Alexander N. G. Kirschel

Automated species recognition of antbirds in a Mexican rainforest using hidden Markov models

Behavioral and ecological studies would benefit from the ability to automatically identify species from acoustic recordings. The work presented in this article explores the ability of hidden Markov models to distinguish songs from five species of antbirds that share the same territory in a rainforest environment in Mexico. When only clean recordings were used, species recognition was nearly perfect, 99.5%. With noisy recordings, performance was lower but generally exceeding 90%. Besides the quality of the recordings, performance has been found to be heavily influenced by a multitude of factors, such as the size of the training set, the feature extraction method used, and number of states in the Markov model. In general, training with noisier data also improved recognition in test recordings, because of an increased ability to generalize. Considerations for improving performance, including beamforming with sensor arrays and design of preprocessing methods particularly suited for bird songs, are discussed. Combining sensor network technology with effective event detection and species identification algorithms will enable observation of species interactions at a spatial and temporal resolution that is simply impossible with current tools. Analysis of animal behavior through real-time tracking of individuals and recording of large amounts of data with embedded devices in remote locations is thus a realistic goal.

The evolutionary consequences of interspecific aggression

Competition has always been a cornerstone of evolutionary biology, and aggression is the predominant form of direct competition in animals, but the evolutionary effects of aggression between species are curiously understudied. Only in the past few years, existing theoretical frameworks have been extended to include interspecific aggression, and significant empirical advances have been made. After arguing that agonistic character displacement (ACD) theory provides the most suitable theoretical framework, we review new empirical evidence for ACD and the results of mathematical models of the process. We consider how ACD can be distinguished empirically from ecological and reproductive character displacement and the additional challenges posed by developmental plasticity. We also provide the first taxonomically broad review of theoretical and empirical work on the effects of interspecific aggression on species coexistence and range limits. We conclude by highlighting promising directions for future research on the evolutionary effects of interspecific aggression.

Acoustic localization of antbirds in a Mexican rainforest using a wireless sensor network

Acoustic localization is a promising method to passively observe vocal animal species, but remains difficult and time consuming to employ. To reduce the labor intensity and impact of deployment, an acoustic localization system has been developed consisting of battery powered wireless sensor nodes. The system also has the ability to perform an acoustic self-survey, which compares favorably in accuracy to global positioning system survey methods, especially in environments such as forest. The self-survey and localization accuracy of the system was tested in the neotropical rainforest of Chiapas, Mexico. A straight-forward and robust correlation sum localization computation method was utilized and is described in detail. Both free-ranging wild antbird songs and songs played from a speaker were localized with mean errors of 0.199 m and 0.445 m, respectively. Finally, additional tests utilizing only a short segment of each song or a subset of sensor nodes were performed and found to minimally affect localization accuracy. The use of a wireless sensor network for acoustic localization of animal vocalizations offers greater ease and flexibility of deployment than wired microphone arrays without sacrificing accuracy.

TESTING ALTERNATIVE HYPOTHESES FOR EVOLUTIONARY DIVERSIFICATION IN AN AFRICAN SONGBIRD: RAINFOREST REFUGIA VERSUS ECOLOGICAL GRADIENTS

Geographic isolation in rainforest refugia and local adaptation to ecological gradients may both be important drivers of evolutionary diversification. However, their relative importance and the underlying mechanisms of these processes remain poorly understood because few empirical studies address both putative processes in a single system. A key question is to what extent is divergence in signals that are important in mate and species recognition driven by isolation in rainforest refugia or by divergent selection across ecological gradients? We studied the little greenbul, Andropadus virens, an African songbird, in Cameroon and Uganda, to determine whether refugial isolation or ecological gradients better explain existing song variation. We then tested whether song variation attributable to refugial or ecological divergence was biologically meaningful using reciprocal playback experiments to territorial males. We found that much of the existing song variation can be explained by both geographic isolation and ecological gradients, but that divergence across the gradient, and not geographic isolation, affects male response levels. These data suggest that ecologically divergent traits, independent of historical isolation during glacial cycles, can promote reproductive isolation. Our study provides further support for the importance of ecology in explaining patterns of evolutionary diversification in ecologically diverse regions of the planet.

USING SONGS TO IDENTIFY INDIVIDUAL MEXICAN ANTTHRUSH FORMICARIUS MONILIGER: COMPARISON OF FOUR CLASSIFICATION METHODS

ABSTRACT This study compares the ability of four classification methods to distinguish between songs of individual Mexican Antthrush Formicarius moniliger: self-organizing maps (SOMs), discriminant function analysis, fuzzy logic and hidden Markov models. Recordings were made under field conditions in a Mexican rainforest. Two types of data were analysed—recordings from birds that had been ringed and identified to sex, and recordings from birds that had been identified based on their recording location and song timing. An event detector extracted song features and SOMs were used to confirm the number of individuals recorded. The SOM separated all five ringed birds successfully, and also differentiated two other birds that were not identified while vocalising. The three supervised learning methods correctly classified over 97% of songs to individual from the set of identified recordings. Tests with songs for predicted, rather than known, individuals yielded more variable results across methods, with results ranging from 77.8% to 93.9% correctly identified. The respective merits of the three supervised classification procedures are discussed for automated recording, detection and classification.

