Provide an approximate precision with observations?

[frousseu]

Hi!

Would it be possible and are there any plans for eBird to provide an estimate of precision with observations or checklist data (and add that to what is passed on to GBIF)? From what I understand, observations from traveling checklists are given what seems to be the centroid of the traveling path. The distance traveled is given along with the checklist information, but this distance is not easily converted to an estimate of precision because of the varying path shapes. What I think could be useful is to have a distance that spans the furthest point of the path from the centroid used for the checklist. This would provide a slightly more useful estimate of precision that can differentiate between a loop and a linear path of the same length for example. When hotspots are used instead of traveling, I guess it would not be possible to give any estimate of precision (unless hotspots are associated with defined polygons?). My main interest here is to determine an appropriate level of resolution for predictors when modeling species distributions. I'm well aware that relatively low resolutions ( > 2.5km) have to be used when modeling distributions with eBird data, but I think that having slightly more information on the precision of each observation or checklist could be helpful in determining an appropriate resolution for a given application. For example, in certain cases, it could be possible to model distributions at a finer scale if the local density of "precise" checklists is higher. Overall, I think this information would be useful for end users trying to get a sense of an appropriate scale for modeling the data. Let me know if something is not clear or if somewhere else would be more appropriate for this discussion.

Cheers, François

[frousseu]

Just to be clear, if the red dot is the centroid of the traveling path (blue), the precision could be the radius of the black circle.

[mstrimas]

Hi François, this is a great and idea and something we've been thinking about recently. We're unable to share the tracks themselves for privacy reasons, but I think it may be possible to provide some information about the track. It may take some time for us to figure this out, but I'll post here if we are able to provide a metric of precision.

Also, you may be interested in this new version of the Best Practices I've been working on: https://ebird.github.io/ebird-best-practices/index.html

[frousseu]

This may be a bit more involved, but I wonder if some functionalities could be added to ebird so that birders have the ability to make their track public or obscured, with a default set to obscured for example. A bit like with iNaturalist observations that can be opened or obscured. This may not be very useful though since in most cases a bird recorded in the app at a certain time along the track may actually have been seen at a previous time/location. For the precision, I guess it might also be important to consider the GPS precision along with the track precision. In any case, I'm looking forward to some precision data and I will update my links for the best practises.