One-quarter of freshwater fauna threatened with extinction

Data compilation

We used the IUCN Red List categories and criteria^37,61, together with methods developed in other global assessment efforts^28,31,33, to assess the extinction risk of freshwater species. The detailed processes for the assessment of freshwater fishes and odonates are described below. The global assessment efforts for freshwater decapod crustaceans (focused on primary (that is, those that spend their entire life cycle in freshwaters) freshwater crab, crayfish and shrimp species) have been reported previously^38,39,40. However, some freshwater decapod species have been assessed or reassessed since these global assessment efforts were first completed. Therefore, updated Red List assessment figures are provided below to reflect the version of the dataset used in this analysis.

For all taxonomic groups considered in this analysis, we used the 2022-2 (December 2022) version⁶² of the tabular and spatial data from the IUCN Red List website, which was downloaded in March 2023 (see the ‘Red List assessment process’ and ‘Distribution maps’ sections below for a description of the data). This included data on 13,259 freshwater fish species in the classes Actinopterygii, Cephalaspidomorphi, Chondrichthyes, Myxini and Sarcopterygii. Freshwater fish species are defined here as those that spend all or a critical part of their life cycle in freshwaters, and are identified with the ‘Freshwater =Inland waters’ systems code on the IUCN Red List. In addition, we included preliminary (defined below) IUCN Red List categories for 1,577 freshwater fish species, including 208 reassessments. Together, this covered 14,628 freshwater fish species, representing 79% of the 18,432 formally described freshwater fish species as of December 2022 (ref. ⁶³). The IUCN Red List dataset also included other freshwater-dependent species: 6,223 odonate (dragonfly and damselfly) species (order Odonata), representing 97% of the 6,392 formally described species as of December 2022 (ref. ⁶⁴); and 2,645 decapod crustacean (primary freshwater crab, crayfish and shrimp) species (families Alpheidae, Astacidae, Atyidae, Cambaridae, Desmocarididae, Euryrhynchidae, Gecarcinucidae, Palaemonidae, Parastacidae, Potamidae, Potamonautidae, Pseudothelphusidae, Trichodactylidae, Typhlocarididae and Xiphocarididae), representing 81% of the 3,267 formally described species as of December 2022 (ref. ⁶⁵). Together, the assessed species represent 84% of the formally described species in all three taxonomic groups (see the ‘Missing species’ section below for a description of the gaps in coverage). These groups were selected for assessment because they cover both vertebrates and invertebrates, illustrate a range of life histories, occupy various wetland habitats, are not restricted to particular continents and are groups for which there was thought to be a reasonable level of existing information, such that their comprehensive assessment would provide a good indication of the status of freshwater fauna globally.

This freshwater species dataset was compared with that of the tetrapods: amphibians (class Amphibia; 7,468 species), birds (class Aves; 11,188 species), mammals (class Mammalia; 5,973 species) and reptiles (class Reptilia; 10,222 species).

Red List assessment process

We worked with species experts and members of the IUCN Species Survival Commission (SSC) to compile data, draft and review Red List assessments.

For the freshwater fishes, the Red List assessments were primarily completed through a series of IUCN-led regional sub-projects, through which global IUCN Red List assessments of all species native to the region were completed. There were 40 regional sub-projects between 2003 and 2023, which included close to 60 workshops (Supplementary Table 1a). Some families (that is, Acipenseridae, Anguillidae and Salmonidae) were assessed primarily through separate processes facilitated by the relevant IUCN SSC Specialist Group (that is, sturgeon, anguillid eel and salmon, respectively). Overall, the freshwater fish Red List assessments were drafted by 820 species experts (Supplementary Note 1a) with additional individuals contributing through facilitation and review.

For the odonates, the Red List assessments were also primarily completed through a series of regional sub-projects (led by either IUCN, members of the IUCN SSC Dragonfly Specialist Group or related groups, for example, the Sociedad de Odonatología Latinoamericana), through which global IUCN Red List assessments of all species native to the region were completed. There were 26 regional sub-projects between 2003 and 2023, which included close to 50 workshops (Supplementary Table 1b). In addition, between 2006 and 2008, the sampled approach to the Red List Index project⁴³ randomly selected and assessed 1,500 odonate species. Finally, work through the IUCN-Toyota Partnership⁶⁶ filled remaining gaps (primarily in Latin America and Southeast Asia) to complete the global dragonfly assessment. Overall, the odonate Red List assessments were drafted by 112 species experts (Supplementary Note 1b) with additional individuals contributing through facilitation and review.

Global assessment efforts for freshwater decapod crustaceans have been previously reported^38,39,40. Overall, these assessments were drafted by 78 species experts (Supplementary Note 1c) with additional individuals contributing through facilitation and review.

