Materials and methods
Study sitesnext section
Field work took place between March and October, 2007 and 2008, in the Liouguei Experimental Forest (LEF) surrounding the Shanping Field Station (23°55’N, 120°41’E, ca. 993 ha in area and 550~1200 m in elevations; Taiwan Forestry Research Institute, TFRI) in Kaohsiung, southern Taiwan. The area is characterized by monthly mean temperature of around 16 °C in January and 24 °C in July-August. Rainfall occurs mainly in the monsoon season and often is accompanied by typhoons from May to September, with an accumulated amount of nearly 3000 mm that accounts for over 85% of the total annual precipitation (about 3500 mm; Shanping Weather Station data, TFRI).
Secondary forests constitute the vegetation of the study sites, composed of 134 families, 71 species of pteridophytes, 23 species of gymnosperms, 128 species of monocotyledons, and 463 species of dicotyledons (Wang 1991). Dominant vegetation on high hills included mainly Lauraceae [e.g., Beilschmiedia erythrophloia (Hayata), Machilus japonica Siebold & Zucc., and Cinnamomum camphora (L.) J. Presl.)] and Fagaceae [e.g., Castanopsis carlesii (Hemsl.) Hayata, C. indica (Roxb. ex Lindl.) A. DC., and Cyclobalanopsis glauca Thunb.], whereas Moraceae (e.g., Ficus superba Miq., F. septica Burm.f., F. benjamina L.1767) took over along with Broussonetia papyrifera (L.) Vent., Mallotus paniculatus (Lam.) Mull.Arg., and Trema orientalis (L.) Blume. in riparian valleys.
Dawn chorus sampling
We monitored dawn chorus at eight replicate sites that were in similar vegetation structure. Any two proximate sites were beyond 250 m apart in distance, and all sites were away from the main pass or any trails, and thus from potential human disturbance or noise. In each month we conducted two bi-weekly sessions of 4-day dawn sampling, one site each day. The sampling in each session occurred in consecutive four days as possible, and in clear stable weather conditions or nearly so. We randomly alternated the order of site among census days within each month. At each site, we chose a relatively open forest gap that was at least 20 m in radius to allow for auditory monitoring, and visual observations when light level permitted.
Sampling started 1.5 hr before sunrise at local times (the Central Weather Bureau data, Taiwan), and lasted until dawn singing ended, defined as 20 min after the last singing bout of the last species heard, usually near 08:00. We recorded the onset and end times of each bout of dawn chorus sequences to seconds using a stop watch (Thompson et al., 1994), and the species that made songs, following Severinghaus et al. (2012) and BirdLife International (2014) for the nomenclature. We also tallied song numbers in each bout by a counter. Two consecutive bouts from the same species were distinguished by a silent period, mostly over 10 seconds, or occasionally (less than 3.7%, n = 5510) when two bouts with a pause timed shorter than 10 seconds but were from two identifiable source directions. We acknowledged that our measurements might have been underestimations, but only in a few occasions (3%, n = 5510) where singing by two birds of the same species overlapped in time. Prior to our auditory sampling observers were trained and tested for accuracy and consistency by data collected using a super-cardioid condenser microphone (ME66/K6, Sennheiser Electronic, Wedemark, Germany) and a field recorder (Fostex FR-2, Foster Electric Corporation, Tokyo, Japan). The microphone was positioned on a tripod 1.5 m above the ground and pointed to the most proximate direction of bird sounds, where the sound with maximum intensity was heard. We recorded light level at the onset of each bout of singing using a light meter (1339 Pro, TES Electronic Corporation, Taipei, Taiwan), and humidity and temperature in the beginning and the end of each census by a combined Humidity and Temperature Indicator (WISEWIND, Centenary Materials, Hsinchu, Taiwan).
Eye size and morphometric measurements
We estimated the eye size of bird species that were recorded in our sites by measuring skull specimens of museum collections (Chen, 2010). We measured orbital morphometric parameters, including the maximum length along the long and short axes of the orbit socket, and the length and width of skull. Eye size in volume (ES, mm3) was estimated using ES = 2 × 1.33πa 2 b (Garamszegi et al., 2002), where a and b each represents the long the short axial length of an eye socket. This estimate was strongly correlated (r = 0.97, p < 0.05; Chen, 2010) to that obtained from the approach of applying a series of spherical plastic balls of different sizes to best fit with the eye socket of the birds (Brooke et al., 1999; Thomas et al., 2006). For few species where proper skull specimens were inadequate or completely lacking, we adopted live bird measurements as a substitute. Birds were mist-netted along trails surrounding our study area, and we measured the maximum lengths along the long and short axes of eye socket (to 0.1 mm) using an electronic caliper (SV-03 Digital Vernier Caliper, E-BASE, Yunlin, Taiwan). Our netting complied with the legal requirements of Taiwan and the guidelines for the use of animals in research (Sherwin, 2006) throughout our procedures. The orbital measurements (mm) from skull specimens were correlated to direct eye measurements of live birds in both the long axis (r = 0.52, p < 0.05) and the short axis (r = 0.80, p < 0.05; Chen, 2010), respectively. We corrected our eye measurements using these correlations, but kept the use of this measure to a necessary minimum for a better consistency. We used log-transformed eye size (LES) and its relative value (RES), log (eye size)/log (body mass), for further analyses. The latter considered the allometric effect and was corrected for body size (Berg et al., 2006).
