Welcome to TaxonBodyMassML, where our goal is to provide accurate body mass estimations for any species you desire!
Below are some Frequently Asked Questions for an idea of how to use our site; if these don't quite answer your questions, there is a user path designed to get you on the right path. Finally, if you are still struggling with an unanswered query, please type it out in the comment box at the bottom of the page, and a team member will reply with an answer!
Q: Where is the input bar?
A: When you first open the application, there is an introduction modal welcoming you to the website; if you click 'Get Started!' at the bottom of this modal, it will reveal the input bar for you to type into. This input bar should have placeholder text that reads: "Input species name(s)..."
Q: What sort of input do I supply?
A: You will input a scientific species name. The name is resolved against the GBIF Backbone Taxonomy (with an NCBI fallback) to obtain its kingdom-to-genus classification, which the model uses to predict body mass. Names must be spelled correctly: the web interface does not correct misspellings. If the name cannot be resolved, no prediction is returned.
Q: Does the site handle typos or misspelled species names?
A: No. The web interface looks up the name exactly as typed, so typos will usually result in no prediction. Fuzzy name correction (via the GBIF species-match service) is available in the R and Python packages by setting fuzzy_match_name = TRUE in predict_mass(); the corrected name is then reported alongside the result.
Q: What are some examples of inputs I could try?
A: Here are a list of scientific species names that have direct matches in our dataset:
Q: Why is my first result slow to appear?
A: The very first prediction request after the service starts up may take a little while to complete as the system initializes. Subsequent requests will be significantly faster.
Q: What do the lower and upper bounds mean?
A: For model-predicted masses they are a 90% conformal prediction interval: across taxa queried at the same level of taxonomic resolution (e.g. genus known, species not), about 90% of true body masses fall inside the interval. The interval is calibrated separately for each level of taxonomic resolution, so it widens when only the family or order of a name can be resolved. For species whose measured mass is held in the database, the measured value is returned and no interval is shown. The R and Python packages let you choose other coverage levels.
1. I am a student writing a paper for my zoology class about a deep-sea organism; there are no known body mass records for the organism. I stumble across the TaxonBodyMassML website while searching the web for body mass information. I click the 'x' in the upper right-hand corner of the box welcoming me to the website to close it. I click on the input bar that reads "Input species name(s)..." and type in the scientific name of the deep-sea species. The "single" option appears to already be selected, so I click the button that says "Go!". I watch as the output box appears, at first saying "Checking species name...", then reading "Species not found". I realize I made a typo when typing in the species name, so I click on the input bar again, which is still populated with the incorrectly-typed scientific name, fix my mistake, and press the "Go Again!" button. This time, the output box contains a number which is labeled "mass". I am curious about how the website came up with this number, so I click the "Learn More?" button under the mass output and learn about the algorithm and predictive process. When I include this information in my zoology paper, I make sure to say that this is a projected mass, and cite the website I used.