Hydrodynamics rather than water temperature shapes the macroinvertebrate communities in alpine rivers on the Qinghai–Xizang Plateau

Abstract

Climate change is rapidly altering global glacier-fed rivers, with prevailing models assuming that water temperature increases dictate aquatic community succession. However, this thermal-centric paradigm struggles to explain biodiversity patterns within the extreme topographic and hydrologic gradients of the Third Pole. By analyzing macroinvertebrate assemblages across the Yarlung Tsangpo Basin, we show that hydrodynamic condition, rather than water temperature, primarily influences biodiversity and community assembly in these high-altitude alpine rivers. With hydrodynamic intensity indicated by specific stream power, taxa richness exhibits a universal unimodal response, consistently peaking under moderate hydrodynamic intensity (1 to 10 W m⁻²). Extreme stream power acts as a severe physical filter, driving a stark taxonomic shift from filamentous-gilled species in moderate flows to robust, flat-bodied specialists capable of resisting ultra-high-power environments. These findings demonstrate that localized hydrodynamic modulation—rather than unattainable temperature control—provides a highly viable, climate-resilient strategy to safeguard biodiversity in the Third Pole and analogous alpine river networks worldwide.

Graphical abstract

Highlights

• •Hydrodynamics rather than water temperature drives macroinvertebrate communities.
• •Taxa richness shows a unimodal response to specific stream power, peaking at moderate intensity (1–10 W m⁻²).
• •High stream power acts as a physical filter, selecting for robust, flat-bodied specialists in extreme flows.
• •Modulating hydrodynamic intensity offers a viable, climate-resilient strategy to sustain alpine biodiversity.

1. Introduction

The Qinghai–Xizang Plateau (QXP), which is often referred to as the Third Pole, contains the largest volume of frozen water outside the polar regions [1]. The QXP is also known as the Asian Water Tower, as its abundant water resources serve as the primary source of many of Asia's major rivers [2], e.g., Yangtze River, Yellow River, and Brahmaputra River, which form critical links between nature and human society in both the surrounding areas and downstream regions. The QXP is among the most climate-sensitive regions globally, and its accelerated glacier mass loss represents one of the most pronounced environmental changes worldwide in recent decades [3]. The situation is particularly acute in the southeastern QXP, where glacier mass loss is occurring even more rapidly than elsewhere across the plateau [4]. Indeed, the alpine rivers that flow through the southeastern QXP have experienced significant shifts in their hydrological and sediment regimes [5], as well as increased flood magnitude and frequency [6,7]. These rivers are projected to experience more fluctuating hydrological conditions [8,9] and decreasing meltwater runoff [10] before twenty two century. Given their critical ecological and economic importance, there is an urgent need for effective strategies to protect the QXP's alpine rivers and enhance their ecological functions. This will require a comprehensive understanding of the mechanisms driving the aquatic communities in the region.

Glacial mass loss is one of the most significant factors affecting alpine riverine ecosystems by altering runoff [11,12], as evidenced by responses in macroinvertebrates, often regarded as key indicators of riverine ecosystem health. In fact, some species are losing their habitats due to glacier shrinkage [13], while others are projected to face extinction following the complete disappearance of glaciers [14]. In this regard, the majority of studies were conducted in temperate to arctic regions (average latitude ranging 42° N–79° N, Supplementary Table S1), focusing on the effects of meltwater variation from the perspective of water temperature (Fig. 1). These studies identified a consistent significant increase in local taxa richness as the water temperature increased [[15], [16], [17], [18], [19]]. However, high-altitude tropical alpine rivers between 30° S and 30° N showed no significant relationship between taxa richness and water temperature [20,21], and our investigation in the southeastern margin of the QXP suggested a decreasing pattern distinct from that noted in the aforementioned studies. If the water temperature is not the dominant factor influencing the macroinvertebrate communities in the southeastern QXP, what is shaping these macroinvertebrate communities?

Fig. 1.