For all freshwater species, species experts and/or project staff first sourced information for each species from the published and grey literature, online databases, and expert knowledge and collections, and then entered it in the IUCN Species Information Service database (https://sis.iucnsis.org/apps/org.iucn.sis.server/SIS/index.html). Following the supporting information guidelines for the IUCN Red List⁶⁷, information was recorded on the geographical distribution, population status and trend, habitats and ecological requirements, use and trade, threats, and research and conservation actions relevant to the species, with supporting sources cited. In addition, a distribution map was produced for each species (see ‘Distribution maps’ below). The Red List criteria³⁷ were then applied to assign a Red List category indicating the extinction risk of the species: extinct (EX), extinct in the wild (EW), critically endangered (CR), endangered (EN), vulnerable (VU), near threatened (NT), least concern (LC) or data deficient (DD). Threatened species are those categorized as critically endangered, endangered or vulnerable, and are considered to be facing a high to extremely high risk of extinction in the wild. Species assessed as critically endangered may further be tagged as possibly extinct or possibly extinct in the wild. The former refers to species that are likely to be extinct, but for which there is a small possibility they are still extant, and the latter refers to species that still survive in captivity⁶¹.

Each assessment then underwent two reviews. First, at least one independent scientist familiar with each species reviewed the assessment to ensure the data presented were correct and complete, and that the Red List criteria had been applied appropriately. Once each assessment had passed this first stage of the review, including revision (if necessary), staff from the IUCN Red List Unit reviewed the assessments to ensure that the Red List criteria had been applied appropriately, and the documentation standards had been met. Once each assessment had passed this second stage of the review, again including revision (if necessary), they were considered finalized and set for publication on the IUCN Red List website.

The preliminary Red List assessments of 1,577 freshwater fish species used in this analysis had undergone only the first step of the review process described above at the time of analysis. Of these assessments, 87% (1,370 species) are now published, with only seven species (0.5%) changing Red List category before publication as a result of the second stage of review. We expect the remaining 207 species (13%) with preliminary assessments to have completed the assessment process outlined above by October 2024.

Taxonomy

For freshwater fishes, we used Eschmeyer’s Catalog of Fishes⁶³ as a taxonomic source, and for odonates, we used the World Odonata List⁶⁴. In both cases, we diverged only to follow well-justified taxonomic standards as recommended by the relevant IUCN SSC Specialist Group. We were not able to revisit new descriptions for regions or families after the end of the original regional sub-projects. Therefore, the final species lists are not fully consistent with any single release of either of the above taxonomic sources.

Distribution maps

Where data allowed, we produced polygon distribution maps for each species following the IUCN Red List mapping standards⁶⁸. These distribution maps represent the best-available depiction of the historical, present and inferred distribution of a species. They represent the limits of the distribution of a species, indicating that the species probably only occurs within the polygon, but not necessarily everywhere within the polygon. Each polygon was coded according to the species presence, origin and seasonality in the corresponding area. For freshwater species, these polygons were based on river and lake catchments, which are generally accepted as the most appropriate management unit for inland waters⁶⁹. These catchments are delineated using HydroBASINS, a globally standardized hydrological framework⁷⁰, with level 8 HydroBASINS as the default resolution (mean sub-basin area of 576 km²). Assessors are recommended to use higher-resolution level HydroBASINS (that is, levels 10 and 12) for species with restricted distributions.

Threats

All known past, current and future threats to a species were coded using the IUCN Threats Classification Scheme (version 3.3; https://www.iucnredlist.org/resources/threat-classification-scheme).

Habitat preferences

Where known, species habitats were coded using the IUCN Habitats Classification Scheme (version 3.1; https://www.iucnredlist.org/resources/habitat-classification-scheme). Species were assigned to all habitat classes in which they are known to occur.

Data limitations

Although we made an extensive effort to complete assessments for all freshwater fishes and odonates, some data gaps remain. These data gaps are discussed below, along with limitations with the published freshwater decapod data.

Missing species

As of December 2022, 3,804 freshwater fishes (21% of the formally described species), 169 odonates (3% of the formally described species) and 622 freshwater decapods (19% of the formally described species) were omitted from the study.

For freshwater fishes, the majority of these species were omitted because they were described after the relevant regional sub-projects had already been completed. The earliest sub-projects on the freshwater fishes focused on Pan-Africa⁷¹ (2003–2011; Supplementary Table 1a), and therefore, we would expect the most new descriptions for this region given that the greatest time period has passed. However, a number of regional sub-projects on highly diverse African sub-regions (for example, the Lake Victoria basin⁷², the Lake Malawi/Nyasa/Nyassa basin⁵⁹ and western Africa⁷³) have been completed more recently and will have captured much of this new diversity. In addition, the regions with highest rates of description (that is, South America and Asia)^74,75 have been the focus of more recent regional sub-projects.

