Introduction

Freshwater systems occupy a small fraction of the Earth’s surface but contribute disproportionately to global biodiversity due to the high number of species they host1. Such biodiversity is threatened by several interacting pressures (e.g. land use change, climate change, flow alterations, loss of connectivity), resulting in strong declines worldwide2,3,4,5,6. The loss of longitudinal connectivity in rivers, due to the construction of barriers, is a particularly relevant pressure7,8. The dendritic structure of rivers renders them especially sensitive to habitat connectivity losses, as even a single barrier can isolate entire sections of the river network9,10,11. When multiple barriers are placed on a river system, the result is a sequence of isolated sections where the movement of water, sediments, energy and organisms is impeded12. Fragmented rivers are widespread7,13, and are expected to further increase in the future due to the global proliferation of dams under construction and planned14,15.

Riverscape fragmentation impacts the geographic distribution of fish biodiversity. Due to the presence of barriers, potamodromous and diadromous species may be prevented from reaching habitats that are necessary for the completion of their life cycles16,17,18,19. Even non-migratory fish species are exposed to local extinctions if they occupy isolated river fragments which do not offer enough habitat to support viable populations and prevent rescue effects from neighboring populations15,20. As a result, river habitat fragmentation impacts the structure of freshwater fish communities21,22,23,24, with cascading socioeconomic effects, for instance concerning fisheries declines25,26. Such impacts are expected to increase in the future, given the projected increases in river fragmentation worldwide15. Comprehensive, spatially explicit assessments of fish habitat fragmentation are urgently needed in large river basins to facilitate future fish habitat management and protection27.

The Mekong River lies within the Indo-Burma biodiversity hotspot, supporting the highest fish diversity in Asia, with more than 1,000 freshwater fish species, hundreds of which are endemic28. The Mekong basin is also the largest inland fishery in the world, with an estimated mean annual catch of 2.1 million tonnes, about 15% of the world’s total inland fish catch29. The river suffers from intensive anthropogenic pressures including river barrier construction, overexploitation, pollution, water abstraction, and habitat destruction28,30,31,32. Among all anthropogenic threats, barrier construction is the most concerning due to its profound impacts on flow regimes, sediment transportation and freshwater biodiversity and its rapid projected growth in the near future32,33,34,35. Existing analyses conducted using global datasets of georeferenced dams highlighted the poor connectivity status of the upper Mekong33, and have warned about the detrimental effect of future hydropower development on connectivity36. However, there is ongoing debate over whether few large dams or numerous small barriers are more detrimental to habitat connectivity and consequently, biodiversity15,20.

Existing analyses of habitat fragmentation effects in the world’s largest river basins15,20,33,36,37, including the Mekong, employed global databases or regional databases where barrier data had poor completeness and/or resolution. In particular, these studies used datasets describing large dams located in the main river channel and larger tributaries. While large dams are more likely to occur in the main channel, other barriers like small dams, weirs, and sluice gates are likely to be located in low-order streams and are more likely to be unreported13,38. Estimating the impact of barriers in more distal regions of river networks is also relevant because, while generally having fewer aquatic species, these sites usually host more distinct communities due to their more isolated position in the river network21,39,40. Recent developments in AI-assisted identification of barriers in the Mekong Basin has led to the compilation of a comprehensive dataset with more than 13,000 barriers that also include smaller barriers, such as small dams and sluice gates41. Although small barriers can have major impacts on fish distribution42,43, the impacts of such structures on fragmentation of fish populations in the Mekong basin have not been investigated so far.

Based on the highlighted knowledge gaps, in this paper, we used a comprehensive barriers dataset to investigate the spatial distribution of fragmentation and its link with fish population longitudinal connectivity status in the Mekong Basin. We first reconstructed the potential distribution range of all fish species in the Mekong Basin by overlaying fish dispersal abilities on a fish occurrence database28,44. We then assessed all species’ habitat fragmentation status using the most complete barrier database available to calculate a Population Connectivity index (PCI) and compared the differences in connectivity reduction between migration types and conservation status at the species level. Subsequently, we evaluated which types of barriers contributed most to the degradation of river connectivity by using a Fragmentation Index (FI) that compares the population connectivity with and without barriers, and its normalized version ^FI, that accounts for the mean contribution of a single barrier. With this approach, we identified fish species and regions most threatened by the construction of river barriers in the Mekong Basin, to support future management plans including connectivity restoration and habitat conservation.

Results

Fish distribution within the Mekong Basin

Fish species richness differed between seven ecological regions (i.e., Himalaya, Upper Mekong, Middle Mekong, Chi-Mun, 3S [the Sesan, Sre Pok, and Sekong rivers], Tonle Sap and Delta30,33) of the Mekong Basin (Kruskal-Wallis H test, χ2 (6) = 521.9, P < 0.001), reducing with increasing elevation. Two species were found in the upper reaches of the Himalaya region whereas 413 occurred in sub-basins near the boundary between the Tonle Sap and Delta region (Fig. 1). On the other hand, the Upper Mekong exhibited the most unique communities, as measured by the local contribution to beta diversity index (LCBD), higher than other ecological regions except the Delta (Kruskal-Wallis H test, χ2 (6) = 332.7, P < 0.001).

