HARVARD UNIVERSITY l©») Ernst Mayr Library of the Museum of Comparative Zoo logy OCCASIONAL PAPERS of the '^ NATURAL HISTORY MUSEUM The University of Kansas Lawrence, Kansas^ number 176, pages 1-59 28 february 1996 Natural History and Resource Use of Four Amazonian Tadpole Assemblages Erik Russell Wild Division ofHerpetology Natural History Museum and Department of Systematics and Ecology The University of Kansas Lawrence, Kansas 66045-2454, USA ABSTRACT Four tadpole assemblages were studied during the 1989-90 rainy season at the Reserva Cuzco Amazonico located on the Rio Madre de Dios, in Amazonian Peru. The ponds varied in size, depth, permanency, light penetration, alkalinity, hardness, and number of anuran and non-anuran species present but were similar in water temperature, dissolved oxygen, and pH. While sets of these characteristics appeared correlated among ponds, none of the characteristics was related to tadpole species richness, evenness, or diversity. Netting sampled 3820 tadpoles of 19 species, plus 1413 individuals of one species of crab, two species of aquatic insects, and 10 species of fish. Patterns of differential use among anuran species were found for most of the resource dimen- sions. Macrohabitat and time within the rainy season appeared to be the most important resource dimensions, followed by food (represented by ecomorphological guild), microhabitat (level in the water column), and diel time period. Some species demonstrated associations between microhabitat, diel time period, and develop- ment. Tadpoles of 25 other anuran species known to occur at Cuzco Amazonico were not encountered. These species probably use other macrohabitats, but not other time periods, because most aquatic habitats exist only during the rainy season. Rainfall seems to be the most influential factor in detennining the patterns of resource utilization by tadpoles, because it determines the degree to which the macrohabitats can be utilized temporally. © Natural History Museum. The University of Kansas. Lawrence. ISSN;009 1-7958 2 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Key words: Tadpoles; Assemblages; Communities; Resource utilization; Natural history; Amazonian Peru; Anurans. RESUMEN Se estudiaron cuatro agregados de renacuajos durante la estacion lluviosa de 1989-90 en la Reserva Cuzco Amazonico, localizada en el Rio Madre de Dios, Peru. Las charcas diferfan en tamaiio, profundidad, permanencia, penetracion de la luz, alcalinidad, dureza y numero de especies de anuros y no anuros presentes. pero eran similares en temperatura del agua, concentracion de oxigeno disuelto y pH. Mientas que ciertos conjuntos de estas caracteristicas mostraron correlacion entre las charcas, ninguna de estas variables estaba relacionada con la riqueza, equitabilidad o diversidad en especies de renacuajos. Se capturaron con red 3820 renacuajos de 19 especies, asf como una especie de cangrejo, dos especies de insectos acuaticos y 10 especies de peces. totalizando 1413 individuos. Se encontraron diferencias en los patrones de uso para la mayoria de los recursos utilizados por los renacuajos. El macrohabitat y el tiempo dentro de la estacion lluviosa parecen ser las dimensiones mas importantes, seguidas por el alimento (representado por gremios ecomorfologicos), microhabitat (nivel en la columna de agua), y ritmos de actividad diaria (noche o dia). Algunas especies demostraron asociaciones entre microhabitat, ritmos de actividad diaria, y estadio de desarrollo. No se encontraron renacuajos de otras 25 especies de anuros que habitan en Cuzco Amazonico. Dichas especies probablemente usan otros macrohabitats, pero no otros periodos de tiempo. pues la mayon'a de los habitats acuaticos existen solo durante la estacion lluviosa. La precipitacion parece ser el factor mas influyente en la determinacion de los patrones de uso de los recursos por parte de los renacuajos, porque determina hasta que grado los macrohabitats pueden ser utilizados temporalmente. Palabras cloves: Renacuajos; Agregados; Comunidades; Utilizacion de recursos; Historia natural; Amazonia peruana; Anuros. Most studies of aniiran communities have focused on the adult repro- ductive period. Many have addressed reproductive strategies and attempted to quantify adult community structure but have given tadpole assemblages little more than anecdotal attention (e.g., Inger, 1969; Crump, 1971, 1974; Creusere and Whitford, 1976; Duellman, 1978; Toft and Duellman, 1979; Aichinger, 1987; and the many studies cited in Gascon, 1991). This reflects a long outdated view that anuran communities must be organized in the adult stage because tadpoles are so siinilar (Inger and Greenberg. 1966). The little attention accorded tadpole assemblages is surprising, because it has been suggested that among anurans with a biphasic life cycle, selection is strongest during the larval stage (Blair, 1961; Heyer, 1973, 1979). Whereas both — i.e., groups of adults and of tadpoles — have been treated frequently as independent communities, factors acting at both levels can influence the anuran composition at a given locality. For this reason, throughout this study the term "community" will be reserved for references AMAZONIAN TADPOLE ASSEMBLAGES 3 to the entire set of aniiran species at the study site. The term "assemblage" will be used to refer to the set of tadpole species in a given aquatic habitat in order to emphasize that the composition of a particular assemblage is not independent of factors (including stochastic ones) acting on the adult species present. Toft (1985) reviewed studies of resource partitioning in amphibians (including larval assemblages) and reptiles, and listed six resource dimen- sions that encompass most known ways organisms can differ ecologically. As applied to tadpoles, these dimensions are: macrohabitat — distribution of tadpoles among aquatic habitats; microhabitat — spatial distribution of tadpoles within the aquatic habitat; food type — tadpole food type con- sumed; food size — tadpole food size consumed; diel time period — diel period of tadpole activity; and seasonal time — temporal distribution of tadpole occurrence, in the aquatic habitat. Any discernible nonrandom pattern in the distribution of species among resource dimensions is consid- ered assemblage structure and is indicative of differential resource usage or resource partitioning. Many studies of naturally occurring tadpole assem- blages addressed some aspect of resource partitioning; however, most examined a limited number of resource dimensions, usually seasonal time and microhabitat, and thus provide incomplete descriptions (Blair. 1961; Dixon and Heyer, 1968;Calef, 1973;Heyer, 1976, 1979; Heyer et al, 1975; Walters, 1975;WiesL 1982; Alford and Crump, 1982; Smith, 1983; Berger, 1985; Hero, 1990; Gascon, 1991). A few studies have focused specifically on resource partitioning among tadpoles (Alford, 1986; Inger et al., 1986), but only Heyer (1973, 1974) examined all resource dimensions in a single study. Most studies found habitat partitioning by seasonal time to be the most important dimension (Dixon and Heyer, 1968; Wiest, 1982), but microhabitat (Heyer, 1974. 1976; Alford and Crump. 1982; Alford. 1986; Hero, 1990) and/or macrohabitat (Heyer, 1973; Smith, 1983; Gascon, 1991) also may be important adjuncts. Food is thought to be of minor importance in resource partitioning of tadpole assemblages (Dixon and Heyer, 1968; Calef, 1973; Heyer, 1973, 1974, 1976). Only Inger et al. (1986) found seasonal time to be unimportant; spatial occurrence and food were the most important dimensions at their uniquely aseasonal stream locality. In summary of these studies on resource partitioning of amphibian assemblages. Toft reported that they differ from all other vertebrate assem- blages because seasonal time was the most important dimension in nearly every study conducted. The sequence of mean ranks of importance of each dimension among all studies summarized by Toft are: ( 1 ) seasonal time, (2) microhabitat, (3) macrohabitat, (4) food type, (5) food size, and (6) diel time period. Because only Heyer (1973; 1974) examined all of these 4 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 resource dimensions, the sequence may be biased by the frequency with which each dimension was investigated. The results of most studies on tadpole resource partitioning subsequent to Toft's review conform to her ranking. Resource partitioning has been addressed in studies of tadpole assem- blages in both temporary (Texas — Blair, 1961; Texas — Wiest, 1982) and permanent (British Columbia — Calef, 1973; Maryland — Heyer, 1979) aquatic habitats in temperate climates. Tadpole assemblages have been studied in a wide variety of tropical aquatic habitats (e.g., Mexico — Dixon and Heyer 1968; Thailand— Heyer, 1973, 1974; Costa Rica— Heyer et al, 1975; Borneo— Ingeretal., 1986; Brazil— Hero, 1990; Gascon, 1991), and comparatively between tropical and temperate habitats (e.g., Kansas/Ecua- dor— Berger, 1985; Maryland/Panama — Heyer, 1976). However, there have been few studies of tadpole assemblages of the most diverse frog fauna of the world — that found in the Amazonian Basin (e.g., Berger, 1985; Hero, 1990; Gascon, 1991), where resource partitioning has been shown to be more pronounced in comparison to temperate regions (Berger, 1985). Toft attributed resource partitioning patterns to three causes — viz., com- petition, predation, and factors that operate independently of interspecific interactions. Both descriptive studies (Heyer, 1973, 1976; Heyer et al., 1975; Cecil and Just, 1979; Smith, 1983) and experimental investigations under artificial conditions (Brockelman, 1969; Calef, 1973; Debenedictis, 1974; Walters, 1975; Woodward, 1983; Gascon, 1992) frequently have cited predation as an important influence on natural tadpole assemblages. Most descriptive studies of naturally occurring tadpole assemblages found competition to be nonexistent or present but of minor influence (Heyer, 1973, 1974, 1976; Alford, 1986; Inger et al., 1986). However, competition has been demonstrated frequently in manipulative or artificial experimental studies and, therefore, is considered potentially important in structuring tadpole assemblages (Brockelman, 1969; Wilbur, 1972, 1976; Wilbur and Collins, 1973; Debenedictis, 1974; Smith-Gill and Gill, 1978; Steinwascher, 1978; Travis, 1980; Scale, 1980). Furthermore, experimental studies also have shown that predation and competition are interrelated (Morin, 1981, 1983, 1986, 1987; Wilbur, 1982; Smith, 1983; Scott, 1990; Sredl and Collins, 1992). In addition to predation and competition, abiotic factors influence tad- pole assemblages. Of these, seasonality, (especially with regard to rainfall) clearly explains seasonal partitioning (Dixon and Heyer, 1968; Heyer, 1973; Turnipseed and Altig, 1975; Creusere and Whitford, 1976; Toft et al., 1983; Berger, 1985; Pechman et al., 1989). Other climatic and aquatic factors can be influential as well (e.g., Alford, 1986; Inger et al., 1986; Gascon, 1991 ) including water pH, fluctuations of air and water tempera- AMAZONIAN TADPOLE ASSEMBLAGES 5 ture. and humidity (Gosner