robgur
commented
1 month ago John we need to talk it over on chat because there are issues here beyond what you have listed. Here is a clear listing of what each field means and whether we NEED it (REQUIRED) or not. Please, we need the fields listed as required and you should ask questions here re: what is confusing. One issue is name and rank, which is a little different from other data we've ingested.
(likely can be deleted) -- file_name - the photo file name; a concatenation of the photo_id and image type (.jpg or .jpeg) (maybe we can delete this field since it is linked directly to photo_id)
(REQUIRED) photo_id: The photo_id assigned by iNaturalist. There can be multiple photo_ids assigned to inaturalist observation ids i.e. observation_uuid.
(I think keep but see note) observation_uuid: iNaturalist's observation id. We can maybe delete this since it doesn't itself resolve to anything.
(REQUIRED) inat_URL: This is basically "https://www.inaturalist.org/observations/" + "observation_uuid". This will resolve to a record.
(REQUIRED) latitude: same definition as DwC:latitude (REQUIRED) longitude: same definition as Dwc:longitude
positional_accuracy: This is mapped to DwC:coordinateUncertaintyInMeters so we may as well use that mapping and definition too.
(likely delete) taxon_id: an iNaturalist internal identifier for a taxon. We can likely drop this field.
observed_on: Mapped to dwc:eventDate and may as well use that mapping here.
(REQUIRED) name: The taxonomic name of the identified occurrence to the lowest level so far obtained.
(REQIUIRED) rank: The taxonomic rank of the identification, often but not always species.
(likely delete) family_id: an iNaturalist internal identifier for family names. Can likely delete.
(REQUIRED) family: the name of occurrence at the taxon rank of family according to iNaturalist
(REQUIRED) count_family: The number of records that were annotated at the taxon rank of family associated with the identified occurrence
(REQUIRED) trait: the name of the trait being classified by ML model. For the first tranche of data, this is either "flower" or "fruit".
(REQUIRED) prediction_prob: "A score from 0-1 which represent a scaled probability of presence or absence of a trait"
(REQUIRED) prediction_class: a categorical description of the annotation outcome, in this case either "detected" or "not detected"
(REQUIRED) equivocal: A categorical description of whether the annotation was equivocal or unequivocal. For now all the data ingested into Phenobase is "unequivocal".
(REQUIRED) proportion_equivocal_family: The proportion of the number of equivocal annotations divided by all annotations for iNaturalist observations in the taxonomic family. A metric of how successful the annotation process for observations in that family.
(REQUIRED accuracy_excluding_equivocal_family: A metric of accuracy for the annotations of only unequivocal records based on expert, gold standard observations.
(REQUIREF) accuracy_family: A metric of accuracy for the annotations of all records (equivocal and unequivocal) based on expert, gold standard observations.
(REQUIRED) genus: the name of occurrence at the taxon rank of genus according to iNaturalist
jdeck88
rdinnager
The following are questions about particular column names provided in Daijiang's sample data:
@daijiang @rdinnager