Fig. 1: Distribution of river barriers, fish occurrence, and species richness across the Mekong Basin.
figure 1

The main channel barriers a, fish occurrence records b, and cumulative fish species richness in each sub-basin and the location of seven ecological regions (i.e., Himalaya, Upper Mekong, Middle Mekong, Chi-Mun, 3S, Tonle Sap and Delta) of the Mekong Basin c.

Full size image

Sub-basins near the confluence of the Mun River (outlet of the Chi-Mun Region) and the Mekong main channel exhibited the highest fish species richness (157–385 species per sub-basin) as well as the highest river barrier density (8–215 per 100 km; Fig. 2). The ‘betweenness centrality’, which measures the structural importance of sub-basins within the river network, was high within the Middle Mekong region (mean BC and 95% CI, 2974.5 [2172.8–3776.3]; Supplementary Data 1). This basin also exhibited high species richness (42–385 species per sub-basin) and moderate barrier densities (1–8 per 100 km). The Tonle Sap region also exhibited high species richness (86–345 species per sub-basin) but lowest mean BC (233.2 [156.8–309.7]). The Upper Mekong exhibited relatively low species richness (20–240 species per sub-basin) but high mean BC (2909.6 [2079.4–3739.8]); a few areas in this region had relatively high barrier densities (Fig. 2).

Fig. 2: Relationship between river barrier density and biodiversity patterns in the Mekong Basin.
figure 2

Bivariate choropleth plot showing the relationship between barrier density (ranged from 0–2.2 per km) and total species richness (ranged from 0–952 species; a), and local contributions to beta diversity (LCBD) b.

Full size image

Potamodromous species (n = 135, see Supplementary Data 2), undertaking migrations solely in freshwater, were found to mostly inhabit the main channel in the Middle Mekong, the Tonle Sap, and the Chi-Mun regions. Within the distribution ranges of potamodromous species, high barrier density mostly occurred in the Chi-Mun region, and moderate barrier density was found in sub-basins along the main Mekong channel. Diadromous species (n = 125), which migrate between freshwater and the sea, were mainly found in the Lower Mekong between the 3S (i.e., the Sesan, Sre Pok, and Sekong rivers) confluence and the Delta. Barrier densities were moderate, covering much of the Mekong Delta and reduced further upstream. The listed threatened species (vulnerable [n = 42], endangered [n = 25], and critically endangered [n = 13]) mostly inhabit the main channel between the middle and lower Mekong, the Tonle Sap Lake, and sub-basins next to them (Fig. 3). Similarly, high and moderate barrier densities along the main channel pose threats to those species.

Fig. 3: Relationships between river barrier density and fish species richness across migration and threatened categories in the Mekong Basin.
figure 3

Bivariate choropleth plots showing the relationship between barrier density and potamodromous species richness (ranged from 0–135 species; a, diadromous migrant fish species richness (ranged from 0–125 species; b threatened fish species richness (IUCN status: Vulnerable (low threat), Endangered (moderate threat), and Critically Endangered (highest threat) c, in each sub-basin of the Mekong River.

Full size image

Habitat fragmentation of Mekong fishes

Across all 952 fish species for which data was available, 881 (92.5%) of them suffered from decreased population connectivity (PCIF) comparing with their barrier-free population connectivity (PCIO see methods). When considering barriers, the Population Connectivity Index (PCI) pooled for all the fish species significantly decreased from 4.35 [3.81–5.02] (median and 95% CI, natural status) to 3.70 [3.06–4.21] (moderate passage status). The Fragmentation Index (FI, modeled under moderate barrier passability) ranged from 0.01 to 76.20, with a median value of 19.37, and was positively related to available fish habitat length (Spearman’s correlation, r = 0.566, P < 0.001; Fig. S1). Among all fish families (for circumstances where there were more than five species present), the Pangasiidae (n = 19) experienced the most serious habitat fragmentation (average Fragmentation Index, FI = 51.91), followed by Notopteridae (n = 5; average FI = 51.27) and Channidae (n = 10; average FI = 48.84, Supplementary Data 2). Among all species, Bagarius suchus (Near Threatened conservation status) suffered from the most profound habitat fragmentation level (FI = 76.20), followed by Wallago attu (Vulnerable conservation status; FI = 74.43) and Phalacronotus micronemus (Least Concern conservation status; FI = 74.37).

