Taxonomic revision of the Savanna Nightjar (Caprimulgus affinis) complex based on vocalizations reveals three species

Abstract

Background

The Savanna Nightjar (Caprimulgus affinis) is a widespread, polytypic species which was previously treated as two or three species. It is currently treated as a single species based on superficial similarity of their songs but no detailed comparisons of the songs in this complex have been made.

Methods

A total of 15 acoustic variables were measured for the songs of 86 individuals representing 8 of the 10 subspecies in the complex.

Results

Three major groups can be distinguished based on univariate and multivariate analyses: a northern group consisting of the subspecies C. a. monticolus, C. a. amoyensis and C. a. stictomus; a southern group consisting of C. a. affinis, C. a. kasuidori, C. a. timorensis and C. a. propinquus; and a third group in the Philippines consisting of C. a. griseatus.

Conclusions

It is here argued that these groups are best treated as species, and that Franklin’s Nightjar (C. monticolus) and Kayumanggi Nightjar (C. griseatus) are reinstated as separate species.

Background

Most species of nightjars and owls have a cryptic plumage which has long hampered taxonomic study of their species limits. During the last two decades, quantitative comparisons of songs have helped clarify species limits in several groups, including pygmy owls [Glaucidium [Howell and Robbins 1995; Gwee et al. 2019)], scops owls [Otus (Rasmussen et al. 2000; Sangster et al. 2013)], screech owls [Megascops (Krabbe 2017; Dantas et al. 2021)], hawk owls (Ninox [Rasmussen et al. 2012; Gwee et al. 2017]) and nightjars [Caprimulgus ([Sangster and Rozendaal 2004)]. Three aspects make songs in these groups useful for taxonomic purposes. First, in nearly all groups of non-passerines songs, including owls and nightjars, are not known to be learned (Kroodsma 2004). Variation is therefore likely inherited and may provide information about evolutionary relationships. Second, in some species of owls and nightjars songs are known to be involved in intra- and interspecific communication (reviewed by Sangster and Rozendaal 2004). This makes their songs a useful indicator of species limits (Marshall 1978). Third, songs in both groups are rather simple and stereotypical (Marshall 1978), which makes homology assessment easy. Vocalizations are therefore a useful avenue for clarifying and refining species limits in other species of nightbirds.

The Savanna Nightjar (Caprimulgus affinis Horsfield, 1821) is widely distributed in the Oriental region, ranging from northern Pakistan to Indonesia and Timor-Leste (Fig. 1). The song of the species is distinctive and can be described as a rasping “tschreep” note. Whereas geographic variation in the vocalizations of Large-tailed Nightjar (C. macrurus Horsfield, 1821) has long been known (Marshall 1978) and has been used to delimit species (Mees 1985; Rozendaal 1990; Sangster and Rozendaal 2004), no such knowledge exists for C. affinis.

Fig. 1

Range map of Caprimulgus affinis (based on BirdLife International and Handbook of the Birds of the World 2017), indicating subspecies limits (sensu Cleere 1998). The taxa undulatus Mayr, 1944 and kasuidori Mayr, 1944 were included in C. a. affinis by Mees (2006) and Dickinson and Remsen (2013)

In the early twentieth century, taxonomic authorities recognized C. monticolus Franklin, 1831 and C. affinis as separate specis, the former occurring on mainland Asia and the latter in southern Peninsular Malaysia, Singapore and Indonesia east to Timor-Leste (Sharpe 1901; Peters 1940). Sharpe (1901) also recognized C. griseatus as a species. While discussing a letter from Erwin Stresemann on the birds of Yunnan, China, Rothschild (1927) noted that he disagreed with Stresemann that C. monticolus and C. affinis were conspecific. Mayr (1944) noted that the plumage of the Philippine taxon C. a. griseatus Walden, 1875 was intermediate between that of C. monticolus and C. affinis and regarded them as a single species. Sibley and Monroe (1990) ackowledged the occasional treatment of C. monticolus and C. affinis as species but noted that their calls are identical. These three opinions have formed the basis for recognizing a single species, a treatment which is now universally adopted in field guides (King et al. 1975; Robson 2000; Rasmussen and Anderton 2005; Allen 2020; Eaton et al. 2021), handbooks (Cleere 1998; Holyoak 2001) and taxonomic lists (Wolters 1976; Inskipp et al. 1996; Clements 2007; Dickinson and Remsen 2013; del Hoyo and Collar 2014; Gill et al. 2020). The only exception were Howard and Moore (1991), who presumably followed Peters (1940) in treating C. monticolus as a distinct species.

