVIST. 1996. Asymetric oviposition behaviour and the influence of larval competition in the two pyralid moths Ephestia kuehniella and Plodia interpunctella. Oikos. 76:47-56.

ANDREONE, F., J. E. CADLE, N. A. COX, F. GLAW, R. A. NUSSBAUM, C. J. RAXWORTHY, S. N.

STUART, D. VALLAN, and M. VENCES. 2005. Species review of amphibian extinction risks in Madagascar: conclusions from the global amphibian assessment. Conservation Biology. 19:1790-1802.

ARAÚJO, M. B., P. H. WILIAM, and R. J. FULLER. 2002. Dynamics of extinction and the selection of nature reserves. Proceedings of the Royal Society Biological Sciences Series B. 269:1971- 1980.

BERVEN, K. A. 1988. Factors affecting variation in reproductive traits within a population of wood frogs (Rana sylvatica). Copeia. 1988:605-615. —. 1990. Factors affecting population fluctuations in larval and adult stages of the wood frog (Rana sylvatica). Ecology. 71:1599-1608.

BLAUSTEIN, L., and J. MARGALIT. 1996. Priority effects in temporary pools: nature and outcome of mosquito larva - toad tadpole interactions depend on order of entrance. Journal of Animal Ecology. 65:77-84.

CRONIN, J. T., and J. TRAVIS. 1986. Size-limited predation on larval Rana areolata (Anura: Ranidae) by two species of backswimmer (Insecta: Hemiptera: Notonectidae). Herpetologica. 42:171- 174.

CRUMP, M. L. 1991. Choice of oviposition site and egg load assessment by a treefrog. Herpetologica. 47:308-315.

DUNN, O. J. 1964. Multiple comparisons using rank sums. Technometrics. 6:241-252.

EDGERLY, J. S., M. MCFARLAND, P. MORGAN, and T. LIVDAHL. 1998. A seasonal shift in egg-laying behaviour in response to cues of future competition in a treehole mosquito. Journal of Animal Ecology. 67:805-818.

FIELDING, A. H., and P. F. HAWORTH. 1995. Testing the generality of bird-habitat models. Conservation Biology. 9:1466-1481.

GLAW, F., M. VENCES, and W. BÖHME. 1998. Systematic revision of the genus Aglyptodactylus Boulenger, 1919 (Amphibia: Ranidae), and analysis of its phylogenetic relationships to other Amphibian communities in Madagascar Julian Glos 100

Madagascan ranid genera (Tomopterna, Boophis, Mantidactylus and Mantella). Journal of Zoological Systematics and Evolutionary Research. 36:17-37.

GLOS, J. 2003. The amphibian fauna of the Kirindy dry forest in western Madagascar. Salamandra. 39:75-90.

GLOS, J., and K. E. LINSENMAIR. 2004. Description of the tadpoles of Aglyptodactylus laticeps and A. securifer from Western Madagascar, with Notes on Life History and Ecology. Journal of Herpetology. 38:131-136.

GLOS, J., and A. T. VOLAHY. 2004. An inventory of amphibians and reptiles in the Central Menabe region. Report to DWCT (Durrell Wildlife Conservation Trust), DPZ (Deutsches Primatenzentrum), and FANAMBY, Antananarivo.

GOSNER, K. L. 1960. A simplified table for staging anuran embryos and larvae with notes on identification. Herpetologica. 16:183-190.

HOLM, S. 1979. A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics. 6:65-70.

HUTH, C. J., and O. PELLMYR. 1999. Yucca moth oviposition and pollination behaviour is affected by past flower visitors: evidence for a host-marking pheromone. Oecologia. 119:593-599.

JOHNSON, J. R., and R. D. SEMLITSCH. 2003. Defining core habitat of local populations of the gray treefrog (Hyla versicolor) based on choice of oviposition site. Oecologia. 137:205-210.

LAMPERT, K. 2001. Alternative life history strategies in the West African reed frog, Hyperolius nitidulus. Würzburg University, Würzburg.

MCDIARMID, R. W., and R. ALTIG. 1999. Tadpoles: The Biology of Anuran Larvae. The University of Chicago Press, Chicago.

MORIN, P. J., S. P. LAWLER, and E. A. JOHNSON. 1990. Ecology and breeding phenology of larval Hyla andersonii: the disadvantages of breeding late. Ecology. 71:1590-1598.

MORRISON, M. L., B. G. MARCOT, and R. W. MANNAN. 1998. Wildlife-habitat relationships - concepts and applications. University of Wisconsin Press, Madison.

NELSON, R., and N. HORNING. 1993. AVHRR-LAC estimates of forest area in Madagascar, 1990. International Journal of Remote Sensing. 14:1463-1475.

NEWMAN, R. A. 1987. Effects of density and predation on Scaphiopus couchii tadpoles in desert ponds. Oecologia. 71:301-307. —. 1998. Ecological constraints on amphibian metamorphosis: interactions of temperature and larval density with responses to changing food level. Oecologia. 115:9-16.

REICH, P., and B. J. DOWNES. 2003. Experimental evidence for physical cues involved in oviposition site selection of lotic hydrobiosid caddis flies. Oecologia. 136:465-475.

RESETARITS, W. J., and H. M. WILBUR. 1989. Choice of oviposition site by Hyla chrysocelis: role of predators and competitors. Ecology. 70:220-228. Amphibian communities in Madagascar Julian Glos 101

RUDOLF, V. H. W., and M.-O. RÖDEL. 2005. Oviposition site selection in a complex and variable environment: the role of habitat quality and conspecific cues. Oecologia. 142:316-325.

SACHS, L. 1984. Angewandte Statistik. Springer, Berlin Heidelberg.

SCHRÖDER, B. 2000. Zwischen Naturschutz und theoretischer Ökologie: Modelle zur Habitateignung und räumlichen Populationsdynamik für Heuschrecken im Niedermoor, p. 1-256. In: Landschaftökologie und Umweltforschung 35. Technical University of Braunschweig, Braunschweig.

SEMLITSCH, R. D. 1987. Density-dependent growth and fecundity in the paedomorphic salamander Ambystoma talpoideum. Ecology. 68:1003-1008. —. 1990. Effects of body size, sibship, and tail injury on the susceptibility of tadpoles to dragonfly predation. Canadian Journal of Zoology. 68:1027-1030.

SEMLITSCH, R. D., and H. REYER. 1992. Modification of anti-predator behaviour in tadpoles by environmental conditioning. Journal of Animal Ecology. 61:353-360.

SMITH, D. C. 1987. Adult recruitment in chorus frogs: effects of size and date at metamorphosis. Ecology. 68:344-350.

SORG, J.-P., J. U. GANZHORN, and P. M. KAPPELER. 2003. Forestry and research in the Kirindy Forest/Centre de Formation Professionelle Forestrière, p. 1709. In: The Natural History of Madagascar. S. M. Goodman and J. P. Benstead (eds.). The University of Chicage Press, Chicago and London.

SPIELER, M., and K. E. LINSENMAIR. 1997. Choice of optimal oviposition site by Hoplobatrachus occipitalis (Anura: Ranidae) in an unpredictable and patchy environment. Oecologia. 109:184- 199. —. 1998. Migration patterns and diurnal use of shelter in a ranid frog of a West African savannah: a telemetric study. Amphibia-Reptilia. 19:43-64.

VENCES, M., and F. GLAW. 2001. When molecules claim for taxonomic changes: new proposals on the classification of Old World treefrogs. Spixiana. 24:85-91.

WILBUR, H. M. 1977. Density-dependent aspects of growth and metamorphosis in Bufo americanus. Ecology. 58:196-200. —. 1987. Regulation of structure in complex systems: experimental temporary pond communities. Ecology. 68:1437-1452.

WILBUR, H. M., and R. A. ALFORD. 1985. Priority effects in experimental pond communities: responses of Hyla to Bufo and Rana. Ecology. 66:1106-1114.

Amphibian communities in Madagascar Julian Glos 102

Chapter 8

Mixed-species social aggregations in Madagascan tadpoles – determinants and species composition

Julian Glos1, Kathrin H. Dausmann2, K. Eduard Linsenmair1

ABSTRACT

The frequency, species composition and determinants of mixed-species tadpole aggregations were analysed under natural conditions in a dry forest of western Madagascar. Most aggregations (73 %) were formed by more than one species, with up to four species per individual aggregation. Dyscophus insularis (Microhylidae) and Aglyptodactylus securifer (Mantellidae) were the most abundant species in these aggregations. Using a logistic regression model we analysed to what extent the presence and absence of aggregations in a pond can be predicted by its biotic and abiotic habitat variables. Aggregations are more likely to occur in profound ponds with low cover of leaf litter and with clear water, while the overall density of invertebrate predators in the pond seems to play a minor role. Our observations suggest that the formation of mixed-species aggregations in tadpoles of the Kirindy Forest in our system is primarily a reaction to vertebrate predators. Therefore, aggregation behaviour may play a key role in the ability of several anuran species to utilize these spawning waters that are unpredictable in regard to the presence of fish and other vertebrate predators.

Amphibian communities in Madagascar Julian Glos 103

INTRODUCTION

The phenomenon that animals temporarily group together is found in many animal taxa. Groups are termed social when ‘individuals actively seek the proximity of each other instead of co-occurring in the same spot because of an attraction to the same environmental condition’ (the latter is termed non- social aggregation) (Krause and Ruxton, 2002). Different benefits can accrue from forming social aggregations. Among others animals may become more efficient in foraging (Foster, 1985), save water via reduced evaporation (Cook, 1981) and energy by reduced investment into thermoregulation (Arnold, 1988), reduce the energetic costs of movement (Abrahams and Colgan, 1985), may have higher reproductive success (Alatalo et al., 1992) or be better protected from predators (Riipi et al., 2001). In tadpoles, social aggregative behaviour is found in several anuran families from different geographic regions: Microhylidae (Africa), Bufonidae (Africa, N-America, Europe), Pipidae (Africa), Pelobatidae (N-America), Ranidae (Europe, N-America) and Leptodactylidae (S-America). It is mainly seen as a means of reducing the high predation pressures most species are confronted with in their diverse habitats. In the field, there is a good correlation between the formation of tadpole aggregations and the direct presence of predators (Rödel and Linsenmair, 1997; Watt et al., 1997) and/or chemical compounds released by the predators or by injured tadpoles (Kats and Dill, 1998; Spieler and Linsenmair, 1999). However, also abiotic variables are thought to promote and to modulate the formation of tadpole aggregations, e.g., water transparancy (Spieler, 2001), temperature (Brattstrom, 1962), and daytime (Beiswenger, 1977; Spieler, 2001). Studies on tadpole aggregative behaviour have so far focused on single-species aggregations, and in addition often on aggregations consisting of related individuals (e.g., Waldman, 1991). However, further intriguing questions on the reason for their existence and on costs and benefits for the single group members arise when investigating mixed-species aggregation, as done in this study. This is in particular true when group members are ecologically dissimilar in respect to feeding and microhabitat preferences. Diamond (1981) proposes five theories to explain the existence of mixed- species groups. Being in a large group can be an anti-predator device on the one hand (convoy theory), or may, on the other hand, permit group members access to additional resources (gang theory, beater theory, pirate theory, feeding efficiency theory). The Kirindy dry forest in western Madagascar represents a highly dynamic and heterogeneous habitat for amphibians concerning pond persistence and predator presence. 15 amphibian species breed in forest ponds and in rock-pools of the riverbed before the river is running. A maximum of eight tadpole species can be found syntopically in the same pond (Glos, 2003). Within these water bodies, temporary, mobile mixed-species tadpole aggregations can be regularly observed. In this study, we aim to present an extensive analysis of mixed-species tadpole aggregations in the stagnant, ephemeral waters of this forest. We examined the frequency of aggregations under natural conditions and their species composition. Furthermore, we analysed to what extent the Amphibian communities in Madagascar Julian Glos 104 presence of aggregations in a pond can be predicted by its biotic and abiotic habitat variables using a logistic regression model.

MATERIALS AND METHODS

Study site The Kirindy Forest is a deciduous dry forest at the west coast of Madagascar, 60 km north of Morondava and about 20 km inland (44°39’ O, 20°03 S; 18 - 40 m above sea level; Sorg and Rohner, 1996). The area of the Kirindy Forest covers about 12.000 ha and thus has the dubious reputation of being one of the largest remaining continuous forests in western Madagascar (Nelson and Horning, 1993). The forest is intersected by the Kirindy River and is surrounded by a tree-shrub savanna of anthropogenic origin. The climate is characterized by a marked seasonality. Almost all rain falls in the austral summer from November to March, followed by eight months of virtually no rain (annual mean rainfall: 800 mm; Sorg and Rohner, 1996). The study was conducted in the rainy seasons (Nov – Feb) of 1999/00 and 2000/01. Breeding sites for amphibians exist in three different habitat types: the closed forest, the bed of the Kirindy river (before the river is running), and the surrounding savannah. All breeding ponds apart from a few pools in the riverbed dry out completely during the dry season. They are successively filled with the onset of rain in November and December. The first waters that are used by amphibians for spawning usually form in the rocky parts of the Kirindy riverbed. As a rule, these ponds offer a low risk of desiccation and a low density of invertebrate predators (unpubl. data). During the course of the rainy season, these waters eventually merge and fish immigrate from the few permanent pools in the river bed, exerting then a very high predation pressure on the amphibian larvae. Usually in December, the first breeding ponds arise in the closed forest. Depending on amount and distribution of precipitation and habitat parameters of the ponds themselves, the persistence of these waters varies from three days to five months.

Frequency and composition of aggregations Tadpole aggregations were defined as follows: at least four individuals were not more than 2 cm apart from each other (‘dense” aggregations sensu O'Hara and Blaustein, 1985), and this grouping behaviour persisted for more than 100 sec. Ponds where tadpole density was so high that we could not differentiate between social aggregations and aggregations forced by diminishing water volume due to desiccation were excluded from the analysis. To examine aggregation frequency under natural conditions, 126 ponds were visually checked for the presence and absence of tadpole aggregations, respectively. Depending on pond properties (e.g., underground substrate, exposition, water depth) and the extent of phases with low precipitation, Amphibian communities in Madagascar Julian Glos 105 some study ponds dried out during the rainy season. When they were refilled after rainfalls, they were repeatedly used for breeding by amphibians, leading to a new composition of tadpoles each time. On account of this, each pond was checked for aggregation presence after the initial filling of the pond, and additionally after each cycle of drying-out and refilling. Each control check was done between one and three weeks after the filling and each refilling, respectively. In total, each pond was checked between one and ten times (2.35 ± 1.66; mean ± SD; npond controls = 296) over two field seasons (1999/2000 and 2000/2001). The inspection was done by carefully pacing along the margin and observing the pond to a depth of 15 cm below the water surface. In larger ponds, additional observations were done from within the pond by carefully crossing the pond. We did not find any indications that this method affected the aggregation behaviour of the tadpoles in any way. Minimum observation time was 5 min, maximum observation time (for large ponds) was 20 min. Observations were done exclusively during daylight between 9:00 a.m. and 4:00 p.m. This method was chosen because it could be applied in all ponds irrespective of pond size and water clarity. To determine species composition and aggregation size in the field, aggregations were dip- netted in the closed forest and the riverbed. Tadpole number of 12 randomly selected aggregations and species affiliation of each tadpole of 27 randomly selected aggregations was determined. Each aggregation was taken from a different pond. Tadpole number in small aggregations was determined by dip-netting the complete swarm in one dip-net stroke (dip-net dimensions: 30 x 30 x 30 cm). In large aggregations, not all individual tadpoles of the aggregation could be captured in one dip-net stroke. In such cases, one dip-net stroke was performed and the tadpoles captured therein were counted, the relative proportion of this dip-net stroke on total aggregation volume was estimated, and subsequently total aggregation size was calculated from these two parameters. The average number of individuals that were caught in one of these dip-net strokes and analysed, and that was therefore taken as the basis for these calculations, was 148 ± 90 (range 29 – 351; ntotal = 3,995).

Predictability of aggregations by habitat variables In order to extract the key (biotic and abiotic) habitat factors that predict the presence of aggregations in the field we constructed habitat models using multiple logistic regression (Hosmer and Lemeshow, 1989). By a preliminary data survey, we eliminated high collinearity within the environmental variables (exclusion of variables in cases of Spearman rho ≥ 0.7, see suggestions by Fielding and Haworth, 1997). Using a backward stepwise logistic regression (LR – method) we included binary presence/absence (of aggregations) data, coded as 1 and 0, as the dependent variable. In order to measure classification accuracy of the models irrespective of threshold criteria, Receiver Operating Characteristics (ROC) – plots were constructed, and Area Under Curve (AUC) – values were calculated providing a single quantitative index of the diagnostic accuracy of the model (Zweig and Campbell, 1993). We included eight habitat variables as independent variables in the initial model (Table 1). Habitat variables were recorded parallel to the control checks for the presence of Amphibian communities in Madagascar Julian Glos 106

TABLE 1. Definition and data acquisition of independent habitat variables.