and Black, 1957; Blair, 1961; Wiest, 1982). Such physical factors become more influential in structuring tadpole as- semblages with increased seasonality (i.e., latitude; Heyer, 1973). The effects of abiotic factors may be interrelated in a complex way, especially in determining the timing of adult reproduction (Savage, 1961) and, thus, temporal partitioning of the tadpole habitat. Abiotic factors also may influence interspecific associations or work in combination with them to determine tadpole assemblage structure (e.g., Heyer et al., 1975; Smith, 1983; Woodward, 1983; Scott, 1990; Sredl and ColHns, 1992). Toft (1985) concluded that no single factor explains resource partitioning among tad- poles; rather the interaction among physical factors, predation, and compe- tition produces observed patterns of resource partitioning. Moreover, the nature of the interaction among causal factors seems to vary among sites (although there hav.e been few studies of tadpole assemblages in the Ama- zonian Basin for comparison) and probably from year to year. Disparity in conclusions regarding the relative importance of the vari- ous resource dimensions and causal factors, particularly with regard to predation and competition and especially between descriptive and experi- mental studies, highlights the inherent complexity of tadpole assemblages and the difficulty of studying them (Scott and Campbell, 1982). More information is needed about naturally occurring tadpole assemblages, par- ticularly in the Amazon Basin, (Heatwole, 1982; Toft, 1985; Altig and Johnston, 1989). Initial studies should be descriptive so that patterns can be recognized (Toft, 1985). This information will provide a basis for experi- mental investigations of the underlying causes of the patterns and their interrelationships (Heyer, 1979; Scott and Campbell, 1982). The study reported herein focuses on tadpole assemblages in Amazo- nian Peru in order to address the general question of how 54 species of anurans with life cycles that include aquatic tadpoles can coexist in a tropical community. The study is primarily descriptive and the specific objectives are to (1) provide baseline ecological data and descriptive natural history of selected tadpole assemblages at a single site in Amazo- nian Peru throughout the duration of an entire rainy season; (2) examine simultaneously all resource dimensions (macrohabitat, microhabitat, sea- sonal time, diel time period, and food) previously suggested in the litera- ture to be important in inteipreting patterns of differential utilization among tadpole species; (3) examine the interrelationships among tadpole develop- ment, diel activity, and microhabitat for species with sufficient numbers; (4) examine evidence suggestive of those causal factors (competition, predation, and abiotic) potentially responsible for resource utilization pat- terns; and (5) compare these findings with other studies. 6 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 METHODS AND MATERIALS Description of Study Area The study was conducted from 10 December 1989 to 31 March 1990 during the rainy season at the Reserva Cuzco Amazonico. a tourist lodge and reserve on the north bank of the Rio Madre de Dios, about 15 km ENE of Puerto Maldonado, Provincia de Tambopata, Departamento de Rio Madre de Dios. Peru (12°33' S, 69°03' W). The western boundary of the reserve is a slow-moving stream, the Quebrada Mariposa, that empties into the Rio Madre de Dios. In the adjacent forest, there was a network of about 10 km of trails that included two study zones of marked contiguous quadrats (Fig. 1 ). A detailed description of Cuzco Amazonico and a discus- sion of the biological investigations conducted there are provided by Duellman and Koechlin (1991). The climate at Cuzco Amazonico is seasonally tropical. Records from the nearest continuously operating weather station in Puerto Maldonado during 1971-1989 indicate that the rainy season extends from October through March, with the heaviest rainfall in January and February; the least amount of rain falls in June and July. The mean annual rainfall is 2416 mm with annual extremes of 1844-3718 mm. Temperature records from 1978- 1989 reveal that October is the hottest month, with a mean monthly maximum of 32.2°C. A distinct cool season corresponds to the driest months of the year (May-July). Ambient humidity peaks at night, reaching 92-99% in both seasons, and is lowest near midday. The region is mapped as humid tropical forest, but is situated near the transition between humid tropical forest and dry tropical forest (Tosi, 1960). The mostly evergreen forest comprises terra firma and seasonally inundated forests, which differ floristically. There are fewer ferns and less herbaceous ground cover in inundated forest than in terra firma forest, but more Heliconia and Calathea. Except in areas of dense growths of these plants, the understory is more open in inundated forest than in terra firma forest (Duellman and Koechlin, 1991). Studies by Duellman and his colleagues have produced a thorough inventory of the fauna at Cuzco Amazonico. The herpetofauna includes 81 species of reptiles and 64 species of anuran amphibians (Duellman and Salas, 1991 ). Of the species of frogs, 54 possess life cycles that include an aquatic tadpole stage. To insure the presence of tadpoles and to permit consistent sampling and accessibility, the choice of study ponds was based on three general require- ments. First, there had to be breeding anurans. Second, the pond had to possess, or exhibit the potential to possess, water for a long enough period of time to allow complete development of at least some tadpole cohorts. AMAZONIAN TADPOLE ASSEMBLAGES cuzco AMAZONICO '^Sg Slanding water ^^S in rainy season 0 100 200 300 meters *"*«^:Va, Fig. \. Map of Cuzco Amazonico showing locations of study ponds: Camp Cocha, Swamp A-1 130, Swamp A- 1250. and Pool B-350. (Modified from a draw- ing by Linda Trueb.) Third, the pond had to be reasonably close to camp and accessible. Some of the ponds initially chosen eventually failed to meet these requirements and proved unsuitable. Four ponds of varied characteristics proved adequate for long-term study (Fig. 1). These are: (1) Camp Cocha (Fig. 2) — a permanent Heliconia swamp drained by a small creek. This pond is situated 25 m from the edge of the camp clearing, where it creates a large (75 x 400 m) opening in the terra firma forest canopy. Three 15 x 1-m transects were cleared for sampling in the center of the south end of the swamp. (2) Swamp A-1150 (Fig. 3) — a large ( 150 x 800 m), permanent swamp in inundated forest. There are few large trees in this swamp and the canopy is incomplete. The understory varies from large open areas to impenetrable dense growths of shrubs and vines. Three 15-m long transects were se- lected for sampling; these nearly traversed the southern third of the swamp. (3) Swamp A-1250 (Fig. 4) — a medium-sized (15 x 100 m), temporarily filled depression in inundated forest. The vegetation resembles at of the adjacent forest with an almost completely enclosed canopy of trees and an understory consisting of scattered shrubs. Three 15-m long transects ex- tending lengthwise down the middle of the swamp were sampled. (4) Pool B-550 (Fig. 5) — a small (2x5 m), ephemeral pool situated 20 m from the bank of the Quebrada Mariposa. The pool is located in terra firma forest UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Fig. 2. The Heliconia swamp. Camp Cocha, on 10 March 1990 near the location of the study transects. Fig. 3. Swamp A- 1150 on 10 March 1990 showing site of the sampling transect through the deep pool. AMAZONIAN TADPOLE ASSEMBLAGES Fig. 4. Swamp A- 1250 showing part of sampling transect. (Photo by William E. Duellman. 2 February 1986.) Fig. 5. Pool B-550 on 1 1 March 1990. 10 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 with an almost completely enclosed canopy and moderately dense under- brush. Samples were taken from three 1-m long transects across the width of the pool. Field Procedures The sampling regime followed herein is modified from that of Heyer (1976). Each pond was sampled at three transects with a dipnet near midday and midnight every second day. Sampling of a transect consisted of three passes, one each at the surface, the midwater, and the bottom. The surface usually was sampled first to minimize disturbance of the lower levels; the sample was made by skimming about the top 10 cm of the water column. The second and third passes were made midway between the surface and bottom, and along the bottom, respectively; there was some overlap in net passes when the water was low. Each of the transects at Camp Cocha, Swamp A- 1150, and Swamp A- 1250 was 15 m long. Each pass consisted of five 3-m sweeps with a pentagonal dipnet (base 40 cm, sides 21 cm) composed of 5-mm mesh. It became apparent early in the study it became apparent that smaller tadpoles were not being captured; therefore, a cheese-cloth covering was sewn inside the coarse mesh to reduced the mesh size to approximately 1 mm. Because of the smaller size of Pool B-550 each transect was 1 m long and each pass through a level could be spanned in a single sweep with a 12 x 15-cm rectangular net with mesh of 1 mm. The sampling regime for Pool B-550 differed from that for other ponds and no attempt was made to standardize volumes sampled or account for sampling bias. Tadpoles netted in each sweep were identified, counted, and staged. All tadpoles were released unless their identity was unknown, in which case some individuals were preserved as vouchers in 10% buffered formalin (the method of Altig, 1970), whereas others were reared to identifiable froglets. Tadpole stage was estimated using the staging table of Gosner (1960) reproduced in Duellman and Trueb (1986); because of the inaccu- racy inherent in field-staging, Gosner stages were combined into the fol- lowing categories: 25. 26-28, 29-31. 