A method to implement continuous characters in digital identification keys that estimates the probability of an...

Area of Science:

• Botany
• Taxonomy
• Computational Biology

Background:

• Species identification is crucial across scientific disciplines.
• Digital tools enhance identification, but continuous character data remains underutilized.
• Existing methods often rely on discrete characters, limiting identification accuracy.

Purpose of the Study:

• To develop a classifier leveraging continuous morphological characters for species identification.
• To provide a quantitative measure of confidence (posterior probability) in taxonomic assignments.
• To estimate the strength of evidence supporting candidate identifications.

Main Methods:

• A species model was defined using continuous morphological characters.
• A naive Bayesian classifier algorithm was developed for identification.
• A method for estimating the strength of evidence for candidate species was implemented.

Main Results:

• The method was successfully applied to differentiate native vs. invasive *Myriophyllum* in North America.
• It was also used to identify vegetative *Rhipidocladum* bamboos in Mexico.
• The classifier provided probabilities and evidence strength, enhancing assignments for challenging taxa.

Conclusions:

• Naive Bayesian classifiers effectively utilize continuous morphological data for identification.
• This approach advances the use of digital technology in plant taxonomy.
• It offers improved interactive taxonomic identification capabilities.