Variable Variable definition Unit Methods name Pond size Absolute surface area of m2 Calculated from pond length, pond when maximum width, and shape water-filled Pond depth Maximum pond depth when cm Measured at the deepest point pond maximum water-filled in the pond Water Transparency of the pond 5 categories The visibility of a metal stick transparency water influenced by I (low) to V (high) was estimated in a depth of 15 dissolved and suspensed cm at randomly chosen points matters within the pond and averaged Leaf litter Relative area of pond 6 categories Measured at 12 plots of 0.25 bottom covered with dead I: 0 %, II: 1 – 20 %, m2 within the pond and leaves III: 21 – 40 %, IV: averaged 41 – 60 %, V: 61 – 80 %, VI: 81 – 100 % Surface Relative area of pond 6 categories; see Estimated water plants surface covered with plants Leaf litter (mainly Nymphea lotus, Salvinia sp.) Submerged Relative volume of pond 6 categories; see Measured at 12 plots of 0.25 plants filled with submersed water Leaf litter m2 within the pond and plants (mainly averaged Lagarosiphon madagascariensis, Najas sp.) Invertebrate Density of invertebrate 5 categories Measured at 12 plots of 0.25 predators predators, including: I (low) to V (high) m2 within the pond using dip- Dytiscid beetles > 0.5 cm netting and the box-method Larval dytiscid beetles > 1 (Heyer et al., 1994) and cm averaged; determined to higher Water bugs taxonomic levels in the field (Belostomatidae) > 0.5 cm and subsequently released Water scorpions (Nepidae) > 1 cm Larval dragonflies (Anisoptera) > 0.5 cm Larval damselflies (Zygoptera) > 2 cm Fish Presence of fish (mainly 2 categories Observation, dip-netting Oreochromis sp.) (presence / absence)

aggregations. When a pond was checked for aggregations more than once and tadpole aggregations were never found, we calculated the second order mean of the habitat variables measured during all control checks. When a pond was checked for aggregations more than once and we did find tadpole aggregations, we calculated the second order mean of the habitat variables of only the control checks when aggregations were present. Amphibian communities in Madagascar Julian Glos 107

To show up differences in environmental characteristics between ponds with aggregations and those without, we tested whether ponds, where we found aggregations, differed in respect to their habitat variables from those without aggregations. We used the same habitat variables as in the logistic regression and applied non-parametric statistics. In this analysis, we compared the second order means measured during all control checks in both groups.

RESULTS

Frequency and composition of aggregations In 23.9 % of all visual control checks (n = 296) at least one tadpole aggregation was found. Considering individual ponds, in 49 (38.8 %) of 126 study ponds tadpole aggregations were present at least once. The aggregations varied in size and species composition. The number of individuals in one connected tadpole swarm ranged from 200 to more than 5.000 individuals (1060.0 ± 1386.1; mean ± SD; n = 12). Tadpoles of six different anuran species were found aggregating, with a maximum of four different species within one particular aggregation. A species was defined as being present in an aggregation, when at least one individual of that species was found in the aggregation. 27 % of the aggregations consisted of one species, 37 % of two species, 22 % of three species, and 15 % of four species (n = 27). The tadpoles found aggregating are members of two anuran families: Mantellidae and Microhylidae. Among those, Aglyptodactylus securifer (Mantellidae) and Dyscophus insularis (Microhylidae) were present in most of the aggregations and were by far the most abundant species (Table 2).

TABLE 2. Composition of tadpole aggregations in the field. Total number = number of individuals that were captured and determined to species level; Presence = proportion of aggregations with the respective species present (n = 27); Frequency = mean frequency of the species in the particular aggregation sample (n = 27); Man = Mantellidae; Mic = Microhylidae.

Species Total number Presence (%) Frequency (%) Aglyptodactylus securifer Man 2.094 88.9 56.4 Dyscophus insularis Mic 1.487 70.4 35.3 Laliostoma labrosum Man 397 40.7 8.0 Boophis doulioti Man 4 7.4 0.2 Scaphiophryne calcarata Mic 9 11.1 0.1 Scaphiophryne brevis Mic 4 7.4 0.1

Amphibian communities in Madagascar Julian Glos 108

Predictability of aggregations by habitat variables A multiple logistic regression model was constructed for the occurrence of tadpole aggregations. The independent variables were tested for normal distribution and six of them (pond size, maximum pond depth, dead leaf cover, underwater vegetation, floating water plants, water transparency) were BoxCox-transformed (Box and Cox, 1964). The parameters of the initial and of the final model are listed in Table 3. Their importance for the model predictions is reflected by the correlation coefficient (r). In this analysis, only a subset of 70 ponds was included. These ponds were chosen on the basis of their geographic location and all constitute breeding waters within a 3 km2 area around the field camp, and on the basis of practical reasons, i.e. where we recorded all habitat variables.

TABLE 3. Logistic regression analysis of tadpole aggregations. Initial model: n = 70, χ2 = 19.33, Nagelkerke r2 = 0.35, p = 0.007; Final model (stepwise backwards analysis): n = 70, χ2 = 18.72, Nagelkerke r2 = 0.34, AUC = 0.75, p < 0.001. Coefficient = regression coefficient B.

Variables Coefficient SE Wald P Exp(B) Initial logistic regression model Pond size 0.132 0.338 0.152 0.696 1.141 Pond depth 0.538 0.348 2.387 0.122 1.713 Dead leaf cover -0.453 0.214 4.481 0.034 0.636 Underwater vegetation -0.997 3.658 0.074 0.785 0.369 Floating water plants 0.039 3.369 0.000 0.991 1.039 Water transparency -4.259 1.807 5.557 0.018 0.014 Invertebrate predators -0.178 0.290 0.377 0.539 0.837 Constant -0.584 1.480 0.156 0.693 0.557 Final logistic regression model Pond depth 0.588 0.276 4.533 0.033 1.801 Dead leaf cover -0.465 0.204 5.205 0.023 0.628 Water transparency -4.378 1.627 7.244 0.007 0.013 Constant -0.939 1.352 0.483 0.487 0.391

Four variables were backwards and stepwise eliminated from the model: Floating water plants (step 2), Underwater vegetation (step 3), Pond size (step 4), and Invertebrate predators (step 5). In the final model, the observed and predicted presence of tadpole aggregations rises with increasing pond depth and increasing water transparency. It decreases with cover of dead leaves at the bottom of the pond. The overall percentage of correctly predicted presences and absences is high with 80.0 % (for threshold value P = 0.5). The AUC – value of 0.75 indicates a good reliability to give correct predictions for all possible classification thresholds. The binary variable Fish was not included in the model because all 70 ponds analysed in the model were fish-free ponds. Out of a total of 126 ponds in the area only three contained fish. These ponds were exclusively situated in the Kirindy riverbed. Fish, in particular cichlids (Oreochromis sp.) are obviously causing tadpoles to aggregate. In two instances we observed fish newly immigrating into formerly fish free pools using a small water bridge that had formed after a heavy rainfall. Within less than 30 seconds, the majority of the tadpoles that had been randomly distributed before formed dense Amphibian communities in Madagascar Julian Glos 109 aggregations (JG, G. Erdmann, unpubl. data). Our observations suggest that these aggregations were formed before the fish actually fed on the tadpoles, and therefore before an alarm substance (“Schreckstoff”) could have been released. We found aggregations over a wide range of sizes and depths of ponds. However, ponds with aggregations were significantly deeper, had greater Water transparency, and less Leaf litter (Table 4). This is consistent with the results from the logistic regression which states that these variables play an important role in the probability of the formation of an aggregation.

TABLE 4. Descriptive data and differences in habitat variables between ponds with aggregations and those without. 25%-75% = percentiles. As not all variables could be recorded for all ponds, N differs between the analyses.

Without aggregations With aggregations Median 25%-75% N Median 25%-75% N Z P Pond size (m2) 9.7 3.0-35.0 68 20.0 6.7-47.5 45 -1.8 0.07 Pond depth (cm) 20.0 11.2-32.2 64 33.5 21.0-40.5 45 -3.3 0.001 Water transparency (cat) 1.5 1.0-2.5 67 1.0 1.0-1.4 48 -3.0 0.003 Leaf cover (cat.) 4.0 3.0-5.0 64 3.6 2.9-4.0 48 -2.2 0.02 Surface water plants (%) 0.0 0.0-0.0 66 0.0 0.0-0.0 49 -0.7 0.50 Submerse plants (cat.) 0.0 0.0-0.1 67 0.0 0.0-0.2 41 -0.1 0.93 Invertebrate predators (cat.) 2.7 2.0-3.5 52 2.5 1.5-2.8 19 -1.3 0.19

General observations Tadpole swarms were found in forest ponds as well as in stagnant rock pools of the riverbed, both around the ponds’ edges and in the ponds’ centres. The aggregations were mobile, three-dimensional swarms with the upper part of the swarm usually right below the water surface (range 5 – 80 cm, median = 5 cm, n = 10). Swarm members were observed to face in the same direction while swimming slowly, being 1 to 3 cm apart from each other. The swarms were temporary. They persisted between 30 min and several hours, and they were usually diurnal. The maximum extent of an aggregation was over 5 m in length, 1.5 m in width, and 0.3 m in height. The relative volume of the tadpole aggregation in relation to that of the respective pond depended on the water level. It was usually very small (<< 3 %) in ponds that were completely water filled. However, it could attain more than 20 % in ponds that were nearly dried up. In these ponds social aggregations were not distinguishable from non-social aggregations, or from stochastic aggregations promoted by high tadpole densities, and such aggregations were excluded from the analysis. In general, tadpoles of all sizes and developmental stages were found aggregating. However, within one aggregation, most tadpoles were about the same size and developmental stage irrespective of species affiliation. This was the case even though tadpoles of different sizes and developmental stages were usually present at the same time in the respective ponds. In D. insularis, the majority of tadpoles within aggregations were either in an early or in a late stage of development (Fig. 1). In A. Amphibian communities in Madagascar Julian Glos 110 securifer, on the other hand, most tadpoles in aggregations were at an early stage in their development. Tadpoles shortly before metamorphosis (Gosner 40-44) were not found aggregating.

100

ations 60 g re

gg 40

20 % of a 0 % of aggregations 25 - 29 30 - 34 35 - 39 40 - 44 Stage class

FIG. 1. Relative frequency of aggregations with D. insularis (grey bars, N = 19, n = 1.487) and A. securifer (black bars, N = 9, n = 737) tadpoles in four developmental stage classes (after Gosner, 1960). An aggregation was assigned to a stage class when > 50 % of tadpoles had been in that stage class. N = number of aggregations, n = number of tadpoles.

DISCUSSION

Surprisingly few studies have so far addressed questions on the range and frequency of con- and heterospecific group sizes under natural conditions (Krause and Ruxton, 2002). The available data are mainly restricted to fish (Seghers, 1981; Bonabeau and Dagorn, 1995) and large mammals (Lott and Minta, 1983; Wirtz and Löscher, 1983). In the Kirindy Forest, mobile, social aggregations of tadpoles were frequently observed. The tadpole aggregations consisted of up to several thousand individuals, and thus are paralleled in size only by large Bufo-aggregations (Beiswenger, 1975; Wassersug et al., 1981; Waldman, 1991), and being much larger than those of e.g., the microhylid Phrynomantis microps (range 6 – 580; Spieler, 2003) or north American ranids (e.g., 4 – 40 in Rana cascadae; O'Hara and Blaustein, 1985). The aggregations resembled “true schools”, in the sense ichthyologists use this term, as they not only aggregated but also exhibited parallel orientation and preferential bearing to nearest neighbours (Wassersug et al., 1981). Tadpoles of a broad range of sizes and developmental stages were found in aggregations, although within one aggregation most tadpoles were about the same size and developmental stage. Accordingly, we found aggregations of different size classes simultaneously in one pond (JG, unpubl. data). Associations of size-matched individuals is known from numerous studies, mainly on fish, and both in con- and heterospecific groups (reviewed by Ranta et al., 1994). There, size-assorted shoaling is explained by fitness gains of individual group members, either in the context of predation (e.g., through the ’oddity effect’; Landeau and Terborgh, 1986) or foraging (Ranta et al., 1994). In contrast to these active mechanisms based on decisions of Amphibian communities in Madagascar Julian Glos 111 individuals, size-assortment in fish, as in tadpoles, might be the result of passive sorting mechanisms based e.g., on differential swimming speeds (Gueron et al., 1996) or differential habitat preferences between different-sized individuals. Continuative, experimental studies are needed to disentangle the mechanisms that are responsible for the observed pattern. Aggregation size varied between ponds. Group size is therefore not fixed, but varies with biotic and/or abiotic factors (Spieler, 2003), or might be directly related to the total number of tadpoles within a respective pond. Furthermore, the result that aggregations differed in the size of their individual members indicates that there is no clear-cut ontogenetic shift in sociality, as found by Butler (1999) and Ratchford and Eggleston (1998) in lobsters. However, our data indicate that the readiness to shoal might decrease with ongoing development in Aglyptodactylus securifer, in contrast to Dyscophus insularis where we found as many aggregations consisting of tadpoles in later stages as in younger stages. The ultimate reason for the tadpoles to aggregate might be predator avoidance, as is generally assumed (e.g., Rödel and Linsenmair, 1997; Watt et al., 1997; Kats and Dill, 1998). This is supported by the observation that tadpoles instantly aggregate when predatory fish enter formerly fish-free ponds, even before the first tadpoles were eaten by the fish and alarm substances could possibly have been released (JG, G. Erdmann, unpubl. data). Aggregation behaviour might therefore play a key role for the ability of amphibians to utilize predator-rich spawning waters (Spieler and Linsenmair, 1999). The density of invertebrate predators seems to be less important for the probability of aggregation formation. However, the logistic regression model shows that the threshold of aggregation formation seems to be also influenced by abiotic factors (Pond depth), structures inside the ponds (Leaf litter), and water characteristics (Water transparency). This is notably supported when ponds with and without tadpole aggregations were directly compared. Aggregations were more likely to be found in more profound ponds, with low cover of leaf litter, and high water transparency. Animals tend to prefer structurally complex microhabitats as a reaction to high predation risk, often combined with a reduction in activity (Kats et al., 1988; Magurran and Higham, 1988). Therefore, the absence of sufficient cover against potential predators by leaf litter on the ground might facilitate the change of strategy of tadpoles to seek protection in large, conspicuous groups. Furthermore, low visibility in the water, either caused by water muddiness or by diminishing daylight, is known to negatively affect aggregation size and density, and will eventually lead to its dissolution (Beiswenger, 1977; Rödel and Linsenmair, 1997; Spieler, 2001), as found in our study. The majority of the most important predators of tadpoles are known to hunt visually (JG, unpubl. data; Rödel and Linsenmair, 1997; Rödel, 1999). Therefore, predation risk most likely decreases when prey detectability is low. Thus the high costs in respect to growth and development of being in an aggregation (Spieler and Linsenmair, 1999) do no longer pay when the muddiness of the water increases above a certain level and the benefits of aggregating vanish (Siegfried and Underhill, 1975; Watt et al., 1997). Thus for tadpoles in muddy water high costs in respect to growth and development might outweigh decreasing benefits of being in Amphibian communities in Madagascar Julian Glos 112 an aggregation as soon as a critical level of turbidity is reached. Since the water transparency in ponds often changes quickly due to disturbance by animals or recent rainfall, the flexible behaviour of aggregation formation seems to be an excellent adaptation to the heterogeneous environment of this habitat. Three out of four aggregations in the ponds of the Kirindy Forest consisted of two or more species, with three species being abundant in those aggregations and further three species being present only sporadically. The two most abundant species were Dyscophus insularis (Microhylidae) and Aglyptodactylus securifer (Mantellidae). To our knowledge this is the first detailed study on mixed-species aggregations in tadpoles (see Vences and Glaw, 2003, for an observational note). Dyscophus insularis and A. securifer were frequently found together in one aggregation, despite their ecological dissimilarity. The tadpoles of Dyscophus insularis are filter-feeding, and completely microphageous. They are generally found in all vertical positions of the water column and usually float horizontally in midwater (JG, unpubl. data; Glaw and Vences, 1994). Aglyptodactylus securifer tadpoles, on the other hand, feed on plant and/or microbial matter of detritus. They are usually only found on the bottom of the pond (Glos and Linsenmair, 2004). It is therefore likely that possible costs of aggregative behaviour are asymmetrical between the two species. As aggregations were usually found right below the water surface, tadpoles of A. securifer obviously changed their preferred microhabitat to join these aggregations. Therefore, they presumably bear the costs of a greatly reduced food input, while D. insularis as a filter feeder can continue feeding. However, food input by D. insularis tadpoles might also be reduced as a consequence of food depletion by other group members, in particular for those tadpoles that stay in the swarm centre or in the back part of the swarm. The position of the aggregations beneath the water surface might be an adaptation to reduce the predation risk by those fish that are preferably foraging at the bottom or in the middle of the water column. Moreover, only the bright ventral side of the tadpole is exposed to aquatic predators, when swimming close to the water surface. This might camouflage the tadpoles against the sunlight gleaming onto the water surface and further reduce predation risk by visual hunters as most fish are. On the other hand, this behaviour might render the tadpoles more conspicuous for species foraging from outside the pond, in particular kingfishers. However, kingfishers only consume a negligible proportion of the tadpoles whereas fish usually quantitatively eliminate tadpoles in a pond. The predation pressure by kingfisher will in any case be smaller than that by fish. Benefits for group members within a social mixed-species aggregation may differ (Diamond, 1981; Wolters and Zuberbühler, 2003). They may have access to resources within a territory whose owner would expel solitary competitors (“gang theory”; Diamond, 1981). Group members may facilitate feeding of other group members in the course of their own foraging (“beater theory”). Group members may seize food that another group member has caught (“pirate theory”). Group members benefit in terms of foraging from being in a large group, even when all species have similar diets, by Amphibian communities in Madagascar Julian Glos 113 raising the probability of finding a good feeding patch or by learning new foraging techniques from heterospecifics (“feeding efficiency theory”). These theories all predict that one or more species within a mixed-species group can raise their food input by the presence of other species. However, in tadpole aggregations of Kirindy Forest, this is not very likely. First, competing for food in these tadpoles is exploitative rather than direct (excluding gang and pirate theory). Second, as their feeding ecology is very different, they most unlikely profit from the presence of other species as the beater theory and the feeding efficiency theory suggest. However, the formation of large, temporal aggregations including more than one species may be an adaptation to reduce predation risk (“convoy theory”; Diamond, 1981; Gibson et al., 2002). In several tadpole species, this is seen as the main ultimate cause for social aggregative behaviour (e.g., Rödel and Linsenmair, 1997; Spieler and Linsenmair, 1999; Spieler, 2003). Within mixed-species aggregations, all group members may benefit from mere group-size related anti-predator benefits, such as the dilution effect (Hamilton, 1971), confusion effect (Heller and Milinski, 1979), Trafalgar effect (Siegfried and Underhill, 1975) and the learning effect, when there are unpalatable individuals in the group (Brodie and Formanowicz, 1987). In addition, group members may benefit from other species’ abilities to detect and react to predators, as has been suggested in other vertebrate groups (e.g., tamarin monkeys; Peres, 1993). Different tadpole species are known to react to different proximate cues that indicate the presence of a predator or a high actual predation risk (e.g., Rödel and Linsenmair, 1997; Kats and Dill, 1998¸ Spieler and Linsenmair, 1999). A study by Glos et al. (Chapter 9) shows that these proximate cues are indeed different for D. insularis and A. securifer. Therefore, members of a mixed-species group may be better protected than those of single-species groups as a consequence of differential anti-predatory abilities within a group. Accordingly, members of one species may join an aggregation based solely on the visual presence of other heterospecific individuals (Wassersug and Hessler, 1971). The formation of mixed-species aggregations of tadpoles in the Kirindy Forest is mainly a reaction to vertebrate predators, and is more likely to occur in profound ponds with low leaf cover and clear water, while the density of invertebrate predators in the pond seems to play a minor role. Therefore, aggregation behaviour may play a key role in the ability of several anuran species to utilize spawning waters that are unpredictable in the presence of fish and other vertebrate predators (e.g. turtles, kingfisher). It is not yet clear what the proximate cues are that lead to the formation of aggregations in the different species and what the fitness costs of this behaviour are. This needs to be investigated further. Amphibian communities in Madagascar Julian Glos 114

LITERATURE CITED

ABRAHAMS, M.V., AND W. COLGAN. 1985. Risk of predation, hydrodynamic efficiency and their influence on school structure. Environ. Biol. Fish. 13:195—202.

ALATALO, R.V. 1981. Interspecific competition in tits Parus spp. and the goldcrest Regulus regulus: foraging shifts in multispecies flocks. Oikos 37:335—344.

ALATALO, R.V., J. HOEGLUNG, A. LUNDBERG, AND W.J. SUTHERLAND. 1992. Evolution of black grouse leks: female preferences benefit males in larger leks. Behav. Ecol. 3:53—59.

ARNOLD, W. 1988. Social thermoregulation during hibernation in alpine marmots (Marmota marmota). J. Comp. Physiol. B 158:151—156.

BEISWENGER, R.E. 1975. Structure and function in aggregations of tadpoles of the american toad, Bufo americanus. Herpetologica 31:222—233.

______. 1977. Diel patterns of aggregative behavior in tadpoles of Bufo americanus, in relation to light and temperature.Ecology 58: 98—108.

BONABEAU, E., AND L. DAGORN. 1995. Possible universality in the size distribution of fish schools. Phys. Rev. E 51:R5520—5523.

BOX, G.E., AND D.R. COX. 1964. An analysis of transformations. J. Roy. Stat. Soc. B 26:211—252.

BRATTSTROM, B.H. 1962. Thermal control of aggregational behavior in tadpoles. Herpetologica 18:36—48.

BRODIE, E.D., AND D.R. FORMANOWICZ, JR. 1987. Antipredator mechanisms of larval anurans: Protection of palatable individuals. Herpetologica 43:369—373.

BUTLER, M.J. 1999. The cause and consequence of ontogenetic changes in social aggregation in New Zealand spiny lobsters. Mar. Ecol-Prog. Ser. 188:179—191.

CARPENTER, C.C. 1953. Aggregation behavior of tadpoles of Rana pretiosa. Herpetologica 9:77—78.

COOK, A. 1981. Huddling and the control of water loss by the slug Limax pseudoflavus Evans. Anim. Behav. 29:298—298.

DIAMOND, J.M. 1981. Mixed-species foraging groups. Nature 292:408—409.

DOWNES, S., AND A.M. HOEFER. 2004. Antipredatory behaviour in lizards: interactions between group size and predation risk. Anim. Behav. 67:485—492.

DUNLAP, D.G., AND C.K. SATTERFIELD. 1982. Habitat selection in larval anurans: early experience and substrate pattern selection in Rana pipiens. Dev. Psychobiol. 18:37—58.

EGUCHI, K., S. YAMAGISHI, AND V. RANDRIANASOLO. 1992. The composition and foraging behaviour of mixed-species flocks of forest-living birds in Madagascar. Ibis 135:91—96.

FIELDING, A.H., AND H. HAWORTH. 1997. A review of methods for the assessment of predicting errors in conservation presence/absence models. Environ. Conserv. 24:38—49.

FOSTER, S.A. 1985. Group foraging by a coral reef fish: a mechanism for gaining access to defended resources. Anim. Behav. 33:782—792. Amphibian communities in Madagascar Julian Glos 115

GIBSON, R.M., A.S. ASPBURY, AND L.L. MCDANIEL. 2002. Active formation of mixed-species grouse leks: A role for predation in lek evolution? Roy. Soc. Lond. B. Bio. 269:2503—2507.

GLAW, F., AND M. VENCES. 1994. A Fieldguide to the Amphibians and Reptiles of Madagascar. M. Vences and F. Glaw Verlags GbR, Köln.

GLOS, J. 2003. The amphibian fauna of the Kirindy dry forest in western Madagascar. Salamandra 39:75—90.

GLOS, J., AND K.E. LINSENMAIR. 2004. Description of the tadpoles of Aglyptodactylus laticeps and A. securifer from Western Madagascar, with notes on life history and ecology. J. Herpetol. 38:131—136.

GUERON, S., S.A. LEVIN, AND D.I. RUBENSTEIN. 1996. The dynamics of herds: From individuals to aggregations. J. Theor. Biol. 182:85—98.

HAMILTON, W.D. 1971. Geometry for the selfish herd. J. Theor. Biol. 31:295—311.

HELLER, R., AND M. MILINSKI. 1979. Optimal foraging of sticklebacks on swarming prey. Anim. Behav. 27:1127—1141.

HEYER, W. R., M. A. DONNELLY, R. W. MCDIARMID, L.-A. HAYEK, AND M. S. FOSTER. 1994. Measuring and Monitoring Biological Diversity: Standard Methods for Amphibians. Smithonian Institution Press, Washington.

HOSMER, D.W., AND S. LEMESHOW. 1989. Applied logistic regression. Wiley, New York.

KATS, L.B., AND L.M. DILL. 1998. The scent of death: chemosensory assessment of predation risk by prey animals. Ecoscience 5:361—394.

KATS, L.B., J.W. PETRANKA, AND A. SIH. 1988. Antipredator defenses and the persistence of amphibian larvae with fishes. Ecology 69:1865—1870.

KRAUSE, J., AND G.D. RUXTON. 2002. Living in groups. Oxford University Press, New York.

LANDEAU, L., AND J. TERBORGH. 1986. Oddity and the confusion effect in predation. Anim. Behav. 34:1372—1380.

LOTT, D.F., AND S.C. MINTA. 1983. Random individual association and social group instability in American bison (Bison bison). Z. Tierpsychol. 61:153—172.

MAGURRAN, A.E., AND A. HIGHAM. 1988. Information transfer across fish shoals under predation threat. Ethology 78:153—158.

NELSON, R., AND N. HORNING. 1993. AVHRR-LAC estimates of forest area in Madagascar, 1990. Int. J. Remote Sens. 14:1463—1475.

NOLAND, R., AND G.R. ULTSCH. 1981. The roles of temperature and dissolved oxygen in microhabitat selection by the tadpoles of a frog Rana pipiens and a toad Bufo terrestris. Copeia 1981:645— 652.

O'HARA, R.K., AND A.R. BLAUSTEIN. 1985. Rana cascadae tadpoles aggregate with siblings: an experimental field study. Oecologia 67:44—51. Amphibian communities in Madagascar Julian Glos 116

PERES, C.A. 1993. Anti-predator benefits in a mixed species group of amazonian tamarins. Folia Primatol. 61:61—76.

PEREZ-TRIS, J., J.A. DIAZ, AND J.L. TELLERIA. 2004. Loss of body mass under predation risk: cost of antipredatory behaviour or adaptive fit-for-escape? Anim. Behav. 67:511—521.

RANTA, E., N. PEUHKURI, AND A. LAURILA. 1994. A theoretical exploration of antipredatory and foraging factors promoting phenotype-assorted fish schools. Ecoscience 1:99—106.

RATCHFORD, S.G., AND D.B. EGGLESTON. 1998. Size- and scale-dependent chemical attraction contribute to an ontogenetic shift in sociality. Anim. Behav. 56:1027—1034.

RIIPI, M., R.V. ALATALO, L. LINDSTRÖM, AND J. MAPPES. 2001. Multiple benefits of gregariousness cover detectability costs in aposematic aggregations. Nature 413:512—514.

RÖDEL, M.-O. 1999. Predation on tadpoles by hatchlings of the freshwater turtle Pelomedusa subrufa. Amphibia-Reptilia 20:173—183.

RÖDEL, M.-O., AND K.E. LINSENMAIR. 1997. Predator-induced swarms in the tadpoles of an African savanna frog, Phrynomantis microps. Ethology 103:902—914.

SEGHERS, B.H. 1981. Facultative schooling behavior in the spottail shiner (Notropis hudsonius): possible costs and benefits. Environ. Biol. Fish. 6:21—24.

SIEGFRIED, W.R., AND L.G. UNDERHILL. 1975. Flocking as an anti-predator strategy in doves. Anim. Behav. 23:504—508.

SORG, J.-P., AND U. ROHNER. 1996. Climate and tree phenology of the dry deciduous forest of the Kirindy forest, p. 57—80. In: Ecology and Economy of a Tropical Dry Forest in Madagascar, Primate Report 46-1. J. U. Ganzhorn and J.-P. Sorg (eds.). DPZ, Göttingen, Germany.

SPIELER, M. 2001. Tadpole aggregation behaviour reduces predation risk. Zoology 103:52. ______. 2003. Risk of predation affects aggregation size: A study with tadpoles of Phrynomantis microps (Anura: Microhylidae). Anim. Behav. 65:179—184.

SPIELER, M., AND K.E. LINSENMAIR. 1999. Aggregation behaviour of Bufo maculatus tadpoles as an antipredator mechanism. Ethology 105:665—686.

SULLIVAN, K.A. 1984. The advantages of social foraging in downy woodpeckers. Anim. Behav. 32:16—22.

VENCES, M., AND F. GLAW. 2003. Laliostoma, p. 921. In: The Natural History of Madagascar. S. M. Goodman and J. Benstead (eds.). The University of Chicago Press, Chicago and London.

WALDMAN, B. 1991. Kin recognition in amphibians, p. 162—219. In: Kin Recognition. G. Hepper (ed.). Cambridge University Press.

WASSERSUG, R., AND C.M. HESSLER. 1971. Tadpole behaviour: aggregation in larval Xenopus laevis. Anim. Behav. 19:386—389.

WASSERSUG, R.J., A.M. LUM, AND M.J. POTEL. 1981. An analysis of school structure for tadpoles (Anura: Amphibia). Behav. Ecol. Sociobiol. 9:15—22. Amphibian communities in Madagascar Julian Glos 117

WATT,J., S.F. NOTTINGHAM, AND S. YOUNG. 1997. Toad tadpole aggregation behaviour: evidence for a predator avoidance function. Anim. Behav. 54:865—872.

WIRTZ,, AND J. LÖSCHER. 1983. Group sizes of antelopes in an East African park. Behaviour 84:135— 155.

WISENDEN, B.D., K.A. VOLLBRECHT, AND J.L. BROWN. 2004. Is there a fish alarm cue? Affirming evidence from a wild study. Anim. Behav. 67:59—67.

WOLLMUTH, L.P., L.I. CRAWSHAW, R.B. FORBES, AND D.A. GRAHN. 1987. Temperature selection during development in a montane anuran species, Rana cascadae. Physiol. Zool. 60:472— 480.

WOLTERS, S., AND K. ZUBERBÜHLER. 2003. Mixed-species associations of Diana and Campbell's monkeys: the costs and benefits of a forest phenomenon. Behaviour 140:371—385.

ZWEIG, M.H., AND G. CAMPBELL. 1993. Receiver-operating characteristic (ROC) plots: A fundamental evaluation tool in clinical medicine. Clin. Chem. 39:561—577.

Amphibian communities in Madagascar Julian Glos 118

Chapter 9

Causes and costs of social aggregation: a comparative study on two tadpole species from western Madagascar

Julian Glos1, Georgia Erdmann1, Kathrin H. Dausmann2 and K. Eduard Linsenmair1

ABSTRACT

In the dry forest of western Madagascar, mixed-species social tadpole aggregations are frequent (chapter 8). In two species, naturally often found together within one aggregation, Aglyptodactylus securifer (Mantellidae) and Dyscophs insularis (Microhylidae), we experimentally tested which proximate mechanisms lead to the formation of tadpole aggregations. We show that aggregations are induced by the direct presence of predators, or by indirect chemical cues indicating a predation risk. However, the specific cues that initiated the formation of aggregations differed between the two species. Aglyptodactylus securifer reacted (predator unspecifically) to con- and heterospecific tadpole homogenisate (“Schreckstoff”), while the reaction of D. insularis was predator-specific to fish, as it reacted directly to fish and indirectly to chemical cues released by fish. Although the ultimate benefit of this behaviour is thought to be reduced predation, it also causes fitness costs. Accordingly, tadpoles of A. securifer in aggregations showed reduced growth and retarded development as opposed to non- aggregating tadpoles. Amphibian communities in Madagascar Julian Glos 119

INTRODUCTION

Predation is a crucial factor influencing individual fitness, and animals will often react to a predation threat with behavioural adaptations. In many cases, the predation risk for individual prey animals is thought to be reduced by congregating with conspecific or heterospecific individuals in groups (Diamond, 1981; Siegfried and Underhill, 1975). Accordingly, studies on different animal species under natural conditions have shown that the formation of aggregations is often positively correlated to the presence of predators (e.g., in fish: Johannes, 1993; hemiptera: Foster and Treherne, 1981; birds: Cresswell, 1994). Aggregative behaviour, however, also involves costs for the aggregating individuals. Animals in groups may have reduced food intake and consequentially reduced growth and/or slowed development compared to solitarily foraging animals (Krause and Ruxton, 2002). Additionally, they might have increased parasite or other pathogen burdens (e.g., Brown and Brown, 1986). Alarm cues that are used to detect predators and subsequently initiate predator avoidance behaviour include visual detection, alarm calls by con- and heterospecifics (e.g., in birds: Elgar et al., 1984; in primates: Wolters and Zuberbühler, 2003), substrate vibrations caused by the predator (e.g., in spiders: Hodge and Uetz, 1993), and chemicals released by predator or prey (reviewed by Kats and Dill, 1998). In tadpoles, the proximate mechanisms leading to aggregation are addressed in very few studies. The visual detection of predators (e.g., Rödel and Linsenmair, 1997), mechanical stimuli (e.g., Spieler and Linsenmair, 1999) and alarm substances, originating from the body fluid of injured conspecific or heterospecific tadpoles (e.g., Rödel and Linsenmair, 1997; Hokit and Blaustein, 1995), or a combination of two or more stimuli appear to account for this reaction (Spieler and Linsenmair, 1999). In the dry forest of western Madagascar, temporary associations of tadpoles are frequent (chapter 8). The two most abundant species in these aggregations, Aglyptodactylus securifer (Mantellidae) and Dyscophus insularis (Microhylidae), are very different in respect to feeding ecology and microhabitat choice (Glaw and Vences, 1994; Glos and Linsenmair, 2004). Nevertheless they form mixed-species aggregations. In a comparative study, we examined the effects of predation risk on aggregative behaviour, activity and microhabitat choice in tadpoles of these two species. We experimentally assessed the proximate cues used by tadpoles to detect and react to predators, and we analysed the costs on growth and development of behavioural reactions to non-lethal predator presence in A. securifer. Amphibian communities in Madagascar Julian Glos 120

METHODS

Study site The Kirindy Forest, a deciduous dry forest, is situated near the west coast of Madagascar, 60 km north of Morondava and about 20 km inland (44°39’ O, 20°03 S; 18 - 40 m NN; Sorg and Rohner, 1996). The climate is characterized by a marked seasonality. Almost all rain falls in the austral summer from November to March, followed by seven months of virtually no precipitation (annual mean rainfall: 800 mm; Sorg and Rohner, 1996). The first waters that are used by amphibians for spawning usually arise in the rocky parts of the Kirindy riverbed. As a rule, the density of invertebrate predators in these pools is low at this early stage (J. Glos, unpubl. data). During the course of the rainy season, these pools eventually become connected with each other and fish immigrate from the few permanent pools in the riverbed, resulting then in a very high predation risk for amphibian larvae. The first breeding ponds in the closed forest usually fill in December. There are no fish in the forest ponds throughout the rainy season. In contrast to the river bed ponds, the density of invertebrate predators (Dytiscidae, Belostomatidae, Anisoptera, Zygoptera) in larger forest ponds (> 200 m2) can be very high (mean 37 / m2; range 7 – 160 / m2; J. Glos, unpubl. data).

Study species

Aglyptodactylus securifer (Mantellidae: Laliostominae) GLAW,VENCES &BÖHME 1998 This frog is limited to two localities in western Madagascar. As a strictly explosive breeding species it reproduces only after the very first heavy rains at the beginning of the rainy season. Breeding takes place mainly in rock pools, before the river is flowing. Tadpoles are benthic and feed primarily on plant detritus, but also on carcasses of con- and heterospecific tadpoles (Glos and Linsenmair, 2004). Their mode of locomotion is characterized by long periods of low activity, followed by short bursts of swimming movement. Together with their brown dorsal colouration this leads to a rather cryptic appearance. Predation pressure on A. securifer tadpoles is potentially very high. Waters that contain fish are avoided for breeding. However, strong rainfall may raise the water level in the riverbed, interconnecting ponds already spawned in, and thus enable the immigration of fish. Birds (kingfisher Alcedo vintsioides, paradise flycatcher Terpsiphone mutata) and turtles (Pelomedusa subrufa, Rödel, 1999) are further major predators of tadpoles. Aquatic insects occur only in low densities in these ponds and have a relatively small size at the time of spawning and tadpole development. Invertebrate predators therefore constitute a relatively low predation risk for A. securifer tadpoles.

Dyscophus insularis (Microhylidae: Dyscophinae) GRANDIDIER, 1872 This frog occurs in dry habitats all over western Madagascar (Glaw and Vences, 1994). It is an explosive breeder, reproducing only after heavy rains. However, contrary to A. securifer, reproduction Amphibian communities in Madagascar Julian Glos 121 occurs throughout the whole rainy season although it peaks at its beginning. It uses a wide variety of habitats and pond types for reproduction (Glos, 2003). Hence, predation pressure is presumably very variable. The filter feeding D. insularis tadpoles mainly swim in mid-water and move almost constantly.

Proximate factors causing aggregations Experimental setup Experiments were conducted from December-January of 2000-2001 and 2001-2002, respectively. To avoid genetic effects by related siblings, tadpoles of A. securifer were taken from eight different clutches that were deposited by eight amplectant couples under controlled conditions. Tadpoles of D. insularis in young developmental stages (25 - 28; Gosner, 1960) were dip-netted from different parts of a fish-free pond to increase the probability of collecting members of different sibships. Tadpoles of both species were raised separately in the field camp in a predator-free environment under natural temperature conditions. They were kept in mixed-sibling groups in stock tanks (55 cm in diameter, 25 cm in depth) and fed ad libitum with commercial fish food (TetraMinTabs£). The aggregation experiments were conducted in green polyethylene arenas (45 x 26 cm L x B), the bottom of which was subdivided into 18 equal squares (7.5 x 8.7 cm each). The arenas were filled with rainwater to a depth of 12 cm. To control for light regime, and to exclude possible effects of rain dropping into the arenas or of ground vibrations emitted by the observer, all arenas were positioned on elevated roofed frames. The arenas were covered with mesh wire to prevent disturbances by other animals. All tadpoles were used only once in the experiment. Tadpoles within one species were of similar size and developmental stage. A randomly chosen sample of A. securifer tadpoles at the time of the experiments had a body size of 7.0 ± 0.6 mm (mean ± SD; n = 44) and a developmental stage of 31.0 ± 1.0 (n = 20); D. insularis tadpoles had a body size of 8.4 ± 1.1 mm and a developmental stage of 29.8 ± 3.5 (n = 90). Tadpoles of these sizes and stages were found aggregating under natural conditions (chapter 8). The following protocol was used for all experimental trials: Tadpoles were chosen at random from the stock tanks to ensure mixing of sibling groups amongst treatments. 30 tadpoles were transferred into each test arena. Each test arena represented one replicate. Tadpoles were allowed to acclimatize overnight (> 15 hours) to the arenas. Food (one TetraMinTabs£ / arena) was added at night and evenly distributed in the arena. The experimental arenas were arranged in a row and randomly assigned to either experimental or control treatment. In ten daily trials, including all treatments, n = 78 arenas were equipped with A. securifer tadpoles, and n = 78 with D. insularis tadpoles. A glass of rainwater (200 ml) containing either a live predator or a predator’s chemical cue (see below) was added to the experimental arenas at 11:00 a.m. In the simultaneous control treatments, the same mechanical manipulations were performed by adding a glass of pure rainwater. Measurements were taken by an observer from above, every 20 min over the course of two hours (= Amphibian communities in Madagascar Julian Glos 122 six measurements / arena). The mean of six measurements per arena represents one datum point. The effect of the experimental treatment was analysed by one-way ANOVA and Dunnet t-test post hoc comparisons. To acquire information on the persistence time of the aggregations, we continued the measurements in 20 min intervals for two more hours. After each experiment, all used tadpoles were released into their natural habitats.

Response variables Five parameters were measured during the experiments: First, we recorded the presence of aggregations in the arena (0 = absence, 1 = presence). Aggregations were defined as at least four individuals being not more than 2 cm apart from each other. This distance between individuals reflects the natural situation in the field, and was the basis of field studies on natural aggregations in these species (chapter 8) and of earlier studies on tadpole aggregations (O'Hara and Blaustein, 1985). If there was an aggregation present, the number of tadpoles per aggregation was counted (aggregation size). The third parameter, the aggregation index, is a quantitative measure of aggregative behaviour, defined as the number of tadpoles in that square of the arena (of 18 squares) with the highest number of tadpoles. Fourth, we measured swimming activity of the tadpoles. For that, we recorded the number of tadpoles crossing the long axis in the arena within a 10 sec period. Finally, we recorded microhabitat choice of all tadpoles in respect to the vertical position in the water column. We distinguished bottom (0 – 3 cm), middle (3 – 9 cm), and surface (9 – 12 cm) microhabitat. Due to practical reasons, not all parameters were measured for all predators or predatory cues. At the end of the experiment, we counted the number of tadpoles in each arena. This number was used to determine survival. As the predators were free to move within the arenas and to prey on the tadpoles this number regularly dropped below 30 in the experimental treatments. We corrected the variables aggregation index, aggregation size, swimming activity and microhabitat choice according to the number of tadpoles actually present in the arena.

Predatory cues For both species of tadpoles, we tested predators that represent different taxa and predation modes (a – h): (a) juvenile fish (Oreochromis sp.: Cichlidae; body length 36.2 ± 4.1 mm; mean ± SD; n = 6); (b) young turtles (Pelomedusa subrufa; carapax length 68.7 ± 17.3 mm; n = 9); (c) adult diving beetles (Dytiscidae; body length 27.8 ± 1.4 mm, n = 10). For A. securifer tadpoles only, (d) Malagasy kingfisher (Alcedo vintsioides, n = 7) was tested. This was done by allowing these naturally occurring birds to prey in the arenas by removing the mesh wire cover that kept them from doing so otherwise. Generally, the birds were sitting on the edge of the arena about 20 cm above water level, and eventually flew up and snatched tadpoles from the arena. For D. insularis only, (e) giant water bugs (Lethocerus sp., Belostomatidae, body length 72.0 ± 10.5 mm; n = 6) were tested. All these predators are present in the tadpoles’ natural habitats. All predators were allowed to move freely in the arenas Amphibian communities in Madagascar Julian Glos 123 and prey on the tadpoles. Individual predators were used only once, although the kingfisher may have been the same. Additionally, (f, g) we tested in both species the reaction to homogenisate of both, conspecific and heterospecific tadpoles (A. securifer and D. insularis, respectively). To this end, one tadpole / arena was homogenised using a surgical blade and dissolved in 200 ml of rainwater. To test for the reaction to the chemical stimulus of fish (h), “fish water” was added. Fish water was standardized: For each replicate, we kept one different juvenile fish (Oreochromis sp.; total length about 40 mm each) for 24 h in 5 l of fresh rain water. Subsequently, one glass (200 ml) of the water was taken for the experiments.

Costs of aggregations Experimental setup Experiments were conducted on A. securifer during the rainy season 2001-2002. Tadpoles from eight different clutches were pooled and raised in stock tanks (55 cm in diameter, 25 cm in depth) filled with rainwater and fed ad libitum with commercial fish food (TetraMinTabs£). Tadpoles of similar size and developmental stage were randomly collected from these tanks. The experiments were conducted in circular arenas (diameter 40 cm, volume 15 l, water depth 12 cm). A line on the bottom divided each arena into halves. Environmental conditions were similar to those described above. 20 tadpoles were kept in each arena for 12 days (density 1.3 tadpoles / l). One arena represents one replicate (treatment: n = 8, control: n = 8). The following protocol was used: To induce aggregations, A. securifer – homogenisate (one tadpole / basin; for extraction see previous experiment) dissolved in a glass of rainwater was added daily (9:00 a.m.) to the experimental treatments. A glass of pure rainwater was added to the control treatments. Tadpoles were fed with one TetraMinTab£ per day, which was evenly distributed at 7:00 a.m. in each arena. To reduce handling effects, we did not take measurements on these tadpoles before the experiment. Independently thereof, we randomly selected a representative sample of 30 tadpoles from the stock tanks at the beginning of the experiment and measured body length and developmental stage to determine average size and developmental stage at the beginning of the experiment. These tadpoles were subsequently released into their natural habitat.

Response variables To detect whether there were behavioural reactions induced by the addition of homogenisate over longer time periods, aggregation presence, swimming activity and microhabitat choice (as defined above) were recorded four times per day. Times of data acquisition were at least 90 min apart. After 12 days, size (as body length) and developmental stage was measured in all tadpoles in the experimental arenas. The mean of all 20 tadpoles in one arena represents one replicate. Amphibian communitiesinMadagascar TABLE 1. Effects of predatory treatments on response variables in tadpoles of (a) Aglyptodactylus securifer and (b) Dyscophus insularis (mean r SD). Aggregation index = number of tadpoles in that square of the arena with the highest number of tadpoles, aggregation presence = % of control checks with aggregations, aggregation size = number of tadpoles per aggregation, swimming activity = crossing of the lateral axis in the arena / h * animal, microhabitat choice = vertical position in the water column: b = bottom, c = centre, s = surface, survival = number of tadpoles present at the end of the experiment; “-“ = not measured. ANOVA with Dunnet t-test post hoc comparisons (shown are P - values); microhabitat choice was tested with Ȥ2 – goodness of fit tests; significant results are highlighted in bold letters. N = number of tested arenas.

(a)

Aglyptodactylus ANOVA Control Fish Fish water A.securifer - D.insularis - Diving beetle Turtle Kingfisher securifer homogenisate homogenisate Aggregation index P < 0.001 7.59 r 1.81 9.42 r 5.40 7.89 r 1.81 16.53 r 5.87 10.79 r 4.89 6.32 r 1.65 7.43 r 1.26 13.16 r 4.47 F = 9.67 0.91 1.00 < 0.001 0.14 0.92 1.00 < 0.01 N = 77 Aggregation presence P < 0.001 10.6 r 19.5 12.5 r 15.8 4.6 r 7.9 79.6 r 37.1 38.9 r 40.8 10.0 r 14.0 18.4 r 19.3 81.7 r 36.6 F=13.49 1.00 1.00 < 0.001 < 0.05 1.00 0.93 < 0.001 N = 77 Aggregation size P = 0.126 6.16 r 3.65 5.25 r 1.77 4.15 r 0.25 8.09 r 4.09 5.68 r 2.15 3.96 r 0.13 5.27 r 0.84 10.30 r 4.18 F = 2.32 ANOVA n. s. N = 39 Swimming activity P = 0.02 27.48 r 20.60 21.43 r 7.74 24.36 r 13.88 14.23 r 15.55 18.51 r 7.25 18.45 r 13.06 20.36 r 18.16 0.79 r 1.42 F= 2.54 1.00 1.00 0.21 0.63 0.59 0.86 0.001 N = 77 Microhabitat choice b 56.80 b 75.43 b 45.87 b 67.70 b 55.07 b 72.17 b 79.83 b 96.77 (%) c 34.40 c 22.63 c 43.73 c 30.23 c 38.97 c 20.23 c 19.07 c 2.60 s 8.80 s 1.93 s 10.40 s 2.10 s 6.00 s 7.60 s 1.10 s 0.67 > 0.05 > 0.05 > 0.05 > 0.05 > 0.05 > 0.05 < 0.01 F2=2.84 F2=0.72 F2=1.65 F2=0.28 F2=1.70 F2=0.94 F2=13.44 df = 2 df = 2 df = 2 df = 2 df = 2 df = 2 df = 2 Survival 29.61 r 0.77 29.00 r 1.41 28.89 r 2.09 29.86 r 0.38 29.89 r 0.33 29.70 r 0.82 23.25 r 5.60 23.00 r 4.00 Julian Glos124 Amphibian communitiesinMadagascar (b)

Dyscophus insularis ANOVA Control Fish Fish water A.securifer - D.insularis - Diving beetle Water bug Turtle homogenisate homogenisate Aggregation index P < 0.001 6.01 r 1.09 10.60 r 1.98 6.24 r 0.87 6.87 r 0.56 5.56 r 1.36 6.51 r 0.76 5.75 r 0.59 6.16 r 1.24 F = 15.17 < 0.001 0.99 0.41 0.92 0.88 0.78 0.99 N = 78 Aggregation presence P < 0.001 32.0 r 31.1 97.2 r 6.9 68.9 r 37.8 0.0 r 0.0 42.9 r 34.3 - 19.3 r 16.2 6.9 r 12.4 F = 10.60 < 0.001 0.01 0.06 0.91 0.86 0.17 N = 64 Aggregation size P < 0.001 5.84 r 1.51 11.06 r 1.98 8.57 r 2.74 - 6.49 r 0.99 - - 6.25 r 1.77 F = 9.66 < 0.000 0.02 0.91 0.99 N = 33 Swimming activity P = 0.07 19.57 r 9.20 - 11.86 r 15.30 28.50 r 15.39 14.34 r 11.84 - - 11.86 r 7.30 F = 2.38 ANOVA n. s. N = 40 Microhabitat choice b 30.33 - b 56.67 b 43.77 b 40.67 - - b 72.03 (%) c 35.20 c 11.67 c 32.77 c 52.67 c 21.63 s 34.47 s 31.67 s 23.47 s 6.67 s 6.33 - > 0.05 > 0.05 < 0.05 < 0.01 F2=5.98 F2=1.41 F2=7.13 F2=11.88 df = 2 df = 2 df = 2 df = 2 Survival 28.75 r 2.40 26.67 r 3.08 30.00 r 0.00 30.00 r 0.00 30.00 r 0.00 29.29 r 1.11 24.33 r 6.71 22.25 r 4.69 Julian Glos125 Amphibian communities in Madagascar Julian Glos 126

RESULTS

Proximate factors causing aggregations Effects of predatory cues Tadpoles of A. securifer reacted significantly to the presence of some of the potential predators or predatory cues, but not to all (Table 1a). The addition of both, conspecific and heterospecific homogenisate caused tadpoles to aggregate more often (aggregation presence), and their overall aggregative tendency (aggregation index) was increased by conspecific homogenisate. The same was true with birds as predators. In addition, the presence of the birds greatly reduced the swimming activity of the tadpoles and altered their microhabitat choice, with the effect that a greater proportion of tadpoles were found on the bottom of the experimental arenas. The aggregation size was not changed by predators or predatory cues. Dyscophus insularis tadpoles formed significantly larger and significantly more frequent aggregations when fish were present or fish water added (aggregation presence, aggregation size), and the aggregation index was higher in the presence of fish (Table 1b). The microhabitat choice was altered by the addition of conspecific homogenisate and the presence of turtles; significantly less tadpoles were recorded close to the water surface than in the control treatment. There was no change in the swimming activity. Aggregative behaviour in both species did not considerably decline for up to four hours after the initiation of the aggregation (Fig. 1). In general, aggregations persisted for several hours, and dissolved with diminishing day-light.

Costs of aggregations in A. securifer Formation of aggregations over longer periods of time We first tested whether the variables aggregation presence, aggregation size, swimming activity and microhabitat choice differed between the experimental and the control treatment over a period of 12 days. We found significant differences in the aggregation presence and the swimming activity (Table 2). Accordingly, we found aggregations more often and for longer periods of time. The aggregations, initiated by the daily addition of conspecific homogenisate normally persisted until nightfall. Whenever aggregations were found in the control treatment, there was no difference in the aggregation size as compared to the experimental treatment. Therefore, the results of these experiments (when tadpoles were exposed to homogenisate in the long-term) are consistent with the results of the experiment when they were exposed only for a short time as shown above, with the exception that the swimming activity was additionally reduced when predator cues were presented over a longer period of time. Amphibian communities in Madagascar Julian Glos 127

(a)

5 Aggregation index

0 0 20 40 60 80 100 120 140 160 180 200 220 240 Ti me ( mi n)

(b)

10 Aggregation index 5

0 0 20 40 60 80 100 120 140 160 180 200 220 240 Ti me ( mi n)

FIG. 1. Reaction of A. securifer (a) and D. insularis (b) tadpoles to the experimental treatment over time. Shown is the aggregation index (mean r SD) for the control treatment (black bars) and those treatments that proved to be significantly different between predatory and control treatment (grey bars = A. securifer - homogenisate, white bars = kingfisher, hatched bars = fish). For definition of variables see Table 1. Predator or predator stimulus was introduced at 0 min. Amphibian communities in Madagascar Julian Glos 128

TABLE 2. Effects of addition of conspecific homogenisate on response variables in tadpoles of A. securifer (mean r SD). For definition of variables see Table 1. Mann – Whitney U – tests and F2 - test; significant results (Į = 0.05) are highlighted in bold letters. N refers to the number of tested arenas.

Response variable Control (N = 8) Homogenisate (N = 8) Z - value P Aggregation presence (%) 39 r 9 70 r 25 2.53 0.01 Aggregation size 5.39 r 0.47 6.40 r 1.17 1.68 0.11 Swimming activity 12.02 r 3.15 3.92 r 3.62 2.89 < 0.01 Microhabitat choice (%) b 94.40 r 3.90 b 99.05 r 0.55 F2 = 0.71 > 0.05 c 4.90 r 3.20 c 0.90 r 0.50 s 0.70 r 0.80 s 0.05 r 0.05

Costs on growth and development Initial body length of the A. securifer tadpoles before the experiment was 6.63 r 0.31 mm (mean r SD; n = 30) and the developmental stage was 30.93 r 1.05. After the twelve days of the experiment, the size of the tadpoles in the homogenisate treatment (7.98 r 0.45 mm) was significantly smaller than in the control treatment (8.48 r 0.26 mm; t-test: p = 0.018, T = 2.69, n = 8; Fig. 2). Furthermore, the presence of homogenisate had a significant negative effect on development (control: stage 35.88 r 0.53, homogenisate: stage 34.90 r 1.09; t-test: p = 0.039, T = 2.28, n = 8; Fig. 3). Mortality during the experiment was very low (control: 19.62 r 0.70, homogenisate: 18.87 r 0.33 surviving tadpoles / arena at the end of the experiment) and did not differ between treatment and control (Mann-Whitney U- Test: p = 0.65, Z = -0.69, n = 8).

