Seabird image identification in natural scenes using Grabcut and combined features

This paper proposes an automated seabird segmentation and identification method that applies to seabird images taken in natural scenes with a non-uniform and complex background. A variety of different bird postures appeared in natural scenes present different features from different points of view, even for the same posture. At first, the Grabcut method is introduced to segment seabird unit from a complicated background. Then, global features, namely the colour, shape and texture characteristics, and local features are integrated to describe the birds regarding various postures. Later, a combined recognition model, which is built using the k-Nearest Neighbor, Logistic Boost and Random Forest models by a voting mechanism that is designed for seabird identification. Finally, the efficiency and effectiveness of the proposed method in recognising seabirds were experimentally demonstrated. The experimental results on 900 field samples (6 seabird species, and 150 samples of each species) achieved a recognition accuracy of 88.1%, which indicates that the proposed method is effective for automated seabird identification.     

DISCOVERING EXCEPTIONAL DIVERSIFICATIONS AT CONTINENTAL SCALES: THE CASE OF THE ENDEMIC FAMILIES OF NEOTROPICAL SUBOSCINE PASSERINES

The study of continental adaptive radiations has lagged behind research on their island counterparts in part because the mere identification of adaptive radiations is more challenging at continental scales. Here, I demonstrate a new method based on simulations for discovering clades that show exceptionally high phenotypic diversity. The method does not require a phylogeny but accounts for differences in age and species richness among clades and incorporates effects of the phylogenetic structure of data. In addition, I developed a new multivariate measure of phenotypic diversity, which has the advantage over other measures of disparity in that it takes covariation into account. I applied these methods to a clade of endemic Neotropical suboscine passerines, within which the family Furnariidae has been considered an adaptive radiation. I found that the families Thamnophilidae, Furnariidae, and Dendrocolaptidae have experienced a higher rate of cladogenesis than have other clades. Although Thamnophilidae is exceptionally diverse in body size, only Furnariidae and Dendrocolaptidae are exceptionally diverse in shape. The combination of high rates of cladogenesis and high morphometric diversity in traits related to feeding and locomotion suggest that the clade Furnariidae‐Dendrocolaptidae represent an authentic continental adaptive radiation.     

Multi-view features fusion for birdsong classification

As important members of the ecosystem, birds are good monitors of the ecological environment. Bird recognition, especially birdsong recognition, has attracted more and more attention in the field of artificial intelligence. At present, traditional machine learning and deep learning are widely used in birdsong recognition. Deep learning can not only classify and recognize the spectrums of birdsong, but also be used as a feature extractor. Machine learning is often used to classify and recognize the extracted birdsong handcrafted feature parameters. As the data samples of the classifier, the feature of birdsong directly determines the performance of the classifier. Multi-view features from different methods of feature extraction can obtain more perfect information of birdsong. Therefore, aiming at enriching the representational capacity of single feature and getting a better way to combine features, this paper proposes a birdsong classification model based multi-view features, which combines the deep features extracted by convolutional neural network (CNN) and handcrafted features. Firstly, four kinds of handcrafted features are extracted. Those are wavelet transform (WT) spectrum, Hilbert-Huang transform (HHT) spectrum, short-time Fourier transform (STFT) spectrum and Mel-frequency cepstral coefficients (MFCC). Then CNN is used to extract the deep features from WT, HHT and STFT spectrum, and the minimal-redundancy-maximal-relevance (mRMR) to select optimal features. Finally, three classification models (random forest, support vector machine and multi-layer perceptron) are built with the deep features and handcrafted features, and the probability of classification results of the two types of features are fused as the new features to recognize birdsong. Taking sixteen species of birds as research objects, the experimental results show that the three classifiers obtain the accuracy of 95.49%, 96.25% and 96.16% respectively for the features of the proposed method, which are better than the seven single features and three fused features involved in the experiment. This proposed method effectively combines the deep features and handcrafted features from the perspectives of signal. The fused features can more comprehensively express the information of the bird audio itself, and have higher classification accuracy and lower dimension, which can effectively improve the performance of bird audio classification.     

