When sorting by “Date observed”, it would be nice to include the full time precision in the sort so that the sorted observation order matches the order of photos in my camera roll. Otherwise, observations made within the same minute are sorted randomly and inconsistently.
My workflow is generally to upload all my photos to iNaturalist while identifying them as best as I can. A few days later, after other people have added their identifications, I’ll go back to my photos and tag them with the consensus IDs. Thus, it’s easiest for me if the order of observations in iNaturalist matches the order of photos in my computer.
I usually upload my photos in the same order I took them, but sometimes I pick a few out to upload first because I’m curious about the suggested ID, and sometimes a few get missed until a second upload pass later. So sorting by ‘observation date’ generally works better for me than sorting by ‘date added’, except when two observations are within the same minute and so don’t get ordered correctly.
I’ve constructed this artificial search to demonstrate the issue. The search results include only a few observations, sorted by observation date in ascending order (oldest on top):
At 11:20 AM PDT, I observed Castilleja exserta, then Erodium cicutarium, then Eschscholzia california, but that is not the listed order.
At 11:24 AM PDT, I observed Tragopogon porrifolius, then Rumex crispus. The listed order is correct, but if I click at the top of the “Date observed” column to switch to descending order, these two observations don’t reverse order and are therefore now in the wrong order.
Poking through the exported data, I notice that the time_observed_at field includes seconds, but the value is always rounded off to “:00”. I checked the EXIF data of the photos I uploaded, and it has precision to fractions of a second, so I’m guessing that iNaturalist rounds off the value during upload.
Ideal feature update: Record the observation date with precision at least to seconds and use this precision when sorting by observation date. (The time can still be displayed as hours:minutes as usual.)
If it’s too painful to add more precision to the observation date, a partial solution could break ties in the observation date by using the date added field (or observation ID). This would be acceptable to me since I usually submit observations in sequential order.
This is not a high priority request. It’d be a nice improvement for me and perhaps other people as well, but I can continue to work around the issue if nothing changes.