COP16: number of fish species at risk of extinction may be five times higher than previous estimates, new study finds

The loss of biodiversity on our planet is a fact. For example, WWF's Living Planet Report 2022 found that populations of mammals, birds, fish, reptiles and amphibians have declined by an average of 69 percent since 1970.

Much of the species' loss is due to habitat destruction from unsustainable agriculture or deforestation, and the effects of climate change - not a major driver so far - will become a major driver in the coming decades. This is an issue that will be discussed at the United Nations Conference on Biodiversity (COP16) in Cali in October.

It is a constant concern for scientists and institutions dedicated to studying and monitoring the issue, and the reason for the existence of the International Union for Conservation of Nature's (IUCN) Red List of Threatened Species, which monitors more than 150,000 species to guide global conservation efforts in favor of the most endangered. A list they hope will continue to grow.

(Follow all the news of COP16 here)

"Currently, the IUCN Biodiversity Assessment and Knowledge Team manages data on more than 163,000 species, a number that will increase significantly in the coming years. More than 156,500 species are well documented, with ing information on ecology, population size, threats, conservation measures and use," the organization says.

The IUCN Red List has nine categories, three of which relate to threatened species: "critically endangered," "endangered," and "vulnerable." Its criteria are global and based on scientific evidence to assess risk.

But given the sheer number of species on the planet, efforts often fall short. Research recently published in the open-access journal PLOS Biology cautions against species that do not receive an IUCN conservation status due to insufficient data, while proposing a new model to identify groups at risk of extinction using AI and machine learning techniques.

"The IUCN provides criteria for assessing species, such as rate of decline, population size, geographic range, and degree of fragmentation and distribution. These criteria are the most accurate, but they require a lot of data to compile and are therefore very time and resource consuming. As a result, many species are not assessed due to a lack of data," Nicolas Loiseau, a researcher at the French National Center for Scientific Research (CNRS) and leader of the study, told EL TIEMPO.

(Take a look: What is it, why is it important, and what is COP16 leaving behind for Colombia?)

In their work, the researchers predict that 12.7 percent of marine teleost fish species (the most numerous) are at risk of extinction, five times higher than the 2.5 percent previously estimated by the IUCN. The report includes 4,992 species that have not received an official conservation status due to insufficient data. This represents 38 percent of marine fish species.

To better target conservation efforts to species in need, Loiseau and colleagues combined a machine learning model with an artificial neural network to predict extinction risk for data-poor species. The models were trained using data on the presence, biological characteristics, taxonomy, and human use of 13,195 species.

"Thus, AI offers a unique opportunity to provide a rapid, comprehensive, and cost-effective assessment of species extinction risk and to identify which species traits and geographic regions should be prioritized for assessment," Loiseau said.

78.5 percent of the 4,992 species were classified as not threatened or threatened (which includes the IUCN categories of Critically Endangered, Endangered and Vulnerable). Endangered species increased by a factor of five (from 334 to 1,671) and vulnerable species by a factor of three (from 7,869 to 10,451).

According to the study, endangered species are primarily characterized by a small geographic range, a relatively large body size, and a low growth rate. They also found that pockets of these threatened groups are concentrated in the South China Sea, the Philippine Sea, the Celebes Sea, and the west coast of Australia and North America.

The researchers also noted a significant shift in the ranking of conservation priorities following the IUCN species projections, and recommended prioritizing the Pacific Islands and the polar and subpolar regions of the Southern Hemisphere to for newly threatened species.

"We propose the integration of multifactorial ensemble learning to assess species extinction risk and provide a more complete picture of threatened taxonomic groups to ultimately achieve global conservation goals, such as expanding the coverage of protected areas where species are most vulnerable," they said in the study.

They also point out that many of the species for which there is no data are found in the Coral Triangle, indicating that more research is needed in this area.

"We found that small, hidden fish species such as gobies and blennies are at risk. These cryptobenthic fish play a vital role in reef ecosystems, particularly in food chains and overall reef health. Because they are difficult to detect and live in specific habitats, it is difficult to assess their populations, meaning that some may be quietly heading towards extinction. This highlights the urgent need to pay more attention to these species," Loiseau told the newspaper.

And while they note that models like the ones they used cannot replace direct assessments of endangered species, AI offers a unique opportunity to provide a rapid, comprehensive and cost-effective assessment of species' extinction risk, while pinpointing species on which to prioritize data collection and conservation efforts.

"Artificial intelligence makes it possible to reliably assess the extinction risk of species that have not yet been assessed by the International Union for Conservation of Nature. Our analysis of 13,195 marine fish species shows that the extinction risk is significantly higher than initial IUCN estimates. We propose to incorporate recent advances in predicting species extinction risk into a new synthetic index, the IUCN predicted status. This index can serve as a valuable complement to the current 'measured IUCN status'," said Loiseau.

In their research, the experts also note that although several studies have already proposed automated methods for making a preliminary assessment of the conservation status of species based on their attributes or remotely sensed predictors, to their knowledge these have not yet been incorporated into the official Red List assessment. However, to their knowledge, these have not yet been incorporated into the official Red List assessment.

"We believe that ensemble learning is relevant because it is accurate and conservative. The performance of machine learning algorithms is known to vary depending on factors such as the dimensionality of the data set. To address this variability, we propose a multi-model strategy that combines different algorithms to leverage their strengths and mitigate their weaknesses," they said in the study.

ALEJANDRA LÓPEZ PLAZAS

Science journalist

@malelopezpl