Significant differences were found in FI between different migration types (Fig. 4, Kruskal-Wallis H test, χ2 (5) = 143.9, P < 0.001). Potamodromous fish had the worst FI (42.45 [38.61– 46.62]), significantly higher than anadromous fish (14.55 [2.18– 27.29]), amphidromous fish (7.30 [5.98–10.45]), and species with unknown migration type (18.13 [16.64–19.92]) (pairwise post hoc, P < 0.05 in all cases). Significant differences in ^FI were also observed (Kruskal-Wallis H test, χ2 (5) = 19.42, P < 0.001), with potamodromous fish (0.09 [0.06–0.10]) and unknown migration type (0.12 [0.11–0.14]) significantly lower than catadromous fish (0.67 [0.52–1.16]) (pairwise post hoc, P < 0.05 in both cases). The FI exhibited a statistically significant positive relationship with mean habitat betweenness centrality (Spearman’s correlation, r = 0.183, P < 0.001); however, the effect size suggests a practically negligible monotonic association.

Fig. 4: Variation in fragmentation index across fish migration types and conservation statuses.
figure 4

Boxplots showing the fragmentation index (FI) and mean fragmentation index caused per barrier (^FI) across different migration types (potamodromous, catadromous, anadromous, amphidromous, non-migratory, and unknown [NA]) and conservation status (Critically Endangered [CR], Endangered [EN], Vulnerable [VU], Near Threatened [NT], Least Concern [LC], Not Evaluated [NE], and Data Deficient [DD]). Letters above each box (e.g., “a”, “b”, “c”) denote statistical groupings based on post-hoc comparisons.

Full size image

Significant differences in FI occurred across IUCN categories (Fig. 4, Kruskal-Wallis H test, χ2 (5) = 30.13, P < 0.001). Critically Endangered species experienced the greatest FI (34.48 [19.46–53.52]). The FI status of Least Concern species was higher than Not Evaluated and Data Deficient (combined) species (pairwise post hoc, P < 0.001). Significant differences in ^FI were observed among conservation status (Kruskal-Wallis H test, χ2 (5) = 28.1, P < 0.001); CR species had higher ^FI (0.33 [0.18–3.13]) than LC species (0.10 [0.08–0.11]).

Contribution of different barrier types to fragmentation

Across four major types of barriers, small dams and sluice gates cumulatively contributed most significantly to habitat fragmentation as measured by FI for all fish species - more than weirs and large dams (ANOVA, F (3, 3804) = 42.6, P < 0.001; pairwise post hoc, P < 0.001 in all cases; Fig. 5) under all passability values (modeled passability ranged between 0.1 and 0.9). A similar pattern was observed for all tested passability values, here only values for passability = 0.3 are presented. This pattern was evident for potamodromous species, where small dams and sluice gates had a profound impact on habitat fragmentation compared to weirs and large dams (ANOVA, F (3, 536) = 43.1, P < 0.001). For diadromous species, the impact of sluice gates on habitat fragmentation was significantly greater than that for other barrier types (ANOVA, F (3, 496) = 57.4, P < 0.001). For non-migratory species, no significant differences were found across the four types of barriers (ANOVA, P > 0.05). For species with unknown migratory status, small dams contributed more to habitat fragmentation than the other barrier types under all passability values (ANOVA, F (3, 2720) = 10.6, P < 0.001).

Fig. 5: Contribution of different barrier types to habitat fragmentation across fish migration categories and barrier passability levels.
figure 5

Comparison of habitat fragmentation by barrier type across different fish migration categories (potamodromous, diadromous, non-migratory, and unknown [NA]) and passability scenarios. This set of graphs displays the Fragmentation Index (FI) and average fragmentation index (^FI) for four barrier types (large dam, small dam, weir, and sluice gate) across various passability values, stratified by migratory behavior of fish species. Each point represents the mean FI, and the error bars indicate the 95% confidence interval (CI). Letters above each box (e.g., “a”, “b”) denote statistical groupings based on post-hoc comparisons, shown at 0.3 passability only, as patterns across all modeled passability levels were similar.

Full size image

In comparison, large dams individually contributed more to habitat fragmentation as measured by ^FI, compared to other barrier types, were observed when considering all fish species (ANOVA, F (3, 3804) = 4.14, P = 0.006) and unknown migration types (ANOVA, F (3, 2720) = 4.7, P = 0.003) species (Fig. 5), suggesting a single large dam can have a large impact on such migration life history categories. For potamodromous (ANOVA, P > 0.05), diadromous (ANOVA, F (3, 496) = 1.5, P = 0.004), and non-migratory species (ANOVA, P > 0.05), sluice gates and small dams contributed more to ^FI, indicating that these barrier types have a greater localized impact on fragmentation for these groups.

Discussion

By utilizing the most complete fish occurrence database alongside the most comprehensive basin-scale barrier database, the impacts of various barrier types on habitat fragmentation for all fish species in the Mekong Basin have been assessed for the first time. Our results demonstrate that high species richness predominantly occurs in the middle and lower Mekong Basin along the main channel, which is also the region where a high number of barriers have been constructed28. On the other hand, sub-basins with a high ecological uniqueness, as measured by the local contribution to beta diversity (LCBD) were identified in the upper Mekong where the Chinese cascade dams are located. Across the entire basin, the most fragmented region appears to be the Chi-Mun sub-basin, characterized by the highest numbers of small dams and sluice gates45. In contrast, the Mekong headwaters and most of the upper regions were relatively less fragmented and still provide accessible habitats for certain highland fish species. Nevertheless, among all fish species, 93% were affected by barrier construction, strongly suggesting that most Mekong fish species are already threatened by habitat fragmentation.

