Back

Modeling American Kestrel Decline Using Spatiotemporal Subsampling to Improve eBird Data Reliability

Sharma, A.; Bednarz, J. C.; Kwon, J.

2025-01-02 ecology
10.1101/2024.10.31.621376 bioRxiv
Show abstract

American Kestrel (Falco sparverius) numbers have plummeted over 40% since 1980, despite many other raptor species, including the closely related Merlin (Falco columbarius) rebounding in the same period, post the ban on the use of DDT. One plausible reason for this decline is the increase in Coopers Hawks (Astur cooperii), a predatory species. To model the impact of Coopers Hawk on American Kestrels, one can use citizen science data such as eBird. However, eBird contains a heavy sampling bias towards recent data in large cities as it has gained popularity over time. A novel framework testing different spatiotemporal subsampling resolutions was employed in seeing how to make the data more reliable and to enjoy a finer time resolution compared to the Breeding Bird Survey (BBS) standard. The ecological relationship between two species is then modeled using a time-series framework that accounts for both current and past interactions, with the model order determined based on criteria such as BIC. The creation and analysis of synthetic data are also tested to verify that our framework reflects the ground truth of the bird abundance trend. Immediate forecast results based on the eBird data suggests a decline of American Kestrels at the current pace.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.