Periodic Reporting for period 1 - HyperCitizen (Breaking the wall between professional science and citizen science by hyperautomation)
Reporting period: 2023-07-01 to 2024-12-31
Citizen science has become an integral part of biodiversity research with enormous potential for both improving biodiversity monitoring and involving citizens in scientific projects. However, the full integration of data generated with citizen science in professional research programs has not been fully reached. This is largely due to methodological challenges that compromise the value of the data that most citizen science projects have up to date generated. This project aimed to create an innovative solution that enables citizen science projects to generate data that are essentially equally informative as data generated by professional scientists. At the same time, this project aimed to provide additional value for the citizen scientists themselves, engaging them more profoundly to scientific research than has been earlier possible. The project aimed to generate the proof of concept of enhanced digital citizen science using audio-based monitoring of Finnish birds as a case study. Specific aims involved (i) tailoring an innovative mobile phone citizen science application (Research for JYU Mobile) to audio-based bird monitoring, (ii) using hyperautomation to interact with citizen scientists anonymously to both guide data collection to locations and times suggested by Bayesian adaptive design and to provide motivating feedback, (iii) using convolutional neural networks for automated species classification from the audio data, (iv) using joint species distribution models to continuously update predicted maps of species distributions and their vocalization activity, (v) utilizing high-performance computing to enable essentially real-time flow of information from user recording to biomonitoring outputs, (vi) and running the citizen science campaign in close collaboration with the Finnish Broadcasting company YLE to ensure high visibility of the project on social media, internet, television, and radio, and attracting a sufficient number of participants.
The activities performed followed closely those described in the project proposal. Out of the specific aims, i, iii and vi were fully achieved (more comprehensively than described in the proposal). Namely, we tailored the innovative mobile phone citizen science application (Research for JYU Mobile) to audio-based bird monitoring, and published the "Muuttolintujen kevät" (spring of migratory birds) phone app, currently available in app stores for Finnish users. The phone app submits the raw audio to computational backend that uses convolutional neural networks for automated species classification from the audio data; the model we have developed performs much better than e.g. BirdNet and is well calibrated to these data. We have run the citizen science campaign in close collaboration with the Finnish Broadcasting company YLE, which approach has attracted a very large number of users (270,000 users; ca. 5% of the Finnis population) who have submitted >20M bird vocalizations. To integrate the citizen science approach to professional research, we have implemented permanent point count routes into two Finnish cities (Helsinki and Jyväskylä) as well to ten Finnish National Parks. Citizens have performed in these pre-selected locations >10,000 standardized 5-minute point counts with their smart phones. All raw audio is stored, enabling rigorous validation and reuse of the data. Out of the specific aims, ii, iv and v were partially achieved. Namely, while we have implemented high-performance computing workflows of joint species distribution modelling to enable essentially real-time flow of information from user recording to biomonitoring outputs, these workflows are currently implemented only as beta-versions and they will be finalized and published in future work. As these developments have taken more time than we expected, we were not able to utilize them in in hyperautomation to incorporate citizen scientists in Bayesian adaptive design. We have however provided the citizens encouraging feedback through the app. This has been successful, as evidenced by our large and committed userbase.
The main two results of this project that go beyond the state of the arts are the following:
(1) Developing an approach where all citizens can meaningfully participate in producing scientifically valid data on bird biodiversity, whether they can identify birds or not. This is achieved by AI-based species identification and by storing not only the identifications, but also the raw audio.
(2) Integration of citizen science by systematic monitoring by the implementation of citizen-science based point count routes in collaboration with Finnish cities and National parks.
Further uptake and success will require finishing the workflow for real-time monitoring and hyperautomation.
Overall, we consider this project highly successful. If suitable resources and collaborators emerge, we aim to scale up the approach to European level.
(1) Developing an approach where all citizens can meaningfully participate in producing scientifically valid data on bird biodiversity, whether they can identify birds or not. This is achieved by AI-based species identification and by storing not only the identifications, but also the raw audio.
(2) Integration of citizen science by systematic monitoring by the implementation of citizen-science based point count routes in collaboration with Finnish cities and National parks.
Further uptake and success will require finishing the workflow for real-time monitoring and hyperautomation.
Overall, we consider this project highly successful. If suitable resources and collaborators emerge, we aim to scale up the approach to European level.