search

jessluo / PSSdb

Workflow for the Pelagic Size Structure database (PSSdb)
https://PSSdb.net
2 stars 0 forks source link

Flagging and filtering out bad quality IFCB samples #4

Open mdugenne opened 2 years ago

mdugenne commented 2 years ago

Known issues (leak, air bubbles, camera focus, and settings) with the IFCB instrument may alter the detection of plankton and result in inaccurate count and/or size estimates, affecting the observed size distribution and predicted slope of the size spectrum.

Given that (1) IFCB projects uploaded on Ecotaxa have not been flagged, (2) only a small percentage of IFCB projects have been validated (3) some predicted artefacts are actually good quality images of plankton, is there a reliable way to flag and filter out bad quality IFCB samples?

Few ideas: Within the possible artefacts (e.g. bubbles, beads, badfocus), bubbles are likely the best predicted, followed by beads.

MarCorralesU commented 2 years ago

considering that we don't want to get rid of data (rows) for the "cleaned" version of the tsv files, this part of the file reading function https://github.com/jessluo/PSSdb/blob/cd3116e484aba24f4bf99849bfc65a86bd06bba7/scripts/Size_spectra_functions.py#L75-L82

needs to be included after the data is read and standardized but before the data is used for PSS analysis