A Comprehensive Quantitative Assessment of Bird Extinction Risk in Brazil

In an effort to avoid species loss, scientists have focused their efforts on the mechanisms making some species more prone to extinction than others. However, species show different responses to threats given their evolutionary history, behavior, and intrinsic biological features. We used bird biological features and external threats to (1) understand the multiple pathways driving Brazilian bird species to extinction, (2) to investigate if and how extinction risk is geographically structured, and (3) to quantify how much diversity is currently represented inside protected areas. We modeled the extinction risk of 1557 birds using classification trees and evaluated the relative contribution of each biological feature and external threat in predicting extinction risk. We also quantified the proportion of species and their geographic range currently protected by the network of Brazilian protected areas. The optimal classification tree showed different pathways to bird extinction. Habitat conversion was the most important predictor driving extinction risk though other variables, such as geographic range size, type of habitat, hunting or trapping and trophic guild, were also relevant in our models. Species under higher extinction risk were concentrated mainly in the Cerrado Biodiversity Hotspot and were not quite represented inside protected areas, neither in richness nor range. Predictive models could assist conservation actions, and this study could contribute by highlighting the importance of natural history and ecology in these actions.     

Testing acoustic versus physical marking: two complementary methods for individual-based monitoring of elusive species

Individuals of some species differentiate each other on the basis of the acoustic features of their vocalizations, and this can be used in individual‐based population monitoring studies. No research has tested for the effectiveness of individual marking through voice recognition as compared to traditional monitoring methods relying on physical marks. We compared voice recognition and physical marking using the Dupont's lark Chersophilus duponti as a study species. This bird needs to be attracted with playback in order to be seen (or captured). We first demonstrated that the territorial calls from a sample of banded males were individually distinctive and constant over time by means of discriminant function analysis, which correctly classified 100% of marked males. Then, we applied similarity techniques on call spectrotemporal features to define a threshold value of similarity within banded individuals, to be combined with qualitative spectrogram inspection for the classification of all recorded birds. Eventually, we compared the voice and the capture samples, to test for differences in relation to re‐location rate, territory fidelity and dispersal movements both within and between years. Voice recognition was less time‐consuming than capture‐recapture method in the field, but it was useless for monitoring yearlings in call development stage. The two methods provided the same results in terms of territory fidelity and dispersal movements, but differed in re‐location rates, which were significantly greater in the case of voice recognition method. By means of physical captures we possibly trapped a large sample of young and silent floaters, with low probability of recapture or recording. This mismatch between methods could bias the estimates of annual survival, which strongly depend on re‐location rates. We suggest considering the two methods as complementary rather than alternatives for monitoring populations. Each technique offers unique information, and the two sources should be combined to provide correction factors that would eventually sharpen our knowledge on bird population ecology.     

Western Mediterranean Wetland Birds dataset: A new annotated dataset for acoustic bird species classification

The deployment of an expert system running over a wireless acoustic sensors network made up of bioacoustic monitoring devices that recognize bird species from their sounds would enable the automation of many tasks of ecological value, including the analysis of bird population composition or the detection of endangered species in areas of environmental interest. Endowing these devices with accurate audio classification capabilities is possible thanks to the latest advances in artificial intelligence, among which deep learning techniques stand out. To train such algorithms, data from the sources to be classified is required. For this reason, this paper presents the Western Mediterranean Wetland Birds (WMWB) dataset, consisting of 201.6 min and 5795 annotated audio excerpts of 20 endemic bird species of the Aiguamolls de l'Empordà Natural Park. The main objective of this work is to describe and analyze this new dataset. Moreover, this work presents the results of bird species classification experiments using four well- known deep neural networks fine-tuned on our dataset, whose models are also made public along with the dataset. These results are aimed to serve as a performance baseline reference for the community when using the WMWB dataset for their experiments.     

