TL;DR: many observations appear to be improperly located in a one-block area of San Francisco due to a combination of aggressive auto-completion and a bogus accuracy value. Some of these may be intended to be placed in the overall San Francisco area, but many appear to be completely misplaced.
I’ve noticed quite a few observations that appear to be misplaced at the corner of Market & Van Ness in San Francisco, CA, USA. E.g.
This observer has 121 observations in Europe, and then this odd one out:
https://www.inaturalist.org/observations/136802781
A map search reveals ~1000 observations arranged in a pair of N-S lines across the intersection:
https://www.inaturalist.org/observations?nelat=37.77531873399651&nelng=-122.41863948964091&place_id=any&subview=map&swlat=37.77473571276069&swlng=-122.42008520029994
I notice that if I upload an observation in the web interface, then type “S” in the location field and click the second auto-completion for “San Francisco, CA, USA”, the pin is placed exactly at this street corner, although the accuracy ring for my upload encompasses the entire city while the misplaced obesrvations have a 90m accuracy.
That encouraged me to check San Diego, which is the top auto-completion for “S” in the web interface. There are ~2000 observations at the corner of Broadway and Fourth:
Interestingly, however, the San Diego observations I looked at all had very large accuracy rings that encompassed the whole city. So perhaps this is a difference between the web import auto-completing to San Diego with a large accuracy ring and some other import method auto-completing to San Francisco with a small accuracy ring.
Does anyone have any more insight? Should auto-completion require more than one typed letter to make this sort of thing less likely? Does an app need to be fixed to use the correct-size accuracy ring for an auto-completed region?