Depending on your processing capabilities after Lightroom, you might be able to make use of hierarchical keywords for this.

Lightroom allows keywords to have duplicate names if they're in different parts of the hierarchy.  And they can contain any characters, such as digits.  So you can do e.g.:

Creating all the necessary tags can be a little laborious, but you need only do it on demand (as you need each species or plant count) and it amortises over time.

When you export your images, you just need to ensure you have "Write Keywords as Lightroom Hierarchy" checked (it's off by default).

The resulting files will contain a "Hierarchical Subject" metadata entry ("EXIF") which encodes the list of keywords as their full hierarchy, e.g.:

Family|Genus A|Species 1|1
Family|Genus A|Species 2|5
Family|Genus B|Species 3|2

You can use various tools, e.g. exiftool, to extract that, in a variety of formats.  e.g. exiftool -j -HierarchicalSubject Example.jpg.

An extension of this - if you want to be able to specify where in the image each individual is - is to use "Face Detection" / "People" regions like I suggested earlier.  You can even apply hierarchical keywords to those if you wish, although it is a bit clumsy and error-prone in Lightroom (basically, you drag out a region, click on it to focus on the bounding box rather than the text input field, then right-click on a keyword in the Keyword List (right side panel) and choose "Add this Keywords to Selected Face".  I haven't found any way to just type in the keyword if it's part of a hierarchy.

With "people" defined like that, you'll find an additional entry in the image metadata, "Region Info".  You can extract that like e.g. exiftool -j -struct -RegionInfo Example.jpg:

[{
  "SourceFile": "/Users/me/Pictures/Example.jpg",
  "RegionInfo": {
    "AppliedToDimensions": {
      "H": 203,
      "Unit": "pixel",
      "W": 293
    },
    "RegionList": [{
      "Area": {
        "H": 0.10380,
        "W": 0.21127,
        "X": 0.84447,
        "Y": 0.32851
      },
      "Name": "Species 3",
      "Rotation": 0.00000,
      "Type": "Face"
    },{
      "Area": {
        "H": 0.05365,
        "W": 0.31607,
        "X": 0.54884,
        "Y": 0.45270
      },
      "Name": "Species 2",
      "Rotation": 0.00000,
      "Type": "Face"
    },{
      "Area": {
        "H": 0.06180,
        "W": 0.31239,
        "X": 0.45768,
        "Y": 0.55634
      },
      "Name": "Species 2",
      "Rotation": 0.00000,
      "Type": "Face"
    },{
      "Area": {
        "H": 0.05876,
        "W": 0.20918,
        "X": 0.40603,
        "Y": 0.66991
      },
      "Name": "Species 3",
      "Rotation": 0.00000,
      "Type": "Face"
    },{
      "Area": {
        "H": 0.06814,
        "W": 0.29979,
        "X": 0.44986,
        "Y": 0.78018
      },
      "Name": "Species 3",
      "Rotation": 0.00000,
      "Type": "Face"
    }]
  }
}]

Note that it doesn't directly include the keyword hierarchy in this "Region Info", but it includes these "people" keywords in the same "Hierarchical Subject" as before, so you can match them up based on leaf name (e.g. species) as long as those leaf names are unique.

This does open up some interesting possibilities for further processing, since you can now do things like extract the subsection of each image for each individual of each species, as separate images, which could then e.g. be added to the training data for an ML model.

Interesting, Iʻll give it a shot. Looks like this could save me some time.

Thanks for the explanation @wade!