Relationships between life history traits and habitat fragmentation

Since the start of the Anthropocene, many migratory fish species have experienced reductions in habitat range46. According to the Living Planet Index method, among 184 monitored freshwater migratory fish species globally, 81% exhibited a decreasing abundance trend between 1970 and 202047,48. When migratory fish traverse multiple sub-basins containing various habitats, they encounter a broader array of anthropogenic threats including a higher number of constructed barriers49. Habitat loss due to barriers constructed at key locations (e.g., stopover sub-basins, spawning and foraging habitat) may contribute to population decline by limiting migration success50,51. On the other hand, when larger areas of habitat are fragmented into smaller, less accessible patches of river channel, fish populations become more isolated, therefore making it more difficult to maintain meta-populations42. This results in decreased rates of immigration and emigration, hindering gene flow49.

In the Mekong Basin, we found that migratory species with larger habitat range and longer dispersal distances suffered more profound habitat fragmentation compared to species with narrower habitat ranges and limited dispersal ability. The greater dispersal distance, the higher the likelihood that species would encounter more barriers during movement49,52, and result in reduced population exchange. The habitat connectivity status of potamodromous species was significantly worse compared to other migration types, particularly for the middle and lower parts of the main Mekong channel and major sub-basins linked to it (e.g., the 3S, Chi-Mun, and Xe Don [located at the bottom of the Middle Mekong region] sub-basins). Small barriers such as sluice gates and small earthen dams were found to have particularly strong impacts on potamodromous species in the Chi-Mun and Middle Mekong regions53. Although certain sluice gates (e.g., the Pak Mun Dam [located at the outlet of the Mun River]) in the Chi-Mun region are regulated to open during the rainy season, which could provide temporal passage for migratory species such as Helicophagus waandersii54, thousands of other sluices gates are managed by various authorities, with unclear operational regimes. Further evaluation is needed to quantify their effects on fish migration dynamics.

Diadromous fish species such as Pangasius krempfi, Arius maculatus and Osteogeneiosus militaris, depend on seasonal migrations between freshwater, estuarine, and marine habitats for their life cycles55. However, increasing development of tidal barriers in the Mekong Delta has significantly disrupted these migration routes. Otolith microchemistry studies have revealed complex and diverse migration patterns in these species, which are now at risk due to physical barriers that block or delay upstream and downstream movements56. These disruptions pose serious threats to the sustainability of diadromous fish populations in the Mekong and underscore the urgent need for integrated river management strategies that account for their full migratory range.

Among the taxa exhibiting greatest habitat fragmentation risks, the Osteoglossiformes contains many highly threatened species according to the IUCN Red List, including the Endangered Asian bony-tongue Scleropages formosus, the Near Threatened royal featherback Chitala blanci, and the now Extinct giant featherback Chitala lopis. The population of S. formosus has experienced a significant decline over the past decades due to multiple human disturbance, with catch statistics indicating a 90% reduction in Cambodia and Thailand since 201057. In the Mekong, the freshwater habitats of the aforementioned species are evaluated as under pressure due to dam construction57 according to the IUCN Red List, and their conservation status could be further worsened when considering all proposed barriers as well as those omitted from existing databases58. With regard to fish families of particular concern, the most impacted Pangasiidae includes many flagship and megafauna freshwater species, such as the Critically Endangered Mekong giant catfish Pangasianodon gigas, Giant pangasius Pangasius sanitwongsei and Endangered striped catfish Pangasianodon hypophthalmus, all of which are ecologically and economically important. This family mostly consists of potamodromous species that undertake long-distance migrations along the main channel of the middle and lower Mekong, which suffers from main channel dams and sluice gates59. The population of P. gigas has declined by 80% over the past two decades in the Mekong River, with annual catches dropping from 40–50 individuals in the early 1900s to just 1–2 individuals in recent years60. The population of Pangasius sanitwongsei has undergone a drastic decline; annual catches of this species have dropped from around 100 individuals in the late 1980s to just 5–20 individuals by the late 1990s, with the species virtually disappearing from local fisheries thereafter59. Further research is urgently needed to re-assess the conservation status of all these threatened species within the Mekong, using the complete barrier database58. Additionally, the passability of most Mekong fish species at semi-passable barriers (barriers including weirs or sluice gates that allow fish to pass under certain conditions) remains unclear due to the lack of monitoring and performance standards. Future studies are required to address these gaps and to develop highly efficient bidirectional fish passes.

Large vs. small barriers: which causes greater habitat fragmentation?