Molecular systematics of New World suboscine birds

Phylogenetic relationships among New World suboscine birds were studied using nuclear and mitochondrial DNA sequences. New World suboscines were shown to constitute two distinct lineages, one apparently consisting of the single species Sapayoa aenigma, the other made up of the remaining 1000+ species of New World suboscines. With the exception of Sapayoa, monophyly of New World suboscines was strongly corroborated, and monophyly within New World suboscines of a tyrannoid clade and a furnarioid clade was likewise strongly supported. Relationships among families and subfamilies within these clades, however, differed in several respects from current classifications of suboscines. Noteworthy results included: (1) monophyly of the tyrant-flycatchers (traditional family Tyrannidae), but only if the tityrines (see below) are excluded; (2) monophyly of the pipromorphine flycatchers (Pipromorphinae of ) as one of two primary divisions of a monophyletic restricted Tyrannidae; (3) monophyly of the tityrines, consisting of the genus Tityra plus all sampled species of the Schiffornis group (), as sister group to the manakins (traditional family Pipridae); (4) paraphyly of the ovenbirds (traditional family Furnariidae), if woodcreepers (traditional family Dendrocolaptidae) are excluded; and (5) polyphyly of the antbirds (traditional family Formicariidae) and paraphyly of the ground antbirds (Formicariidae sensu stricto). Genus Melanopareia (the crescent-chests), although clearly furnarioid, was found to be distant from other furnarioids and of uncertain affinities within the Furnarii. Likewise, the species Oxyruncus cristatus (the Sharpbill), although clearly tyrannoid, was distantly related to other tyrannoids and of uncertain affinities within the Tyranni. Results of this study provide support for some of the more novel features of the suboscine phylogeny of, but also reveal key differences, especially regarding relationships among suboscine families and subfamilies. The results of this study have potentially important implications for the reconstruction of character evolution in the suboscines, especially because the behavioral evolution of many suboscine groups (e.g., Furnariidae) is of great interest.     

Classifying and ranking audio clips to support bird species richness surveys

Advances in programmable field acoustic sensors provide immense data for bird species study. Manually searching for bird species present in these acoustic data is time-consuming. Although automated techniques have been used for species recognition in many studies, currently these techniques are prone to error due to the complexity of natural acoustics.In this paper we propose a smart sampling approach to help identify the maximum number of bird species while listening to the minimum amount of acoustic data. This approach samples audio clips in a manner that can direct bird species surveys more efficiently. First, a classifier is built to remove audio clips that are unlikely to contain birds; second, the remaining audio clips are ranked by a proxy for the number of species. This technique enables a more efficient determination of species richness.The experimental results show that the use of a classifier enables to remove redundant acoustic data and make our approach resilient to various weather conditions. By ranking audio clips classified as “Birds”, our method outperforms the currently best published strategy for finding bird species after 30 one-minute audio clip samples. Particularly after 60 samples, our method achieves 10 percentage points more species. Despite our focus on bird species, the proposed sampling approach is applicable to the search of other vocal species.     

Automatic classification of electrocardiogram signals based on transfer learning and continuous wavelet transform

Classification and subsequent diagnosis of cardiac arrhythmias is an important research topic in clinical practice. Confirmation of the type of arrhythmia at an early stage is critical for reducing the risk and occurrence of cardiovascular events. Nevertheless, diagnoses must be confirmed by a combination of specialist experience and electrocardiogram (ECG) examination, which can lead to delays in diagnosis. To overcome such obstacles, this study proposes an automatic ECG classification algorithm based on transfer learning and continuous wavelet transform (CWT). The transfer learning method is able to transfer the domain knowledge and features of images to a EGG, which is a one-dimensional signal when a convolutional neural network (CNN) is used for classification. Meanwhile, CWT is used to convert a one-dimensional ECG signal into a two-dimensional signal map consisting of time-frequency components. Considering that morphological features can be helpful in arrhythmia classification, eight features related to the R peak of an ECG signal are proposed. These auxiliary features are integrated with the features extracted by the CNN and then fed into the fully linked arrhythmia classification layer. The CNN developed in this study can also be used for bird activity detection. The classification experiments were performed after converting the two types of audio files containing songbird sounds and those without songbird sounds from the NIPS4Bplus bird song dataset into the Mel spectrum. Compared to the most recent methods in the same field, the classification results improved accuracy and recognition by 11.67% and 11.57%, respectively.     