Global evidence for temperature–richness relationships in alpine macroinvertebrate communities. Distribution of existing studies examining macroinvertebrate responses to environmental changes in the alpine regions. Green points indicate sites from previous studies, and purple points indicate sites from the present Yarlung Tsangpo River Basin (YTB) study; glacierized areas are shown in blue. Colored lines in the insets show fitted linear relationships between macroinvertebrate taxa richness and water temperature in representative high-altitude alpine regions, including YTB, Peru [21], and Ecuador [20]; temperate alpine regions, including Italy [15], the Rocky Mountains, USA [17], and New Zealand [16]; and Arctic regions, including Fennoscandia, Canada, and Greenland [18,19]. The purple line is based on data from the YTB, and the green lines are based on data from other regions surveyed in previous studies. Taxa richness significantly increased as water temperature rose in the temperate and Arctic regions (P < 0.05), whereas no significant positive relationship was detected in high-altitude regions (P > 0.05), with the YTB exhibiting an inconsistent decreasing pattern. References are shown in the Supplementary Table S1.

To characterize the intrinsic properties of the aquatic communities and elucidate their mechanistic linkages with environmental changes in the QXP's alpine rivers, we used macroinvertebrates as indicator species and conducted comprehensive field surveys in the middle–lower Yarlung Tsangpo Basin (YTB). The YTB, the upper course of the Brahmaputra River, remains largely pristine along the southeastern margin of the QXP. Its meltwater runoff varies spatially due to the heterogeneous glacier distribution across the QXP [22], making it an ideal basin for studying how aquatic communities respond to environmental changes associated with changing meltwater runoff. Biological indices were calculated to characterize macroinvertebrate biodiversity and community structure and to assess their ecological responses to environmental changes. A suite of statistical analyses was conducted to determine the differences in macroinvertebrate assemblages and environmental conditions among the rivers. Community assembly processes were further assessed via the neutral community model (NCM) and modified stochasticity ratio (MST). To identify the primary drivers shaping the macroinvertebrate communities, we performed the Mantel test and permutation multivariate analysis of variance (perMANOVA). Generalized additive mixed models were applied to explore the response pattern of macroinvertebrates to key environmental drivers.

2. Materials and methods

2.1. Overview of the study area

We surveyed macroinvertebrate communities in the middle–lower YTB during late April in 2017 and 2023, when most species were in the larval stage and pupation had little effect on assemblage structure. The Yarlung mainstem and two major tributaries of the middle-lower YTB—the Nyang River (Nyang) and Parlung Tsangpo River (Parlung)—were selected as representative rivers (Fig. 2a). During the study period, runoff sources differed among the three rivers. The Yarlung mainstem was primarily fed by a combination of rainfall and groundwater, while the Nyang and Parlung were primarily fed by meltwater, indicating an increasing gradient of meltwater influence from the Yarlung to the Nyang and Parlung (Fig. 2b) [[23], [24], [25]]. These differences led to pronounced variations in physicochemical conditions among the rivers (Fig. 2d, Supplementary Fig. S1). For example, the meltwater-dominated Parlung had the lowest pH (7.11 ± 0.38) and water temperature (10.03 ± 2.92 °C), consistent with its stronger meltwater influence, whereas the rainfall-dominated Yarlung had the highest electrical conductivity (241.34 ± 44.58 μs cm⁻¹).

Fig. 2.

Study region and environmental setting of the middle–lower Yarlung Tsangpo River Basin (YTB). a, Sampling sites in the middle–lower YTB, a glacier-influenced region of the southeastern Qinghai–Xizang Plateau. The study included 19 sites in the Yarlung Tsangpo River (Yarlung), 15 in the Nyang River (Nyang), and 12 in the Parlung Tsangpo River (Parlung). b, Mean April runoff components of the Yarlung, Nyang, and Parlung rivers, based on published hydrological records for 2003–2016 for the Yarlung [24] and Nyang [23], and 2003–2018 for the Parlung [25]. c, Representative channel morphology and substrate conditions along the longitudinal profiles of the three rivers. d, Median elevation and environmental conditions of the Yarlung, Nyang, and Parlung rivers. Radar charts show median values for each parameter. DO, dissolved oxygen; J, hydraulic slope; D₅₀, grain size of the 50th percentile; ω, specific stream power; S_C, sieving coefficient.

