Science1 publisher3 min readPublished
Who Eats Whom builds a searchable food web from 17,000 tagged iNaturalist sightings
The site holds about 17,000 tagged feeding observations covering some 5,700 species, contributed by roughly 2,000 observers in more than 100 countries. The team is still building tools to validate submissions.
The Scientist · Science desk
What happened
- An interdisciplinary team has put an open-source site called Who Eats Whom online, drawing feeding records from the worldwide citizen science platform iNaturalist.
- Contributors log an interaction by tagging both ends of it in a photo or video, and the tagged predator can be a herbivore while the prey can be a plant.
- The database so far holds roughly 17,000 observations of trophic interactions involving about 5,700 species, gathered from around 2,000 observers in more than 100 countries.
- The paper behind the site, titled "Who Eats Whom? A global food web derived from citizen science," publishes Sept. 22 in the open-access journal PLOS Biology.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- capability A researcher who wants every recorded predator of one beetle can query a single searchable index instead of working through that taxon's papers one at a time.
- constraint At roughly three records per species, the set can support a claim that a feeding link exists. It cannot say how much of a diet that link carries, so anyone modeling interaction strength still needs their own fieldwork.
- decision With the validation tooling still in development, a user who treats a record as evidence is accepting responsibility for confirming both identifications.
- contradiction Allf frames the problem as knowledge scattered across the literature, and the site answers it with fresh sightings while the published records stay where they are, so it grows a second archive beside the first.
Divide about 17,000 tagged interactions by about 5,700 species and the average is roughly three records per species [1]. Three records establish presence: somewhere, at some point, a person photographed one organism eating another and labeled both ends [2]. Frequency of the link, and the share of a predator's diet it accounts for, are not in the record. A model that needs interaction strength, or needs to know that a plausible link was looked for and not found, will not get those numbers from opportunistic photographs [2].
The contributor base is thin in the same way. Around 2,000 observers in more than 100 countries [4] works out to roughly eight and a half records each [2]. What those observers chose to photograph is its own filter, favoring interactions that happen in the open and hold still long enough to frame [2]. Aditi Mallavarapu, an assistant professor of computer science at NC State and a co-author [8], said the submissions have ranged past the obvious cases: "One of the interesting observations at this point is the extent to which the public is interested not only in charismatic predators like lions or wolves, but in smaller trophic relationships such as the insects that eat persimmon leaves" [9].
Bradley Allf, the corresponding author, who began the project as a graduate student at NC State and is now a postdoctoral researcher at Colorado State University [7], described the underlying problem: "Our understanding of what different organisms eat is incomplete and is often scattered across the academic literature" [6]. Scatter is a genuine cost to anyone assembling a food web, and a searchable index of interactions does address it [16]. The published records themselves stay where they are. The site collects new sightings from iNaturalist users who tag predator and prey themselves [2]. Allf called the result a "parallel database that can complement the academic literature by compiling as much of this information in one place as possible" [12].
Validation is not part of the pipeline yet. The team is developing tools to help validate submissions to the database on iNaturalist [13]. In the meantime, a user who wants to cite a record has to check the identifications at both ends independently.
Mallavarapu said more is coming on the visual side: "There is a simple version online now, but we're in the process of adding more features that could support exploration of feeding patterns and answer questions like, 'What do I plant to attract hummingbirds in North Carolina?'" [14]. She also said the team has built educational games on the database, "aimed at helping children understand the food web and the dependencies it captures," and is interested in expanding them if it can secure funding [15].
What to watch
- Whether the Sept. 22 paper reports how many distinct predator-prey pairs the 17,000 records resolve to, and how duplicates were handled.
- Whether the validation tooling ships, and what share of submitted interactions survives it.
- Whether a novel trophic relationship found on the site reaches the literature, as Allf said the team intends.