An automatic bird sound recognition system is a useful tool for collecting data of different bird species for ecological analysis. Together with autonomous recording units (ARUs), such a system provides a possibility to collect bird observations on a scale that no human observer could ever match. During the last decades, progress has been made in the field of automatic bird sound recognition, but recognizing bird species from untargeted soundscape recordings remains a challenge. In this article, we demonstrate the workflow for building a global identification model and adjusting it to perform well on the data of autonomous recorders from a specific region. We show how data augmentation and a combination of global and local data can be used to train a convolutional neural network to classify vocalizations of 101 bird species. We construct a model and train it with a global data set to obtain a base model. The base model is then fine‐tuned with local data from Southern Finland in order to adapt it to the sound environment of a specific location and tested with two data sets: one originating from the same Southern Finnish region and another originating from a different region in German Alps. Our results suggest that fine‐tuning with local data significantly improves the network performance. Classification accuracy was improved for test recordings from the same area as the local training data (Southern Finland) but not for recordings from a different region (German Alps). Data augmentation enables training with a limited number of training data and even with few local data samples significant improvement over the base model can be achieved. Our model outperforms the current state‐of‐the‐art tool for automatic bird sound classification. Using local data to adjust the recognition model for the target domain leads to improvement over general non‐tailored solutions. The process introduced in this article can be applied to build a fine‐tuned bird sound classification model for a specific environment.

Citizen science provides large amounts of biodiversity data. Key challenges in unlocking its full potential include engaging citizens with limited species identification skills and accelerating the transition from data collection to research and monitoring outputs. Here we use a large dataset from Finland to show how even citizens who cannot identify birds themselves can contribute to real-time predictions of avian distributions. This is achieved through a digital twin that combines smartphone-based citizen science with long-term knowledge in a continuously updating model. The app submits raw audio to a backend that classifies birds with machine learning, reducing variation in data quality and enabling validation and reclassification by continuously improving classifiers. We counteracted spatiotemporal sampling biases by interval recordings and permanent point count networks. Over 2 years, the app generated 15 million bird detections. Independent test data show that the digital-twin-informed models are more accurate at predicting bird spatiotemporal distributions. Because our approach is highly scalable and has the potential to generate biomonitoring data even in understudied areas, it could accelerate the flow of reliable biodiversity information and increase inclusivity in citizen science projects.

Across the world, human (anthropophonic) sounds add to sounds of biological (biophonic) and geophysical (geophonic) origin, with human contributions including both speech and technophony (sounds of technological devices). To characterize society’s contribution to the global soundscapes, we used passive acoustic recorders at 139 sites across 6 continents, sampling both urban green spaces and nearby pristine sites continuously for 3 years in a paired design. Recordings were characterized by bird species richness and by 14 complementary acoustic indices. By relating each index to seasonal, diurnal, climatic and anthropogenic factors, we show here that latitude, time of day and day of year each predict a substantial proportion of variation in key metrics of biophony—whereas anthropophony (speech and traffic) show less predictable patterns. Compared to pristine sites, the soundscape of urban green spaces is more dominated by technophony and less diverse in terms of acoustic energy across frequencies and time steps, with less instances of quiet. We conclude that the global soundscape is formed from a highly predictable rhythm in biophony, with added noise from geophony and anthropophony. At urban sites, animals experience an increasingly noisy background of sound, which poses challenges to efficient communication.

Bats are ecologically important mammals whose monitoring increasingly relies on acoustic data. However, many tools for bat call identification remain subscription‐based, closed‐source, region‐specific or limited in scalability, creating barriers to global data integration and method development. We present BSG‐BATS, an open‐access annotation portal and convolutional neural network (CNN)‐based classifier for bat calls. The portal enables researchers and practitioners to annotate calls at species or phonic group level and contribute directly to the iterative improvement of the classifier. As a proof of concept, we trained the first BSG‐BATS model using over 4000 annotated recordings of 21 European species. It gave promising results and outperformed many commonly used commercial and academic tools by providing higher AUC (area under the receiver operating characteristic curve) scores for test data. All annotations are transparently annotated in the portal ( https://bsg.laji.fi/bats/identification/instructions ) and the trained model, code and documentation are available at www.zenodo.org/records/15495676 . BSG‐BATS offers a foundation for community‐driven development of bat sound identification tools. By integrating annotation and model retraining on an open‐access platform, it enables collective improvement and adaptation to new regions, species and sound types. We invite all bat researchers and experts to join this collaborative effort, whether by contributing data ( bsg-bat@helsinki.fi ), annotating sounds ( https://bsg.laji.fi/bats/identification/instructions ) or testing the model in their own work ( https://www.zenodo.org/records/15495676 ).