Due to the uneven uplift of the QXP, the middle–lower YTB exhibits spatial heterogeneity in its topographic and hydrodynamic characteristics (Fig. 2c). Alternating narrow and wide valleys characterize the three rivers. The narrow valleys are constrained within the bedrock gorges, with boulders and cobbles primarily constituting the substrate. In the wide valleys, the channels adopt braided patterns, although the substrate type differs among rivers: the Yarlung is characterized by a mixture of fine sand and gravel, while the tributaries are characterized by cobble and gravel. The lower reaches of the tributaries also differ in topography, channel morphology, and substrate composition. The Nyang becomes increasingly braided downstream, with cobble as the dominant substrate, whereas the Parlung develops deeply incised channels with knickpoints, where both boulder and cobble are the primary substrates. Along its longitudinal profile, the Parlung has more knickpoints than the other two rivers [26] and a larger hydraulic slope (0.01 ± 0.01). In terms of the hydrodynamic intensity, the Parlung exhibits the highest specific stream power (14.02 ± 15.66 W m⁻²), while the Nyang shows a moderate level (6.25 ± 14.74 W m⁻²), and most sites in the middle reaches of Yarlung exhibit a low level (1.17 ± 1.65 W m⁻²). The downstream of the confluence of the Parlung and Yarlung, known as the Lower Yarlung, also exhibits strong hydrodynamic conditions, with an average specific stream power of 292.1 W m⁻².

2.2. Field investigation of macroinvertebrates and environmental conditions

2.2.1. Macroinvertebrate sampling

Macroinvertebrates were sampled at 46 sites across the middle–lower YTB (Fig. 2a), including 19 sites along the Yarlung, 15 along the Nyang, and 12 along the Parlung. Sampling protocols followed substrate type [27]. At boulder- or cobble-dominated sites, three replicate macroinvertebrate samples were collected using a 420 μm mesh kick net, each covering 1 m². At sand-dominated sites, three replicate samples were collected using a Peterson grab, each covering 1/16 m². Replicates from each site were summarized and divided by the total sampled area to represent the macroinvertebrate assemblage per unit area.

Specimens were filtered through a 250 μm mesh funnel, preserved in 75% ethanol in the field, and counted and identified at the genus or family level in the laboratory based on their morphological features. All taxa were classified into five functional feeding groups based on their feeding strategies and behaviors [27], and further categorized into 34 trophic groups based on their taxonomic positions (Supplementary Table S2). Biological integrity indices were then calculated for each site (Supplementary Table S3), covering biodiversity (seven indices), taxonomic structure (twelve indices), functional structure (nine indices), and tolerance capacity (five indices).

2.2.2. Environmental conditions survey

Elevation at each sampling site was recorded using a real-time kinematic positioning device (iRTK2, Hi-Target, Guangzhou, China). The physicochemical conditions, including the water temperature, electrical conductivity, pH, and dissolved oxygen, were measured using a YSI EXO water quality platform (Xylem, Washington D. C., USA). Hydrodynamic variables were measured in the field. The average hydraulic slope (J) was calculated from the difference in water surface elevation over a 3 m distance using a tube. The average velocity (v) was measured using a portable acoustic Doppler velocimeter (FlowTracker2, SonTek, SanDiego, USA), while the water depth (h) was measured using a sliding staff. Substrate characteristics were assessed using grain-size distributions (GSDs), measured by different methods depending on the substrate type. For boulder or cobble beds, we divided a 5 × 5 m² area into nine equal squares, and grain counts were conducted within five squares at the center and edges of the area. GSDs were then calculated from the volume proportions of grains in various diameters, using average counts. For sandy beds, the substrate was collected from the field, dried, and sieved in the laboratory. Representative grain size percentiles—16th (D₁₆), 50th (D₅₀), and 84th (D₈₄)—were obtained from the GSD, with D₅₀ representing the median particle size. The sieving coefficient, which indicates the uniformity of the substrate grain size, was calculated as follows:

$$S_C=\sqrt{\frac{D_{84}}{D_{16}}}$$ (1)

where S_C is the sieving coefficient.

To characterize the hydrodynamic intensity, we calculated specific stream power (ω), a compound variable of hydraulic slope and discharge [28]. Stream power is related to the sediment transport rate and grain size [29] and reflects the channel sensitivity to erosion and deposition processes [30]. Specific stream power was calculated [28,31]:

$$\omega =\frac{\mathrm{\gamma }QJ}{w}=\mathrm{\gamma }qJ$$ (2)

where γ is the specific weight of the water (9800 N m⁻³), Q is the discharge crossing the section, w is the width of the river, q is the unit discharge, and J is the average hydraulic slope.