In this study, we revisit species limits in C. affinis using bioacoustic data on eight of the ten subspecies recognized by Cleere (1998) and Holyoak (2001).

Methods

Recordings were obtained from the Xeno-Canto (http://www.xeno-canto.org/) and AVoCet (http://avocet.zoology.msu.edu/) databases, and the bird sound collections of the Florida Museum of Natural History (https://www.floridamuseum.ufl.edu/bird-sounds/) and the British Library Sound Archive (http://cadensa.bl.uk/). The data set was supplemented by published recordings (Marshall 1978; Ranft and Cleere 1998; Jännes 2002; Supari 2003; Scharringa 2005; Chappuis et al. 2009) and our own recordings, which we submitted to the Macaulay Library (https://www.macaulaylibrary.org/). In total, songs of 86 individuals of 8 of the 10 subspecies of C. affinis (sensu Cleere 1998; Holyoak 2001) were available for analysis: C. a. monticolus (34 individuals), C. a. amoyensis (2), C. a. stictomus (11), C. a. affinis (21), C. a. kasuidori (1), C. a. timorensis (5), C. a. propinquus (1) and C. a. griseatus (11). A list of recordings with localities and recordists is provided in Additional file 1: Table S1.

A total of 15 variables was defined on the basis of sonagrams (Fig. 2). The following measurements were recorded: (1) F1, frequency at the start of the song; (2) F2, frequency at the first low; (3) F3, frequency at the second peak; (4) F4, frequency at the second low; (5) F5, frequency at the third peak; (6) F6, maximum frequency, which is the highest frequency present; (7) F7, minimum frequency, which is the lowest frequency present; (8) DF1, the frequency drop between the second peak and the second low; (9) DF2, the frequency drop between the second and third peaks; (10) DF3, frequency range, which is the difference between the maximum and minimum frequency; (11) DT1, total song duration; (12) DT2, the duration of the first downward element at the point where the song begins to increase in frequency; (13) DT3, the interval between the second peak and the end of the song; (14) DT4, the interval between the second and third peaks; and (15) DT5, the interval between the first and second peaks. The first ten of these (F1 to DF3) are frequency-related variables, whereas the last five (DT1-DF5) are time-related variables.

Fig. 2

Measurement of song variables. The left sonagram shows the frequency parameters and the right sonagram shows the time parameters

Principal Component Analysis (PCA) was used to reduce the 15 acoustic variables to a limited number of uncorrelated variables. ANOVA was used to test whether the groups defined by PCA differed from each other.

Canonical Discriminant Function Analysis (DFA) was applied to the acoustic variables of individuals to test whether the individuals could be correctly assigned to the groups defined by PCA. DFA generates a set of criteria to assign individuals to groups that are defined prior to the analysis. Prior to DFA analysis, a tolerance test was conducted to assess the independence of each variable. Variables that failed the tolerance test, i.e. which are an almost linear combination of other variables, were excluded from the analyses. Two DFAs were performed: (i) a “descriptive” DFA, in which the observations used to develop the criteria are then subjected to these criteria; (ii) a “predictive” DFA, which uses a jackknife procedure to obtain a more accurate test of the predictive performance of the DFA. In the jackknife procedure, the DFA is recalculated using the combination of variables of the initial DFA with one individual removed from the data set. The criteria are then used to classify the removed individual. This process was repeated for all individuals of the data set.