(a) (b) body length (mm) developmenttal stage developmenttal

Control + Homogenisate Control + Homogenisate

FIG. 2. Costs of aggregation formation, shown as body length (a) and developmental stage (b) of A. securifer tadpoles. Tadpoles in the control treatment (N = 8) are significantly larger than those in the homogenisate treatment (N = 8), and development is faster. Shown are median, 25 % - and 75 % percentiles, and minima and maxima. Amphibian communities in Madagascar Julian Glos 129

DISCUSSION

Most anurans in the dry forest of western Madagascar use ephemeral pools as breeding sites. Together with the desiccation of breeding ponds, predation is thought to contribute most to tadpole mortality (e.g., Rödel, 1998; Hero et al., 1998). Selection pressure on the development of predation avoidance by morphological, chemical and behavioural adaptations should therefore be very strong. This study shows that a high predation risk induces tadpoles of A. securifer as well as of D. insularis to form aggregations, and that this behaviour is often associated with a decrease in swimming activity and a shift in their microhabitat. We found reactions not only to the immediate presence of predators (fish, bird, turtle), and therefore possibly to a combination of visual, mechanical, and olfactory cues, but also to purely chemical stimuli. However, the specific proximate cues that initiated the formation of aggregations, of activity decrease, or of microhabitat change were different between the two species. Aglyptodactylus securifer responded strongly to the homogenisate of conspecifics and still significantly but less intensive to that of heterospecific tadpoles. The presence of chemical cues deriving from injured tadpoles (called “Schreckstoff” by Pfeiffer, 1966) presents an indirect but reliable signal of immediate predation risk, but is not predator specific. Dyscophus insularis, on the other hand, also uses indirect signals to assess predation risk and release aggregation behaviour. In contrast to A. securifer, its reaction is specifically to fish or to chemicals released by fish, respectively. Prey defences to chemical cues emitted by predators have been demonstrated across a wide range of taxa, and the ability to recognize potential predators through chemicals cues clearly has adaptive value (see Kats and Dill, 1998, for review). In general, the formation of aggregations was triggered by the presence of vertebrate predators, but not of invertebrate predators (diving beetle, giant water bug). It is known, that the formation of aggregations as a predator avoidance behaviour is a reaction to active, mostly visually hunting predators (Krause and Ruxton, 2002). Aquatic invertebrates, however, are predominantly sit- and-wait predators, and aggregating might not pay off the costs of this behaviour in this case. Furthermore, A. securifer and to a lesser extent also D. insularis breed at the very beginning of the rainy season, directly after the filling of the breeding ponds (Glos, 2003). Invertebrate size and density, and hence predation probability from this group is low at that time of the season (J. Glos, unpubl. data), and presumably does not exert any strong selection pressure. Accordingly, we did not find a correlation of invertebrate predator density and the presence of aggregations under natural conditions (chapter 8). Depending on environmental conditions and predator identity, behavioural defences that are triggered predator-specifically (as shown by D. insularis vis-à-vis to fish) are advantageous in comparison to those reactions that are triggered by more general cues (as it is tadpole homogenisate for A. securifer). In the former cases, the prey’s reaction can specifically oppose the predator’s sensory abilities and foraging mode, and defence might thus be more efficient. Dyscophus insularis is exposed Amphibian communities in Madagascar Julian Glos 130 to fish in only some of its natural habitats, while Aglyptodactylus securifer lives sympatrically with the cichlid Oreochromis sp. over much of its range. Although fish-free waters are initially preferred as breeding sites (Glos and Linsenmair, 2004), Oreochromis sp. regularly immigrates into the breeding waters of A. securifer presenting a potentially very high predation threat. Therefore, at first sight it appears paradoxical that it is D. insularis and not A. securifer that reacts directly to fish. However, Oreochromis sp. is exotic to Madagascar (Reinthal and Stiassny, 1991). Hence, predator avoidance behaviour to this fish might simply have not yet evolved. The intriguing question remains; why do these two species aggregate in mixed-species associations, as was found under natural conditions at the same study site (chapter 8), while both species react to different predatory cues. Several explanations are feasible. If predation risk is high, multiple cues might be present simultaneously in the water. For example, if a fish captures several tadpoles, alarm substances from injured or killed tadpoles might be released and cause tadpoles of A. securifer to aggregate and join a D. insularis aggregation that was already formed as a reaction to the presence of the fish itself. Furthermore, the presence of an already existing aggregation might lower the threshold (e.g., of the concentration of alarm substance) at which further aggregations are initiated or at which individuals join an aggregation. Alternatively, joining an aggregation might be induced solely by the presence of an already existing aggregation (Wassersug and Hessler, 1971). By doing so, group members may benefit from other species’ abilities to detect and react to predators earlier (Diamond, 1981; Peres, 1993). Therefore, under some circumstances members of a mixed-species group may be better protected than those of single-species groups. The ultimate benefit of predator mediated formation of aggregations, activity decrease and microhabitat change in tadpoles is primarily seen as an increase in individual survival chances by reduction of the predation risk. All group members within aggregations may benefit from mere group- size related anti-predator benefits, such as the dilution effect (Hamilton, 1971; Riipi et al., 2001), the confusion effect (Heller and Milinski, 1979), the Trafalgar effect (Siegfried and Underhill, 1975) and also the learning effect, if the group includes unpalatable individuals (Brodie and Formanowicz, 1987). Furthermore, a reduction in swimming activity may lower encounter rate of or detectability by predators and shifting the microhabitat may help in predator avoidance. Both behaviours have been shown to increase survivorship of individual tadpoles (e.g., Watt et al., 1997; Spieler, 2001). Aglyptodactylus securifer and D. insularis tadpoles reacted flexibly to predation threats. In the experiments, aggregations were formed only when a predator cue existed, and they usually dissolved with the diminishing daylight, as is the case in Phrynomantis microps (Rödel and Linsenmair, 1997). Accordingly, aggregations of these species in natural ponds are temporary (chapter 8). These inducible reactions may return higher fitness benefits than constitutive defences would provide. Constitutive reactions would not allow for shifts in foraging strategies to increase growth and developmental rates when predation pressure is low (Sih, 1987). Indeed, aggregation probability decreases when water transparency and therefore visibility is low (chapter 8, see also Rödel and Linsenmair, 1997). In a Amphibian communities in Madagascar Julian Glos 131 natural situation, the predation risk is presumably strongly decreased under low light conditions in particular by vertebrate predators (fish, birds, turtles), as these are predominantly visually hunting. This flexibility in behaviour also indicates that fitness costs are associated with aggregating, reduced swimming activity or change of microhabitat, or any combination of these. In fact, tadpoles of A. securifer that were induced to perform these behaviours over a prolonged time had a reduced growth and a retarded development as opposed to control tadpoles. As the timing of and size at metamorphosis is seen as an important factor in amphibian life-history (Wilbur and Collins, 1973; Rowe and Ludwig, 1991), affecting e.g., reproduction (Berven, 1981; 1982), survival during hibernation (Lyapkov, 1998), or desiccation and predation risk (Wilbur, 1997). Aglyptodactylus securifer tadpoles that have a prolonged larval development in combination with a reduced metamorphic size are predicted to suffer negative fitness consequences. Mechanisms that generate reduced growth and retarded development in tadpoles that react behaviourally to predators include increased food competition in groups (reviewed by Krause and Ruxton, 2002), decreased foraging activity as a result of decreased swimming activity (Werner, 1986; Skelly and Werner, 1990; Skelly, 1992), staying in less profitable food patches as a result of microhabitat change, and hormonal-physiological effects associated with high density. Temporary mixed-species social associations of tadpoles are frequent in the dry forest of western Madagascar. They are formed as a reaction to predatory threats, although generating fitness costs for the species involved. This indicates that aggregation behaviour may play a key role in the ability of amphibians to successfully develop in predator-rich breeding sites. Amphibian communities in Madagascar Julian Glos 132

LITERATURE CITED

BERVEN, K. A. 1981. Mate choice in the wood frog, Rana sylvatica. Evolution. 35:707-722. —. 1982. The genetic basis of altitudinal variation in the wood frog Rana sylvatica. I. An experimental analysis of life history traits. Evolution. 35:962-983.

BRODIE, E. D., and D. R. FORMANOWICZ,JR. 1987. Antipredator mechanisms of larval anurans: Protection of palatable individuals. Herpetologica. 43:369-373.

BROWN, C. R., and M. B. BROWN. 1986. Ectoparasitism as a cost of coloniality in cliff swallows (Hirundo phyyhonota). Ecology. 67:1206-1218.

CRESSWELL, W. 1994. Flocking is an effective anti-predation strategy in redshanks, Tringa totanus. Animal Behaviour. 47:433-442.

DIAMOND, J. M. 1981. Mixed-species foraging groups. Nature. 292:408-409.

ELGAR, M. A., P. J. BURREN, and M. POSEN. 1984. Vigilance and percption of flock size in foraging house sparrows (Passer domesticus L.). Behaviour. 90:215-223.

FOSTER, J. P., and J. E. TREHERNE. 1981. Evidence for the dilution effect in the selfish herd from fish predation on a marine insect. Nature. 295:466-467.

GLAW, F., and M. VENCES. 1994. A Fieldguide to the Amphibians and Reptiles of Madagascar. M. Vences & F. Glaw Verlags GbR, Köln.

GLOS, J. 2003. The amphibian fauna of the Kirindy dry forest in western Madagascar. Salamandra. 39:75-90.

GOSNER, K. L. 1960. A simplified table for staging anuran embryos and larvae with notes on identification. Herpetologica. 16:183-190.

HAMILTON, W. D. 1971. Geometry for the selfish herd. Journal of Theoretical Biology. 31:295-311.

HELLER, R., and M. MILINSKI. 1979. Optimal foraging of sticklebacks on swarming prey. Animal Behavior. 27:1127-1141.

HERO, J.-M., C. GASCON, and W. E. MAGNUSSON. 1998. Direct and indirect effects of predation on tadpole community structure in the Amazonian rainforest. Australian Journal of Ecology. 23:474-482.

HODGE, M. A., and G. W. UETZ. 1993. Antipredator benefits of single- and mixed-species grouping by Nephila clavipes (L.) (Araneae, Tetragnathidae). Journal of Arachnology. 20:212-216.

HOKIT, D. K., and A. R. BLAUSTEIN. 1995. Predator avoidance and alarm-response behaviour in kin- discriminating tadpoles (Rana cascadae). Ethology. 101:280-290.

JOHANNES, M. R. S. 1993. Prey aggregation is correlated with increased predation pressure in lake fish communities. Canadian Journal of Fisheries and Aquatic Sciences. 50:66-73. Amphibian communities in Madagascar Julian Glos 133

KATS, L. B., and L. M. DILL. 1998. The scent of death: chemosensory assessment of predation risk by prey animals. Écoscience. 5:361-394.

KRAUSE, J., and G. D. RUXTON. 2002. Living in groups. Oxford University Press.

LYAPKOV, S. M. 1998. Peculiarities of growth in Rana temporaria and Rana arvalis in the first years of terrestrial life. Influence of size on survival during wintering. Zoologicheskii Zhurnal. 76:356-363.

O'HARA, R. K., and A. R. BLAUSTEIN. 1985. Rana cascadae tadpoles aggregate with siblings: an experimental field study. Oecologia. 67:44-51.

PERES, C. A. 1993. Anti-predator benefits in a mixed species group of amazonian tamarins. Folia Primatologica. 61:61-76.

PFEIFFER, W. 1966. Die Verbreitung der Schreckreaktion bei Kaulquappen und die Herkunft des Schreckstoffes. Zeitschrift für vergleichende Physiologie. 52:79-98.

REINTHAL, P. N., and M. J. L. STIASSNY. 1991. The freshwater fishes of Madagascar: a study of an endangered fauna with recommendations. Conservation Biology. 5:231-243.

RIIPI, M., R. V. ALATALO, L. LINDSTRÖM, and J. MAPPES. 2001. Multiple benefits of gregariousness cover detectability costs in aposematic aggregations. Nature. 413:512-514.

RÖDEL, M. O. 1999. Predation on tadpoles by hatchlings of the freshwater turtle Pelomedusa subrufa. Amphibia Reptilia. May. 20:173-183.

RÖDEL, M.-O. 1998. Kaulquappengesellschaften ephemerer Savannengewässer in Westafrika. Edition Chimaira, Frankfurt a.M.

RÖDEL, M.-O., and K. E. LINSENMAIR. 1997. Predator-induced swarms in the tadpoles of an African savanna frog, Phrynomantis microps. Ethology. 103:902-914.

ROWE, L., and D. LUDWIG. 1991. Size and timimg of metamorphosis in complex life cycles: time constraints and variation. Ecology. 72:413-427.

SIEGFRIED, W. R., and L. G. UNDERHILL. 1975. Flocking as an anti-predator strategy in doves. Animal Behavior. 23:504-508.

SIH, A. 1987. Predators and prey lifestyles: an an evolutionary and ecological overview, p. 203-224. In: Predation: direct and indirect impacts on aquatic communities. W. C. Kerfoot and A. Sih (eds.). University Press of New England, Hanover, NH, USA.

SKELLY, D. K. 1992. Field evidence for a cost of behavioral antipredator response in a larval amphibian. Ecology. 71:704-708.

SKELLY, D. K., and E. E. WERNER. 1990. Behavioral and life-historical responses of larval american toads to an odonate predator. Ecology. 71:2313-2322.

SPIELER, M. 2001. Tadpole aggregation behaviour reduces predation risk. Zoology. 103:52. Amphibian communities in Madagascar Julian Glos 134

SPIELER, M., and K. E. LINSENMAIR. 1999. Aggregation behaviour of Bufo maculatus tadpoles as an antipredator mechanism. Ethology. 105:665-686.

WASSERSUG, R., and C. M. HESSLER. 1971. Tadpole behaviour: aggregation in larval Xenopus laevis. Animal Behavior. 19:386-389.

WATT, P. J., S. F. NOTTINGHAM, and S. YOUNG. 1997. Toad tadpole aggregation behaviour: evidence for a predator avoidance function. Animal Behavior. 54:865-872.

WERNER, E. E. 1986. Amphibian metamorphosis: growth rate, predation risk, and the optimal size at transformation. American Naturalist. 128:319-341.

WILBUR, H. M. 1997. Experimental ecology of food webs: complex systems in temporary ponds. Ecology. 78:2279-2302.

WILBUR, H. M., and J. P. COLLINS. 1973. Ecological aspects of amphibian metamorphosis. Science. 182:1305-1314.

WOLTERS, S., and K. ZUBERBÜHLER. 2003. Mixed-species associations of Diana and Campbell's monkeys: the costs and benefits of a forest phenomenon. Behaviour. 140:371-385. Amphibian communities in Madagascar Julian Glos 135

Appendix

Aquatic Zebras? - The tadpoles of the Madagascan treefrog Boophis schuboeae, compared to those of B. ankaratra

J. Glos1,2,4, M. Thomas3, and M. Vences2.

1 Department of Animal Ecology and Tropical Biology, Biocenter, Würzburg University, Am Hubland, 97074 Würzburg, Germany 2 Technical University Braunschweig, Zoological Institute, Spielmannstr. 8, 38106 Braunschweig 3 University of Cologne, Department of Genetics, Zülpicher Str. 47, 50674 Cologne, Germany

ABSTRACT

The treefrogs Boophis schuboeae and B. ankaratra occur at different elevations in the mountain rainforest of central Madagascar and are considered to be sibling species within the Boophis luteus group. Adults within this group differ in call characteristics and the mitochondrial 16S rRNA gene but, considering external criteria, are very similar to one another. The key to identifying sibling species can also be morphological characters of other life stages. Therefore, we analysed the morphology and ecology of larval stages to distinguish between these closely-related species and discuss the unusual pigmentation pattern of B. schuboeae in the context of predator avoidance.

Amphibian communities in Madagascar Julian Glos 136

INTRODUCTION

The genus Boophis is a highly diverse group of treefrogs endemic to Madagascar. Most of the approximately 50 Boophis species are restricted to rainforests, but several are found in open or disturbed areas (e.g., B. doulioti, B. goudoti) or dry forests (e.g., B. xerophilus). Most species occur at

600–1200 m rather than in low or high elevations (CADLE 2003). Boophis has been subdivided into a number of phenetic species groups (BLOMMERS-SCHLÖSSER 1979, BLOMMERS-SCHLÖSSER & BLANC

1991, GLAW & VENCES 1994, VENCES & GLAW 2005); some probably are monophyletic while others are not (VENCES et al. 2002, VENCES & GLAW 2005). The Boophis luteus group contains green- coloured treefrogs that, considering external criteria, are very similar to one another. Only a single species was recognized by BLOMMERS-SCHLÖSSER & BLANC (1991). Since bioacoustic methods were applied for the last fifteen years, this species group has experienced a dramatic increase in recognized species diversity, and ten new species have been described (GLAW & VENCES 1992, ANDREONE 1993,

GLAW & THIESMEIER 1993, ANDREONE et al. 1994, GLAW & VENCES 1994, ANDREONE 1996). The validity of these species has been confirmed by applying molecular methods (VENCES et al. 2002 and unpublished data), and these methods have demonstrated a rather high genetic differentiation among these taxa. Within the B. luteus group, B. ankaratra and B. schuboeae are considered to be sibling species. They occur at different elevations in fairly intact parts of the mountain rainforest of central Madagascar. B. ankaratra is known from seven localities (Manjakatompo/Ankaratra = type locality, Ambohitantely, Mandraka, Col des Tapias, Itremo, Andringitra, Antoetra), all of which are high elevation sites distributed over central Madagascar. B. schuboeae appears to be restricted to mid- elevations. This frog is only known from neighbouring sites (Ambatolahy and Vohiparara) in the Ranomafana National Park in central Madagascar. The two species were observed at night along streams or brooks and are morphologically very similar as adults. The species differ in call characteristics and have a relevant differentiation in the mitochondrial 16S rRNA gene (GLAW &

VENCES 2002). The analysis of the morphology and ecology of larval stages can help to clearly distinguish between closely-related species that are similar in adult morphology. In the majority of Madagascan anurans, information on morphology and ecology of tadpoles is scarce. In this paper, we describe the morphology of the tadpoles of B. ankaratra and B. schuboeae and discuss the unusual pigmentation pattern of B. schuboeae.