2.3. Environmental conditions and macroinvertebrate community characteristics analysis

All statistical analyses were conducted using R software (version 4.3.3). Differences in environmental conditions and biological integrity indices among the three rivers were evaluated using standard protocols, following pre-test for independence, normality, and homogeneity of variance. Variables and indices that satisfied these assumptions were analyzed using the one-way analysis of variance function in stats [32]. When overall differences were significant, we used Fisher's least significant difference test function in the PMCMRplus package to pairwise differences among rivers [33]. Variables that did not meet parametric assumptions were analyzed using the Kruskal–Wallis test with the kruskal.test function in the stats package [32], followed by pairwise Mann–Whitney U tests with the wilcox.test function in the stats package [32].

We analyzed macroinvertebrate assemblages in terms of the biodiversity, community structures, and assembly processes. Biodiversity was quantified at multiple spatial scales, including local α-diversity, basin-scale β-diversity, and basin-scale γ-diversity. These were represented by local taxa richness (S_α), the Whittaker index, and the total taxa richness of the basin's species pool (S_γ), respectively. The Whittaker index of each river was calculated as ${S_{\gamma }/\overline{S_{\alpha }}}^{-}1$ [34], where $\overline{S_{\alpha }}$ is the average local taxa richness of the river basin. To minimize the scale effects induced by the basin size and the potential sampling insufficiency, we applied site-based rarefication to the Whittaker index and the total taxa richness of the basin's species pool for the β- and γ-diversity, respectively. Regional richness was rarefied using the iNEXT function in the iNEXT package [35,36], while the Whittaker index was rarefied using the Michaelis–Menten equation [37].

We analyzed community structure from both the taxonomic and functional perspectives. Differences among the Yarlung, Nyang, and Parlung were visualized using principal coordinates analysis (PCoA) based on Bray–Curtis dissimilarity, implemented with the pcoa function in the ape package [38]. We tested among-river differences using perMANOVA with 999 permutations, implemented with the adonis2 function in the vegan package [39]. When the overall difference was significant (P < 0.05), we conducted pairwise comparisons to identify which river pairs differed significantly after adjustment for multiple testing (adjusted P < 0.05).

For the assembly process, we first used the NCM [40] to assess the potential influence of stochastic processes on macroinvertebrate assemblages in rivers. The migration ratio parameter of the NCM was estimated using the nlsLM function in the minpack.lm package [41]. A positive R² for the NCM indicates that the stochastic process contributed to assembly. To quantify the relative role of the deterministic and stochastic processes in the macroinvertebrate assembly, we then calculated a null model-based MST [42] using the taxonomic normalized stochasticity ratio function in the NST package [43]. An MST value below 0.5 indicates that the deterministic process dominates the assembly; otherwise, the stochastic process is dominant.

2.4. Ecological response of macroinvertebrate communities to key drivers

We used correspondence analysis to identify the primary environmental variables associated with the differences in macroinvertebrate community structures among the three rivers. Genus-level taxonomic data were first assessed using detrended correspondence analysis (DCA), implemented with the decostand function from the vegan package [39]. Because the longest DCA gradient exceeded the threshold value of 4, canonical correspondence analysis (CCA) was used for subsequent analysis, using the cca function in the vegan package [39]. Before CCA, environmental variables were selected in two steps to reduce collinearity and improve model parsimony. First, we used perMANOVA to select environmental variables that could significantly explain the dissimilarity in community structures among the rivers [39]. Second, to minimize collinearity, we performed Spearman's rank correlation analysis and retained variables with pairwise correlation coefficients below 0.4, using the cor function in the stats package [32]. The results of the perMANOVA and Spearman correlation analyses are shown in Supplementary Fig. S2. We then performed CCA to identify the key environmental factors explaining differences in community structure among the rivers. Mantel tests, implemented with the mantel function in the vegan package [39], were performed to assess the correlations between macroinvertebrate assemblage and environmental variables for each river.

Site-level ecological responses of macroinvertebrate assemblages, including taxon richness and community structure, to key environmental variables were analyzed with a series of generalized additive mixed models, implemented with the gamm function in the mgcv package [[44], [45], [46], [47], [48]]. When modeling the ecological response at the scale of the entire middle–lower YTB, the river basin was included as a random effect (Supplementary Table S4). Taxa richness was modeled using a Poisson distribution with a log link. The relative abundance of dominant groups, including Ephemeroptera, Plecoptera, and Trichoptera (EPT), chironomids, and functional feeding groups, was modeled using a Gaussian distribution with an identity link. Responses of the representative taxa, defined as those with mean relative abundance greater than 10% across the middle–lower YTB, to the primary environmental drivers were analyzed via generalized additive models implemented with the gam function in the mgcv package [[44], [45], [46], [47], [48]]. These generalized additive models were fitted to taxa occurrence records using a Poisson distribution with a log link.

