can anyone please explain the relation between PA and PO dataset
", "messages": [ { "id": "2725466", "postDate": "03/31/2024 15:03:02", "content": "can anyone please explain the relation between PA and PO dataset
", "rawMarkdown": "can anyone please explain the relation between PA and PO dataset", "votes": null }, { "id": "2725491", "postDate": "03/31/2024 15:27:43", "content": "Hi @shiwani14 ,
\nBellow I include one more info copied from the dataset section:
\n\n\nPresence-Absence (PA) surveys: including around 90 thousand surveys with roughly 10,000 species of the European flora. The presence-absence data (PA) is provided to compensate for the problem of false-absences of PO data and calibrate models to avoid associated biases.
\n
\n Presence-Only (PO) occurrences: combines around five million observations from numerous datasets gathered from the Global Biodiversity Information Facility. This data constitutes the larger piece of the training data and covers all countries of our study area, but it has been sampled opportunistically (without standardized sampling protocol), leading to various sampling biases. The local absence of a species among PO data doesn't mean it is truly absent. An observer might not have reported it because it was difficult to \"see\" it at this time of the year, to identify it as not a monitoring target, or just unattractive.
Hope this helps. Let me know if you need anything else.
\nBest,
\nLukas
Hello, @picekl. Thank you for your kind explanation. I also have a question about the relationship between PA and PO datasets. It seems that there are some species which do not exist in PA dataset but exist in PO dataset. Could the true labels(speciesIds) of test dataset include speciesId which do not exist in PA dataset but only exist in PO dataset?
", "rawMarkdown": "Hello, @picekl. Thank you for your kind explanation. I also have a question about the relationship between PA and PO datasets. It seems that there are some species which do not exist in PA dataset but exist in PO dataset. Could the true labels(speciesIds) of test dataset include speciesId which do not exist in PA dataset but only exist in PO dataset?", "votes": null }, { "id": "2795528", "postDate": "05/05/2024 20:04:51", "content": "Hi @swinak,
\nYes, some PA test species occur only in the PO data. Therefore, the use of PO data is vital for better performance, more precisely, for the Recall.
\nAnyway, the use of the PO data is not necessary, you can \"easily\" achieve around 0.33 just with the PA data :)
\nBest,
\nLukas
Thank you very much for your informative answer. :)
", "rawMarkdown": "Thank you very much for your informative answer. :)", "votes": null } ], "comments": [ { "id": 2725491, "author_name": "picekl", "author_url": "", "post_date": "03/31/2024 15:27:43", "content": "Hi @shiwani14 ,
\nBellow I include one more info copied from the dataset section:
\n\n\nPresence-Absence (PA) surveys: including around 90 thousand surveys with roughly 10,000 species of the European flora. The presence-absence data (PA) is provided to compensate for the problem of false-absences of PO data and calibrate models to avoid associated biases.
\n
\n Presence-Only (PO) occurrences: combines around five million observations from numerous datasets gathered from the Global Biodiversity Information Facility. This data constitutes the larger piece of the training data and covers all countries of our study area, but it has been sampled opportunistically (without standardized sampling protocol), leading to various sampling biases. The local absence of a species among PO data doesn't mean it is truly absent. An observer might not have reported it because it was difficult to \"see\" it at this time of the year, to identify it as not a monitoring target, or just unattractive.
Hope this helps. Let me know if you need anything else.
\nBest,
\nLukas
Hello, @picekl. Thank you for your kind explanation. I also have a question about the relationship between PA and PO datasets. It seems that there are some species which do not exist in PA dataset but exist in PO dataset. Could the true labels(speciesIds) of test dataset include speciesId which do not exist in PA dataset but only exist in PO dataset?
", "votes": null, "replies": [ { "id": 2795528, "author_name": "picekl", "author_url": "", "post_date": "05/05/2024 20:04:51", "content": "Hi @swinak,
\nYes, some PA test species occur only in the PO data. Therefore, the use of PO data is vital for better performance, more precisely, for the Recall.
\nAnyway, the use of the PO data is not necessary, you can \"easily\" achieve around 0.33 just with the PA data :)
\nBest,
\nLukas
Thank you very much for your informative answer. :)
", "votes": null, "replies": [] } ] } ] } ] } ], "raw_markdown_by_id": { "2725466": "can anyone please explain the relation between PA and PO dataset", "2725491": "Hi @shiwani14 , \n\n- **The PA data** originates from the scientific \"surveys\" where experts carefully explore what species are present at a selected squared region (the size of the region is provided in the metadata in a column named \"areaInM2\"). \n- **The PO data** originate from citizen science platforms and mobile applications for automatic image-based species identification. Therefore just a single species is reported even though there are (most likely) multiple different species at that location.\n\nBellow I include one more info copied from the dataset section:\n\n>**Presence-Absence (PA) surveys**: including around 90 thousand surveys with roughly 10,000 species of the European flora. The presence-absence data (PA) is provided to compensate for the problem of false-absences of PO data and calibrate models to avoid associated biases.\n**Presence-Only (PO) occurrences**: combines around five million observations from numerous datasets gathered from the Global Biodiversity Information Facility. This data constitutes the larger piece of the training data and covers all countries of our study area, but it has been sampled opportunistically (without standardized sampling protocol), leading to various sampling biases. The local absence of a species among PO data doesn't mean it is truly absent. An observer might not have reported it because it was difficult to \"see\" it at this time of the year, to identify it as not a monitoring target, or just unattractive.\n\nHope this helps. Let me know if you need anything else.\n\nBest,\nLukas", "2795204": "Hello, @picekl. Thank you for your kind explanation. I also have a question about the relationship between PA and PO datasets. It seems that there are some species which do not exist in PA dataset but exist in PO dataset. Could the true labels(speciesIds) of test dataset include speciesId which do not exist in PA dataset but only exist in PO dataset?", "2795528": "Hi @swinak,\n\nYes, some PA test species occur only in the PO data. Therefore, the use of PO data is vital for better performance, more precisely, for the Recall.\n\nAnyway, the use of the PO data is not necessary, you can \"easily\" achieve around 0.33 just with the PA data :)\n\nBest,\nLukas", "2796620": "Thank you very much for your informative answer. :)" }, "source": "meta" }