Handling duplicate observations

[andrewrodrigues]

Currently, cube downloads make no indication to what extent occurrences within a grid cell are repeat observations/duplicates which would inflate counts. Could related records be removed from downloads? Or, could thee be an additional column added to a download that could provide an indication of the likelihood of their being repeat observations within a specific grid cell? NY understanding is that related records could work when the occurrences are in different datasets, however, when repeat observations are in the same dataset it is more diffciult to identify them. I am thinking of some of the large bird datasets where you might repeat observations of the same individual at the same time by different observers.

[peterdesmet]

Is "related records" information available in the occurrence table?

Note that users currently have (indirect) ways to assess repeat observations, such as UNIQUE(datasetKey). That gives the number of datasets that contribute to an aggregation row. The higher the less likely those are repeat observations.

species	cell	occ count	unique(datasetKey)
A	X	300	2
A	Y	150	11

Cell X contains 300 observations for species A, but only coming from 2 datasets: this can indicate repeat observations.

Also, raw occurrence records are always to be taken with a grain of salt. It is mainly useful for occupancy (is a grid cell occupied or not), where both very low and very high numbers can be disregarded.