The effect size, expressed as Cohen’s d, was calculated to show the strength of the acoustic differences between taxa. For interpretation of effect size data, we used the classification of Cohen (1988), which was updated and expanded by Sawilowsky (2009). Thus, we regard an effect size of d ≥ 0.1 as “very small”, d ≥ 0.2 as “small”, d ≥ 0.5 as “medium”, d ≥ 0.8 as “large”, d ≥ 1.2 as “very large” and d ≥ 2.0 as “huge”.

SPSS version 27.0 (IBM Corp 2020) was used to calculate all descriptive statistics and perform analyses of variance (ANOVA), Mann-Whitney U-tests, Principal Components Analyses, and Discriminant Function Analyses.

Results

Principal component analysis

The songs of 86 individuals were used in the PCA. The results of the PCA on the 15 measurements are summarized in Table 1. Four components with eigenvalues > 1 were extracted from the data set. The first principal component (PC1) accounted for 46.0% of the variance. PC2, PC3 and PC4 accounted for an additional 24.0, 12.0, and 9.1% of the variance, respectively. PC1 was represented by most frequency variables, especially F3 and F6, and DF1. PC2 was determined mostly by F2 and F3, and PC3 mostly by DT1 and DT5.

Table.1 Factor loadings of 15 acoustic variables on the 4 principal components in the Caprimulgus affinis complex. Eigenvalues and percentage of variance explained by the respective components are given at the bottom of the table

Plotting individuals on PC1 versus PC2 revealed three distinct clusters, corresponding to songs from the affinis-group (subspecies C. a. affinis, C. a. kasuidori, C. a. timorensis and C. a. propinquus), the monticolus-group (subspecies C. a. monticolus, C. a. amoyensis and C. a. stictomus) and the griseatus-group (subspecies C. a. griseatus) (Fig. 3). One-way ANOVA showed that the three groups identified by PCA differed in all four principal components (Table 1).

Fig. 3

PCA (left) and DFA (right) scatterplots of acoustic variables measured for songs of the Caprimulgus affinis complex (n = 86)

Discriminant function analysis

The songs of the three groups identified by PCA were used in the DFA. Most variables passed the tolerance test, except F7, DF1, DF2, DF3 and DT5 which were excluded from the test. The descriptive DFA was highly significant (Wilks’ lambda = 0.004; Chi Square₂₀ = 435.6; P < 0.001). The variables most important in the discrimination were F2, F3, F4, F6 and DT4 (Table 2). The initial DFA led to a 100% correct classification of the individuals into the three groups. The jackknife procedure also provided a high degree of predictive discrimination, with 85 of 86 (98.8%) individuals being correctly assigned to their group defined by PCA.

Table.2 Standardized canonical discriminant function coefficients examining trends in variance of ten acoustic variables measured for songs of three groups of Caprimulgus affinis

Univariate analysis

Song characteristics of the three groups identified by PCA are given in Table 3 and illustrated in Fig. 4. All 15 variables differed significantly in comparisons of the monticolus-group with the affinis-group, and some of these also showed non-overlapping ranges (DF1 and DF2 in monticolus-group vs. affinis-group. Similarly, 14 variables differed significantly in comparisons of the monticolus-group with the griseatus-group, and 4 (DF1, DF3, DT3 and DT4) showed no overlap. Comparisons of the affinis-group with the griseatus-group revealed seven significant differences and five variables that showed no overlap between the 2 groups (F2, F7, DF3, DT3 and DT4).

Table.3 Descriptive statistics of 15 variables measured for songs of 3 species in the Caprimulgus affinis complex (mean ± SD, range)

Fig. 4

Sonagrams of songs of the monticolus-group, affinis-group and griseatus-group illustrating the differences among the three groups

The effect size of the differences between the three groups is given in Table 3. The three groups showed multiple “very large” (Cohen’s d > 1.2) or “huge” (Cohen’s d > 2.0) differences in both frequency-related and time-related variables (Table 3).

