The goal is to develop and prove a method to map seabird nests quickly, with low disturbance to the birds, and with higher accuracy and repeatability.
In addition, a secondary goal is to count number of chicks in the colony with similar methods, making it possible to gather breeding success data currently not available.
Lastly, we aim to use the same techniques to map and count seabirds and ducks outside the breeding season.
Drones are flown in from outside the colonies, usually mapping the areas by flying over in a grid pattern. Images are then orthorectified and analysed. Counting of birds in the images are done by a machine learning model trained for the purpose, and the data is stored in a database, all happening on Sigma2.
The work done through SeaBee has revolutionized the way we count seabird colonies in Norway. By deploying state-of-the-art drone technology, we are now able to conduct large-scale, non-intrusive bird counts with a higher degree of accuracy and repeatability then before. With the AI-based pipeline we are also able to map and document different species and their activity quicker then ever.
We have found that in most scenarios, our drone-based approach drastically reduces the human disturbance. The autonomous flight modes of the drones allowed us to monitor seabird colonies from a distance, which minimizes our impact on the seabird colonies.
So far, we have gathered drone images of many colonies and species. The data are currently being processed on Sigma2 through a pipeline that orthorectifies the images in OpenDroneMap before publishing it to Geonode.
See current data resources for this application: geonode.seabee.sigma2.no
A database designed to store both annotations of individual birds and automatic detections has been developed and is about to be deployed into this pipeline. Data are then intended to be aggregated and flow out into different databases automatically, reducing time used to maintain multiple databases.
We have deployed and tested drones for multiple tasks that previously have been almost impossible to do accurately. The most important one being breeding success (counting how many chicks the colony produces). Using drones have already become the standard for how colony size and breeding success is measured for black-headed gulls in Norway.
We have recently tested heat-seeking infrared sensors for detecting chicks in the colonies with a higher degree of accuracy. So far, the tests show very promising results.
We will continue to develop the pipeline to get the data from images into production at the national seabirds database ran by NINA. The database is the basis for data visualised on Seapop (nina.no). Data from the 2023 field campaign must be analysed, and we will need to do more annotation and to train a new model.
There will also be some work done to make sure that data captured in open waters can be analysed in somewhat the same way, and that we are ready to scale up to even bigger or more missions than currently.
SeaBee is already a central part of the current offshore wind power race, as the 2023 field campaign was only possible due to SeaBee. In the future there is also necessary to have as precise information as possible on both breeding populations and breeding success, something SeaBee will easily provide for species that are not monitored today, as common gull and terns.