The odonates are the most comprehensively assessed of the three freshwater groups with the few omitted species being new descriptions, reflecting the geographical distribution of the order overall.

For freshwater decapods, the majority of omissions were because of new descriptions⁶⁵ after the global assessment efforts^38,39,40 were completed. New descriptions also reflect the geographical distribution of the group overall and have primarily been in Asia, followed by the Neotropics and Afrotropics, for crabs; in Asia, followed by Oceania and Latin America, for shrimps; and in North America, followed by Australia, for crayfishes. Therefore, we do not expect a geographical bias in the results for the freshwater species due to recent descriptions, although we underestimate overall species richness⁶⁵.

Recently described species are often not well known and may be rare or occur in very restricted or poorly surveyed areas that are often subject to high levels of human impacts. As such, recent descriptions are more likely to be assessed as data deficient or in a threatened category⁷⁶, than to be assessed as least concern. The effect on our analyses is likely to be an underestimate of the number of threatened species and lower surrogacy levels than reported here.

For the freshwater fishes, a secondary reason for the omission of species was that some of the regional sub-projects were still under way at the time of data export in March 2023 (Supplementary Table 1), such that not all species had assessments that had sufficiently progressed through the Red List assessment process to be considered. The omissions for this reason are primarily freshwater fishes native to China (about 850 species), South America (approximately 670 species), India (about 190 species) and the Korean Peninsula (approximately 30 species). With the exception of China, where only approximately 50% of freshwater fishes are assessed, coverage of freshwater fish species in the other countries or regions is considered sufficiently comprehensive (that is, 80% or more) to evaluate the status of the taxonomic group^77,78.

Data deficient species

An assessment of data deficient indicates that there is insufficient information available on the distribution and/or population status of a species to make a direct or indirect assessment of its risk of extinction³⁷. Eighteen per cent of freshwater fishes (2,634 species), 29% of odonates (1,830 species) and 39% of freshwater decapods (1,042 species) were assessed as data deficient.

Spatial data

Although we made extensive efforts to map the current known distribution of each species, this is missing or incomplete for some species. Polygon map availability for all freshwater species was as follows: freshwater decapods (94%), freshwater fishes (95%) and odonates (77%). Polygon map availability for threatened freshwater species was as follows: freshwater decapods (89%), freshwater fishes (96%) and odonates (87%). Polygon map availability for data deficient freshwater species was as follows: freshwater decapods (93%), freshwater fishes (92%) and odonates (77%). Polygon map availability for threatened tetrapod species was as follows: amphibians (100%), birds (100%), mammals (100%) and reptiles (96%). Species missing maps generally had older Red List assessments where the supporting data requirements were lower or only had non-polygon spatial data (that is, point locality data).

It should be noted that species occurrence is unlikely to be spread evenly or entirely throughout the area depicted in species distribution maps, for example, with gaps expected in areas without suitable habitat.

Time span of assessments

Reassessment of species on the IUCN Red List is recommended every 10 years⁷⁹, with assessments older than this tagged as ‘needs updating’. The freshwater fish assessments were completed between 1996 and 2023, with 4,834 assessments (33%) completed before 2013. The odonate assessments were completed between 1996 and 2021, with 1,663 (27%) completed before 2013. The freshwater decapod assessments were completed between 1996 and 2022, with 2,482 (94%) completed before 2013. Approximately the same proportion of freshwater fishes were assessed as threatened in the pre-2013 (27%) and post-2013 (26%) datasets. However, a higher proportion of odonates and freshwater decapods were assessed as threatened in the post-2013 dataset (17% and 42%, respectively) than the pre-2013 dataset (12% and 29%, respectively). This could be because the later assessments focused more on the tropical Americas and Asia, where species have smaller ranges on average than in Africa and the Holarctic, which is likely to lead to a higher percentage threatened. Given the continuing decline of biodiversity globally⁶, it is probable that species with assessments older than 10 years are more likely to be assessed in a higher threat category today, indicating a potential underestimation of the extinction risk in these groups.

Analyses

Proportion of species threatened with extinction

To capture the uncertainty in the proportion of species threatened with extinction that is introduced by data deficient species, we report three values for percentage threatened as follows.