Our analysis revealed that, across all species, the cumulative effect of small dams and sluice gates nearly doubled the habitat fragmentation compared to large dams at all modeled levels of passability. This finding aligns with previous research suggesting that the cumulative impacts of small-sized barriers are comparable to those of large dams15,20,61. Although the relationship between the fragmentation index (FI) and mean habitat betweenness centrality was statistically significant, the weak effect size suggests that tributary-dwelling species may experience fragmentation levels comparable to those in the mainstem, highlighting the need for greater attention to connectivity in tributary habitats. Such results underscore the importance of maintaining an openly accessible comprehensive barrier database to support ecological conservation and management efforts41, and highlight the need for standardized comprehensive databases at the river basin scale or, even, global scale.

All river barriers should be considered both individually, and in terms of cumulative effects when planning and conducting ecological impact and restoration activities6,62. For example, our analysis suggests that small dams and sluice gates are the main reason for habitat fragmentation of potamodromous fish species, which suffered from the most significant habitat fragmentation among all migration types. By contrast, large dams, which have generally received more attention in previous studies, contribute less to habitat fragmentation for potamodromous species compared with other barrier types. So, impacts of small-sized barriers should be carefully evaluated during connectivity impact assessment and restoration processes62. In comparison, the habitat of diadromous species in the Mekong is largely fragmented due to the construction of sluice gates56,63. As a large number of tidal gates and flaps have been built in the Mekong estuary region in the last 50 years to support regional agriculture45, these have blocked routes between the freshwater environments and the South China Sea45, necessitating increased connectivity restoration in the near future. For non-migratory species (resident fish), all types of barriers contributed equally to the fragmentation effects. Although the impacts of barriers on fish dispersal may be less severe for these species, their short distance movements between meta-populations, as well as egg and larvae drifting phases, still require protection64,65.

While small and semi-passable barriers contribute significantly to habitat fragmentation in the Mekong Basin, the impacts of large dams should never be underestimated, and particularly for dams located in sub-basins hosting unique fish communities, such as those identified by high LCBD in our analysis. For example, a new large hydropower station is set to be constructed at Ganlanba (in the lower reaches of the Lancang River), where the highest LCBD within the basin has been observed. Long-term monitoring will be essential to assess its ongoing impacts and guide adaptive management efforts to minimize ecological impacts. Together, these findings emphasize the need for integrated management strategies that account for both the type and spatial context of barriers to effectively safeguard freshwater biodiversity.

Strategic restoration plan needed to mitigate barrier impacts

River management decisions to prioritize passability improvement of a large dam or remove multiple small barriers must carefully balance ecological benefits, economic costs, and the presence of key species66,67. Large dams, while having the potential to cause widespread flow modification and ecological disruption, are often critical for power generation and regional development. Installing fish passes or bypass channels at such dams can reconnect migratory routes and mitigate some ecological impacts but is often difficult or ineffective at large dams in the tropics68. These solutions can be expensive, with installation and maintenance costs running into millions of dollars, and their effectiveness is site- and species-dependent. For example, species with limited swimming ability may still struggle to navigate even well-designed fish passes69. Moreover, fish passage measures cannot address broader impacts such as habitat degradation (transformation of lotic to lentic habitats), modified sediment transport, altered flow regimes, and degraded water quality67.

In contrast, small barriers often have lower socioeconomical benefits but are numerous and cumulatively can cause extensive habitat fragmentation, especially in areas located in remote tributary sections far from the main channel where high fish community differentiation might occur9,40,53. Removing multiple small barriers can significantly restore connectivity across a broad spatial scale, benefiting a wide range of aquatic species and enhancing ecosystem functions70. However, this approach involves substantial logistical and financial challenges, as the costs of identifying, assessing, and dismantling barriers across an entire basin, or multiple sub-basins, can be high. Moreover, some small barriers serve local needs, such as irrigation or flood control, which complicates their removal due to stakeholder opposition71. In such contexts, where resources are limited, a low-cost method for prioritizing barrier remediation, such as those developed from a fish passage perspective in previous studies can offer a practical solution for guiding restoration efforts70,72.

The presence of species with specific ecological requirements further complicates the decision regarding the optimal management strategy for mitigating impacts of habitat fragmentation from anthropogenic river barriers67. For instance, if a region or sub-basin supports a flagship or economically significant migratory species that requires access to upstream habitats, improving the passability of a large dam might be the most effective action when the ^FI value is high. On the other hand, for sub-basins with diverse, localized species (i.e., regions with high LCBD values) that depend on interconnected habitats, removing small barriers may yield greater cumulative benefits (for species with high FI and relative lower ^FI values).

A cost-effective and ecologically sound river management strategy often requires an integrated approach37. This may involve improving fish passage at key large dams to benefit migratory species while systematically prioritizing and removing small barriers to restore the dendritic structure of the river network21. Decisions should be guided by ecological modeling, cost-benefit analyses, and species-specific data, ensuring that investments align with ecological priorities and stakeholder needs, while minimizing unnecessary expenses37.

