ramonawalls
commented
2 months ago We have had multiple discussions on this topic. At the PPO in person meeting today, we decided the following:
For images, if a structure is absent, we will create a characteristic of the image that is "structure absent from image". We will include that information in the database. It will be connected to the observation, which connects to the whole plant, but you will not be able to make a concrete inference about whether or not the structure is present or absent in the plant. However, by keeping the information in our DB, researchers can choose for themselves if they want to interpret that information as absent from the plant.
We also decided today (although it will need to be tested) that we should apply the same principle to herbarium specimens.
When an image or herbarium specimens lacks flowers/leaves/fruits and contains part of a plant (often the case), we cannot definitively say that flowers/leaves/fruits were absent from the whole plant. However, we know from practice of collecting herbarium specimens and photographing images for iNaturalist, that if flowers/leaves/fruits are present, they will be collected or photographed. Therefore, it is likely that absence from the image or specimen does mean absence from the whole plant (with some error which has never been quantified).
We have been discussing this issue for a long time, but at our in person meeting today, @robgur, @daijiang, and I decided that we should score those cases as absent. Since our data also include the source and type of observing process, users should know that those records come with some error. We will need to document it well on our portal.
At some point, we should do a study to quantify how often images in iNat or herbaria are scored absent when in fact the whole plant had present (e.g., based on label notes or other observations of the same plant).