Native plantations as an important element for biodiversity in vanishing forested landscapes: A study of the near threatened araucaria tit spinetail (Leptasthenura setaria,Furnariidae)

Although forest loss is still a problem worldwide, estimated rates of deforestation have declined in the last decade, primarily because of an increase in the area of tree plantations. This leads to the central question of how suitable plantations are for indigenous species. Native plantations are thought to have higher value for biodiversity than plantations of non‐native trees; however, not all studies support this view. We assessed occupancy and density of the araucaria tit spinetail (Leptasthenura setaria, Furnariidae), a near threatened species, in the highly endangered araucaria forests of north‐eastern Argentina and in araucaria plantations, which comprise 90% of the remaining habitat for this species. All natural forest remnants were occupied by araucaria tit spinetails. Only 85% of the plantations were occupied; however, density was almost threefold higher in plantations compared with natural forests. Our models indicated that stand age was the most important factor in determining occupancy and density of this bird species in plantations. Plantations <10 years old exhibited lower densities than older plantations. This species does not occur in plantations of non‐native trees, but our results indicate that native plantations may provide important habitat for the araucaria tit spinetail, particularly given that most native forest has been removed. Restoration of natural remnants and conservation of old, connected plantations may assure the protection of significant populations of spinetails. The role of native araucaria plantations as habitat for other species merits further examination.     

Indicator species for avian biodiversity hotspots: Combination of specialists and generalists is necessary in less natural environments

In this work, I tested the premise that the distribution of a group of few common bird species can be used to predict bird species hotspots in Central Italy. The data on bird observations were collected on 530 sampled sites (150 in cultivated, 150 in forest, 150 in grassland and 80 in urban and peri-urban environments). In each environment, sampled sites with values of bird species richness in the upper than third quartile were classified as high species richness spots (HSRS), while sites with lower bird species richness were classified as non-HSRS (binary classification system).Generalized Linear Models (GLM) were applied using HSRS or non-HSRS as binomial response variable and bird species occurrence was used as the predictor variable. All selected models showed “fair” or “good” capacities to predict the avian hotspots, using only few common birds (4–6) species. However, bird species selected as predictors were different on each environment. In more natural environments (grassland, forest), specialist species were selected, while in most disturbed environments (cultivated and urban) both generalist and specialist species were selected. The results are in agreement with other studies which show how homogenization of bird communities is strongly correlated to landscape disturbance. The findings supports the hypothesis that indicators have to incorporate both specialists and generalist’s species simultaneously. Furthermore, the groups of birds selected as surrogates are easy to detect and this makes it possible to involve citizen-science programmes in obtain data. This approach can be a cheap and efficient and can help to significantly speed up the process of assessing ecosystems that might be under threat.     

Ecological opportunity and diversification in a continental radiation of birds: climbing adaptations and cladogenesis in the Furnariidae