Limitations in current study and future direction

Several limitations in current study deserve consideration. First, our analysis is based on modeling approaches. We made assumptions regarding barrier passability in the model, which were informed by previous studies and expert knowledge (see Methods). However, properly assessing the passability of individual barriers requires fish tagging and monitoring experiments at key sites under the range of environmental conditions, particularly flow, that are commonly experienced69. Future validation work should aim to compare model outputs with observed species distributions and movement patterns across selected locations. Additionally, the uneven spatial distribution of barriers across the basin (as shown in Fig. 1) may result in region-specific impacts on fish populations. This spatial bias highlights the need for further investigation into how fish species distributions have changed in response to localized high barrier densities, as reductions in richness and connectivity may be strongly dependent on regional contexts32. Such validation would enhance the robustness of the methodology and support the application of these tools in conservation planning and policy development. In the present study, all fish species occurrence data were used collectively to reconstruct potential distribution ranges, without separating records by time periods. However, historical records of species presence in localities within watersheds can serve as valuable baselines for assessing the impacts of river barrier construction31,73. When combined with information on barrier age, location, and restoration time (if applicable), such records allow for inference of connectivity loss and recovery over time, identification of range contractions, and estimation of species' pristine dispersal potential, offering useful benchmarks for guiding barrier removal or mitigation strategies74.

Second, many fish species in the Middle Mekong require lateral dispersal during the high flow period (June–October) to complete their life cycle, particularly to access inundated floodplains, wetlands, and off-channel habitats for spawning, feeding, and juvenile development72. They also need to be able to return to main channels or survive in isolated lentic waters as floodwaters recede. Yet lateral connectivity onto the Mekong floodplain is increasingly threatened by different types of barriers such as levees, embankments, and pumping stations53,65,72. While our current study focused on longitudinal connectivity, future research should also assess the impacts of such structures on lateral movement of fishes and other aquatic biota. Incorporating lateral connectivity into fragmentation assessments will be essential for capturing the full ecological consequences of river infrastructure development and for designing effective conservation and restoration strategies.

Last, while the present study focuses primarily on barriers as physical impediments to fish movement, we recognize that their ecological consequences often extend beyond fragmentation alone. River barriers, especially larger ones, also modify natural flow regimes and local habitat conditions75,76,77, which can disrupt critical ecological processes such as fish spawning, hatching, and juvenile development19,53,78. For example, many fish species in the Mekong floodplains spawn at the beginning of the flood season53. The flow-regime-adapted spawning behavior can be easily disturbed when dams or sluice gates are operational, potentially delaying migratory or spawning timings4,78. Additionally, many species depend on the drifting of eggs (e.g., Bagarius sp.), larvae, and juvenile stages to downstream nursery and feeding areas to complete their life cycles79, but elevated flow velocity is seldom maintained in impounded regions, and the minimum flow speed required to keep larval fish in the water column is rarely met. Although these factors were not directly addressed in this study, they represent critical dimensions for future research in the Mekong Basin. Furthermore, barrier construction derived habitat modifications may facilitate the invasion of non-native species80, further exacerbating conservation challenges. To advance effective management strategies in the Mekong Basin, future assessments should incorporate a broader suite of hydrological, geomorphological, and ecological variables, such as flow regimes, seasonal hydrological variability, and barrier characteristics19.

Conclusion

We demonstrate that while the effects of individual large dams on habitat fragmentation and fish distribution in the Mekong Basin are greater than for small barriers, the large number of small barriers potentially causes greater cumulative impacts on fragmentation and resultant risks for effective conservation especially for non-migratory fish. We showcase the importance of using complete databases with small, as well as large, river barriers for evaluating conservation risks for barrier development and connectivity restoration in large river basins. We encourage the development of thorough barrier and faunal databases in other large, biodiverse river basins and the application of similar analyses of the ecological impacts of existing and projected barrier distributions.

Materials and methods

River barrier data processing

River barrier data of the Mekong Basin were gathered from the Mekong River Barrier Database41. This database comprises 13,054 barriers classified into five types following standardized terminology (dam, sluice gate, weir, bridge apron, and others) given in that publication70. Ninety seven percent of the barriers in this database were previously unreported in the Mekong Basin compared with other openly accessible databases. To analyse the impact of barrier distribution on fish dispersal across sub-basins, we first extracted all main channel networks (i.e. the section of a river network that links sub-basins to each other), then manually reviewed all barriers in associated Google Earth satellite (Landsat, NASA-USGS) images, and retained barriers that were located in the main river channel, resulting in 3066 barriers including 1648 dams, 488 sluice gates, 840 weirs, 42 bridge aprons, and 48 ‘other’ type barriers (e.g. ramps, bamboo/wooden fences, Fig. 1). Among the 1,130 sub-basins in the Mekong Basin (HydroBASIN level 8), 604 have barriers constructed along their main channels. Dam were categorized into large dam (≥10 m) and small dam (<10 m) based on available information from the Mekong dam database81. While this database provides detailed information on dams, it lacks records for other types of barriers81. The main channel barrier density for each sub-basin was calculated by dividing barrier numbers by main channel length.