32-34, 35-37, 38-40, 41-43. The data were recorded in the field with a microcassette recorder and later transcribed into a data book. Non-anuran organisms (e.g., fish, aquatic insects, and crabs) thought to be potential predators or competitors, were identified, measured for total length to the nearest millimeter, and counted. Most of these organisms were released, although small samples were collected as vouchers and preserved in 10% buffered formalin for later identification. The preserved tadpoles and non-anuran organisms were deposited in The University of Kansas Natural History Museum, and the Museo de Historia Natural at the Universidad Nacional Mayor de San Marcos in Lima, Peru. AMAZONIAN TADPOLE ASSEMBLAGES 1 1 Additional samples were collected frequently in disparate parts of the study ponds to assess the efficacy of the netting procedure in collecting all species present. These additional tadpoles were included in the data analy- sis only in determining the temporal occurrence of tadpoles. The small size of Pond B-550 permitted a complete and thorough netting at the end of the study period, and all tadpoles were preserved. Ponds other than the four study ponds were visited frequently as potential study sites and to obtain additional specimens to aid in tadpole identification. A variety of abiotic factors was measured in order to characterize each aquatic habitat and to identify factors influencing tadpole distributions among ponds. Water temperature and depth were recorded at each tadpole sampling. Midwater temperature was measured at each transect and aver- aged. Water depth was measured with a wooden stake calibrated in centi- meters that was permanently installed in the deepest part of each pond near the transects. Thus, the mean depths reported are mean depths at these stakes through time. Light penetration was measured in centimeters during each midday sampling at each depth stake using a Secchi disk. These values are a measure of the amount of the water column receiving sunlight (not turbidity of the water) and are reported as percentages of the depth of the pond. The use of percentages allows relative comparisons of light penetration among ponds of different depths and was necessary because the disk was frequently visible on the bottom in some ponds making compari- sons of absolute measurements meaningless (i.e., reporting 25 cm light penetration in two different ponds, one 100 cm deep and another 25 cm deep, hides the fact that the entire water column of the second pond receives light whereas the first pond is only partially illuminated). Water temperature, pH, dissolved oxygen, hardness, and alkalinity were measured with a HachT^'^ water-test kit. Water tests were performed initially on samples taken every dawn and dusk in order to detect diel changes in water chemistry. However, late in January a shift was made to every other dawn and dusk (days opposite of tadpole sampling) in order to test all four ponds for the entire season. Water samples were taken from an arbitrarily determined but consistent location near the middle of each study pond. Samples were taken by submerging 500-ml polyethylene bottle to middepth, letting it fill, and capping it underwater so that no air was trapped in the container. The samples were transported to the field laboratory within 1 hr. for completion of the tests. Pond water temperature was measured at midwater coincidentally with water-sample collection. Daily rainfall was measured with a standard rain gauge located in the camp clearing. To monitor weather fluctuations, ambient air temperature and humidity were recorded continuously throughout the study with a hygrothermograph located in the field laboratory. 12 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Statistical Analyses Inteipond comparisons were made using richness, evenness, and diver- sity indices. Rarefaction, rather than direct species counts, was used as a richness measure because of unequal sample sizes among ponds (Krebs. 1989). Rarefaction values and their standard deviations were calculated with the program accompanying Krebs (1989). Modified Hill's ratio (Alatalo, 1981) and HilFs numbers were used as evenness and diversity measures, respectively; these calculations were performed with programs accompanying Ludwig and Reynolds (1988). The G-test statistic (Sokal and Rohlf, 1981) was used to test distributions among microhabitats and between diel time periods and was performed with BIOM (Rohlf, 1985). Relationships among the variables of developmental stage, level in the water column, and diel activity of tadpoles were examined by Log-Linear Model Analysis using BMDP (Dixon, 1981) at The University of Kansas Computer Center. Niche breadths were calculated using the measure of Levins (1968), as standardized by Hurlbert (1978), because it is one of the more frequently used niche breadth measures. Niche overlap was calcu- lated using the measure of Morista ( 1959) because it minimizes bias caused by variation in sample size and number of resource states (Smith and Zaret, 1982). Both of the niche metrics were calculated using modifications of the programs accompanying Krebs (1989). RESULTS Natural History The rainy season of 1989-90 was much drier than the average for the previous 18 yr. (Fig. 6). The average rainfall for December-March was 1355.4 ± 402.6 mm (820.4-2221.1 for 1971-89), whereas the rainfall for these months during 1989-90 was 978.8 mm. The discrepancy between the average rainfall each month and the rainfall for the months of the study period was least in December, when slightly more than average rain fell in 1989-90 (307.0 mm; .r = 291.0 ± 101.1). The discrepancy between the monthly average and that for the study months increased dramatically during the course of the rainy season, with the greatest difference in March (107.2 mm; J = 287.9 mm), (Fig. 6). The amount of rain falling in March 1990 was exactly the average amount that falls in May (x = 107.2 mm); hence, the rainy season was essentially abbreviated by 2 mo. Although the most rain fell in December, the single 24-hr period with the greatest rainfall was 10 February with 127.6 mm, followed by 28 December with 53.6 mm, 30 January with 49.6 mm, and 13 February with 42.2 mm. Although there were three periods in December with at least three AMAZONIAN TADPOLE ASSEMBLAGES L^ 500- — o- — ■• - - 1971-88 - 1 989-90 -p 400- ^ 300- 'c 'as (14.1)^ / (17.9)1 r ~ ~<>-. [(21) 1 (23) ^^16.0) (20) ^ ,(12.8) °- 200- f J-/ d \ \ N 1 \ 100- f ^^ (20) • Kl 0- 1 1 1 1 — 1 1 1 1 1 o D J Month M M Fig. 6. Mean monthly rainfall tor 1971-1988 at Puerto Maldonado (O) and monthly rainfall for December 1989-March 1990 at Cuzco Amazonico (•). Verti- cal bars represent ± 1 SD. and the number of days with rain are in parentheses (means for Puerto Maldonado). 30 2 20 o Q- E •- lOH -tz±Tizt 1 978-88 : Maximum ! Minimum 1 989-90 ■ — Maximum » Minimum — r- A O D J F I M T A M Month Fig. 7. Monthly mean maximum (□, ■) and minimum (O, •) temperatures for 1978-1988 (averaged over years) at Puerto Maldonado (□, O) and for Decem- ber 1989-March 1990 at Cuzco Amazonico (■. •). consecutive days with substantial rain, because these were among the first rains; the water was absorbed quickly by the soil and did not accumulate in ponds. Rainy periods in late January and mid-February seemed to saturate 14 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Table 1. Summary of tadpoles sampled from the four study ponds. Camp Swamp Swamp Pool Species of tadpoles Cocha A-1150 A- 1250 B-550 Dendrobatidae Colostethus marchesianus — 1 11 Hylidae Hyla brevifrous 1 3 3 Hyla fasciata 1 Hyla koechlini 1 52 Hyla cf. leali 1 2 6 Hyla leucophyllato 1 12 — Hyla pan'iceps — 10 38 Hyla sp. — 2 3 — Phyllomedusa tomopterna — — 653 Phyllomedusa vaillanti 2318 55 Scinax icterica — 1 299 Leptodactylidae LeptodacTylus mystaceiis 14 Microhylidae Altigius alios 1 Chiasmocleis ventrimaculata 109 Ctenoph)-yne geayi 3 Elachistocleis oralis 38 Hamptophiyne boliviana 39 3 76 Unknown Species A 1 — Species D — — 61 Total number of species 7 8 12 5 Total number of tadpoles 2363 78 610 769 the soil, created pools, and raised water levels in the swamps. The number of days with measurable rain was surprisingly higher than average for all months of the study, with January having the most rainy days (Fig. 6). It is possible that this difference is the artificial result of a greater precision of measurement at Cuzco Amazonico or because of a true weather difference between the Reserva Cuzco Amazonico and Puerto Maldonado. Monthly maximum and minimum temperatures were less and greater than average, respectively (Fig. 7). The known anuran community at Cuzco Amazonico consists of 64 species representing seven families (Duellinan and Salas, 1991). Eight species {Eleiitherodactylus spp. and Pipa pipa) have direct development, and two {Adenomeni andreae and A. hylaedactyla) have tadpoles develop- ing in terrestrial foam nests. The remaining 54 species have an aquatic AMAZONIAN TADPOLE ASSEMBLAGES 15 Table 2. Summary of non-anuran species sampled from the four study ponds. Camp Swamp Swamp Pool Species Cocha A- 11 50 A- 1250 B-550 Trichodactylidae sp. 187 11 5 10 Belostomatidae Bclostoma sp. 110 115 1 Nepidae sp. 74 102 3 Callichthyidae Callichthys callichthys 10 11 — Con'doras sp. 2 — Characidae Species A 72 — Species B 73 — Species C 33 — Cyprinodontidae Cynolebias sp. 316 45 1 Pterolebias sp. 144 38 — Gasteropelecidae Carnegiella sp. 38 — Lebiasinidae Pyrhiilina sp. 4 6 1 Species A 1 — Total number of species 7 13 5 1 Total number of individuals 845 547 11 10 tadpole stage. A total of 3820 tadpoles of 19 species (35% of the species with tadpoles) representing four families (Dendrobatidae, Hylidae. Leptodactylidae, and Microhylidae) was sampled from the four study ponds throughout the duration of the study (Table 1 ). Tadpoles of two species {Hyla allenorwn and H. fasciata) that were previously unknown, and a new species of microhylid and its tadpole {Altigius alios) are de- scribed elsewhere (Wild, 1992, 1995). In addition to the tadpoles, several other organisms that are recognized as potential predators or competitors were encountered in the four study ponds throughout the study period (Table 2), as follows: 213 individuals of one species of crab (Trichodactylidae); 405 individuals of two species of aquatic insects (Nepidae and Belostomatidae); and 795 individuals of 10 species of fish representing five families (Callichthyidae, Characidae. Cyprinodontidae, Gasteropelecidae. and Lebiasinidae). The species of fish include known predatory species {Callichthys callichthys, Cynolebias sp., Pterolebias sp. and Pryhiilina sp.) that probably feed on tadpoles, but some of the fish species (Corydoras sp. and Carnegiella sp.) are not predatory at all. 16 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 o o o (0 O ^ ctl ._^ C r ■2 cc :i 1 1 til o -Q 1 -C D -2 1 Q- co 0) (0 ■D c S CD O o 1 tn 3 ■5 is 2 o a. j: j: ^ -C -C G < 0 O a. E ... o - o = Q Depth (mm) o O o o o o o O c\j o o o o CD in S '^ CO CM y— c3 Q o o U £ u 4J C ^c n _o .2 > -o c (U > c c X5 c o N o XI "o c 3 -O CJ 3 O <-' O =" — >-< a o I- <^ C D- c °° s "^ (U c AMAZONIAN TADPOLE ASSEMBLAGES 17 Table 3. Abiotic characteristics of the four study ponds. Values for water tempera- ture, dissolved oxygen, pH, alkalinity, and total hardness are given as means ± 1 SD, and for the latter four characteristics the first value is at dusk and the second at dawn. Camp Swamp Swamp Pool Characteristics Cocha A-1150 A- 1250 B-550 Width X length (m) 75 X 400 150x800 15 X 100 2.3 X 5.2 Depth 452 ± 94 885 ± 178 416 ± 109 144 ±32 No. days with water 103 104 45 88 Light penetration 76 ± 1 8% 55 ± 13% 85 ±21% 90 ± 19% Water temperature 25.4 ± 0.7 25.1 ±0.6 25.3 ± 0.6 25.8 ± 1.0 (°C day/night) 25.2 ± 0.8 25.0 ± 0.7 25.4 ± 0.6 25.4 ± 0.9 Dissolved oxygen 6± 1 7± 1 6± 1 7± 1 (mg/1) 6± 1 6± 1 7± 1 7± 1 pH 6.5 ±0.1 6.6 ±0.2 6.7 ± 0.3 6.6 ±0.2 6.5 ±0.1 6.5 ± 0.2 6.7 ± 0.3 6.7 ±0.3 Alkalinity 37 ±9 30 ±8 63 ±21 43 ± 18 (mg/1 CaC03) 42 ±8 30 ±7 64 ± 17 48 ± 16 Total hardness 49 ± 10 42 ± 10 68 ± 17 49 ± 12 (mg/1 Ca C03) 56 ± 12 41 ± 12 73 ± 14 53 ± 19 Characteristics of Study Ponds Camp Cocha. — Abiotic characteristics (Table 3): Cainp Cocha is a large Heliconia (Musaceae) swamp situated 25 m from the camp clearing where it creates a large opening in the canopy of terra firma forest. The swamp supports a dense growth of Heliconia and a few low shrubs. Although there are few living trees in the swamp there are numerous dead trunks in the water. The swamp is drained by a small creek at the southern end that varied in flow from totally dry to a rushing torrent after heavy rains. Standing water was present for the entire 103 days of the study period (1 1 December 1989-23 March 1990). Although the swamp is per- manent, observations during July 1989, when there were only shallow puddles of standing water, indicate that water is reduced substantially during the dry season. During the rainy season of 1990, the swamp filled to a maximum size of about 75 x 400 m. The mean water depth throughout the study period was 452 ± 94 mm; the shallowest depth was 200 mm on 13 December (the first day of sampling) and the greatest depth was 750 mm on 1 1 February (the day after the heaviest rainfall). Water temperature varied little between day (.f = 25.4 ± 0.7°C) and night (x = 25.2 ± 0.8°C), and throughout the study period varied from 23.0-27. 