We used the following formula as the midpoint of the proportion of species threatened with extinction:

where EW, CR, EN, VU, EX and DD are the number of species in each corresponding Red List category, and N is the total number of species assessed. This formula assumes data deficient species are equally threatened as data-sufficient species. This value is considered the best estimate of extinction risk⁷⁸.

We used the following formula to calculate the lower estimate of the proportion of species threatened with extinction:

where EW, CR, EN, VU and EX are the number of species in each corresponding Red List category, and N is the total number of species assessed. This formula assumes data deficient species are not threatened.

Finally, we used the following formula to calculate the upper estimate of the proportion of species threatened with extinction:

where EW, CR, EN, VU, DD and EX are the number of species in each corresponding Red List category, and N is the total number of species assessed. This formula assumes that all data deficient species are threatened.

Threats

We analysed threats as classified under the IUCN Threats Classification Scheme. Multiple threats can affect a single species. It is possible to assign codes for scope and severity to each threat. However, this is optional⁶⁷ and not available for the majority of freshwater species. Therefore, we did not use scope and severity to distinguish the relative importance of threats to individual species, and all coded threats were included in the analysis. We recommend future reassessments to include these codes, such that major threats can be distinguished from trivial threats in analyses based, for example, on the proportion of the population affected.

The threats analysis presented here focuses on threatened species, plus on extinct and extinct in the wild freshwater species. Threats data availability for threatened species for each taxonomic group was as follows: freshwater decapods (97%), freshwater fishes (98%), odonates (97%), amphibians (100%), birds (100%), mammals (99%) and reptiles (97%). Eighty-two per cent of extinct and extinct in the wild freshwater species had threats coded. Species without threats data either faced no known major threats, faced unknown threats or had no threats data coded because they have older Red List assessments where the supporting data requirements were lower.

Habitats

We first analysed habitat use at the highest level of the IUCN Habitats Classification Scheme, focusing on threatened species. In addition, for freshwater species only, we analysed habitat use within habitat code 5 ‘Wetlands (inland)’, focusing on all species, and then threatened, extinct in the wild and extinct species.

Habitat data availability for threatened species for each taxonomic group was as follows: freshwater decapods (99%), freshwater fishes (99%), odonates (97%), amphibians (100%), birds (100%), mammals (99%) and reptiles (97%). Wetland-specific habitat data availability for all species for each freshwater taxonomic group was as follows: freshwater decapods (100%), freshwater fishes (99%) and odonates (99%). All extinct and extinct in the wild freshwater species had habitats coded. Species without habitat data either had habitats coded as ‘unknown’ or had no habitat data coded because they have older Red List assessments in which the supporting data requirements were lower.

Statistical tests

We used chi-squared tests to assess whether any threats were recorded more for extinct species than would be expected based on threats recorded for threatened species, whether any habitats were recorded more for threatened species than would be expected based on habitats recorded for all species, and whether any habitats were recorded more for extinct species than would be expected based on habitats recorded for threatened species.

Spatial analyses

Analyses of the distribution maps used polygons coded with the following presence, origin and seasonality values (as defined in the IUCN Red List mapping standards⁶⁸):

• Presence denotes extant (code 1), probably extant (code 2) or possibly extinct (code 4)

• Origin refers to native (code 1), reintroduced (code 2) or assisted colonization (code 6)

• Seasonality indicates resident (code 1), breeding (code 2), non-breeding (code 3) or passage (code 4)

In addition, for the surrogacy analyses, we excluded ranges coded as presence code 4 (possibly extinct).

All spatial mapping and subsequent analyses were conducted on a global 0.5 × 0.5 latitude–longitude grid (approximately 50-km resolution; WGS84). We converted all polygon range maps (including HydroBASIN-based maps) tagged with presence, origin and seasonality codes as described above to these grids to have a consistent format for analysis across all taxonomic groups. We mapped the distribution of species as a count of the number of species ranges overlapping each grid cell.

Surrogacy estimation analyses within two conservation strategies

To evaluate the degree to which conservation of threatened amphibians, birds, mammals and reptiles (individually or combined) serve as surrogates for conservation of threatened freshwater decapods, fishes and odonates (individually or combined), we calculated the SAI of surrogate effectiveness³³. A surrogate is selected as a representative of conservation planning, simplifying the process of monitoring and conserving biodiversity. Its effectiveness hinges on how well the surrogate can reflect the presence, abundance and diversity of species in a given area. Here we used species accumulation curves to measure this effectiveness, by comparing the species accumulation curves of surrogates with those of the target group.

We performed the analyses on two main global conservation strategies: (a) maximizing rarity-weighted richness (that is, the aggregate importance of each grid cell to the species occurring there) of threatened species, and (b) maximizing inclusion of the most range-restricted threatened species. The first strategy prioritizes areas containing many threatened species with highly restricted ranges globally, whereas the second prioritizes essential areas for the most globally range-restricted threatened species.