Fish species data processing

Fish occurrence data for the Mekong Basin were gathered from a published systematic database28,82, that compiles all fish occurrence data by combining long-term field sampling data and data gathered from all existing books, peer-reviewed articles, gray literature, and online databases via literature review (covering the time period from 1970s to 2020 s). Non-native fish species, farmed species, and species without occurrence descriptions were excluded from the original database, which resulted in the final database comprising a total of 164,878 occurrence points from 1032 species (Fig. 1). The migration type of each species was gathered from FishBase83 and published literature. Each species was categorized into one of six migration types: potamodromous, catadromous, anadromous, amphidromous, non-migratory, and unknown (defined according to Lucas & Baras84; see Supplementary Data 2). Each species was categorized into one of eight conservation status categories according to the International Union for Conservation of Nature (IUCN) Red List of Threatened Species (https://www.iucnredlist.org/): Not Evaluated (NE), Data Deficient (DD), Least Concern (LC), Near Threatened (NT), Vulnerable (VU), Endangered (EN), Critically Endangered (CR), and Extinct (EX). To analyze the spatial patterns of fish distribution, the cumulative fish species richness of each category was calculated at sub-basin level (HydroBASIN level 885,86) and compared across seven ecological regions30,33 (i.e., Himalaya, Upper Mekong, Middle Mekong, Chi-Mun, 3S, Tonle Sap and Delta; see Supplementary Notes for details). The HydroBASINS level 8 was adopted to align with the resolution used in previous studies61,87.

To reconstruct the potential distribution range of each species, the dispersal distance of each species was calculated88. Total fish length, aspect ratio of the caudal fin, stream order inhabited, and timescale allowed for dispersal were employed in this model88,89. Species-specific fish length and aspect ratio of the caudal fin for each species were measured based on figures from published articles or gathered from FishBase. The highest stream order that each species inhabits was gathered using the HydroRIVERS dataset85. A time window of 10 years was used in the model, as appropriate to long-term population level dispersal89,90. The estimated dispersal distance was used as a radius to create buffer zones for each occurrence point44,91. Then, sub-basins (HydroBASIN level 8) that overlapped with buffer zones of all occurrence points were considered as the predicted distribution for that species31,92. The predicted distribution of each species was further refined by cropping to its native sub-basins to exclude predictions that were out of range, based on the occurrence points and expert opinion44. For example, if one species only moves along the main channel of the Mekong and does not inhabit tributaries according to published literature, when the buffered regions fell within some tributaries, then those overlapped sub-basins were removed during this manual refinement step. Furthermore, if a potential sub-basin was linked with valid habitat only through land rather than river channel, then it was removed during the manual process.

To express the compositional uniqueness of each subbasin, and thus to identify unique biodiversity hotspots within the Mekong basin, we calculated the local contributions to beta diversity (LCBD)93. We calculated the LCBDs based on the beta diversity distance matrix calculated using Sørensen-based indices of the Podani family (‘BAT’ package)94.

Fish habitat fragmentation level assessment

To assess the species’ habitat connectivity, we used the Population Connectivity Index (PCI95), which considers structural connectivity (spatial arrangement of the sub-basins and the barriers) and functional connectivity (fish dispersal ability), distribution range (habitat length), and the passability between sub-basins. The PCI considers that a fish population inhabits multiple  sub-basin:

$${PCI}=\sum\limits_{i=1}^{n}\sum\limits_{j=1}^{n}{B}_{{ij}}{c}_{{ij}}\frac{{l}_{i}}{L}\frac{{l}_{j}}{L}{\delta }_{i}{\delta }_{j}$$
(1)

Where n is the number of sub-basins, Bij measures the dispersal capacity of the given fish species between sub-basin i and sub-basin j, cij is the total cumulative passability of river barriers between sub-basins i and j, and li and lj represent the total length of the river network in sub-basin i and j, which are inhabited by the specific species; \({\delta }_{i}\) and \({\delta }_{j}\) represent the presence of a species in subcatchments i and j (0 if species is absent, 1 if present); L represents the total river length of all populations in the basin. Since the PCI is a network-based index, we generated a spatial network accounting for the geographic distribution of subbasins. We used the 1,130 HydroBASINS level 8 sub-basins as nodes of the network, where outlets/entrances between sub-basins are links.

We defined the cumulative passability cij between sub-basins i and j as:

$${c}_{{ij}}=\prod\limits_{m=1}^{M}{p}_{m}$$
(2)

Where \({p}_{m}\) is the passability of the sub-basin m, defined as:

$${p}_{m}=\,\prod\limits_{k}{p}_{k}^{{n}_{k}}$$
(3)

Where \({p}_{k}\) is the passability of each barrier type k, where k = 1 for large dams, k = 2 for small dams, k = 3 for sluice gates, k = 4 for weirs; nk is the number of k-th fragmentation item in the main channel of sub-basin m.

