Would it smell as sweet by any other name? Credit: R. Evans/Photolibrary

Brian Enquist and his collaborators were delighted with their freshly compiled data set of 22.5 million records on the distribution and traits of plants in the Americas. But their delight turned to horror when they realized that the data set contained 611,728 names: nearly twice as many as there are thought to be plant species on Earth.

Completed in December 2010, the records were intended to help Enquist and his colleagues to discern trends in how forest trees in a wide variety of environments respond to climate change. But the data were clearly full of bogus names, making it impossible to count the species in a particular area, or their relative abundance. "I started to question our ability even to compare something as basic as species diversity at two sites," says Enquist, a plant ecologist at the University of Arizona in Tucson.

This month, Enquist's team will unveil a solution that could help botanists and ecologists worldwide. The Taxonomic Names Resolution Service (TNRS) aims to find and fix the incorrect plant names that plague scientists' records.

"It looks really good," says Gabriela Lopez-Gonzalez, a plant ecologist at the University of Leeds, UK, who curates a database of forest plots. Fixing species lists by hand is arduous, she says. "This should save us a lot of time".

She and others agree that the problem is widespread in botanical databases. "Digitization has made the problem worse," says TNRS co-leader, botanist Brad Boyle, also at the University of Arizona. Boyle explains that as more data are added to digital records, the chance of introducing errors also increases. Even in herbarium specimens, which ought to be the gold standard for plant identification, about 15% of the names are misspelt, he says.

Many of the errors seem to arise because biologists are not as careful as they should be when entering data into digital records. The TNRS team estimates that about one-third of the names entered into online repositories — such as GenBank, the US National Institutes of Health collection of DNA-sequence data, or the Ecological Society of America's VegBank database of plant-plot data — are incorrect.

The other problem is that names change. Old names can be abolished when experts reclassify plants as ideas about evolutionary relationships change, or when they realize the species already had a name — an occurrence almost as old as taxonomy itself. The result is that the same plant can have many names, and not everyone knows which one to use. Such synonyms are a particular problem in the study of medicinal plants, says Alan Paton, a plant taxonomist and bioinformatician at Kew Gardens in London.

The TNRS was built with financial and technical support from iPlant, a project run by the US National Science Foundation to fund cyberinfrastructure for plant science. It corrects names by comparing lists that users feed into it with the 1.2 million names in the Missouri Botanical Garden's Tropicos database, one of the most authoritative botanical databases. If the TNRS cannot find a name in Tropicos, it uses a fuzzy-matching algorithm, similar to a word-processor's spellchecker, to find and correct misspellings. It also hunts through Tropicos's lists of alternative names and supplies the one that is most up to date. When Enquist ran the 611,728 names through the system, just 202,252 came back, showing that two-thirds of them were invalid.

Because Tropicos is less comprehensive for plants outside the Americas, the team hopes to link the TNRS with The Plant List (www.theplantlist.org), a collaborative compilation of databases from Kew and other sources. Launched online in December 2010, it aims to become a global record of plants. The scientists are also working on a tool to correct geographical data — one that knows, for example, that Brazil, Brasil and Brésil are the same place, and can recognize when someone has muddled up longitude and latitude.