AI-assisted practical guide. Examples are hypothetical; proposed workflows are editorial suggestions.
A photograph can outlast your memory of where it came from. Build filenames and notes around stable observations rather than a species name that may later change. An uncertain identification should not become permanent simply because it was typed into a filename.
Separate identity from interpretation
Assign each observation an ID. Include a date, site code, and sequence number in the filename or accompanying index. Keep the original image and record any crop or annotation as a derivative.
Store tentative identification, reviewer, and revision date in separate fields. That lets a later correction improve the interpretation without breaking the connection to the photograph.
Follow a hypothetical correction
Imagine observation G014 was initially labeled “possible bee group A.” A reviewer later narrows the identification but cannot confirm a species. Update the identification field and preserve the earlier label in a change note. Do not rename the original file into a certainty the reviewer did not provide.
Record the plant only to the level you know. An unidentified flower can have its own temporary plant ID and photographs.
Make the archive usable
Keep a short explanation of the naming pattern beside the files. Test whether another person can find an observation from its ID and distinguish originals from edited copies. Back up the collection and verify that a sample can be opened. A tidy folder is useful only if its context survives with it.