The dispersal capacity Bij is defined as follows:

$${B}_{{ij}}={{PD}}^{{d}_{{ij}}}$$
(4)

Where the probability of dispersal (PD) is based on fish size and swimming capability88, and the distance (dij) is the main channel length between two populations located in sub-basins i and j. The dispersal distance of each species was estimated using the ‘fishmove’ R package88. Next, we defined seven classes of dispersal distance of all species of the entire fish community, by setting the quantiles at 0%, 14.3%, 28.6%, 42.9%, 57.1%, 71.4%, 85.7%, and 100%95. Dispersal distance that fell within the quantile ranges was then categorized into one of the seven class boundaries, ranging from PD = 0.3 to PD = 0.9, which represent fish species’ dispersal probability from low to high95 As this index is applicable to species that occupy two or more sub-basins, species that inhabit only one sub-basin (n = 80) were excluded from further analysis, resulting in a total of 952 fish species for which PCI was calculated using the ‘riverconn’ package96.

We calculated PCI for each species and scenario (i.e., different barrier passability status). We defined the fragmentation index (FI) as the percentage decrease of PCI when barriers are considered. While PCI describes the fragmentation status of populations that is due to barriers and isolation, FI describes the effect of barriers on the geographic range occupied by the species.

$${FI}=\frac{({{PCI}}_{o}-{{PCI}}_{F})}{{{PCI}}_{o}}\times 100$$
(5)

Where, PCIO is the population connectivity index calculated when no barrier is present within the species’ distribution range (i.e. calculating PCI by setting cij = 1). PCIF represents the fragmented status (i.e. all the barriers are considered). When calculating the PCIF, pk for dams (both large and small) was set to 0 in all circumstances20. For semi-passable barriers such as weirs, pk was set to 0.5, representing moderate fish passability31. The FI values range from 0 to 100. When fish populations are located in connected (neighboring) sub-basins with no barriers, or barriers are only located at the edge of the fish’s distribution range such that they do not affect the internal dispersal, FI = 0. Conversely, if the specie’s range is greatly fragmented by barriers, preventing fish from freely moving from one sub-basin to another, the FI value would approach 100. The FI integrates both structural connectivity and functional connectivity into a species-specific, quantitative measure that reflects the percentage loss of connectivity due to barriers. Unlike structural and functional connectivity alone, FI provides a direct assessment of fragmentation impacts on fish populations by comparing with ideal connectivity scenarios.

We also calculated the average fragmentation index (^FI) as the ratio between FI and the number of barriers considered in PCIF. ^FI represents the average fragmentation caused by one barrier and is useful when comparing scenarios where different numbers of barriers are considered. ^FI is calculated by dividing total FI by the number of barriers within a given species’ distribution range, making it useful for comparing scenarios with varying barrier counts.

Additionally, as a metric of network centrality, based on the graph used to calculate PCI, we calculated also the betweenness centrality (BC) for each sub-basin (HydroBASIN level 8) using the ‘igraph’ R package97. BC quantifies the structural importance of sub-basins within a network by measuring their role as intermediary pathways between other nodes. High betweenness sub-basins, serving as pivotal pathways for ecological processes (e.g., fish dispersal and migration), are critical to network resilience and are prioritized in conservation planning due to their disproportionate vulnerability to fragmentation. BC differentiates between central and isolated sites and was shown to be a strong predictor of fish alpha and beta diversity in undisturbed river systems21,40,98. Sites with high BC are located in the main channel, while sites with low BC are located in the tributaries. To assess whether habitat fragmentation differs between tributary-dwelling and mainstream-dwelling species, we calculated the mean BC for each species based on their occupied habitats. We then assessed the relationship between mean habitat BC and the associated fragmentation index (FI).

Contribution of different types of barriers to habitat fragmentation

To assess how different types of barriers affect species with different migration types, and which type of barrier contributed most to habitat fragmentation, four major type barriers: large dam (≥10 m), small dam (<10 m), weir, and sluice gate were selected. Then, the FI for each species was calculated at nine different passability status values (from 0.1, 0.2, 0.3 to 0.9), assuming only one type of barrier is present in each occasion17. After that, the difference between FIlarge-dam, FIsmall-dam, FIweir and FIsluice among all migration types of fish was evaluated.

Data analysis

Spearman’s correlation was used to identify the relationship between fish habitat range and its associated fragmentation index (FI), as well as the relationship between mean habitat betweenness centrality (BC) and FI. Kruskal-Wallis H tests were used to identify if there were significant differences in fragmentation index (FI and ^FI) between different migration types and IUCN conservation status99 (https://www.iucnredlist.org/), followed by a Bonferroni post-hoc test to identify the significantly different groups. ANOVA followed by a Turkey post-hoc test were used to assess the difference in fragmentation status (FI and ^FI) contributed by different types of barriers across migration types.