The Xeno-Canto Dawn Chorus (XCDC) Dataset

We construct a dataset of species-rich audio by sampling dawn chorus recordings from Xeno-Canto, a global archive of natural sound recordings. The dawn chorus is a period of intense, multi-species vocal activity that occurs near sunrise, particularly during the breeding season. To isolate these species-rich recordings, we select audio that (1) was recorded during the spring dawn chorus, (2) is at least 3 minutes in duration, and (3) contains annotations for at least 10 distinct species. This filtering yields 576 recordings with associated geographic coordinates and species labels. We refer to this dataset as the Xeno-Canto Dawn Chorus (XCDC).