0°C. The mean depth of light penetration during the study period was 76 ± 18% of the water column. Light penetrated to the bottom (100%) on 1 1 days and penetrated less than 50% of the depth only once (48%). u UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Table 4. Species with adults (A), tadpoles (T), or both (B) encountered at the four study ponds. Camp Swamp Swamp Pool Species Cocha A- 11 50 A- 1250 B-550 Bufonidae Bufo marinus A Dendrobatidae Colostethus marchesianus B B Hylidae Hyki alleiioriiin A Hyla brevifrons B B T Hyla fasciata B A A Hyla koechlini A T B — Hyla cf. leali T B B Hyla leucopliyllata B B — Hyla parviceps A B B Hyla schubarti A — Hyla sp. T T — Osteocephalus taiiriniis A A Phyllomedusa tonioptenui B Phyllomedusa vaillanti B B Scarthyla ostinodactyla A A A Scinax garbei A Scina.x icterica B B Sphaenorhyncluis lacteiis A Leptodactylidae Ceratophrys cornuta A Leptodactyhis leptodactyloides A Leptodactyhis mystaceus T Leptodactyhis petersii A A Microhylidae Altigius alios T Cli iasfiiocleis ventrimaculata B — Ctcnoplvyne geayi — B EUichistocleis ovalis — T Hamptophryne boliviana T T B Unknown Species A T Species D T Total number of species 8 5 4 1 present only as adults (A) Total number of species 3 4 4 1 present only as tadpoles (T) Total number of species 4 4 8 4 present as adults and tadpoles (B) AMAZONIAN TADPOLE ASSEMBLAGES 19 The mean dissolved oxygen content of the water was constant between dawn and dusk (.? = 6 ± 1 mg/1) but varied from 4-10 mg/1 at dusk 29 December and 1 1 March, respectively. The water pH varied little through- out the study and was consistent from dawn to dusk ( J = 6.5 ±0.1). The mean alkalinity was slightly higher at dawn (Z=42 ± 8 mg/1) than dusk (x = 37 ± 9 mg/1), and reached extremes of 20 and 72 mg/1 at dusk 06 January and dawn 16 March, respectively. The mean total hardness was slightly higher at dawn ( J = 56 ± 12) than at dusk (A" = 49 ± 10) and reached extremes of 20 and 90 mg/1 at dusk 11 February and dawn 16 March, respectively. Biotic characteristics: Throughout the study period. 2363 tadpoles of seven species were sampled from Camp Cocha (Table 1 ). Ninety-eight percent of these were Phyllomedusa vaillanti {n = 2318), followed by the microhylid Hamptophryne boliviana with 2% (n = 39). The remaining five species {n - 6) wefe collected only rarely (Hyla fosciata twice, the others only once each). Casual observations of adult anurans present near the pond indicate that the tadpoles do not represent the same species. This may indicate an inefficient sampling technique, but this disparity between the local adult community and an associated tadpole assemblage has been documented elsewhere (Dixon and Heyer, 1968; Heyer, 1973; Gascon, 1991 ). Furthermore, additional netting in disparate parts of the study ponds rarely produced species not found in the sampling transects. These few additional tadpoles were included in the data analysis only in determining the temporal occurrence of tadpoles. Adults, but no tadpoles, of the follow- ing species were present: Biifo marinus, Hyla allenoruni, H. koechlini, H. schubarti, Leptoclactylus petersii, Scarthyla ostinodactyla, Scinax garbei, and Sphaenorhynchus lacteus (Table 4). Although tadpoles of//, boliviana and Altigius alios were found, no adults were observed. One tadpole. Species A, could not be identified. Tadpoles were present for essentially the entire study period (Fig. 8). Tadpoles off. vaillanti were taken on the first (13 December) and last (23 March) days of sampling as well as most days in between. Apparently these represent the product of three breeding bouts, the tadpoles of which demonstrate synchronous development; the first bout started prior to 13 December, another around 03 February, and the third prior to 15 March. Tadpoles of Hamptophryne boliviana were first col- lected on 15 December, after the first of two breeding bouts, the second being around 2 January; //. boliviana tadpoles were not found after 28 January. Tadpoles of each of the remaining species were the result of a single breeding bout and were encountered at various times throughout February and March, but mostly after the heaviest rainfall on 10 February. Seven non-anuran species were collected in net samples from Camp Cocha, as follows: 187 individuals of the trichodactylid crab species; 184 of the two species of insects; and 474 of seven species of fish (Table 2). All of these, except the fishes Callichthys callichthys and Pyrhiilina sp., were present throughout the study period. 20 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Swamp A-1150. — Abiotic characteristics (Table 3): Swamp A- 1 1 50 is a large, permanent swamp in inundated forest. Few large trees are present, and the canopy is incomplete. The understory varies from wide, open areas to impenetrable dense growths of shrubs and vines. The presence of a small (3x5 m) pool about 0.5 m deep in July 1989 indicates that at least part of the swamp contains standing water throughout the year. Standing water was present for the entire 104 days of the study period ( 1 1 December 1989- 24 March 1990). During the rainy season of 1990 the swamp reached a maximum size of about 150 x 800 m. The mean water depth throughout the study period was 885 ± 178 mm. The shallowest depth recorded at this location was 610 mm on 24 March, the last visit, whereas the greatest depth was 1375 mm on 12 February, 2 days after the heaviest rainfall. Water temperature varied little between day ( J = 25.1 ± 0.6°C) and night (.? = 25.0 ± 0.7°C) or throughout the study period (23.0-26.5°C). The mean depth of light penetration was 55 ± 1 3% of the water column. The greatest percentage of depth that light penetrated was 89%. Light penetrated less than 50% of the water depth on 18 days, and the least penetration was 20%. The mean dissolved oxygen content of the water of Swamp A- 1 150 did not vary significantly between dawn {x - 6 ± 1 mg/1) and dusk {x = 7± 1 mg/1), and reached extremes of 4 and 1 1 mg/1 at dawn 16 December and dusk 16 March, respectively. The water pH varied little throughout the study and between dawn (.f = 6.5 ± 0.2) and dusk (J = 6.6 ± 0.2), and reached extremes of 6.0 and 7.0. The mean alkalinity was nearly identical at dawn ( .F = 30 ± 7) and dusk ( J = 30 ± 8), and reached extremes of 14 and 46 mg/1 at dawn on 12 and 21 December, respectively. The mean total hardness was nearly the same at dawn ( J = 41 ± 12) and dusk ( J=42± 10) and reached extremes of 20 and 60 mg/1 on several occasions. Biotic characteristics: During the study period 78 tadpoles of 8 species were sampled from Swamp A-1150 (Table 1 ). Seventy-one percent (n = 55) were Phyllomedusa vaillanti and 15% {n - 12) were Hyhi leucophyllata. The remaining six species (n - 11) were only taken rarely. Casual observa- tions of adults present at the site indicates that there were fewer species of anuran larva than adults. Adult Hyla fasciata, H. parviceps, Leptodactyhis petersii, Osteocephalus taurinus, and Scarthyla ostinodactyla were present, but tadpoles of none of these were collected (Table 4). Although tadpoles of Hamptophryne boliviana, Hyla koechlini, and H. cf. leali were collected, no adults were seen. One tadpole, Hyla sp., could not be identified. Tad- poles were not present for the entire study period (Fig. 9). Tadpoles of H. boliviana were the first encountered on 13 January and none was collected after 08 February. Phyllomedusa vaillanti larvae were taken on 1 5 and 1 7 January, but not again until 1 8 February; apparently these were the result of two separate breeding bouts. Hyla leucophyllata was collected periodically between 08 February and 14 March, and H. brevifrons was found once on AMAZONIAN TADPOLE ASSEMBLAGES 21 o in o If) < a E (0 if) ^ c .^^ •5 c- i» m o 1 1 0 c 1 s it "o o .2 c CD 0) E 5 ^ "c5 0) CL CO fc "S, n 3 CC CO Ct! -5 .c tc -r ^ is is is is CI j: 0. J : -c -C J ^ j: CO Depth (mm) o o o o o o o C\J ■.- O Oix) »— 1— T- O) 00 J I I L^ O O o o CO r- c^ ^ ^1 (n '^ r^ o ^ --^ ^o u _r ^ a = O "^ r CO a, - CO ■a c a :: JD ^ ., Q) c/l ^ U- 03 30 X) 1 -o c« CJ 1^1 Ui OJ "-, > ?J ^-^ , 1- I c^ ■a c JO O (U > C C3 C3 o N ■a o T3 .2 '*. Oh U 3 o O O ^ c ca P . 9J o E Q 24 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 ventrimaculata (18%, // = 109), Hamptophryne boliviana (12%, n = 76), Hyla koechlini (9%, n = 52), and Elachistocleis ovalis (6%, n = 38). Other species that were collected rarely and accounted for 6% {n = 36) are Colostethus marchesianus, Ctenophryne geayi, Hyla brevifroiis, H. cf. leali, H. pan'iceps, H. sp., and Leptodactylus mystaceus. Adult Ceratophrys coniHta, Leptodactylus leptodactyloides, Scarthyla ostinodactyla, and Hyla fasciata were present, but no larvae of these species were collected (Table 4). Although tadpoles of Elachistocleis ovalis, Hyla brevifrons, and Leptodactylus mystaceus were found, no adults were seen in the vicinity. One tadpole, Hyla sp., could not be identified. Adults of Hyla leali and H. rhodopepla were present, but their tadpoles are indistinguishable; thus, the tadpole referred to as H. cf. leali could be either or both of these taxa. A large chorus on the night of 25 January when Ctenophryne geayi, Ceratophrys cornuta, Hamptophryne boliviana, Hyla koechlini, H. pan'iceps, Leptodactylus leptodactyloides, and Scinax icterica were active seems to have signaled the initial and primary breeding activity that led to the tadpole fauna in Swamp A- 1250. Prior to February, only Leptodactylus mystaceus larvae were encountered, but by 08 February, all of the afore- mentioned species of the chorus, except Ceratophrys cornuta and Leptodactylus leptodactyloides, were present as tadpoles in addition to Chiasmocleis ventrimaculata and Elachistocleis ovalis. There were subse- quent choruses but none of the magnitude of the first one. Tadpoles were found whenever water was present (Fig. 10). The first tadpoles that ap- peared were Leptodactylus mystaceus on 01 January; these were the only anuran larvae present prior to February and they were absent after 04 February. It seems as though tadpoles of L. mystaceus resulted from two breeding bouts, because young tadpoles were present on 01 January and again on 25 January; all other species seem to have resulted from a single breeding bout. Chiasmocleis ventrimaculata, Ctenophryne geayi, Elachistocleis ovalis, Hamptophryne boliviana, Hyla koechlini, H. parviceps, and Scinax icterica all appeared in net samples in the first week of February. Colostethus marchesianus, Hyla brevifrons, H. cf. leali, and Hyla sp. appeared later; the latest was H. brevifrons on 24 February. Only Colostethus marchesianus, Hamptophryne boliviana, Hyla par\'iceps, and Scinax icterica were collected in March; the last species encountered were H. boliviana and H. parviceps on 08 March. Five non-anuran species were collected in net samples of Swamp A- 1250, as follows: five individuals of the trichodactylid crab species; three individuals of Nepidae sp.; and one each of Belostoma sp., Cynolebias sp., and Pyrhulina sp. (Table 2). The crab was found only in late January, whereas the remaining non-anuran species were encountered between 14 - 24 February. Two colubrid snakes {Liophis reginae, Helicops sp.), a chelid turtle {Platemys platycephala), and a crocodylid {Paleosuchus trigonatus) also were found in Swamp A- 1250. AMAZONIAN TADPOLE ASSEMBLAGES 25 T3 (l> J3 M a a o in >o r 1 PQ ON o o '~~, o- ■J^i O =^ 05 2 1) c x: a. r^. (U _ •73 r-^ c (~i )^i c "-, o ' ' r , a - CD > .