{ "id": 323286, "title": "Species co-occurrence matrices? (Update: Public notebook available)", "url": "/competitions/birdclef-2022/discussion/323286", "author_name": "", "post_date": "2022-05-05T18:35:49.322789100Z", "votes": 11, "comment_count": 3, "views": 0, "content": "

Has anyone tried to categorize which species are likely/unlikely to occur in the same soundscape? For instance, i'iwi and apapane are upland forest birds, while dunlin and ruddy turnstones are shorebirds. Some birds, like palila, are so endangered that any recordings are going to come from a very limited area with (likely) only a limited amount of co-occurring species.

", "messages": [ { "id": "1778845", "postDate": "05/05/2022 18:35:49", "content": "

Has anyone tried to categorize which species are likely/unlikely to occur in the same soundscape? For instance, i'iwi and apapane are upland forest birds, while dunlin and ruddy turnstones are shorebirds. Some birds, like palila, are so endangered that any recordings are going to come from a very limited area with (likely) only a limited amount of co-occurring species.

", "rawMarkdown": "Has anyone tried to categorize which species are likely/unlikely to occur in the same soundscape? For instance, i'iwi and apapane are upland forest birds, while dunlin and ruddy turnstones are shorebirds. Some birds, like palila, are so endangered that any recordings are going to come from a very limited area with (likely) only a limited amount of co-occurring species.", "votes": null }, { "id": "1778928", "postDate": "05/05/2022 22:43:24", "content": "

I did a co-occurrence analysis on secondary label in EDA[1] (see “Co-occurrence of second labels” section), and there was a tendency for \"birds with large number of inhabitants seem to appear in the background more often\". However, due to the small number of samples with second labels, no other characteristics could be found.

\n

I think there are quite a few recordings that do not have secondary labels attached, so doing the same thing with the trained model's predictions may yield different results.

\n", "rawMarkdown": "I did a co-occurrence analysis on secondary label in EDA[1] (see “Co-occurrence of second labels” section), and there was a tendency for \"birds with large number of inhabitants seem to appear in the background more often\". However, due to the small number of samples with second labels, no other characteristics could be found.\n\nI think there are quite a few recordings that do not have secondary labels attached, so doing the same thing with the trained model's predictions may yield different results.\n\n- [1] https://www.kaggle.com/code/tatamikenn/birdclef22-eda-on-the-scored-species#Sedondary-labels", "votes": null }, { "id": "1779160", "postDate": "05/06/2022 06:50:47", "content": "

This is an interesting question. I have not tried to categorize which species are likely/unlikely to occur in the same soundscape, but it would be a interesting exercise. It would be especially interesting to try to predict which species are likely to co-occur in a given soundscape based on their ecological niche.

", "rawMarkdown": "This is an interesting question. I have not tried to categorize which species are likely/unlikely to occur in the same soundscape, but it would be a interesting exercise. It would be especially interesting to try to predict which species are likely to co-occur in a given soundscape based on their ecological niche.", "votes": null }, { "id": "1780762", "postDate": "05/07/2022 20:21:31", "content": "

I created a co-occurrence matrix using data from eBird:

\n

https://www.kaggle.com/code/lobrien/species-co-occurrence-from-ebird-checklists

\n

Species in score_birds.json in Y, species co-occurring with them in X:

\n

\"score-vs-found.png\"

", "rawMarkdown": "I created a co-occurrence matrix using data from eBird: \n\nhttps://www.kaggle.com/code/lobrien/species-co-occurrence-from-ebird-checklists\n\nSpecies in `score_birds.json` in Y, species co-occurring with them in X:\n\n[![score-vs-found.png](https://i.postimg.cc/QMXqT9jw/score-vs-found.png)](https://postimg.cc/nXP78LjG)", "votes": null } ], "comments": [ { "id": 1778928, "author_name": "tatamikenn", "author_url": "", "post_date": "05/05/2022 22:43:24", "content": "

I did a co-occurrence analysis on secondary label in EDA[1] (see “Co-occurrence of second labels” section), and there was a tendency for \"birds with large number of inhabitants seem to appear in the background more often\". However, due to the small number of samples with second labels, no other characteristics could be found.

\n

I think there are quite a few recordings that do not have secondary labels attached, so doing the same thing with the trained model's predictions may yield different results.

\n", "votes": null, "replies": [] }, { "id": 1779160, "author_name": "", "author_url": "", "post_date": "05/06/2022 06:50:47", "content": "

This is an interesting question. I have not tried to categorize which species are likely/unlikely to occur in the same soundscape, but it would be a interesting exercise. It would be especially interesting to try to predict which species are likely to co-occur in a given soundscape based on their ecological niche.

", "votes": null, "replies": [] }, { "id": 1780762, "author_name": "lobrien", "author_url": "", "post_date": "05/07/2022 20:21:31", "content": "

I created a co-occurrence matrix using data from eBird:

\n

https://www.kaggle.com/code/lobrien/species-co-occurrence-from-ebird-checklists

\n

Species in score_birds.json in Y, species co-occurring with them in X:

\n

\"score-vs-found.png\"

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1778845": "Has anyone tried to categorize which species are likely/unlikely to occur in the same soundscape? For instance, i'iwi and apapane are upland forest birds, while dunlin and ruddy turnstones are shorebirds. Some birds, like palila, are so endangered that any recordings are going to come from a very limited area with (likely) only a limited amount of co-occurring species.", "1778928": "I did a co-occurrence analysis on secondary label in EDA[1] (see “Co-occurrence of second labels” section), and there was a tendency for \"birds with large number of inhabitants seem to appear in the background more often\". However, due to the small number of samples with second labels, no other characteristics could be found.\n\nI think there are quite a few recordings that do not have secondary labels attached, so doing the same thing with the trained model's predictions may yield different results.\n\n- [1] https://www.kaggle.com/code/tatamikenn/birdclef22-eda-on-the-scored-species#Sedondary-labels", "1779160": "This is an interesting question. I have not tried to categorize which species are likely/unlikely to occur in the same soundscape, but it would be a interesting exercise. It would be especially interesting to try to predict which species are likely to co-occur in a given soundscape based on their ecological niche.", "1780762": "I created a co-occurrence matrix using data from eBird: \n\nhttps://www.kaggle.com/code/lobrien/species-co-occurrence-from-ebird-checklists\n\nSpecies in `score_birds.json` in Y, species co-occurring with them in X:\n\n[![score-vs-found.png](https://i.postimg.cc/QMXqT9jw/score-vs-found.png)](https://postimg.cc/nXP78LjG)" }, "source": "meta" }