Amphibian communities in Madagascar Julian Glos 137

MATERIAL AND METHODS

Tadpoles were collected in the field, euthanised by immersion in chlorobutanol solution, and immediately sorted into homogeneous series based on morphological characters. From each series one specimen was selected and a tissue sample from its tail musculature or fin preserved in 99% ethanol. This specimen is here named "DNA voucher". After tissue collection, all specimens were preserved in 4% formalin. Specimens were deposited in the Zoologische Staatssammlung München, Germany (ZSM); comparative specimens were examined from the herpetological collection of the Zoological Museum Amsterdam, Netherlands (ZMA). Tadpoles were identified using a DNA barcoding approach based on a fragment of the mitochondrial 16S rRNA gene known to be sufficiently variable among species of Malagasy frogs

(THOMAS et al. 2005). The 550 bp fragment was amplified using primers 16Sa-L and 16Sb-H from

PALUMBI et al. (1991) and standard protocols, resolved on automated sequencers, and compared to a nearly-complete database of sequences of adult Madagascan frog species (VENCES et al. 2005). Identification was considered to be unequivocal when the tadpole sequence was 99-100% identical to an adult specimen from the same geographical region and not more similar to any sequence from another species. DNA sequences were deposited in Genbank (accession numbers DQ068393- DQ068398; accession numbers of comparative adult specimens included in the sequence sets AY847959-AY848683 and AJ315909-AJ315913). Morphological features of the tadpoles were compared following the terminology of

MCDIARMID & ALTIG (1999). For comparative purposes we give also the labial tooth row formula

(LTRF) according to Dubois (1995). Developmental stages are defined according to GOSNER (1960). Measurements of preserved specimens were taken with a stereo microscope and a measuring ocular, and drawings were done with a camera lucida. For measured variables see Table 1. For the preparation of tadpole mouthparts for the SEM (scanning electron microscope), the front parts of the tadpoles’ bodies were detached with a surgical blade, and samples were dried (critical point drying). SEM was done at the ZMA. Amphibian communities in Madagascar Julian Glos 138

TABLE 1. Measurements (mm) of tadpoles of Boophis schuboeae and B. ankaratra (mean ± SD, range). Staging after GOSNER (1960), measurements follow MCDIARMID & ALTIG (1999). BH = body height, BL = body length, BW = body width, ED = eye diameter, IND = internarial distance, IOD = interorbital distance, MTH = maximum tail height, NED = naris – eye distance, ODW = oral disc width, SED = snout – eye distance, SND = snout – naris distance, SSD = snout – spiracle distance, TAL = tail length, TL = total length, TMH = tail musculature height, TMHM = tail musculature height at midlength of tail, TMW = tail muscle width. Measurements and ratios were compared using Mann– Whitney U–Tests; Z = test statistic, P = probability. For α = 0.05, significant differences are indicated in bold. There are no significant differences when Bonferroni correction of α is applied. One specimen of B. schuboeae and two specimens of B. ankaratra had damaged tails; in these specimens, TL, TAL, TMH, TMW, MTH, Dorsal and ventral fin height, and TMHM was not determined.

B. schuboeae (n = 12) B. ankaratra (n = 9) Z P Stage 25.0 ± 0.0 25.0 - 25.0 25.3 ± 1.0 25.0 - 28.0 -1.11 0.27 TL 18.2 ± 2.5 14.8 - 23.3 20.3 ± 4.4 16.6 - 28.5 -1.41 0.16 BL 7.1 ± 1.4 5.9 - 10.6 8.3 ± 2.5 5.7 - 11.7 -1.1 0.27 TAL 11.5 ± 1.7 9.0 - 15.0 13.0 ± 2.5 10.9 - 17.4 -1.5 0.13 BL/TL 0.37 ± 0.02 0.35 - 0.40 0.36 ± 0.02 0.32 - 0.39 -1.69 0.09 BL/BW 1.83 ± 0.13 1.64 - 2.07 1.82 ± 0.07 1.65 - 1.91 -0.04 0.97 BL/BH 2.13 ± 0.11 1.97 - 2.32 2.11 ± 0.21 1.79 - 2.40 -0.04 0.97 BW/BH 1.16 ± 0.07 1.00 - 1.25 1.16 ± 0.09 1.05 - 1.31 -0.36 0.72 TMH 2.03 ± 0.31 1.41 - 2.52 2.81 ± 0.96 1.91 - 4.31 -1.88 0.06 TMW 1.69 ± 0.33 1.23 - 2.46 2.32 ± 0.73 1.29 - 3.38 -2.21 0.03 TMHM/MTH 0.59 ± 0.05 0.50 - 0.69 0.52 ± 0.05 0.44 - 0.58 -2.59 0.01 Dorsal/ventral fin 1.68 ± 0.44 1.00 - 2.40 1.57 ± 0.67 1.00 - 2.80 -0.68 0.49 TMH/BH 0.64 ± 0.05 0.53 - 0.70 0.71 ± 0.09 0.57 - 0.85 -1.76 0.08 TMW/BW 0.45 ± 0.04 0.38 - 0.52 0.51 ± 0.04 0.42 - 0.57 -2.47 0.01 ED 1.04 ± 0.23 0.87 - 1.72 1.06 ± 0.30 0.78 - 1.50 -0.71 0.48 NED 0.95 ± 0.16 0.74 - 1.23 1.04 ± 0.41 0.62 - 1.60 -0.32 0.75 NED/BL 0.14 ± 0.01 0.13 - 0.16 0.12 ± 0.01 0.11 - 0.14 -2.35 0.02 NED/SND 0.40 ± 0.03 0.36 - 0.46 0.36 ± 0.05 0.28 - 0.43 -1.95 0.05 SND/SED 0.78 ± 0.07 0.70 - 1.00 0.79 ± 0.06 0.72 - 0.88 -0.36 0.72 IOD/IND 1.83 ± 0.12 1.59 - 2.00 1.95 ± 0.25 1.64 - 2.47 -1.43 0.15 ODW 3.24 ± 0.51 2.77 - 4.18 3.90 ± 1.31 2.63 - 5.84 -0.64 0.52

Amphibian communities in Madagascar Julian Glos 139

RESULTS

Boophis schuboeae Glaw and Vences 2002

Two tadpole series were genetically identified as belonging to Boophis schuboeae: ZSM 823/2004 (one specimen; original field number FGMV 2002.1790; DQ068393) and ZSM 812/2004 (many specimens), 817/2004 (11 specimens), 826/2004 (one specimen) (all with original field number FGMV 2002.1804; DQ068395). All specimens were collected in one pool of a big cascade right under a relatively high waterfall in the Ranomafana National Park, diameter about 7 m and depth more than 3 m (21°15.77’S 47°24.78’E, 846 m above sea level). The pool has a stony bottom and rocky walls, and tadpoles were attached to rocks in a very strong current. The following description is based on 12 tadpoles in ZSM 817/2004 and 826/2004. SEM pictures in Fig. 1 are based on specimen ZSM 826/2004. In Table 1, we summarize data for the 12 tadpoles measured, which all were in stage 25. The general appearance is that of a typical suctorial brook-living tadpole. Its body is elongated ovoid in lateral and dorsal view (Fig. 1A, 1B), narrow and relatively flat (“streamlined”). The tail is compressed, long and muscular, with only low dorsal and ventral fins. The oral disc is very prominent. Body length 0.37 ± 0.02 (mean ± SD) of total length, body 1.83 ± 0.13 times longer than wide, 2.13 ± 0.11 times longer than high, 1.16 ± 0.07 times wider than high, snout rounded in dorsal and lateral profile, eyes medium sized and directed dorsolaterally. Nostrils dorsal, apertures directed dorsolaterally, nostrils closer to eyes than to snout tip (ratio naris-eye distance/snout-naris distance 0.40 ± 0.03), spiracle sinistral, at midbody, just below midline, inner wall free from body, aperture opens posteriorly. Medial vent tube without lateral displacement, short. Tail fins convex, higher from the body to the middle of the tail, dorsal fin higher than ventral fin (ratio dorsal fin height/ventral fin height 1.68 ± 0.44). Tail tip rounded. Origin of dorsal fin distinctly behind the base of tail. Oral disc ventral, Labial Tooth Row Formula (LTRF) 8(5-8)/3 (Fig. 1C). LTRF according to Dubois (1995) 4:4 + 4/3. Tooth density 85/mm in anterior labium, 76/mm in posterior labium, no obvious difference in tooth density between tooth rows within the anterior and posterior labium. Marginal papillae conical, with rounded tips, not pigmented; length of marginal papillae 0.11 mm, density 23 papillae/mm. Several rows of marginal papillae interrupted anteriorly, not interrupted posteriorly. Upper jaw sheath M-shaped, lower jaw sheath V-shaped, both sheaths fully pigmented. Jaw sheath appears to be composed of a series of fused columns; the surfaces of these sheaths are irregular. In life as well as in preservation a distinct banded pigmentation pattern is visible. There is one heavily pigmented band at eye level, one band at the posterior part of the body, and two bands on the tail. Caudal musculature is heavily pigmented in these parts and not pigmented otherwise. Dorsal and Amphibian communities in Madagascar Julian Glos 140 ventral tail fins are not pigmented. In life, the belly has partly a distinct golden shine, and the gut is black coloured.

A D

B E

C F

FIG. 1. Lateral view (A), dorsal view (B) and oral disc (C) of Boophis schuboeae tadpole and of Boophis ankaratra tadpole (D, E, F) at stage 25. Scale bar represents 5 mm (A, B, D, E) and 1 mm (C, F).

Boophis ankaratra Andreone 1993

Three tadpole series were genetically identified as belonging to Boophis ankaratra: ZSM 814/2004 (two specimens; original field number FGMV 2002.1697; DQ068396), ZSM 816/2004 (one specimen; original field number FGMV 2002.1698; DQ068397), ZSM 819/2004 (six specimens; original field number FGMV 2002.1699; DQ068398). All specimens of these series were collected in Imaitso forest close to the village Ambalamarina, Andringitra Massif at an elevation of about 1400 m (coordinates Amphibian communities in Madagascar Julian Glos 141 not taken in the forest but at a nearby locality: 22°07.09’S 46°55.53’E). Tadpoles were found in a stony, rushing brook with a lot of pools and rapids and a width about 3 m. As summarized in Table 1, eight tadpoles were in stage 25 and one tadpole was in stage 28. As in B. schuboeae, the general appearance is that of a typical suctorial brook-living tadpole. Its body is elongated ovoid in lateral and dorsal view (Fig. 1D, 1E), narrow and relatively flat (“streamlined”). The tail is compressed, long and muscular, with only low dorsal and ventral fins. The oral disc is very prominent. Body length 0.36 ± 0.02 of total length, body 1.82 ± 0.07 times longer than wide, 2.11 ± 0.21 times longer than high, 1.16 ± 0.09 times wider than high, snout smoothly rounded in dorsal and rounded in lateral profile, eyes medium sized, and directed dorsolaterally. Nostrils dorsal, apertures directed dorsolaterally, nostrils closer to eyes than to snout tip (ratio naris-eye distance/snout-naris distance 0.36 ± 0.05), spiracle sinistral, at midbody, just below midline, inner wall free from body, aperture opens posteriorly. Medial vent tube without lateral displacement, rather short. Tail fins convex, higher from the posterior end of vent tube to the middle of the tail, dorsal fin higher than ventral fin (ratio dorsal fin height/ventral fin height 1.57 ± 0.67. Tail tip rounded. Origin of dorsal fin distinctly behind the base of tail. Oral disc ventral, LTRF 8(5-8)/3 (Fig. 1F). LTRF according to Dubois (1995) 4:4 + 4/3. Tooth density 68/mm in anterior labium, 58/mm in posterior labium, no obvious difference in tooth density between tooth rows within the anterior and posterior labium. Marginal papillae conical, with rounded tips, largely not pigmented; faint pigmentation of papillae anteriorly; length of marginal papillae 0.17 mm, density 14 papillae/mm. Several rows of marginal papillae interrupted anteriorly, not interrupted posteriorly. Upper jaw sheath M-shaped, lower jaw sheath V-shaped, both sheaths fully black pigmented. Jaw sheath appears to be composed of a series of fused columns; the surfaces of these sheaths are irregular. Caudal musculature is moderately pigmented at the base of the tail, pigmentation diminishing towards the tail tip, dorsal tail fin very faintly pigmented, ventral tail fin not pigmented.

Significant interspecific differences. At the tail base, the tail muscle in B. ankaratra is absolutely and relatively wider than in B. schuboeae. At midlength of the tail, however, tail muscle is higher in B. schuboeae. Additionally, the eyes are relatively closer to the nostrils in B. ankaratra (Table 1).

Amphibian communities in Madagascar Julian Glos 142

DISCUSSION

Boophis schuboeae and B. ankaratra are sibling species that are restricted to different elevations on the high plateau of central Madagascar. Adults of these species differ in call and genetics, but morphological differentiation is low. However, the key to identifying sibling species can also be morphological characters of other life stages. For example, adults of the neotropical skipper butterfly genus Astraptes are very similar, but the caterpillars are easily distinguishable (HEBERT et al. 2004). In B. schuboeae and B. ankaratra, in contrast, we found a high similarity among larval characters. Tadpoles of both species are exotrophic and suctorial, and are morphologically typical for brook- inhabiting species. They have elongated bodies that are narrow and flat (“streamlined”). Their tail is compressed, long and muscular, with only low dorsal and ventral fins. The oral disc of both species is very prominent, as is characteristic for suctorial species living in brooks. Its large size, high number of tooth rows and uninterrupted rows of marginal papillae is seen as an adaptation for feeding and staying in place in this habitat (MCDIARMID & ALTIG 1999). The jaw sheaths appear to be composed of a series of fused columns, and the irregular surfaces of these sheaths are in stark contrast to the smooth surfaces of typical forms usually found in this area. Within the genus Boophis, this peculiar feature occurs only in suctorial forms, and its function is unknown (MCDIARMID & ALTIG 1999). Even though we did detect significant statistical differences in larval morphology between the two species, in particular concerning the tail muscle, we consider these differences as small, and they may be undetectable in naturally varying populations. Most prominently, however, there are striking differences in pigmentation patterns between the two species. While that of B. ankaratra is not unusual for members of this genus (BLOMMERS-SCHLÖSSER & BLANC 1991), Boophis schuboeae tadpoles are characterized by four heavily pigmented bands that are visible both in the lateral and dorsal view. This conspicuous pattern might be very useful for determining preliminary species identity in the field. The unusual pigmentation presumably represents an adaptation for predator avoidance. Banded patterns are known to break up body outlines to avoid detection and subsequent pursuit by visual predators (e.g., in African zebras, reviewed by RUXTON 2002, see ALTIG &

CHANNING 1993 for hypotheses referring to tadpoles). This is in particular true for poor light conditions as are typical for aquatic habitats. Moreover, this effect is not necessarily restricted to vertebrate predators. Recent experiments support the theory that a primary function of stripes (in zebras) is to provide cryptic protection not from big cat predators but from insects (tsetse flies;

RUXTON 2002). Accordingly, the striped pigmentation pattern in B. schuboeae tadpoles might not only protect from vertebrate predators (fish, turtles) but also from aquatic invertebrates (e.g., water beetles, dragonfly and damselfly larvae). Tadpoles of B. ankaratra and of B. schuboeae have a strikingly different coloration, although the two species are closely related (GLAW & VENCES 2002). In B. ankaratra, we observed the non- striped colour in specimens of rather different size from the Andringitra Massif and as well in Amphibian communities in Madagascar Julian Glos 143 specimens very probably belonging to this species collected at Manjakatompo, the type locality. The

Manjakatompo tadpoles had been superficially described and figured by GLAW & VENCES (1994). In B. schuboeae, we examined many specimens from the Ranomafana area (series ZSM 812/2004), and the peculiar pattern was seen in all individuals independent of size. This indicates that the differences are indeed constant and species-specific. As known so far, B. ankaratra and B. schuboeae inhabit generally the same type of habitat that is streams or brooks in primary forests of higher altitudes. The tadpoles’ morphology of both species seems to be well adapted to this habitat. However, the known localities of Boophis schuboeae are on lower altitudes than those of B. ankaratra. There are currently no data available on differences of predator identities and/or density between these two habitats. It is therefore conceivable, that the peculiar zebra pattern of pigmentation in B. schuboeae tadpoles is an adaptation to a specific predator or to a general higher predation risk in this lower elevation habitat.

Amphibian communities in Madagascar Julian Glos 144

REFERENCES

ALTIG R. & CHANNING A. 1993. Hypothesis: Functional significance of colour and pattern in anuran tadpoles. Herpetological Journal 3: 73-75.

ANDREONE F. 1993. Two new treefrogs of the genus Boophis (Anura: Rhacophoridae) from central- eastern Madagascar. Bollettino del Museo Regionale di Scienze Naturali di Torino 11: 289- 313.

ANDREONE F. 1996. Another new green treefrog, Boophis anjanaharibeensis n. sp. (Ranidae: Rhacophorinae), from northeastern Madagascar. Aqua (Journal of Ichthyology and Aquatic Biology) 2: 25-32.

ANDREONE F., NINCHERI R. & PIAZZA R. 1995. Un nouveau Boophis vert (Ranidae: Rhacophorinae) des forets pluviales du S. Madagascar. Revue Française d’Aquariologie et Herpétologie 21: 121-127.

BLOMMERS-SCHLÖSSER R.M.A. 1979. Biosystematics of the Malagasy frogs. II. The genus Boophis (Rhacophoridae). Bijdragen Dierkunde 49: 261-312.

BLOMMERS-SCHLÖSSER R.M.A. & BLANC C.P. 1991. Amphibiens (première partie). Faune de Madagascar 75: 1-379.

CADLE J.E. 2003. Boophis, pp. 916-919. In: Goodman S.M. & Benstead J.P. The Natural History of Madagascar. Chicago and London: University of Chicago Press.

DUBOIS A. 1995. Keradont formulae in anuran tadpoles: proposals for a standardization. Journal of Zoological Systematics and Evolutionary Research 33: 1-15.