^, 1- -o ^^ o c ' 0) a — u. ^ :?- r^ X) v-i "■ C3 O v^ ^. UJ oo > '" ' — ' J-.' , C3 U— ?l C ^3 ^ QO CO -a vC _) c c_ cS -r CT3 ^> -3 C/3 rt X) oc 13 c 'T o r-, N 2: ^ ^ E - O = Q o o- T3 C3 t+- O T3 O c u« , o ■n o -5 o I/) m ;-i L« f5 o (^ n. £ -4—1 a, 13 C (D — 1 (IJ ^" s t* C/2 u- U c 26 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Pool B-550. — Abiotic characteristics {Table 3): Pool B-550 is a small, ephemeral pool situated 20 m from the bank of the Quebrada Mariposa. The pool is located in terra firma forest with an almost completely closed canopy and moderately dense understory of vines and many shrubs. Pool B-550 began to fill a few days prior to 27 December, the date sampling started, and contained standing water until sampling was terminated on 23 March (88 days). The mean water depth during this time was 144 + 32 mm; the water depth reached a maximum of 270 mm on 01 February and was never less than 100 mm. At its average depth, the pool was 2.3 x 5.2 m, and at its greatest depth. 2.6 x 6.5 m. Prior to late December, there was no indication that this pool existed other than a slight depression in the forest floor. Water temperature varied little between day {x = 25.8 ± 1.0°C) and night (x = 25.4 ± 0.9°C) and reached extremes of 23 and 27°C. Because of the shallowness of the pool, light penetrated to the bottom most of the time, and was less than 70% of the water column only once (38%). The dissolved oxygen content of the water in Pool B-550 was the same at dawn and dusk ( .v = 7 ± 1 mg/1), and reached extremes of 5 and 10 mg/ 1 at dusk 3 1 January and 03 March, respectively. The mean water pH varied little between dawn (x - 6.7 ± 0.3) and dusk {x = 6.6 ± 0.2), and reached extremes of 6.5 and 7.0. The mean water alkalinity was slightly higher at dawn (.^= 48 ± 16) than at dusk (.F = 43 ± 18) and reached extremes of 18 and 88 at dusk on 1 1 February and dawn 20 March, respectively. Total hardness was slightly higher at dawn ( j = 53 ± 9) than dusk (^ = 49 ± 12), and reached extremes of 30 and 100, the first on several occasions and the second at dawn on 20 March. Biotic characteristics: Throughout the study, 769 tadpoles of five spe- cies were sampled from Pool B-550 (Table 1 ). Phyllomedusa tomopterna was the most abundant species with 85% (/? = 653) of all tadpoles sampled, followed by Species D with 8% {n = 61). Hyla parviceps {n = 38), Colostethus marchesianus (/; = 1 1 ), and Hyla cf. leali (/? = 6) accounted for the remaining 7%. Adults of all species, except for the unknown Species D, were present or heard calling near Pool B-550 at some time during the study (Table 4). The most frequently encountered adults were Hyla par\'iceps and Phyllomedusa tomopterna. Tadpoles were present for the entire sampling period (Fig. 1 1 ). Phyllomedusa tomopterna and Species D were the only species present for the entire period; because these larvae were present on the first day of sampling, breeding must have occurred previously. On 14 January, there was an additional hatching of Phyllomedusa tomopterna that resulted in the presence of smaller tadpoles and individual eggs in the water. On 24 January, there were two size classes of P. tomopterna larvae present; the presence of young tadpoles indicates an additional hatching. Larvae of Species D seemed to have resulted from three additional hatchings; one in mid-January, another in mid-February, AMAZONIAN TADPOLE ASSEMBLAGES 27 and the third in early March. Colostethus marchesianus was found first on 16 January, and subsequently, was encountered infrequently. Hyla pannceps was present from 26 January until 23 March, and H. cf. leali was present only in March. Tadpoles of all species, except H. cf. leali. plus an addi- tional unknown Species A, were present when the entire tadpole fauna was removed by thorough netting on 23 March. The abundances of the tadpoles of this final netting were: Phyllomedusa tomopterna — 32, Species D — 31, Colostethus marchesianus — 26, Hyla pan'iceps — 22, and Species A — 5. The only non-anuran species sampled from Pool B-550 was the species of trichodactylid crab (Table 2). Crabs were first found on 16 January and were occasionally taken thereafter until 19 March. Achelid turtle {Phrynops gibbus) and a colubrid snake {Helicops polylepis) were also encountered at Pool B-550. Comparisons of Study Ponds Abiotic Factors (Table 3). — The study ponds differed markedly in vegetation, pond structure, permanency, size, and depth. Camp Cocha and Pool B-550 were both located in terra firma forest; however, they differed in many respects, including vegetation. Camp Cocha was densely veg- etated, almost exclusively by Heliconia, creating a large opening in the forest canopy. Thus, it was distinct from the surrounding forest. The vegetation of the small Pool B-550 resembled that of the surrounding forest, having a moderately dense underbrush of vines and woody shrubs under an almost completely closed canopy. In the dry season, Pool B-550 was virtually undetectable because it was simply a slightly depressed area in the forest floor, whereas Camp Cocha remained distinct from the sur- rounding forest throughout the annual cycle. Swamps A- 1 1 50 and A- 1 250 covered large areas in the inundated forest. These swamps were only 100 m apart, and the forest surrounding them was similar. However, the vegetational composition of these two swamps was quite different. The forest canopy over swamp A- 11 50 was incomplete, and the understory varied from wide, open areas to impenetrable, dense growths of woody shrubs and vines. Swamp A- 1250, in contrast, had an almost completely closed canopy, like that of the surrounding forest, and the sparse understory consisted only of scattered woody shrubs. Both swamps had downed trunks, smaller branches, and vines in the water. The permanency of study ponds differed. Both Camp Cocha and Swamp A- 11 50 contained standing water throughout the study period. Swamp A- 1150 contained standing water during the dry season, but Camp Cocha did not although the ground remained wet. Pool B-550 was temporary; after the first several consecutive days of rain and ground saturation, the pond maintained a relatively constant level of standing water. The temporary 28 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Swamp A- 1250 was the most ephemeral and required several days of rain to fill and maintain a constant level of standing water; in the absence of rain, the water level dropped and the swamp dried rapidly. The relative sizes and depths of the ponds does not perfectly corresponded to their permanency. Whereas Swamp A- 11 50 was the largest, deepest, and most permanent, followed by Camp Cocha, Pool B-550 was the smallest and shallowest yet less ephemeral than the larger, deeper Swamp A- 1250. Pool B-550 had the greatest mean percentage light penetration followed by Swamp A- 1250 and Camp Cocha. Swamp A- 11 50 had a much lower average amount of light penetration than the other three ponds. The amount of light penetration is related to water depth and simply reflects the pres- ence or absence of an unlit zone. There was relatively little difference in the water chemistry of the ponds. All had similar water temperatures, dissolved oxygen contents, and pHs, all of which showed little diel variation throughout the study as well. Water alkalinity was lower in Swamp A- 1 1 50 than in the other study ponds, and highest in Swamp A- 1250. All ponds had slightly higher alkalinity at dawn than at dusk, except Swamp A-1 150, in which alkalinity was nearly constant. A similar diel pattern was found for total water hardness. Among ponds, total hardness, like alkalinity, was the lowest in Swamp A-1 150 and the highest in Swamp A- 1250. Total hardness was slightly higher at dawn than at dusk for all study ponds except A- 11 50, in which hardness was nearly constant. In summary of abiotic characteristics, each study pond had different vegetation and structure. There was a continuum of permanency, size, and depth among the ponds. Thus, Swamp A- 11 50 was the most permanent, largest, and deepest pond followed by Camp Cocha in all three aspects. Swamp A- 1250 was larger and deeper than Pool B-550, but was more ephemeral. Percentage light penetration was inversely related to size and depth of the ponds. The ponds were similar in water temperature, dissolved oxygen content, and pH, and showed no diel shifts in these abiotic factors. Alkalinity and total hardness were inversely related to pond permanency. All of the study ponds except A-1 150 showed a slight diel shift, with both alkalinity and total hardness being higher at dawn. Biotic Comparisons. — The greatest number of tadpoles yielded by any pond was 2363 from Camp Cocha. Pool B-550. Swamp A- 1250, and Swamp A-1 150 followed with 769, 610, and 78 tadpoles, respectively. In terms of the number of species. Swamp A- 1250, with 12, was the richest, followed by Swamp A- 11 50, Camp Cocha, and Pool B-550 with eight, seven, and five species, respectively. When sampling effort, reflected in the total number of tadpoles sampled in each pond, is taken into account by using rarefaction for samples of equal sample sizes {n = 78), Swamp A- 1 1 50 is the richest (8.00 ± 0.00), followed by Swamp A- 1 250 (7.98 ±1.12), AMAZONIAN TADPOLE ASSEMBLAGES 29 Pool B-550 (4. 