The differences between the three groups are visible on sonagrams (Fig. 4). The songs of the griseatus-group differ most prominently from the monticolus- and affinis-groups by their lack of a raspy quality (shown on sonagrams as a narrow line in the first upward-inflected element). The differences between the monticolus- and affinis-groups are reflected by (i) the broader basis (i.e. longer duration) of the first downward element of the songs of the monticolus-group than in those of the affinis-group, and in the much larger frequency drops between (ii) F3 and F4 and (iii) F3 and F5 in the monticolus-group.

Discussion

The results of this study show that the northern subspecies C. a. monticolus, C. a. amoyensis and C. a. stictomus, the southern subspecies C. a. affinis, C. a. kasuidori, C. a. timorensis and C. a. propinquus and the Philippine subspecies C. a. griseatus represent separate groups in Principal Component Analysis of variation in vocalizations, and that individuals can be classified correctly at high proportions in Discriminant Function Analysis. The three groups show significant differences in the three principal components and in all univariate variables and there are “very large” to “huge” differences in effect size between the three groups in both frequency-related and time-related variables.

The lack of evidence for vocal learning in most non-passerines, including nightjars, implies that vocal differences are innate and likely have a genetic basis. Thus, population-level differences in vocalizations may reflect evolutionary divergence. The multiple differences documented between the geographically separate monticolus and affinis groups likely means that these populations have been subjected to a long period of genetic isolation. Consequently, the three non-overlapping groups in the DFA plot are therefore best interpreted as evidence that the monticolus-group, the affinis-group and the griseatus-group represent separate evolutionary lineages, i.e. separate branches in the tree of life. According to the General Lineage Concept (de Queiroz 1999, 2007) such lineages are best considered as species: C. monticolus, C. affinis and C. griseatus. As English names we propose to reinstate the name Franklin’s Nightjar for C. monticolus and to maintain Savanna Nightjar for C. affinis. For C. griseatus we suggest the name Kayumanggi Nightjar, in recognition of the Filipino name Kayumanggi for brownish-coloured, particularly with reference to tan coloured skin. The three species reinstate species last recognized by Sharpe (1901) and reflect differences in size and plumage coloration. C. monticolus is larger and browner than C. affinis (Cleere 1998). Indeed, data in Cleere (1998) show no overlap in wing length between C. monticolus (males 181–208 mm; females 177–208 mm) and C. affinis (males 150–172 mm; females 152–170 mm). C. griseatus is greyer than C. affinis and the barring on its underparts is finer and extends lower on the belly (Cleere 1998; Holyoak 2001).

Unfortunately, no recordings were available of the Philippine taxon C. a. mindanensis Mearns, 1905. Thus, it is not clear if this taxon belongs to C. griseatus or to C. affinis, or perhaps represents another vocally distinct group. Pending further analysis, we suggest that C. a. mindanensis be treated as conspecific with C. griseatus on geographic grounds. We are not aware of any reliable recent records of C. a. mindanensis, and we hope our paper provides impetus to find and study this poorly known taxon.

Taxonomic study of the C. affinis complex, and that of other groups of nightjars, could further benefit from molecular phylogenetic and phylogeographic analyses. This could (i) corroborate and refine of species limits based on morphological or bioacoustic patterns, (ii) facilitate the discovery of additional lineages, and (iii) provide a historical perspective on the biogeography of the group. Conversely, modern morphological and bioacoustic studies of species limits may benefit phylogenetic and phylogeographic analyses by indicating which populations should be sampled and where additional cryptic species may be located.

Conclusions

In recent decades, avian species-level taxonomy shows two major trends: improved documentation of species taxa and a refinement of species limits. As a consequence, the scientific underpinnings of avian taxonomy continue to be improved and the number of taxonomically-recognized species increases steadily (Sangster and Luksenburg 2015; Sangster 2018). The increase of species is not a goal of taxonomy but results from the improved understanding of species limits due to new information on groups that often have long been neglected. This process is especially important in birds due to the large-scale lumping of species in the first half of the twentieth century without detailed study (reviewed by Haffer 1992; Sangster 2018). The Savanna Nightjar complex is an example of three valid species that have long been treated as a single species without a solid scientific basis. The results of this study thus underscore the importance of identifying and revisiting poorly-documented taxonomic changes.