GLAW F. & THIESMEIER B. 1993. Bioakustische Differenzierung in der Boophis luteus-Gruppe (Anura: Rhacophoridae), mit Beschreibung einer neuen Art und einer neuen Unterart. Salamandra 28: 258-269.

GLAW F. & VENCES M. 1992. Zur Kenntnis der Gattungen Boophis, Agylyptodactylus und Mantidactylus (Amphibia: Anura) aus Madagaskar, mit Beschreibung einer neuen Art. Bonner zoologische Beiträge 43: 45-77.

GLAW F. & VENCES M. 1994. A Fieldguide to the Amphibians and Reptiles of Madagascar. Köln: M. Vences & F. Glaw Verlags GbR, 480 pp.

GLAW F. & VENCES M. 2002. A new cryptic treefrog species of the Boophis luteus group from Madagascar: bioacoustic and genetic evidence. Spixiana 25: 173-181.

GOSNER K.L. 1960. A simplified table for staging anuran embryos and larvae with notes on identification. Herpetologica 16: 183-190.

HEBERT P.D.N, PENTON E.H., BURNS J.M., JANZEN D.H. & HALLWACHS W. 2004. Ten species in one; DNA barcoding reveals cryptic species in the neotropical skipper butterfly Astraptes fulgerator. Proceedings of the National Academy of Sciences USA 101: 14812-14817. Amphibian communities in Madagascar Julian Glos 145

MCDIARMID R.W. & ALTIG R. 1999. Tadpoles: The Biology of Anuran Larvae. Chicago: The University of Chicago Press, 444 pp.

PALUMBI S.R., MARTIN A., ROMANO S., MCMILLAN W.O., STICE L. & GRABOWSKI G. 1991. The Simple Fool's Guide to PCR, Version 2.0. Privately published, University Hawaii.

RUXTON G.D. 2002. The possible fitness benefit of striped coat coloration in zebra. Mammal Review 32: 237-244.

THOMAS M., RAHARIVOLOLONIAINA L., GLAW F., VENCES M. & VIEITES D.R. 2005. Montane tadpoles in Madagascar: molecular identification and description of the larval stages of Mantidactylus elegans, M. madecassus and Boophis laurenti from the Andringitra Massif. Copeia 2005: 174-183.

VENCES M., ANDREONE F., GLAW F., KOSUCH J., MEYER A., SCHAEFER H.C. & VEITH M. 2002. Exploring the potential of life-history key innovation: brook breeding in the radiation of the Malagasy treefrog genus Boophis. Molecular Ecology 11: 1453-1463.

VENCES, M. & GLAW F. 2005. A new cryptic frog of the genus Boophis from the northwestern rainforests of Madagascar. African Journal of Herpetology 54: 77-84.

VENCES M., THOMAS M., VAN DER MEIJDEN A., CHIARI Y. & VIEITES D.R. 2005. Comparative performance of the 16S rRNA gene in DNA barcoding of amphibians. Frontiers in Zoology 2: article 5.

Amphibian communities in Madagascar Julian Glos 146

Zusammenfassung

EINLEITUNG Madagaskar, die viertgrößte Insel der Welt, ist ein ‘Reliquienschrein der Evolution’, ein unwiederholbares natürliches Experiment (Whitmore, 1993). Seit ihrer Abspaltung vom ostafrikanischen Festland vor etwa 165 Millionen Jahren verhinderte der Indische Ozean weitgehend das Einwandern später in Afrika evoluierter Arten und sorgte somit für einen hohen Grad an Isolation über eine selbst im erdgeschichtlichen Maßstab lange Zeitspanne. Die unabhängige Entwicklung der Fauna und Flora führt bis heute zu einem einzigartigen Endemismusgrad (z. B. 99 % bei Amphibien; Glaw and Vences, 2003) und macht die Insel zu einem ‘Hot Spot’ biologischer Vielfalt (Myers et al., 2000). Daher kommt ökologischen Untersuchungen auf Madagaskar ein besonderes Interesse zu. Jedoch ist die biologische Vielfalt gerade in Madagaskar durch anthropogenen Einfluss extrem gefährdet. Über 85 % des ursprünglichen Waldbestandes sind durch Brandrodung und Feuerholz- gewinnung bereits vernichtet worden und nur 11 % der Insel sind noch bewaldet (Nelson and Horning, 1993). Die Trockenwälder an der Westküste stellen hier wie in anderen Kontinenten eines der am meisten gefährdeten Ökosysteme überhaupt dar (Ganzhorn, 1991; Janzen, 1988; Smith, 1997). Die räumliche und zeitliche Heterogenität stark saisonaler Lebensräume wie des westmadagassischen Trockenwaldes stellt besondere Herausforderungen an Tiere und Pflanzen. Die Aktivitäts- und Fortpflanzungsperiode ist bei vielen Organismen in diesen Gebieten aufgrund des Klimas auf bestimmte Zeitspannen im Jahresverlauf beschränkt. Außerdem müssen unwirtliche Zeiten überbrückt werden, was häufig durch Phasen der Inaktivität geschieht. Betrachtet man beispielsweise Amphibiengemeinschaften, ist für den Reproduktionserfolg ihrer Arten und somit ihre Zusammen- setzung vor allem die Verfügbarkeit von offenem Wasser in Raum und Zeit entscheidend, welche im Jahresverlauf stark variiert. Im westmadagassischen Trockenwald ist die Reproduktion von Amphibien nur in der zwei bis vier Monate dauernden Regenzeit möglich, da nur während dieser Periode offene Gewässer vorhanden sind. In dieser kurzen Zeit müssen sich sämtliche Arten fortpflanzen, die aquatische Larvalentwicklung abschließen und genügend Reserven für das Überdauern der anschließenden Trockenperiode sammeln. Wie artenreiche Gemeinschaften in tropischen Lebensräumen koexistieren können, ist dabei noch vielfach ungeklärt (Linsenmair, 1990; 1995). Zum Verständnis dieser Problematik ist es wichtig zu klären, welche biotischen und abiotischen Faktoren Lebensgemeinschaften strukturieren und durch welche Mechanismen sie aufrechterhalten werden. Da es unmöglich ist, das gesamte Artenspektrum eines Ökosystems detailliert zu untersuchen, ist die modellhafte Betrachtung einzelner Organismengruppen bzw. von Teilhabitaten erforderlich. Aufgrund ihrer überschaubaren Diversität und taxonomischen Zugänglichkeit, geringen Migrationsfähigkeit, ihren kurzen Entwicklungszeiten und nicht zuletzt wegen ihrer leichten experimentellen Manipulierbarkeit, eignen sich larvale Anurengesellschaften in besonderer Weise als Modell für ökologische Fragestellungen. Verglichen Amphibian communities in Madagascar Julian Glos 147 mit anderen Tiergruppen reagieren Amphibien zudem besonders sensibel auf die Degradierung natürlicher Lebensräume und gelten deshalb als geeignete Bioindikatoren für den Zustand von Ökosystemen (z. B. Dunson et al., 1992; Heyer et al., 1994).

Species Behaviour conservation Species Influence of predators level Distribution in on spatial d istribution space and time and behaviour

Community Community ecology level Distributions in space and time

Taxonomy Life-history Basis Natural history General Ecology

Abb.1 : Grafische Darstellung des Aufbaus der Arbeit. Die Grundlage und Voraussetzung der Arbeit bilden taxonomische und beschreibend ökologische Studien. Ausgehend von den Studien auf Gemeinschaftsebene werden naturschutzrelevante und verhaltensökologische Aspekte an einzelnen Arten untersucht.

In meiner Arbeit habe ich mir daher dieses geeignete System zu Nutze gemacht, um taxonomische, gemeinschaftsökologische und autökologische Aspekte im westmadagassischen Trockenwald zu untersuchen (Abb. 1). Ziel dieser Arbeit war es Antworten auf die Fragen zu geben wie die einzelnen Arten die Habitate in Raum und Zeit nutzen, welchen Einfluss abiotische Parameter, Austrocknungsrisiko der Laichgewässer und Mikrohabitat haben und wie Prädatoren die Gemeinschaft und das Verhalten einzelner Arten beeinflussen. Somit trägt diese Arbeit dazu bei die grundlegenden Mechanismen zu verstehen, die die Zusammensetzung einer Lebensgemeinschaft bestimmen. Im Einzelnen untersuchte ich hierzu folgende Fragestellungen:

Aus welchen Arten bestehen die Anurengemeinschaften des westmadagassischen Trockenwaldes, und wie lassen sich diese Arten morphologisch voneinander abgrenzen? Welche Unterschiede finden sich zwischen den Arten bezüglich ihres Paarungssystems, ihrer life- history und ihrer Habitatwahl bzw. den Anpassungen an ihr Habitat?

Gibt es spezifische Kaulquappengemeinschaften, die sich anhand biotischer und abiotischer Umweltvariablen vorhersagen lassen? Unterscheiden sich die Muster der Vorhersagbarkeit von Gemeinschaften zwischen unterschiedlichen Habitattypen innerhalb eines lokalen räumlichen Skalenniveaus?

Amphibian communities in Madagascar Julian Glos 148

Wie beeinflusst das Vorkommen von Raubfeinden die Verteilung von Kaulquappen und deren Verhalten auf der räumlichen Skalenebene einzelner Laichgewässer?

Anhand welcher Umweltvariablen lässt sich die Laichplatzwahl von Anuren in diesem Habitat vorhersagen? Wie lassen sich die Ergebnisse nutzen, um Empfehlungen zum Schutz bedrohter Arten auszusprechen?

DAS SYSTEM Der Kirindy-Wald ist ein laubabwerfender Trockenwald und stellt mit etwa 12.000 ha eines der letzten größeren, zusammenhängenden Waldgebiete Westmadagaskars dar. Der Wald wird von dem Fluss Kirindy durchschnitten und ist von einer Busch-Baum-Savanne anthropogenen Ursprungs umgeben, auf der die einheimische Bevölkerung Zebu-Rinder und Ziegen hält und die regelmäßig abgebrannt wird. Der Kirindy-Wald wurde bis vor etwa zehn Jahren für selektiven Holzeinschlag genutzt. Gegenwärtig wird in dieser Region Ökotourismus etabliert. Im Untersuchungsgebiet befinden sich für Amphibien in drei verschiedenen Habitattypen geeignete Laichgewässer: im geschlossenen Wald, im Bett des saisonal Wasserführenden Flusses Kirindy (vor dem Fließen des Flusses) und in der angrenzenden Savanne (Kapitel 1). Alle Laichgewässer, bis auf wenige Restgewässer im Flussbecken, fallen in der Trockenzeit völlig trocken und füllen sich nach und nach mit den einsetzenden Regenfällen der Regenzeit. In Abhängigkeit von Zeitpunkt und Menge des Niederschlages, aber auch von verschiedenen Parametern der Gewässer selbst, wie z. B. der Größe und Beschaffenheit des Untergrundes, trocknen diese Gewässer auch innerhalb einer Fortpflanzungsperiode immer wieder unterschiedlich schnell aus. Ich konnte im Kirindy-Wald 15 Anurenarten nachweisen (Kapitel 1). Alle bis auf eine Art sind für Madagaskar endemisch (Glaw and Vences, 1994) und vier der Arten haben ein sehr eng begrenztes Verbreitungsgebiet, mit dem Kirindy-Wald als wichtigster Lokalität. In dieser Arbeit beschreibe ich eine Froschart wissenschaftlich neu (Kapitel 2). Diese Art, Scaphiophryne menabensis, ist die seltenste Froschart in ihrem Verbreitungsgebiet, und aus meiner Arbeit resultiert die dringende Empfehlung, sie in ein bestehendes Schutzkonzept für den Kirindy- Wald und seine Umgebung mit einzubeziehen (Kapitel 2; Glos and Volahy, 2004). Weiterhin beschreibe ich wissenschaftlich erstmalig in dieser Arbeit fünf Kaulquappenarten und präsentiere Daten zu Ökologie, life-history und Verhalten dieser Arten (Kapitel 2, 3 und 4). Die wissenschaftliche Beschreibung weiterer Frosch- und Kaulquappenarten ist Gegenstand noch andauernder Studien (Scaphiophryne sp., Heterixalus carbonei und H. tricolor; Revision der Kaulquappen von Scaphiophryne). Die Ergebnisse dieser Arbeit stellen damit die Basis für alle weiteren ökologischen Studien an Fröschen und Kaulquappen dieses Ökosystems dar.

Amphibian communities in Madagascar Julian Glos 149

GEMEINSCHAFTSÖKOLOGIE Eine der faszinierendsten gemeinschaftsökologischen Fragen ist die, wie sich Artengemeinschaften organisieren. Die Häufigkeit und Verteilung von Arten über Raum und Zeit, und damit die Zusammensetzung der Gemeinschaften, wird von vielen Faktoren bestimmt (z. B. Parris, 2004; Woodward, 1983). Gemeinschaften können dabei in deterministischer oder stochastischer Weise strukturiert werden (Huston, 1994). In deterministischen Gemeinschaften werden die Arten- zusammensetzung und andere Eigenschaften durch bestimmte Umweltfaktoren bedingt und können somit vorhergesagt werden. Ähnliche Habitate beherbergen demnach ähnliche Gemeinschaften. Stochastische Gemeinschaften hingegen lassen sich nicht durch Umweltfaktoren vorhersagen, und die Zusammensetzung der Arten wird zu einem großen Teil durch Zufallsereignisse bestimmt. Entscheidend, ob eine Gemeinschaft als deterministisch oder stochastisch eingestuft wird ist aber auch das räumliche Skalenniveau der entsprechenden Studie (z. B. Wright et al., 1993). Muster, die auf lokaler Ebene gefunden werden, können von solchen abweichen, die auf regionaler oder globaler Ebene auftreten. Aber auch innerhalb eines räumlichen Skalenniveaus können deutlich unterscheidbare Habitattypen vorkommen. Dies kann eine weitere Unterteilung nötig machen, da jeder dieser Habitattypen möglicherweise unterschiedliche Muster in der Zusammensetzung der Gemeinschaften zeigt. Wenn also die lokale Situation nicht berücksichtigt wird, kann eine ganz- heitliche Analyse über verschiedene Habitattypen hinweg eventuell vorhandene Muster in einzelnen Habitattypen überdecken. Der Kirindy-Wald repräsentiert ein solches System unterschiedlicher Typen von Laichgewässern für Anuren und damit unterschiedlichen Habitattypen für deren Kaulquappen. Die adulten Frösche pflanzen sich entweder in temporären Gewässern im Wald oder in Tümpeln im Flussbett fort, das durch den Wald mäandert (Kapitel 1). Meine Arbeit zeigt, dass beide Habitattypen sich zum einen in ihrer zeitlichen Verfügbarkeit und ihren biotischen und abiotischen Eigenschaften und zum anderen in der Artenzusammensetzung ihrer Kaulquappengemeinschaften unterscheiden (Kapitel 5). Ob letztere in einer deterministischen oder einer stochastischen Weise organisiert sind, hängt im Kirindy-Wald in der Tat davon ab, welchen Habitattyp man betrachtet. In den Waldtümpeln sind die Artengemeinschaften mithilfe biotischer und abiotischer Habitatvariablen gut vorhersagbar, also deterministisch zusammengesetzt. Die geographische Distanz zwischen den Gewässern hat keinen Einfluss auf die Artenzusammensetzung und den Artenreichtum. Die Gemeinschaften der Gewässer des Flussbettes hingegen sind weder durch Habitatvariablen noch durch die geographische Distanz zu anderen Gewässern vorhersagbar und somit stochastischer Natur. Diese Unterschiede in den Mustern der Gemeinschaften der verschiedenen Habitattypen sind nicht durch das Vorkommen unterschiedlicher Arten in den beiden Habitaten bedingt. Eine Analyse nur derjenigen Arten, die in beiden Habitattypen vorkommen, bestätigte das bereits gefundene Muster der Gemeinschaftszusammensetzung. Dies zeigt, dass dieselben Arten, oder möglicherweise sogar dieselben Individuen, unterschiedliche Entscheidungen bei der Laichplatzwahl Amphibian communities in Madagascar Julian Glos 150 treffen, je nachdem in welchem Habitattyp sie sich befinden. Weiterhin unterstreicht dieser Befund, wie wichtig es ist, die lokale Situation und die Diversität innerhalb eines Habitats in gemeinschafts- ökologischen Studien zu berücksichtigen.