1 5 ± 0.69) and Camp Cocha ( 1 .93 ± 0.62), (Fig. 12). The low values in Camp Cocha and Pool B-550 reflect the predominance of one common species and several rare species. This is better illustrated using modified HilTs ratio for measuring evenness (Table 5). Camp Cocha (E = 0.35) is the least equitable followed by Pool B-550 (E - 0.46), and Swamp A-1150 (E = 0.50), whereas Swamp A- 1250 is the most equitable (E = 0.62). Diversity measured by Hill's numbers combines both evenness and species richness into a measure of the effective number of abundant (Nl) and very abundant (N2) species. Camp Cocha (N2 = 1.04) clearly has a single dominant species, Phyllomedusa vaillanti, and few abundant species (Nl = 1.11), indicating that most species are very rare. Pool B-550 has a slightly higher value for very abundant species (N2 = 1.37), but it is also close to one, with Phyllomedusa tomopterna being the primary species. Compared to Camp Cocha, Pool B-550 has more abundant species (Nl = 1.80). Both Swamps A-1 150 and A- 1250 have a greater number of very abundant (N2 = 1.93 and 3.35, respectively), as well as abundant species (Nl = 2.85 and 4.76, respectively). Pairwise comparisons of the ponds indicated that Swamps A-1 150 and A- 1250, the two ponds nearest each other, had the greatest number of species of tadpoles in common, with six shared species (and the second greatest in number of species as adults). This pair is followed by Swamp A- 1150 and Camp Cocha, the third most proximate pair, with four shared species of tadpoles (and the greatest number of adults in common). Pool B- 550 had three species in common with Swamp A- 1250, one species in -• — Camp Cocha -■ — Swamp A-1150 <}- Swamp A-1 250 -a - Pool B-550 100 200 300 400 500 600 Sample Size 700 800 900 1000 Fig. 12. Rarefaction curves for tadpole species of the four study ponds. Error bars = ± ISD and vertical dashed lines indicate sample sizes of 78 (left) and 500 (right) individuals. 30 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Table 5. Tadpole species richness, abundance, evenness, and diversity in the four study ponds. No. of No. of Rarefaction Evenness Diversity Pond species tadpoles (/; = 78) (E) (Nl) (N2) Camp Cocha 7 2363 1.93 ±0.62 0.35 1.11 1.04 Swamp A- 1150 8 78 8.00 ± 0.00 0.50 2.85 1.93 Swamp A- 1250 12 610 7.98 ± 1.12 0.62 4.76 3.35 Pool B-550 5 769 4.15 + 0.69 0.46 1.80 1.37 Total 19 3820 common with Swamp A- 11 50, but none with Camp Cocha, even though these were the second most proximate pair. Camp Cocha and Swamp A- 1150 had two species of tadpoles in common. Swamp A-1 150 and Camp Cocha had the greatest number of species of adults in common with six, followed by Swamps A- 11 50 and A- 1250 with four. Swamp A- 1250 had three species of adults in common with Camp Cocha and Pool B-550. Pool B-550 and Swamp A- 11 50 had two species in common, whereas Camp Cocha and Pool B-550 had none, in spite of their being the most proximal pair. The greatest number of non-anurans yielded by any pond was 845 from Camp Cocha. Swamps A- 1 1 50 and A- 1 250, and Pool B-550 followed with 547, 11, and 10 individuals, respectively. In terms of only the number of species. Swamp A- 1 1 50, with 1 3, was the richest followed by Camp Cocha, Swamp A- 1 250, and Pool B-550 with seven, five, and one respectively. The only non-anuran species found in Pool B-550 was the single species of crab, which was encountered rarely. Taking into account sampling effort (number of individuals sampled), rarefaction {n = 10) indicated Swamp A- 1150 to be the richest in non-anuran species (5.91 ± 1.05) followed by Swamp A- 1 250 (5.00 ± 0.00) and Camp Cocha (4.27 ± 0.79), (Fig. 13). The evenness measure of modified Hill's ratio shows Camp Cocha to be the least equitable in non-anuran abundance (E = 0.83), followed closely by Swamp A-1 150 (E = 0.85); (Table 6). Swamp A- 1250 was the most equi- table in non-anurans (E = 1.10). Only Pool B-550 was dominated by one non-anuran species, the crab, which was the only species present and very rare. All other three ponds had approximately equivalent numbers of very abundant and abundant species of non-anurans. Swamp A- 11 50 had the greatest of both (Nl = 8.68, N2 = 7.49), followed by Camp Cocha (Nl = 4.76, N2 = 4. 1 7). Swamp A- 1 250 had more very abundant (N2 = 4.23) than abundant (Nl = 3.92) species. Pairwise comparisons of non-anuran species among the study ponds showed that Camp Cocha and Swamp A-1 150 had the greatest number (7) AMAZONIAN TADPOLE ASSEMBLAGES 31 ^ — • — Camp Cocha -■ — Swamp A-1 150 -o - Swamp A- 1250 -a - Pool B-550 100 Sample Size 150 200 Fig. 13. Rarefaction curves for non-anuran species of the four study ponds. Error bars = ± 1 SD and vertical dashed lines indicate sample sizes of 10 (left) and 100 (right) individuals. Table 6. Non-anuran species richness, abundance, evenness, and diversity in the four study ponds. Pond No. of species No. of Rarefaction Evenness Diversity individuals (n = 10) (E) (Nl) (N2) Camp Cocha 7 845 4.27 ± 0.79 0.83 4.76 4.14 Swamp A-1 150 13 547 5.91 ± 1.05 0.85 8.68 7.49 Swamp A- 1250 5 11 5.00 ± 0.00 1.11 3.92 4.23 Pool B-550 1 10 1 .00 ± 0.00 oo 1.00 1.00 Total 13 1413 of species in common, followed by Swamp A- 1250 and Camp Cocha, and Swamp A- 1250 and Swamp A- 11 50, both with five. Pool B-550 had only one non-anuran species, the crab, but this species occurred in all the ponds. In summary of biotic characteristics. Camp Cocha and Pool B-550 are similar in tadpole diversity, with both ponds being dominated by one common species (interestingly, a Phyllomedusa in both cases); however, these two ponds have no species of tadpoles in common. In addition. Camp Cocha had a much more diverse non-anuran fauna than Pool B-550; the two share only the single species of crab. Swamps A- 1 250 and A- 1 1 50 had a much higher tadpole diversity than the other two ponds — Swamp A- 1250 because it had the greatest number of species, and Swamp A- 11 50 because 32 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 it had few tadpoles for the number of species. These two ponds, which were the most proximate pair, also had the greatest number of tadpoles in common. In terms of their non-anuran faunas. Swamp A-1 150 is the most diverse, whereas Swamp A- 1250 is the second lowest. These two ponds have a moderate number of non-anuran species in common. Overall, the four study ponds display a range of abiotic and biotic characteristics. The following are the perfect rank correlations among the four study ponds for any of the biotic and abiotic characteristics. ( 1 ) The number of species of non-anurans, number of abundant species of non- anurans, size of pond, and depth of pond are correlated, and have a negative correlation to percentage light penetration. (2) Permanency and alkalinity/ hardness are negatively correlated among ponds. (3) The number of species of adult frogs is correlated with the number of individual predators, and both are negatively correlated with evenness of predators. (4) The number of abundant species, number of very abundant species, and evenness of tadpoles are correlated. Clearly, many of the variables of these conelations are not independent, and the small sample size (four ponds) severely limits their interpretation. Furthermore, the absence of any correlation of tadpole richness, evenness, or diversity with any of the other aforementioned characteristics precludes any effort to explain the distribution of tadpole species among ponds in terms of these abiotic and biotic characteristics. RESOURCE UTILIZATION Macrohabitat Twenty-nine of the 54 species (54%) at Cuzco Amazonico with life cycles that include an aquatic larval stage were observed at the study ponds as adults or tadpoles (Table 4). Thus, at least 25 species (46%) known in the local community were not present as adults or tadpoles, and thus presum- ably utilize either other macrohabitats or use the study ponds during other years or seasonal times. The latter possibility seems unlikely, however, because only Swamp A- 1 150 contained standing water beyond the study period. Tadpoles of 10 of the 23 species of adults present were not found in the study ponds: These are Bufo nuirinns. Ceratophrys cornuto, Hyla allenontm, H. sclnibarti, Leptodactylus leptodactyloides, L. petersii, Osteocephalus taurinus, Scarthyla ostinodactyla, Sciiiax garhei, and Sphaenorhynchus lacteus. Some of the latter occur at more than one of the study ponds. Adult Leptodactylus petersii were found at Camp Cocha and Swamp A-1 150; likewise, adult Osteocephalus taurinus were found at Swamp A-1 150 and Pool B-550. Adult Scarthyla ostinodactyla occurred at all study ponds except Pool B-550. There are several possible reasons for the absence of larvae of these taxa. The species may not have been breed- AMAZONIAN TADPOLE ASSEMBLAGES 33 ing. They might have bred, but their tadpoles might have been unsuccess- ful, or their tadpoles occurred in microhabitats other than those sampled. I know of only one case in which a species of tadpole occurred in a micro- habitat not included in the transects, despite repeated sampling. Scarthyla ostinodactyla larvae were encountered in net sweeps away from the sam- pling transects amidst duckweed at the surface of Swamp A-1 150. Because some species occurred both as adults and tadpoles, but only as tadpoles in some ponds suggests differential success at macrohabitat utili- zation. Adult Hylo parviceps were found at Swamps A-1 150 and A- 1250, and Pool B-550, but no tadpoles were found at Swamp A-1 150. Adult Hyla fasciato were found at Camp Cocha and Swamps A-1 150 and A- 1250, but tadpoles occurred only at Camp Cocha. Adult Hyla koechlini were seen at Camp Cocha and Swamps A- 1 1 50 and A- 1 250, but no tadpoles were found at Camp Cocha. Overall, Camp Cocha had the greatest number (8) of adults present for which tadpoles were not also found, followed by Swamps A- 11 50 and A- 1250 with four each, and Pool B-550 with one. Of the species of tadpoles collected, none occurred in all four study ponds, and only Hamptophryne boliviano, Hyla brevifrons, and H. cf. leali, occurred in three. Seven species occurred in two ponds, and nine in only one study pond. There seemed to be no consistent pattern in these occur- rences among the ponds, except for the presence of four of those species that occur in a single pond (three of which were microhylids) in Swamp A- 1250. In summary, at Cuzco Amazonico nearly half of the anuran species with tadpoles did not utilize the study ponds during the study period. Species represented only by adults probably indicates the taxa were unsuccessful in recruitment or that they used other macrohabitats for breeding. The two largest ponds had the greatest number of adults present without tadpoles. These also had the greatest number of non-anuran individuals suggesting predation to be an important factor. None of the tadpoles is ubiquitous. In fact, only three species occur in three different ponds. Of the nine species that occur in only one pond, four occur in Swamp A- 1250. There is differential utilization of study ponds, with adults, tadpoles, or both of different species at different ponds with no apparent pattern; this suggests differential use of the macrohabitats. The determinant factor of this differ- ential use of macrohabitat from year to year by adults is beyond the scope of this study. It is possible that adult pond use varies from year to year and may contain a large stochastic component. Phenology Tadpoles of 35 of the 54 species of anurans at Cuzco Amazonico that have aquatic tadpoles were not encountered. Possibly some of these species breed at other times