When French scientists Laurence Gaume and Marion Desquilbet first heard the news about a new international insect decline database, they felt something was off. It suggested that some insect species were actually on the rise – a claim that contradicted years of research.
So, they called each other up.
“We were concerned. We felt their conclusions were over-optimistic,” says Desquilbet, an environmental economist at Toulouse School of Economics. “It went against prior results clearly pointing to a decline in insect biodiversity.”
When I talk to them five years later, old concerns about mistakes in an insect database shine a light on new, big-picture qualms about how biodiversity is measured, and whether scientific discovery should be up for debate.
The database they were initially concerned over was one called InsectChange, which merged various other datasets created by researchers. In 2020, scientists from Germany, Russia, and the United States published an analysis of InsectChange, examining its data and findings.
The main takeaways were that while insects on land aren’t doing well, declining at a rate of 9 per cent per decade, freshwater insects are making a comeback and increasing at 11 per cent per decade. That suggested the picture of insect decline worldwide is much more nuanced than previously thought.
The findings left some scholars scratching their heads, as other research suggests insect decline is continuing at a far greater pace. In fact, more than 60 scientists published letters saying they had concerns with the results.
Read more:
- Why bees are secretly one of Earth's most intelligent creatures
- This adorable frog could change everything we know about animal empathy
- Why talking to animals could soon be a reality
The databases team issued some corrections to the data and chalked others up to misunderstandings or differences in opinions.
But Gaume and Desquilbet didn’t feel the issues had been addressed thoroughly enough. “We kept feeling concerned about the impact of the publication,” says Desquilbet.
After all, the decline paper is cited 1,112 times in scientific journals and books, and their feedback is quoted just 60 times.
They dug into the dataset themselves, head first, and claimed to uncover over 530 methodological and statistical mistakes. Almost all data sets, 161 out of 165, had at least one issue, they said.
Hopping to conclusions
For one, the database gathers datasets with all sorts of different units of measurement but then applies a mathematical transformation that doesn't yield a percentage of insect change per year, as intended, according to Gaume, an insect ecologist at Montpellier University.
Datasets about aquatic insects also include non-insect invertebrates, such as mussels, snails, worms and crustaceans. For instance, data from a lake in Kazakhstan suggest that insects have increased over the past century, but include invasive mussel blooms, which make up 95 per cent of the invertebrates in the lake.
The majority of datasets came from rewilding and restoration projects with the end goal of increasing insect populations, such as building artificial ponds for insect repopulation. These environments are manipulated and so don’t represent a natural insect habitat. “The user is left unaware of the biases,” says Gaume.
The InsectChange team heard the concerns loud and clear.
“There are mistakes in our data,” says Roel Van Klink, an ecologist at the German Centre for Integrative Biodiversity Research who led the creation of the InsectChange database. “Like there are mistakes everywhere.”
The team is working to incorporate corrections and update the database to make it as accurate as possible, as well as expanding it. They aim to publish a new and improved version of the database in the next six months.
They also asked Gaume and Desquilbet to work on the project together. However, they refused.
“The real question is, does it matter for the outcome of the analysis? I say it doesn’t,” says Van Klink. "The results wouldn’t be different if all of these tiny mistakes were weeded out or if a different selection of data were made."
Van Klink notes that not separating insects from invertebrates in the freshwater data collection was a choice his team made.
Still, this InsectChange debacle likely tells a wider story.
The data hive
In the era of the biodiversity crisis, scientists are scrambling to measure what’s happening to nature and determine trends of what might happen next. They’re dealing with lots of data. Efforts collating these huge databases are hard to cross-reference and keep in check as there’s more and more information and less and less time.
This all serves as an example of why science is always technically up for debate and is a process made of constant changes.
“Science is self-correcting by nature,” says Manu Saunders, an ecologist from the University of New England in Australia who was not involved in either of the studies.
Every individual study has flaws, limitations, and biases. These complexities and limitations are rarely discussed because we’ve been convinced that simplified narratives are the desired ‘norm’ for science communication, says Saunders.
That’s why, Saunders says, scientists should continue to engage in debate every time new science comes out. Just because the research was published doesn’t mean its scientific pursuit is over. Science is a method, and an ongoing process, not a finished product.
About our experts
Laurence Gaume is an insect ecologist at Montpellier University in France. Her research focuses on insect-plant interactions and biodiversity conservation.
Marion Desquilbet is an environmental economist at the Toulouse School of Economics. Her work examines ecological economics and biodiversity policies.
Roel Van Klink is an ecologist at the German Centre for Integrative Biodiversity Research. He specialises in insect population trends and large-scale biodiversity datasets.
Manu Saunders is an ecologist at the University of New England in Australia. Her research explores insect conservation, ecosystem services, and science communication.
Read more: