In this page: Digital Twin Model | Resources | Partners
Real-time Bird Monitoring with Citizen-Science Data
The Prototype Biodiversity Digital Twin (pDT) for Real-time Bird Monitoring leverages the power of citizen science and artificial intelligence to track bird populations. Bird populations are highly responsive to environmental changes, making them excellent indicators of ecological health. Climate change, for example, has been shown to shift migratory patterns and disrupt breeding cycles, leading to mismatches between birds and their food resources. Monitoring these changes is crucial for understanding and mitigating the impacts of environmental shifts. Citizen science has a rich history in bird research, providing vast amounts of data, though often plagued by issues such as observer bias and inconsistent sampling efforts. Utilising a mobile application that records bird vocalisations, this pDT combines these recordings with AI-based classification to provide real-time insights into bird distributions and singing activity. This approach aims to overcome traditional challenges in citizen-science data collection and offers a new framework for continuous bird population monitoring.
To address these challenges, this pDT developed a mobile application named MK ("Muuttolintujen kevät" or "Spring of Migratory Birds" in Finnish), which utilises AI to classify bird vocalisations from audio recordings that are uploaded by anyone using the app. The app not only identifies specific species classifications for the users who upload these files but also uploads raw audio files for further analysis. This dual approach allows for continuous improvement of classification models and ensures data quality as well as engaging citizen scientists.
A main objective of this pDT is to evaluate the feasibility of using citizen science for real-time bird monitoring while ensuring the data's scientific robustness. We aim to integrate these data with existing long-term monitoring efforts by developing standardised point count modules and generating calibration data through simultaneous expert and AI-based counts. An additional goal is to increase public awareness of environmental science and the impacts of climate change on biodiversity.
Real-time Bird Monitoring with Citizen-Science Data and Digital Twin Models
Modelling
The pDT employs a sophisticated modelling strategy that combines prior predictions with real-time data from the app:
Long-term monitoring data and environmental predictors, such as land use, climate, and forest structure.
New audio data are continuously integrated and analysed using the hierarchical model of species communities model, which incorporates species classifications and updates predictions of bird distributions.
Predictions of singing activity are adjusted based on weather forecasts and temporal factors.
Data
The pDT integrates several data sources:
- Audio recordings and metadata from the app, including anonymised user information, location, and time of recording.
- Weather Data from the Copernicus Climate Change Service (C3S) Climate Data Store.
- Land-Cover Data from CORINE.
- Data collected through collaboration with the Finnish bird monitoring program.Historical Citizen Science Data from the laji.FI database.
- Audio Recordings from the ERC-synergy project LIFEPLAN.
Finally to enhance performance, BioDT developed Hmsc-HPC, a high-performance computing module that leverages GPU acceleration for faster model fitting. This integration with the LUMI supercomputer significantly reduces computational time, enabling real-time data analysis and prediction updates.
Who can use the pDT?
The user interface of the pDT, designed using R Shiny, allows users to interact with the model, select bird species, and view predictions through updated maps, graphs, and tables. This interface facilitates real-time engagement and provides insights into bird migration patterns, breeding distributions, and vocalisation activity.