during the year (or at other sites as previously dis- 34 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 cussed). The timing restrictions of this study precluded determination of which species were not encountered because they breed during other time periods (i.e., beyond the rainy season), but they are probably few because of lack of aquatic macrohabitats beyond the rainy season at this relatively seasonal site. Some species temporally partition the study ponds during the rainy season. At Camp Cocha, Hamptophryne boliviana and PhyUomedusa vaillanti were the only larvae present from mid-December until February (Fig. 8). PhyUomedusa vaillanti occurs through the entire study period, whereas H. boliviana was not found after January. The remaining species occurred sporadically during February {Hyla leiicophyllata, Altigius alios) or March after the heaviest rains {Hyla brevifrons, H. fasciata, and Species A). A similar pattern was evident in Swamp A- 1 1 50 (Fig. 9). Hamptophryne boliviana and PhyUomedusa vaillanti were the first species to appear in mid-January. PhyUomedusa vaillanti occurred from mid-February to mid- March, whereas H. boliviana was not found after early February. Only one other species, H. brevifrons, occurred in January at Swamp A-1 150, and did not occur again until mid-March, unlike at Camp Cocha where this species only occurred early in March. The remaining species at Swamp A- 11 50 occurred sporadically in February (Hyla koechlini, H. cf. leali, Scinax icterica) or February and March {Hyla leucophyllata, Hyla sp.) at times of the heaviest rains. Camp Cocha and Swamp A-1 150 have some remarkable similarities. For example, Hamptophryne boliviana and PhyUomedusa vaillanti occurred before any other species in both ponds. Whereas P. vaillanti remained until the end of the study period, Hamptophryne boliviana disappeared midway through the study season, just prior to the heaviest rainfall, after which most of the other species appeared. In Swamp A- 1250, Leptodactylus mystaceus was the only species present in January, immediately after the swamp appeared (Fig. 10). Al- though the site nearly dried, these tadpoles were found in the remaining small puddles. At the beginning of February, after rains filled the swamp again, L. mystaceus was absent and, within 1 wk, seven other species appeared; four additional species appeared by the end of February. Most of these species were present well into March when the pond disappeared for the season, with the exception of the microhylids Elachistocleis ovalis, Ctenophryne geayi, and Chiasmocleis ventrimaculata, which were present for just a short period during the peak of the rainy season in the first half of February. The greatest number of larval species in Swamp A- 1 250 occurred in February when as many as seven species were collected on the same day. This may be characteristic of ephemeral swamps such as Swamp A- 1250, where time is constrained and cannot be differentially utilized by species of tadpoles. Only L mystaceus is absent during this time period, but faces the risk of premature drying of the pond earlier in the rainy season. AMAZONIAN TADPOLE ASSEMBLAGES 35 In Pool B-550, Phyllomedusa tomoptema and Species D were the first species to appear in late December (Fig. 1 1 ), and were the only species present until mid-January when Colostethus marchesianus appeared. The other two species, Hyla parviceps and H. cf. leali, first appeared in late January and early March, respectively. All species were present at the end of the study, when the pond was thoroughly netted in an attempt to remove all tadpoles. Like Camp Cocha, Pool B-550 had two species present early. Both of these species remained with additional rain that brought the three remaining species. The first occurrences of these additional species were staggered. All five species, once present, were continually encountered until the end of sampling. At least some temporal partitioning of the aquatic habitats occurs during the rainy season at Cuzco Amazonico. Some species are probably opportu- nistic, breeding early, at first rains, and only once. Other species are present at the first accumulation of water, including Hamptophryne boliviana and Phyllomedusa vaillanti (Camp Cocha and Swamp A-1 150). Leptodactylus mystaceus (Swamp A- 1250), and Phyllomedusa tomoptema and Species D (Pool B-550). Of these, some (e.g.. Phyllomedusa tomoptema, P. vaillanti, and Species D) persist, whereas others (e.g., Hamptophi-yne boliviana and Leptodactylus mystaceus) have transformed and left the ponds (and no further breeding) before additional heavy rains bring other species. Some species tend to have staggered arrival times (e.g., many species in Swamp A-1150 and Pool B-550), whereas others (e.g., Ctenophryne geayi, Chiasmocleis ventrimaculata, and Elachistocleis ovalis in Swamp A- 1250) are present for limited periods of time at the peak of the rainy season. The ephemeral nature of some ponds seems to limit the degree to which temporal partitioning can take place and, thus, causes an increase in the number of temporally co-occurring species; this can be seen in Swamp A- 1150 and Pool B-550. The length of time that ponds where available for tadpoles at Cuzco Amazonico may have been further constrained because of a shorter than average rainy season during the year of the study. MiCROHABITAT, DIEL ACTIVITY, AND DEVELOPMENT Microhabitat. — Tadpoles were sampled from three levels in the water column representing three general microhabitats — surface, midwater, and bottom. These levels are a subset of the available microhabitats, and it should be noted that areas such as the littoral zone and areas of dense vegetation were not included and may be important components. Deviation from an even distribution among the levels was tested for nine species (those with n > 15), including two species (Phyllomedusa vaillanti, Hamptophryne boliviana) that occurred in two study ponds each (Table 7); these nine species accounted for 97% of all tadpoles sampled. All species 36 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Table 7. Distribution of tadpoles among microhabitats. Dagger (t) indicates adjustment for schooling tadpoles. (See discussion in text.) Mid- Species n Surface level Bottom G P Camp Cocha Hamptophryne boliviana 39 13 15 11 0.618 0.7342 Phyllomedusa vaillanti 2318 365 1332 621 631.924 < 0.0001 Phyllomedusa vaillanti'^ 23 4 13 6 6.319 0.0424 Swamp A- 11 50 Phyllomedusa vaillanti 55 4 7 44 51.382 < 0.0001 Swamp A- 1250 Chiasmocleis ventrimaculata 109 30 51 28 8.505 0.0142 Elachistocleis oralis 38 13 12 13 0.053 0.9738 Hamptophryne boliviana 76 40 27 9 21.353 < 0.0001 Hyla koechlini 52 3 28 21 24.391 < 0.0001 Scinax icterica 299 48 108 143 50.454 < 0.0001 Pool B-550 Hyla pan'iceps 38 9 10 19 4.528 0.1039 Phyllomedusa tomopterna 627 54 271 302 216.969 < 0.0001 Species D 61 4 16 41 36.830 < 0.0001 tested had tadpoles occurring at each level. Of the 11 tests, three species had distributions among the microhabitats that are not significantly differ- ent (P < 0.05) from an even distribution. These are Hamptophryne boliviana from Camp Cocha, Elachistocleis ovalis from Swamp A- 1 250, and Hyla parviceps from Pool B-550. The remaining eight tests showed significant deviations from an even distribution among levels. In Camp Cocha the distribution of Phyllomedusa vaillanti among the levels was uneven (P < 0.0001 ), with 16% of the tadpoles occurring at the surface, 57% in the midlevel, and 27% at the bottom. This differs from the distribution of this species among levels in Swamp A- II 50 with 7% at the surface, 13% in the midlevel, and 80% at the bottom, which was also significant {P < 0.0001). The larvae of Phyllomedusa vaillanti are unique among the species of tadpoles of this study because they are the only ones that school (Branch, 1983) and were observed doing this in both the Camp Cocha and Swamp A-1 150. The G-test statistic used herein assumes inde- pendent observations, an assumption which is violated by schooling behav- ior; one should count individual schools and not individual tadpoles. As a conservative adjustment of the data, the tests were repeated with the counts adjusted by dividing by the school size, which was estimated in the field to AMAZONIAN TADPOLE ASSEMBLAGES 37 be approximately 100 individuals. With this adjustment, the distribution among levels in the Camp Cocha still remains significantly different from an even distribution (P = 0.0424). Too few tadpoles of P. vaiUanti were sampled from Swamp A-1 150 to adjust the test in this manner. In Swamp A- 1250, several larval species were unevenly distributed in the water column. In Scina.x icterico {P < 0.0001), 16% of the tadpoles occurred at the surface, 36% in the midlevel, and 48% at the bottom. In Hamptophryne boliviana {P < 0.0001), 53% of the larvae were at the surface, 36% in the midlevel, and 12% at the bottom. This differs from the distribution of H. boliviana in Camp Cocha, where the larvae were evenly distributed through the water column. In Hyla koechlini {P < 0.0001), 6% of the tadpoles were at the surface, 54% in the midlevel, and 40% at the bottom, and in Chiasmocleis ventrimaculata (P = 0.0142), 28% were at the surface, 47% at midlevel, and 26% at the bottom. In Pool B-550, the larvae of Phyllomedusa tomopterna were distributed unevenly among the levels (P < 0.0001) with 9% at the surface, 43% in the midlevel, and 48% at the bottom. Similarly, the distribution of the unknown Species D was uneven {P < 0.0001) with 7% at the surface, 26% in midlevel, and 67% at the bottom of Pool B-550. Of the eight tests with significantly uneven distributions through the water column, only Hamptophryne boliviana in Swamp A- 1250 had the greatest frequency at the surface. All others (except Chiasmocleis ventrimaculata. which was close with the bottom having just two fewer individuals than the surface) frequented the surface microhabitat least. Only Phyllomedusa vaillanti in Camp Cocha and Chiasmocleis ventrimaculata and Hyla koechlini in Swamp A- 1250 occurred most fre- quently in the midlevel. Phyllomedusa vaillanti in the Camp Cocha is particularly interesting because in Swamp A- 11 50 this species frequented the bottom most, although this observation might have been affected by schooling behavior. The remaining four tests with significantly uneven distributions had the greatest frequency of occurrence at the bottom, fol- lowed by midlevel and then surface, the most common distribution pattern. Diel activity. — Tadpoles were sampled by day and night to determine if there was a preference of diel activity. Deviation from an even distribution of occurrence between day and night was tested in 12 species (those with ;/ > 10), including three species {Phyllomedusa vaillanti, Hcunptophryne boliviana, Hyla parviceps) that occurred in two different study ponds each (Table 8); these 12 species accounted for 98% of all tadpoles sampled. Of the 15 tests, the diel distributions of six were the same by day and by night {Elachistocleis ovalis and Hyla koechlini in Swamp A- 1250; Colostethus marchesianus, Hyla pannceps, Phyllomedusa tomopterna, and Species D in Pool