The bird calls in the Xeno-Canto dataset are labeled with a longitude and latitude position. It would be useful to create a density map of the recordings to show which bird species are located. The point data can be smoothed into a heat map. You might also explore fetching external datasets from places like eBird, birdNET, or Xeno-Canto. The distribution maps can act as important priors for classification tasks.
Key Words: geospatial analysis, kernel density estimation, api or web scraping