Ecological theories of adaptive radiation predict that ecological opportunity stimulates cladogenesis through its effects on competitive release and niche expansion. Given that key innovations may confer ecological opportunity, we investigated the effect of the acquisition of climbing adaptations on rates of cladogenesis in a major avian radiation, the Neotropical bird family Furnariidae, using a species-level phylogeny. Morphological specializations for vertical climbing originated in the woodcreepers ～23 million years ago, well before that adaptation occurred in woodpeckers (Picidae) or in other potential competitors in South America. This suggests that the acquisition of climbing adaptations conferred ample ecological opportunity to early woodcreepers. Nonetheless, we found that increases in speciation rates in Furnariidae did not coincide with the acquisition of climbing adaptations and that the relationship between the accumulation of climbing adaptations and rates of speciation was negative. In addition, we did not detect a diversity-dependent decline in woodcreeper diversification rates consistent with saturation of the trunk-climbing niche. These findings do not support the hypothesis that ecological opportunity related to trunk foraging stimulated cladogenesis in this radiation. The negative effect of climbing on diversification may be mediated by an indirect positive effect of climbing on dispersal ability, which may reduce speciation rates over evolutionary timescales.     

Method for passive acoustic monitoring of bird communities using UMAP and a deep neural network

An effective practice for monitoring bird communities is the recognition and identification of their acoustic signals, whether simple, complex, fixed or variable. A method for the passive monitoring of diversity, activity and acoustic phenology of structural species of a bird community in an annual cycle is presented. The method includes the semi-automatic elaboration of a dataset of 22 vocal and instrumental forms of 16 species. To analyze bioacoustic richness, the UMAP algorithm was run on two parallel feature extraction channels. A convolutional neural network was trained using STFT-Mel spectrograms to perform the task of automatic identification of bird species. The predictive performance was evaluated by obtaining a minimum average precision of 0.79, a maximum equal to 1.0 and a mAP equal to 0.97. The model was applied to a huge set of passive recordings made in a network of urban wetlands for one year. The acoustic activity results were synchronized with climatological temperature data and sunlight hours. The results confirm that the proposed method allows for monitoring a taxonomically diverse group of birds that nourish the annual soundscape of an ecosystem, as well as detecting the presence of cryptic species that often go unnoticed.     

Lineage diversification and morphological evolution in a large-scale continental radiation: the neotropical ovenbirds and woodcreepers (aves: Furnariidae)

Patterns of diversification in species-rich clades provide insight into the processes that generate biological diversity. We tested different models of lineage and phenotypic diversification in an exceptional continental radiation, the ovenbird family Furnariidae, using the most complete species-level phylogenetic hypothesis produced to date for a major avian clade (97% of 293 species). We found that the Furnariidae exhibit nearly constant rates of lineage accumulation but show evidence of constrained morphological evolution. This pattern of sustained high rates of speciation despite limitations on phenotypic evolution contrasts with the results of most previous studies of evolutionary radiations, which have found a pattern of decelerating diversity-dependent lineage accumulation coupled with decelerating or constrained phenotypic evolution. Our results suggest that lineage accumulation in tropical continental radiations may not be as limited by ecological opportunities as in temperate or island radiations. More studies examining patterns of both lineage and phenotypic diversification are needed to understand the often complex tempo and mode of evolutionary radiations on continents.     

The effect of windthrow and its management on breeding bird communities in a managed forest

This study aimed to evaluate the impact of windthrow and two types of its management on the bird community of a managed, pine-dominated forest (NE Poland), where a windstorm event occurred in 2002. In 2007, the bird species composition was assessed using the point count method in three types of habitat: windthrow left undisturbed for natural regeneration, managed windthrow (all fallen trees removed), and intact forest (undisturbed by the windstorm, used as the control). In total, 2,365 individual birds of 70 species were recorded in the three analyzed habitats. The mean number of individuals per point and species per point was significantly higher in the natural windthrow than in the two other habitats. The diversity of the bird community was higher in the natural and managed windthrow, as compared to the control. The bird community of the control was significantly more similar to that of the natural windthrow than to the one in the managed windthrow. These results lead to a conclusion that the bird community was affected more by the clearing and artificial replanting following the windthrow than by the windthrow itself.     