B-550). The remaining nine tests showed significant {P < 0.05) deviations from an even diel distribution. Hamptophryne boliviana and 38 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Table 8. Distribution of tadpoles between day and night. Dagger (t) indicates adjustment for schooling tadpoles. (See discussion in text.) Species n Day Night G P Camp Cocha Hamptophnme holiviana 39 6 33 20.578 < 0.0001 Phyllonu'chtsci vaillauti 2318 1388 930 91.092 < 0.0001 Phyllomedusa vaillantit 23 14 9 0.911 0.3398 Swamp A- 11 50 Phyllomedusa vaillanti 55 1 54 66.250 < 0.0001 Hyla leiicophyllato 12 0 12 16.636 < 0.0001 Swamp A- 1250 Chiasmocleis ventrimaculata 109 25 84 33.713 < 0.0001 Elachistocleis avails 38 19 19 0 1.0000 Hamptophiyne holiviana 76 12 64 39.062 < 0.0001 Hyla koechlini 52 20 32 2.794 0.0946 Hyla parx'iceps 10 1 9 7.361 0.0067 Leptodactylus inystaceits 14 2 12 7.925 0.0049 Scinax icterica 299 23 276 252.331 < 0.0001 Pool B-550 Colostethus nuirchesianiis 11 8 3 2.358 0.1246 Hyla parviceps 38 14 24 2.663 0.1027 Phyllomedusa tomopterna 627 330 297 1.738 0.1874 Species D 61 27 34 0.805 0.3696 Phyllomedusa vaillanti from Camp Cocha, Hyla leucophyllata and Phyllomedusa vaillanti from Swamp A- 1 1 50. and five species from Swamp A- 1250 {Chiasmocleis ventrimaculata, Hamptophryne holiviana, Hyla parviceps, Leptodactylus mystaceus, and Scina.x icterica) were captured more commonly at night. Only Hyla leucophyllata occurred exclusively during one time period (night). Phyllomedusa vaillanti in Camp Cocha was the only species that was more frequent (60%) by day. Schooling behavior in tadpoles such as Phyllomedusa vaillanti often is explained as a predator avoidance mechanism (studies cited in Branch, 1983). This could explain the ability of Phyllomedusa vaillanti to exploit the resources during the day because of its lowered risk to visually oriented predators such as fish. However, when the data are adjusted for schooling (as for the test of distribution among microhabitats), the test fails to find significant differences (P = 0.3398) between diurnal and nocturnal larval distributions in Camp Cocha. Of the other species that occur in two ponds, nearly the same proportion occurred at night and day for Hamptophryne holiviana in both Camp AMAZONIAN TADPOLE ASSEMBLAGES 39 Cocha and Swamp A- 1250. Hyla parviceps was more frequent at night in both Swamp A- 1250 and Pool B-550, although the difference was not significant (P = 0. 1027) in the latter. Phyllomedusa vaillanti occurred more frequently by day in the Camp Cocha, but by night in Swamp A-1 150. Interrelationship of microhabitat, diel activity, and development. — In order to determine the presence, and nature, of interaction among level occupied by the larvae in the water column, diel activity, and development, a Log-Linear Model Analysis was performed on data for those individual species from individual ponds with sufficient sample size. Log-Linear Model Analysis tests for interactions among attributes (= nominal vari- ables) of a contingency table by measuring the relative ability of various models, each sequentially including fewer combinations of variables, to fit expected values to those observed. This analysis requires that all cells in the observed data matrix have non-zero frequencies, that the expected cell frequencies of the model with the best fit are no less than one, and that no more than 20^^ of the cells have frequencies of less than five. Most species had inadequate sample sizes, but others were rendered usable by combin- ing categories of developmental stages or day and night. There are sufficient data for five species from three ponds to examine using Log-Linear Model Analysis — Phyllomedusa vaillanti from Camp Cocha; Chiasmocleis ventrimaculata, Hamptophryne boliviana, and Scina.x icterica from Swamp A- 1 250; and Phyllomedusa tomoptema from Pool B- 550 (Tables 9-13). It should be noted that Log-Linear Model Analysis assumes independent observations which may not be true for P. vaillanti. Table 9. Data table used for 3-way Log-Linear Model Analysis of Phyllomedusa vaillanti from Camp Cocha. Level Stage Diel period Surface Midlevel Bottom Total <28 29-34 Day 213 807 194 1214 Night 108 365 285 758 Total 321 1172 479 1972 Day 25 84 43 152 Night 16 56 66 138 Total 41 140 109 290 Day 0 9 12 21 Night 3 10 20 33 Total 3 19 32 54 35-40 Total 365 1331 620 2316 Day 2 0 1 3 Night 11 16 8 35 Total 13 16 9 38 Day 4 11 7 22 Night 13 24 12 49 Total 17 35 19 71 40 UNIV. KANSAS NAT. HIST MUS. OCC. PAP. No. 176 Table 10. Data table used for 3-way Log-Linear Model Analysis of Chiasmocleis ventrimaculata from Swamp A- 1 250. Level Stage Diel period Surface Midlevel Bottom Total <28 29^0 Total Total 30 51 28 109 so this analysis should be interpreted with the knowledge that schooling behavior may have confounded the results. Camp Cocha Phyllomedusa vaillanti: Analysis of 2316 individuals of Phyllomedusa vaillanti from Camp Cocha (Table 9; note one cell with a frequency of 0) reveals a non-significant 3-way interaction {P = 0.1112; Table 14), but significant associations between each pair of variables. This indicates that the degree of association between any pair of variables is the same over (unaffected by) the level of the third, but the third variable is conditionally dependent on each of the other two variables. The associa- tions between each pair of variables are discussed separately. Stage and level association (Fig. J4A): There is a significant association between larval stage and level in the water column {P < 0.0001; Table 14). As previously demonstrated in the microhabitat section, most tadpoles of Phyllomedusa vaillanti in Camp Cocha occurred at midlevel, then bottom, and the fewest at the surface. This was the case for the earliest stage (< 28) with 59% at midlevel, 24% at the bottom, and 16% at the surface. However, Table 1 1 . Data table used for 2-way Log-Linear Model Analysis of Hamptophryne boliviana from Swamp A- 1 250. Level Stage Surface Midlevel Bottom Total <28 23 17 3 43 29-34 9 7 4 20 35^0 8 3 2 13 Total 40 27 9 76 28 49 62 139 1 14 21 36 5 5 12 22 1 4 1 6 48 108 143 299 AMAZONIAN TADPOLE ASSEMBLAGES 41 Table 1 2. Data table used for 2-way Log-Linear Model Analysis of Sciuax icterica from Swamp A- 1250. Level Stage Surface Midlevel Bottom Total <25 13 36 47 96 26-28 29-31 32-34 35-37 Total at later stages, the proportion of tadpoles occurring at the bottom level increases, whereas the proportion at midlevel and surface decreases. At the most advanced developmental stages (35-40), most tadpoles occurred at the bottom (59%), then midlevel (35%), and the least at the surface (6%). Diel and level association (Fig. 14B. C): There is a significant associa- tion between diel activity and level in the water column (P < 0.0001 ; Table 14). At surface and midwater, more tadpoles occurred by day than at night (65 and 68% respectively), whereas at the bottom more occuned at night (60%). Among levels, the greatest number of tadpoles occuiTed at midlevel, then bottom, and least at the surface for both day and night, although the relative proportions differ, with a lower proportion at the surface and midlevel and a greater proportion at the bottom at night versus day. Table 13. Data table used for 3-way Log-Linear Model Analysis oi Phyllomedusa tomopterna from Pool B-550. Level Stage Diel period Surface Midlevel Bottom Total <28 29^0 Total 49 229 278 556 Day 15 117 107 239 Night 29 96 155 280 Total 44 213 262 519 Day 2 9 9 20 Night 3 7 7 17 Total 5 16 16 37 42 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 Table 14. Results of 3-way Log-Linear Model Analyses. The variables of the models are: D = diel, L = level, S = stage. Phyllomeditsa Chiasmocleis Phyllomedusa vdillcinti 1 I'entrimaciilata tomopter no Camp Cocha Swamp A -1250 Pool B-550 Model df G P df G P df G P L 15 3134.15 0.0000 9 57.97 0.0000 9 516.20 0.0000 D 16 3674.25 0.0000 10 32.76 0.0003 10 705.57 0.0000 S 15 921.82 0.0000 10 56.33 0.0000 10 209.40 0.0000 L.D 14 3042.97 0.0000 8 24.26 0.0021 8 513.60 0.0000 D,S 14 830.65 0.0000 9 22.61 0.007 1 9 206.80 0.0000 sx 13 290.55 0.0000 8 47.82 0.0000 8 17.43 0.0259 L,D,S 12 199.38 0.0000 7 14.11 0.0493 7 14.83 0.0382 LD 12 2906.82 0.0000 6 23.57 0.0006 6 501.58 0.0000 LS 9 242.02 0.0000 6 46.53 0.0000 6 16.15 0.0130 DS 12 8 1 2.05 0.0000 8 14.11 0.0790 8 205.91 0.0000 L,DS 10 180.78 0.0000 6 5.60 0.4692 6 13.95 0.0302 D,LS 8 150.85 0.0000 5 12.82 0.0251 5 13.55 0.0187 S,LD 10 63.23 0.0000 5 13.42 0.0197 5 2.81 0.7297 LD,LS 6 14.70 0.0228 3 12.14 0.0069 3 1 .53 0.6763 LS.DS 6 132.25 0.0000 4 4.31 0.3653 4 12.67 0.0130 DS.LD 8 44.63 0.0000 4 4.92 0.2958 4 1.92 0.7503 LD,LS,DS 4 7.51 0.1112 2 3.75 0.1535 2 0.60 0.7404 Stage and diel association (Fig. 14D): There is a significant association between larval stage and diel activity (P = 0.0228; Table 14). As previously demonstrated more P. vaillanti tadpoles occur by day than at night in Camp Cocha. This is true for the earliest stage (< 28), with 62% occurring during the day. However, at later stages, the proportion of tadpoles found at night increases, such that at the highest stage (35-40), most (61%) occur at night. As noted above, this analysis must be interpreted with caution because the assumption of independent observations required by Log-Linear Model Analysis is suspect because of the schooling behavior of Phyllomedusa vaillanti tadpoles. In spite of this restriction, the associations between variables seem to be explained by this schooling behavior. The size-class schools would be expected to become smaller as they age owing to antici- pated mortality of tadpoles. It seems logical that as schools diminish in size, they would perform less effectively as anti-predator mechanisms. Therefore, one might anticipate that older tadpoles not protected by school- ing would be less active by day, and might occur more frequently at the bottom, thereby remaining inconspicuous to visually oriented predators such as fish. Both of these patterns were illustrated by the associations among stage and level, and stage and diel demonstrated by the Log-Linear AMAZONIAN TADPOLE ASSEMBLAGES 43 B 100 -I c o o 0 CL 50 - 0 -28 29-34 35-40 Day Night Stage n surface Diel H Midlevel ■ Bottom 100 -I C 0 O ^_ CD Q_ 50 - 0 n Surface Midlevel Bottom -28 29-34 Level □ Day Stage ■ Night 35-40 Fig. 14. Associations between the variables larval stage and level in water column (A), diel time period and level in water column (B-C), and larval stage and diel time period (D) for Phyllomedusa vaillanti from Camp Cocha. Model Analysis. Another possibility is that because the P. vaillanti tadpole schools apparently disassociate at night (Branch, 1983; although I have seen some schools at night), one might expect proportionately fewer in the midlevel, where schools are active, and more at the bottom. This associa- tion between diel and level also was revealed by the analysis. Again, owing to analytical restrictions, further study that specifically focuses on P. vaillanti schooling at Cuzco Amazonico is necessary before we clearly understand the dynamics of this schooling behavior. 44 UNIV. KANSAS NAT. HIST. MUS. OCC. PAP. No. 176 B 100 C