Predicting bird song from space

Environmentally imposed selection pressures are well known to shape animal signals. Changes in these signals can result in recognition mismatches between individuals living in different habitats, leading to reproductive divergence and speciation. For example, numerous studies have shown that differences in avian song may be a potent prezygotic isolating mechanism. Typically, however, detailed studies of environmental pressures on variation in animal behavior have been conducted only at small spatial scales. Here, we use remote‐sensing data to predict animal behavior, in this case, bird song, across vast spatial scales. We use remotely sensed data to predict the song characteristics of the little greenbul (Andropadus virens), a widely distributed African passerine, found across secondary and mature rainforest habitats and the rainforest‐savanna ecotone. Satellite data that captured ecosystem structure and function explained up to 66% of the variation in song characteristics. Song differences observed across habitats, including those between human‐altered and mature rainforest, have the potential to lead to reproductive divergence, and highlight the impacts that both natural and anthropogenic change may have on natural populations. Our approach offers a novel means to examine the ecological correlates of animal behavior across large geographic areas with potential applications to both evolutionary and conservation biology.

Smithornis broadbills produce loud wing song by aeroelastic flutter of medial primary wing feathers.

ABSTRACT Broadbills in the genus Smithornis produce a loud brreeeeet during a distinctive flight display. It has been posited that this klaxon-like sound is generated non-vocally with the outer wing feathers (P9, P10), but no scientific studies have previously addressed this hypothesis. Although most birds that make non-vocal communication sounds have feathers with a shape distinctively modified for sound production, Smithornis broadbills do not. We investigated whether this song is produced vocally or with the wings in rufous-sided broadbill (S. rufolateralis) and African broadbill (S. capensis). In support of the wing song hypothesis, synchronized high-speed video and sound recordings of displays demonstrated that sound pulses were produced during the downstroke, subtle gaps sometimes appeared between the outer primary feathers P6–P10, and wingtip speed reached 16 m s−1. Tests of a spread wing in a wind tunnel demonstrated that at a specific orientation, P6 and P7 flutter and produce sound. Wind tunnel tests on individual feathers P5–P10 from a male of each species revealed that while all of these feathers can produce sound via aeroelastic flutter, P6 and P7 produce the loudest sounds, which are similar in frequency to the wing song, at airspeeds achievable by the wing tip during display flight. Consistent with the wind tunnel experiments, field manipulations of P6, P7 and P8 changed the timbre of the wing song, and reduced its tonality, demonstrating that P6 and P7 are together the sound source, and not P9 or P10. The resultant wing song appears to have functionally replaced vocal song. Highlighted Article: Smithornis broadbills produce a loud klaxon-like wing song by aeroelastic flutter of wing feathers P6 and P7, and not the outer primaries P9/P10.

Rapid song divergence leads to discordance between genetic distance and phenotypic characters important in reproductive isolation

Abstract The criteria for species delimitation in birds have long been debated, and several recent studies have proposed new methods for such delimitation. On one side, there is a large consensus of investigators who believe that the only evidence that can be used to delimit species is molecular phylogenetics, and with increasing numbers of markers to gain better support, whereas on the other, there are investigators adopting alternative approaches based largely on phenotypic differences, including in morphology and communication signals. Yet, these methods have little to say about rapid differentiation in specific traits shown to be important in reproductive isolation. Here, we examine variation in phenotypic (morphology, plumage, and song) and genotypic (mitochondrial and nuclear DNA) traits among populations of yellow‐rumped tinkerbird Pogoniulus bilineatus in East Africa. Strikingly, song divergence between the P. b. fischeri subspecies from Kenya and Zanzibar and P. b. bilineatus from Tanzania is discordant with genetic distance, having occurred over a short time frame, and playback experiments show that adjacent populations of P. b. bilineatus and P. b. fischeri do not recognize one anothers songs. While such rapid divergence might suggest a founder effect following invasion of Zanzibar, molecular evidence suggests otherwise, with insular P. b. fischeri nested within mainland P. b. fischeri. Populations from the Eastern Arc Mountains are genetically more distant, yet share the same song with P. b. bilineatus from Coastal Tanzania and Southern Africa, suggesting they would interbreed. We believe investigators ought to examine potentially rapid divergence in traits important in species recognition and sexual selection when delimiting species, rather than relying entirely on arbitrary quantitative characters or molecular markers.

Automated localization of antbirds and their interactions in a Mexican rainforest

Tropical rainforests contain diverse avian communities incorporating species that compete vocally to propagate their signals to intended receivers. In order to effectively communicate with birds of the same species, birds need to organize their song performance temporally and spatially. An automated identification and localization system can provide information on the spatial and temporal arrangement of songs. Acoustic sensor arrays were tested for the ability to localize the source of songs of antbirds recorded in a Mexican rainforest. Pilot studies with a five‐node array arranged in a rough circle with a 20‐m diameter located the song of Dusky Antbird (Cercomacra tyrannina) with an error of 73 cm and Mexican Antthrush (Formicarius moniliger) with an error of 65 cm from the location of a source loudspeaker within the array. An additional source 21 m outside was also localized. Results will be presented for experiments and recordings of individuals at the Mexican rainforest site in October 2006. Locations...