NATURSCHUTZ Innerhalb Madagaskars haben illegaler Holzeinschlag, Brandrodung und die damit einhergehende Habitatfragmentierung zu fast komplett isolierten Waldblöcken geführt (Janzen, 1988; Smith, 1997). Die größten verbleibenden Trockenwaldfragmente sind in der Region Menabe im Westen Madagaskars gelegen. Selbst optimistische Prognosen gehen nicht davon aus, dass der heute noch bestehende Wald in seinem jetzigen Zustand vollständig geschützt werden kann. Vielmehr sind weite Teile des Waldes durch den Druck der Bevölkerung aus den umliegenden Dörfern und durch illegal arbeitende Holzfäller aus der nächsten Stadt stark gefährdet (Sorg et al., 2003). In einer pragmatischen Sichtweise sollten sich aus diesem Grund Forschung und Schutz auf diejenigen Gebiete innerhalb des Waldes konzentrieren, die aufgrund besonderer Kriterien als besonders schützenswert gelten. Solche Kriterien könnten beispielsweise allgemeine Artendiversität sein, das Vorkommen ausgewählter Arten von besonderem Interesse, sogenannter „flagship“ Arten bzw. die diesen Arten entsprechenden ökologischen Ansprüche. Ein solches artenorientiertes Naturschutzkonzept wurde innerhalb des letzten Jahrzehnts intensiv von den in dieser Region engagierten Naturschutzorganisationen verfolgt (v. a. DWCT, CI). Im Fokus standen hierbei hauptsächlich Wirbeltierarten, deren Verbreitung auf die Menabe-Region begrenzt ist (Riesenspringratte Hypogeomys antimena, Sommer and Hommen, 2000; Sommer et al., 2002; Flachschwanzschildkröte Pyxis planicauda, Bloxam et al., 1996; Gibson and Buley, 2004; Berthe’s Zwergmausmaki Microcebus berthae). Amphibien wurden in dieses Natur- schutzkonzept dagegen bisher nicht miteinbezogen. Dies ist bedauerlich, da sie hochgradig gefährdet sind, sowohl innerhalb der Menabe-Region als auch weltweit (z. B. Andreone et al., 2005). Außerdem sind sie als Bioindikatoren für den Zustand eines Ökosystems sehr gut geeignet (e.g., Welsh and Ollivier, 1998) und nehmen nicht zuletzt eine wichtige Rolle für das Funktionieren eines Ökosystems ein (z. B. als Räuber oder Beute in der Nahrungskette). Die Habitatansprüche von Amphibien decken sich aber in der Regel nicht unbedingt mit denen größerer Vertebraten. Es ist daher nicht zwingend zu erwarten, dass Amphibien gleichzeitig von Schutzmaßnahmen profitieren können, wenn das Design und Management von geschützten Gebieten auf die Ansprüche größerer Vertebraten ausgerichtet ist. Aus meiner Arbeit resultiert die explizite Empfehlung, Amphibien in die Naturschutz- aktivitäten in der Region Menabe mit einzubeziehen. Aus diesem Grund untersuchte ich die Habitat- ansprüche einer der am meisten gefährdeten Froscharten, Aglyptodactylus laticeps (EN nach IUCN- Kriterien; Andreone et al., 2005). Diese Art ist endemisch innerhalb der Menabe-Region und kommt nur in relativ ungestörten Teilen des Waldes vor (Glos and Volahy, 2004). Dort ereicht sie lokal hohe Dichten und hat deswegen vermutlich eine bedeutende Funktion im Ökosystem. Weiterhin zeichnet sie sich durch eine außergewöhnliche life-history und Ökologie aus (Kapitel 1 und 3). Ihre Amphibian communities in Madagascar Julian Glos 151

Entwicklungszeit bis zur Metamorphose ist mit minimal zehn Tagen eine der kürzesten bei Anuren überhaupt. Die Kaulquappen dieser und einer weiteren Art (Aglyptodactylus securifer) zeigen zudem eine starke Plastizität in der Entwicklungsgeschwindigkeit und im Wachstum bei verschiedenen Umweltbedingungen (Kapitel 3; unpubl. Daten). In meiner Arbeit untersuchte ich die Laichplatzwahl dieser Froschart, da diese einen entscheidenden Faktor für die erfolgreiche Reproduktion bei Amphibien und deswegen auch für den Schutz dieser Art darstellt (Kapitel 6). Mit Hilfe eines logistischen Regressionsmodells analysierte ich die Beziehung der Laichplatzwahl zu abiotischen und biotischen Variablen, die das Laichgewässer beschreiben. Die Auswahl der verwendeten Variablen richtete sich explizit nach praktischen Gesichtspunkten. Es wurden nur solche Variablen in das Modell integriert, die kosten- und zeiteffektiv aufgenommen werden können und deswegen im praktischen Naturschutz sinnvoll anwendbar sind. Das Modell identifizierte Variablen, mit denen sich das Vorkommen von A. laticeps gut vorhersagen lässt, nämlich die Bedeckung des Gewässerbodens mit Laubstreu und die Bedeckung der Oberfläche mit stehender Vegetation und mit Schwimmpflanzen. Diese leicht zu bestimmenden Parameter können somit genutzt werden, um ein geeignetes Habitat für A. laticeps zu identifizieren und Empfehlungen auszusprechen, wie ein weniger geeignetes Habitat verbessert werden kann bzw. durch welche Maßnahmen ein gestörtes Habitat wiederhergestellt werden kann. Diese Ergebnisse zeigen, dass Modelle, die auf grundlegenden ökologischen Daten beruhen, wertvolle Hilfsmittel im angewandten Naturschutz sein können. Auch wenn sich solche deskriptiven Habitatmodelle sehr gut eignen, um Voraussagen über das Vorkommen einer Art zu treffen (voraussagender Aspekt), sind jedoch kausale Rückschlüsse über den Zusammenhang von Habitatwahl und Umweltparametern nicht immer möglich (erklärender Aspekt). So kann in meinem System aufgrund des Modells nicht mit Bestimmtheit festgestellt werden, ob die am besten korrelierten Variablen zur Vorhersage der Laichplatzwahl von A. laticeps auch tatsächlich die entscheidenden für die Auswahl eines Laichgewässers sind oder ob diese z. B. nur mit den ursächlichen Variablen autokorreliert sind. Auch können diese Modelle nur unzureichend dynamische Faktoren wie z. B. die Besiedlungsgeschichte eines Gewässers durch intra- oder interspezifische Konkurrenten berücksichtigen. Um diese Verknüpfungen zu überprüfen, testete ich in einem Folgeversuch (Kapitel 7) experimentell, ob eine der gefundenen Variablen mit hoher Vorhersagefähigkeit, nämlich die Bedeckung des Gewässerbodens durch Laubstreu, tatsächlich entscheidend für die Laichplatzwahl der Frösche ist. Zusätzlich testete ich eine weitere, dynamische Variable auf ihren Einfluss auf die Laichplatzwahl hin, nämlich die Besiedlungsgeschichte der Gewässer, d. h. ob die Gewässer bereits von anderen Kaulquappen der gleichen Art belegt waren. In diesem Versuch zeigte sich, dass der Faktor Laubstreu an sich keinen direkten Einfluss auf die Laichplatzwahl von A. laticeps hat. Offensichtlich ist dieser Faktor mit einem entscheidenden anderen Faktor (z. B. dem Bedeckungsgrad der Baumkronenschicht) autokorreliert, der das Verhalten Amphibian communities in Madagascar Julian Glos 152 der Frösche beeinflusst. Dieses Ergebnis unterstreicht die Notwendigkeit von experimentellen Studien, wenn Aussagen über kausale Zusammenhänge zwischen Modellvoraussagen und Modellvariablen, in diesem Fall Laichplatzwahl und Umweltfaktoren, getroffen werden sollen. Zum anderen zeigte der Versuch, dass Gewässer, in denen sich bereits arteigene Kaulquappen befinden, für die weitere Eiablage gemieden werden. Dies hebt die Bedeutung auch von dynamischen Faktoren in ökologischen Systemen, hier auf die Laichplatzwahl von Fröschen, hervor. Meine Arbeit unterstreicht also, dass experimentelle Studien wichtige Werkzeuge sind, um die erklärende und prädiktive Aussagekraft von Habitatmodellen zu ergänzen und zu erweitern, und dass die aus experimentellen Studien gezogenen Schlussfolgerungen in diejenigen Empfehlungen zum Management und Schutz von Arten einbezogen werden müssen, die aus Habitatmodellen gewonnen werden.

RÄUBER-BEUTE INTERAKTIONEN Prädation beeinflusst entscheidend die Fitness eines Individuums, und viele Tiere zeigen spezielle Verhaltensanpassungen, um Räubern zu entgehen. In vielen Fällen wird es als vorteilhaft für die Raubfeindvermeidung angesehen, wenn sich potentielle Beuteorganismen mit anderen Individuen der gleichen oder einer anderen Art zu größeren Aggregationen zusammenschließen (Diamond, 1981; Siegfried and Underhill, 1975). Verschiedene Mechanismen könnten der Grund dafür sein, warum dies zu einem verminderten individuellen Prädationsrisiko führen könnte, beispielsweise durch den Verdünnungseffekt (Hamilton, 1971), den Verwirrungseffekt (Heller and Milinski, 1979), den Trafalgar-Effekt (Siegfried and Underhill, 1975) oder den Lerneffekt, wenn einige Gruppenmitglieder giftig oder ungenießbar sind (Brodie and Formanowicz, 1987). Bei Kaulquappen ist Aggregationsverhalten aus sieben Anurenfamilien und aus verschiedenen geographischen Regionen bekannt. Feldstudien haben gezeigt, dass dieses Verhalten eine direkte Reaktion auf die Anwesenheit von Raubfeinden ist (Rödel and Linsenmair, 1997; Watt et al., 1997) und bzw. oder auf chemische Substanzen, die von den Raubfeinden selbst oder von verletzten Kaulquappen abgesondert werden (so genannten Schreckstoffen) (Kats and Dill, 1998; Spieler and Linsenmair, 1999). Bisherige Studien über das Aggregationsverhalten von Kaulquappen beschäftigten sich mit Gruppen, die aus Individuen von nur einer Art gebildet werden. Meine Arbeit ist die erste über Kaulquappenaggregationen, die von zwei oder mehr Arten gebildet werden (Kapitel 8). Solche Aggregationen sind außerhalb des Kirindy-Waldes nur von einer weiteren Stelle, die ebenfalls innerhalb Madagaskars liegt, bekannt. Insgesamt bestanden in den Gewässern des Kirindy- Waldes 73 % der gefundenen Aggregationen aus mehr als einer Art und aus maximal vier Arten innerhalb einer Aggregation. Diese Aggregationen bestanden aus bis zu mehreren tausend Individuen und bildeten sich als Reaktion auf räuberische Vertebraten, v. a. Fische. Mit Hilfe eines logistischen Regressionsmodells konnte ich zeigen, dass die Wahrscheinlichkeit für die Bildung einer solchen Aggregation zudem von biotischen und abiotischen Gewässerfaktoren abhängt. So waren Aggregationen häufiger in tiefen, klaren Gewässern mit einer geringen Laubstreuschicht anzutreffen, Amphibian communities in Madagascar Julian Glos 153 d. h., dass dann bevorzugt Aggregationen gebildet wurden, wenn bessere Sichtverhältnisse für Räuber und weniger Versteckmöglichkeiten für Kaulquappen vorherrschten. Die Dichte von invertebraten Raubfeinden spielte hingegen nur eine geringe Rolle. Aglyptodactylus laticeps (Mantellidae) und Dyscophus insularis (Microhylidae) waren mit Abstand die häufigsten Arten, die in Aggregationen aus mehreren Arten und häufig gemeinsam anzutreffen waren. Dies ist umso erstaunlicher, als die beiden Arten in ihrer Nahrungsökologie und Mikrohabitatwahl äußerst verschieden sind. Kaulquappen von Aglyptodactylus laticeps fouragieren bevorzugt einzeln am Gewässerboden an Laubstreu (Kapitel 4), während D. insularis Kaulquappen Filtrierer sind, die sich zumeist in der Mitte der Wassersäule aufhalten (Grosjean etc. in Vorbereitung; Glaw and Vences, 1994). In einer experimentellen, vergleichenden Studie (Kapitel 9) untersuchte ich die proximaten Mechanismen, die eine Aggregationsbildung bei diesen beiden Arten auslösen. Die Reize, auf die beide Arten mit Gruppenbildung reagierten, unterschieden sich deutlich: A. laticeps reagierte nur auf Schreckstoffe, sowohl auf eigene als auch auf artfremde (Dyscophus insularis), nicht jedoch auf die Anwesenheit verschiedener Raubfeinde. Somit ist seine Reaktion nicht spezifisch gegen bestimmte Raubfeinde gerichtet. D. insularis hingegen reagierte spezifisch auf Fische, entweder auf diese direkt oder auf chemische Substanzen, die von ihnen abgesondert wurden. Aggregationsverhalten bietet jedoch nicht nur Vorteile durch ein vermindertes Prädationsrisiko, sondern verursacht auch Kosten. So können Tiere in großen Gruppen vermutlich weniger Nahrung aufnehmen als einzelne Individuen (Krause and Ruxton, 2002) oder sind einer größeren Parasitenbelastung ausgesetzt (Brown and Brown, 1986). Zudem sind Kosten dann wahrscheinlich, wenn mit dem Zusammenschluss zu einer Gruppe ein Wechsel des Mikrohabitats erfolgt, wie ich für A. laticeps zeigen konnte (Kapitel 8 und 9). In einem Experiment untersuchte ich deshalb an A. laticeps die Effekte des Aggregationsverhaltens auf Wachstum und Entwicklung. Es zeigte sich, dass dauerhaft aggregierende Kaulquappen von A. laticeps gegenüber nicht-aggregierten Kaulquappen kleiner waren und sich langsamer entwickelten. Da der Zeitpunkt der Metamorphose und die Metamorphosegröße wichtige Faktoren in der life-history von Amphibien sind (Rowe and Ludwig, 1991; Wilbur and Collins, 1973), zeigen diese Ergebnisse, dass das Aggregationsverhalten bei Kaulquappen in der Tat auch Fitnesskosten verursacht. Kaulquappen verschiedener Arten in den Gewässern des Kirindy-Waldes können also sehr flexibel auf den oft sehr hohen Raubfeinddruck reagieren. Sie bilden große Schwärme als Reaktion auf solche Reize hin, die ein Prädationsrisiko anzeigen, auch wenn die genauen auslösenden Reize von Art zu Art verschieden sind. Der Bildung von Aggregationen könnte demnach eine Schlüsselrolle in der Besiedlung von denjenigen Gewässern zukommen, die ein potentiell hohes Prädationsrisiko bieten. Aggregationen, die aus mehreren Arten bestehen, könnten den Individuen aller beteiligten Arten einen noch höheren Schutz bieten als Schwärme aus nur einer Art, z. B. indem die Gruppenmitglieder von den unterschiedlichen Fähigkeiten der Arten profitieren, einen Raubfeind frühzeitig zu erkennen.

Amphibian communities in Madagascar Julian Glos 154

FAZIT Die Amphibienfauna Madagaskars ist einzigartig, und sie stellt ein aufregendes Feld für ökologische Fragestellungen dar, sowohl als eigenständiges System betrachtet als auch als Modell für andere Systeme. Umso mehr verwundert es, dass bislang kaum detaillierte ökologische Studien an diesem System durchgeführt wurden. Die vorliegende Arbeit schafft zunächst mit der taxonomischen Beschreibung der vorkommenden Arten die Basis für ökologische Fragestellungen und zeigt dann auf den Ebenen sowohl der Gemeinschaft als auch einzelner Arten, wie verschiedene Umweltfaktoren die Verteilung von Anuren in Raum und Zeit beeinflussen. Es zeigt sich, dass sowohl statische Eigenschaften der Gewässer als auch dynamische Faktoren wie Raubfeinde oder das Vorhandensein anderer Kaulquappen die Verteilung der Arten auf verschiedenen räumlichen Skalenebenen sowie deren Verhalten beeinflussen. Somit tragen die Ergebnisse dieser Arbeit dazu bei, die grundlegenden Mechanismen zu verstehen, die die Zusammensetzung der Lebensgemeinschaften in diesem Ökosystem bestimmen. Nicht zuletzt ermöglichen diese Erkenntnisse, geeignete, artenorientierte Schutzkonzepte für diese in ihrer Existenz stark bedrohte Anurengemeinschaft zu entwickeln und die Effekte von Habitatzerstörung auf diese Gemeinschaft aufzuzeigen.

Amphibian communities in Madagascar Julian Glos 155

LITERATUR

ANDREONE, F., J. E. CADLE, N. A. COX, F. GLAW, R. A. NUSSBAUM, C. J. RAXWORTHY, S. N.

STUART, D. VALLAN, and M. VENCES. 2005. Species review of amphibian extinction risks in Madagascar: conclusions from the global amphibian assessment. Conservation Biology. 19:1790-1802.

BLOXAM, Q. M. C., J. L. BEHLER, E. R. RAKOTOVAO, H. RANDRIAMAHAZO, K. T. HAYES, S. J.

BRODIE, E. D., and D. R. FORMANOWICZ, JR. 1987. Antipredator mechanisms of larval anurans: Protection of palatable individuals. Herpetologica. 43:369-373.

BROWN, C. R., and M. B. BROWN. 1986. Ectoparasitism as a cost of coloniality in cliff swallows (Hirundo phyyhonota). Ecology. 67:1206-1218.

DIAMOND, J. M. 1981. Mixed-species foraging groups. Nature. 292:408-409.

DUNSON, W. A., R. L. WYMAN, and E. CORBETT. 1992. A symposium on amphibian declines and habitat acidification. Journal of Herpetology. 26:349-352.

GANZHORN, J. U. 1991. Forstprojekt im Trockenwald Madagaskars. Biologie in unserer Zeit. 21:293- 297.

GIBSON, R. C., and K. R. BULEY. 2004. Biology, captive husbandry, and conservation of the malagasy flat-tailed tortoise, Pyxis planicauda Grandidier, 1867. Herpetological Review. 35:111-116.

GLAW, F., and M. VENCES. 1994. A Fieldguide to the Amphibians and Reptiles of Madagascar. M. Vences & F. Glaw Verlags GbR, Köln. —. 2003. Introduction to amphibians. In: The Natural History of Madagascar. S. M. Goodman and J. P. Benstead (eds.). University of Chicage Press, Chicago and London.

HAMILTON, W. D. 1971. Geometry for the selfish herd. Journal of Theoretical Biology. 31:295-311.

HELLER, R., and M. MILINSKI. 1979. Optimal foraging of sticklebacks on swarming prey. Animal Behavior. 27:1127-1141.

HUSTON, M. A. 1994. Biological Diversity The coexistence of species on changing landscapes. Cambridge University Press, Cambridge. Amphibian communities in Madagascar Julian Glos 156

JANZEN, D. H. 1988. Tropical dry forests: the most endangered major tropical ecosystem, p. 130-137. In: Biodiversity. E. O. Wilson (ed.). National Academy Press, Washington, D.C., Washington, D.C., USA.

KATS, L. B., and L. M. DILL. 1998. The scent of death: chemosensory assessment of predation risk by prey animals. Écoscience. 5:361-394.

KRAUSE, J., and G. D. RUXTON. 2002. Living in groups. Oxford University Press.

LINSENMAIR, K. E. 1990. Tropische Biodiversität: Befunde und offene Probleme. Verhandlungen der Deutschen Zoologischen Gesellschaft. 83:245-261. —. 1995. Tropische Biodiversität - vom ursprünglichen Verständnis der strukturellen Komplexität noch weit entfernt. Rundgespräche der Kommission für Ökologie. 10:115-126.