Factors affecting nesting success of a bird assembly in the central Monte Desert, Argentina

We estimated nesting success of a breeding bird assemblage in the central Monte Desert in four breeding seasons (1995–1999). We analyzed the effect of year, period within the breeding season, nest type, and nest location on daily survival rates (DSR). Averaged nesting success of passeriforms in natural habitats was very low (<20% on average, n=12 species). DSR tended to be lower during the egg-laying stage in Furnariidae, Rhinocryptidae, Mimidae, and Emberizidae. Tyrannidae had the highest DSR for all nesting stages and Emberizidae the lowest. Predation was the main cause of nest failure (>90% on average, n=12 species). DSR differed among years, being higher in the driest year (1995) and lower in the most humid year (1997). Nests initiated in the first half of the breeding season had higher DSR than those initiated in the second half. Closed nests had similar survival rates to open nests. DSR differed according to species of plant that supported the nest. Nest predation in the central Monte Desert may be an important selective pressure in birds' life history, which is in accordance with other results for South American temperate birds.     

Unsupervised classification to improve the quality of a bird song recording dataset

Open audio databases such as Xeno-Canto are widely used to build datasets to explore bird song repertoire or to train models for automatic bird sound classification by deep learning algorithms. However, such databases suffer from the fact that bird sounds are weakly labelled: a species name is attributed to each audio recording without timestamps that provide the temporal localization of the bird song of interest. Manual annotations can solve this issue, but they are time consuming, expert-dependent, and cannot run on large datasets. Another solution consists in using a labelling function that automatically segments audio recordings before assigning a label to each segmented audio sample. Although labelling functions were introduced to expedite strong label assignment, their classification performance remains mostly unknown. To address this issue and reduce label noise (wrong label assignment) in large bird song datasets, we introduce a data-centric novel labelling function composed of three successive steps: 1) time-frequency sound unit segmentation, 2) feature computation for each sound unit, and 3) classification of each sound unit as bird song or noise with either an unsupervised DBSCAN algorithm or the supervised BirdNET neural network. The labelling function was optimized, validated, and tested on the songs of 44 West-Palearctic common bird species. We first showed that the segmentation of bird songs alone aggregated from 10% to 83% of label noise depending on the species. We also demonstrated that our labelling function was able to significantly reduce the initial label noise present in the dataset by up to a factor of three. Finally, we discuss different opportunities to design suitable labelling functions to build high-quality animal vocalizations with minimum expert annotation effort.     

Birds in Anthropogenic Landscapes: The Responses of Ecological Groups to Forest Loss in the Brazilian Atlantic Forest

Habitat loss is the dominant threat to biodiversity and ecosystem functioning in terrestrial environments. In this study, we used an a priori classification of bird species based on their dependence on native forest habitats (forest-specialist and habitat generalists) and specific food resources (frugivores and insectivores) to evaluate their responses to forest cover reduction in landscapes in the Brazilian Atlantic Forest. From the patch-landscapes approach, we delimited 40 forest sites, and quantified the percentage of native forest within a 2 km radius around the center of each site (from 6 - 85%). At each site, we sampled birds using the point-count method. We used a null model, a generalized linear model and a four-parameter logistic model to evaluate the relationship between richness and abundance of the bird groups and the native forest amount. A piecewise model was then used to determine the threshold value for bird groups that showed nonlinear responses. The richness and abundance of the bird community as a whole were not affected by changes in forest cover in this region. However, a decrease in forest cover had a negative effect on diversity of forest-specialist, frugivorous and insectivorous birds, and a positive effect on generalist birds. The species richness and abundance of all ecological groups were nonlinearly related to forest reduction and showed similar threshold values, i.e., there were abrupt changes in individuals and species numbers when forest amount was less than approximately 50%. Forest sites within landscapes with forest cover that was less than 50% contained a different bird species composition than more extensively forested sites and had fewer forest-specialist species and higher beta-diversity. Our study demonstrated the pervasive effect of forest reduction on bird communities in one of the most important hotspots for bird conservation and shows that many vulnerable species require extensive forest cover to persist.