Foraging perch heights and food habits
For each species engaged in dawn chorus, we located and sighted individuals to record the aspect and the perching locations between 08:00 and 10:00 on the same morning. The distances and heights of perches were visually estimated using on site reference marks, and later corrected by a laser distance meter (Leica DISTO A5). The perch heights were further categorized to up- (≥ 7 m), mid- (≥ 1.5 m to < 7 m), and bottom (< 1.5 m) layers. We additionally recorded the species and estimated the height of the perch trees, and measured habitat variables surrounding a perch, including canopy and ground coverage.
We determined birds’ diets first by onsite observations after dawn chorus during the study period (Chen, 2010), but not every species’ foraging was adequately documented due to practical difficulties. We further incorporated data retrieved from Ding et al. (2008), Wu (2008), and those compiled in Severinghaus et al. (2012). We used data that were collected by a similar method in mid-elevation montane areas, prioritizing those with seasons and locations proximate to this study, or adopting their averages when no distinction could be made. We used the item proportions of animal contents at 25% and 75% as dividing levels to further classify diets into three classes: animal-eating (≥ 75%), omnivores (≥ 25 but < 75 %), and plant-eating (< 25%; Remsen and Parker, 1984). Animal contents here refer only to insects and non-insect invertebrates. We excluded true carnivores, and birds that are active nocturnally or often vocalize in the night, such as night herons, were also excluded.
Phylogenetic independent contrasts
We grouped species engaging in the dawn chorus into three rooted and unambiguous clades: Galliformes (Phasianidae), Piciformes (Picidae and Ramphastidae), and passerines. To assure independence of observations, we calculated Felsenstein’s (1985) phylogenetic independent contrasts in character traits. The contrast calculation was conducted using a phylogenetic hypothesis based on Sibley and Ahlquist (1990), and incorporating recent treatments on the Old World babblers (Timaliidae; Baker et al., 2001; Cibois, 2003; Collar and Robson, 2007), and the results from cytochrome oxidase subunit I (COI) sequence analyses (CT Yao and SH Li, unpubl. data) for interspecific relationships within this group. Phylogenetic independent contrasts were calculated using COMPARE, assuming that the character evolution can be described in a random walk mode, equal branch lengths, and negligible within species variation compared to that among species (Martins, 2004).
Data are presented as mean ± standard error (SE) unless otherwise noted. We performed statistical analyses using MINITAB 14.12 (Minitab 2003) and SPSS 12.0.1 (SPSS 2000) for Windows 2000, and set the significance level at α = 0.05. We corrected onset times relative to the local sunrise to obtain the relative onset times of dawn singing, and calculated duration of dawn signing, for each species and each dawn chorus sequence (Lein, 2007; Liu and Kroodsma, 2007; Foote et al., 2008). We adopted multivariate analysis of variance (MANOVA) with Wilks’ Λ to examine the temporal (sampling month and year) and spatial (elevation) variation in onset time, total chorus duration, and mean chorus duration. Analysis of variance (ANOVA) was used to examine the differences in onset time among birds of different eye-size groups and foraging heights. In both analyses, we conducted post-hoc comparisons using Fisher’s least significant difference (LSD) to locate differences. Pearson’s correlation analysis was used to examine the relationships of onset time and chorus duration with number of species engaged in dawn singing. The inter-correlation among trait variables, eye size, perching height, and diet content, in independent contrasts was examined using COMPARE (Martins, 2004), and tested by Student’s t = r/[√ (1 − r 2)/√(n - 2)], where r is correlation coefficient and n is sample size. We restricted this analysis to species that associated ecological and morphological data were fully available and that occurred in all sampling sites. We further adopted the contrasts to examine the effects of eye size, perching height, and food habit on onset of dawn signing among species using stepwise regression analysis, if each of the correlation coefficients between paired variables was less than 0.8, setting α = 0.05 for adding a variable to the model (Zar, 2010).