We implemented both conservation strategies within the spatial conservation planning software Zonation⁸⁰ and the R⁸¹ package ‘zonator’⁸², using the additive benefit function (ABF) and the core-area zonation (CAZ) algorithms for strategy a and strategy b, respectively. The algorithm for the ABF (strategy a) focuses on ranking areas by the sum of the proportion of the overall range size of all species found within a specific grid cell (that is, a quantity similar to weighted species endemism and endemism richness). The grid cells that contain many species occurring exclusively in that cell or in only a few other cells are given the highest priority. In the CAZ algorithm (strategy b), areas are prioritized based on the maximum proportion of the global range size of all species within a specific grid cell. The algorithm assigns the highest priority to cells that contain the greatest proportions of the ranges of the most range-restricted species.

We estimated optimal, surrogate and random curves based on multiple target species-surrogate species combinations. We used 100 sets of random terrestrial grid-cell sequences to generate 95% confidence intervals around a median random curve. We ran five iterations of each spatial prioritization algorithm for each taxonomic group, and optimal and surrogate curves were summarized using the median and 95% confidence intervals across the five iterations.

We derived the SAI of surrogate effectiveness⁸³, which quantifies the rate of inclusion of target biodiversity units across areas selected optimally based on the targets themselves, based on surrogate diversity, or at random, as

where s is the area under the surrogate curve, r is the area under the random curve, and o is the area under the optimal curve. If SAI = 1, the optimal and the surrogate curves coincide (perfect surrogacy); if SAI is between 1 and 0, the surrogate curve is above the random curve (positive surrogacy); if SAI = 0, the surrogate and random curves are the same (no surrogacy); and if SAI < 0, the surrogate curve is below the random curve (negative surrogacy). We used the following descriptors to define SAI performance: 0.01–0.19 as very poor, 0.20–0.39 as poor, 0.40–0.59 as reasonable, 0.60–0.79 as good, and 0.80–0.99 as very good. It should be noted that if SAI = 0.5, for example, this does not mean that 50% of targets are represented and 50% of targets are not represented. For each SAI, we reported the median and 95% confidence intervals based on the five target and surrogate curve iterations and 100 random curve iterations.

In addition, we evaluated whether prioritizing for two widely used hydrological variables (water stress as a measure of water quantity, and eutrophication (nitrate–nitrite) as a measure of water quality) are effective surrogacy strategies for conservation of threatened freshwater species. We used the SAI to evaluate the ability of both variables to identify areas that most efficiently represent threatened freshwater species, again harnessing strategies for both maximizing rarity-weighted richness (ABF) and maximizing inclusion range-restricted species (CAZ). Once again, we used Zonation⁸⁰ to generate the complementarity-based ranking of conservation values of the target, with the respective algorithms, over the landscape of interest. To generate the rank order, we used (1) the baseline water stress layer from the Aqueduct Water Risk Atlas, which measures the ratio of total water demand (for example, domestic, industrial, irrigation, and livestock consumptive and non-consumptive uses) to available renewable surface and groundwater supplies^84,85, and (2) the baseline nitrogen layer from the World Bank catalogue, which provides global predictions of nitrate–nitrite levels⁸⁶. The water stress layer was considered a proxy of a baseline level of water demand compared with available renewable water and groundwater, as used in setting science-based targets for freshwater⁸. Nitrogen levels in water around the world are highly correlated with population density, sanitation practices and agricultural activities. Here the nitrogen layer was predicted globally and provides valuable information about nitrogen concentrations in areas where no previous observations have been made.

We rasterized the baseline water stress and the nitrogen layers to a 0.5 × 0.5 latitude–longitude grids (approximately 50-km resolution; WGS84) to match the species rasters. For the water stress analysis, we excluded cells with missing water stress data across the world’s land (12% of cells excluded). We found that 44% of the world’s cells with water stress data had no threatened freshwater species, but these cells were still included in the analysis. For the nitrogen levels analysis, we excluded cells missing nitrogen data across the world’s land, which accounted for 16% of the cells. Among the remaining cells with nitrogen data, 52% had no threatened freshwater species, but again these were retained in the analysis. Before constructing the curves, we organized sites (grid cells) in the species matrix from those with high abiotic values to low abiotic values for ranking cells. We used 100 sets of random terrestrial grid-cell sequences to generate 95% confidence intervals around a median random curve. We generated five random terrestrial grid cell sequences for constructing the surrogate curves, so we randomly changed the rank order only between those cells that have the same values.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.