3. Results

3.1. Community structures and biodiversity of macroinvertebrates

Chironomids, EPT taxa, and Oligochaeta taxa accounted for 85% of the total macroinvertebrate abundance in all three rivers, with most species being detritivores (gathering collectors and shredders), indicative of a species pool with functional redundancy adapted to the limited food sources on the QXP (Fig. 3a, Supplementary Table S2). However, the species pools were distinct among the rivers, with shared taxa accounting for only 10% of all taxa (Supplementary Fig. S3). From the Yarlung to the Nyang and Parlung rivers, the relative abundance of oligochaetes decreased, whereas that of EPT taxa increased. In the meltwater-dominated Parlung River, the EPT taxa accounted for more than 50% of the taxa abundance. The composition of EPT taxa also varied among rivers: Ephemeroptera and Plecoptera dominated the EPT taxa in the Parlung, while Trichoptera outnumbered Plecoptera in the Nyang and surpassed Ephemeroptera in the Yarlung (Supplementary Fig. S4). Biodiversity at both the local and basin scales showed consistent patterns across the three rivers. The Nyang exhibited the substantially highest γ-diversity at the basin scale, followed by the Parlung and Yarlung. It also had higher α- and β-diversity than the other two rivers, although these differences were less pronounced than for γ-diversity.

Fig. 3.

Functional composition, taxonomic composition, and diversity of macroinvertebrate communities across the Yarlung, Nyang, and Parlung rivers. a, Functional feeding group (upper parts of rings) and taxonomic composition (lower parts of rings) of macroinvertebrates in the middle–lower Yarlung Tsangpo River Basin. Bar size indicates average relative abundance in each river basin. Hydrodynamic intensity among rivers was compared using specific stream power (Supplementary Fig. S1). b–d, Biodiversity in local and basin scales. At the local scale, α-diversity (b) is represented by the richness of taxa at sampling sites. Boxes show the interquartile range, with lower and upper bounds corresponding to the 25th and 75th percentiles. On the basin scale, β-diversity (c) is represented by the Whittaker index, and γ-diversity (d) is represented by the total taxa richness of the basin's species pool. Error bars in panels c and d indicate 95% confidence intervals.

3.2. What primarily influences macroinvertebrate communities in the rivers?

To elucidate the mechanisms underlying the identified variations in taxa abundance across and within the river basin, we examined the macroinvertebrate assembly processes and community structures. The limited fit to the NCM (R² < 0.2), together with the MST of less than 0.5, suggests the primary contribution of the deterministic process to the community assembly across the three rivers (Fig. 4a). This pattern was consistent with the finding of a similar functional structure but distinct taxonomic structures along the environmental gradients (Fig. 4b and c). Among the environmental variables examined, elevation and specific stream power, rather than water temperature, explained differences in taxonomic structures among the rivers. Taxonomic structures were better discriminated by the specific-stream-power gradient, indicating that hydrodynamic intensity exerted a stronger influence on them.

Fig. 4.

Assembly processes and environmental structuring of macroinvertebrate communities. a, Neutral community models (NCMs) fitted to macroinvertebrate assemblages across the Yarlung, Nyang, and Parlung rivers. Solid lines indicate the best-fit NCM, and dashed lines represent 95% confidence intervals around model predictions. Genera occurring more or less frequently than predicted by the NCM are shown in blue and orange. R² indicates the model fit. The modified stochasticity ratio (MST) indicates the relative contribution of deterministic and stochastic processes; MST < 0.5 indicates stronger deterministic assembly, whereas MST > 0.5 indicates stronger stochastic assembly. b, Principal coordinates analysis (PCoA) based on Bray–Curtis dissimilarity showing functional community structures. Sites of each river were circled out with 95% confidence dotted ellipses. Functional structure did not differ significantly among rivers (permutation multivariate analysis of variance [perMANOVA]: R² = 0.034, F = 0.76, P = 0.679). c, Canonical correlation analysis (CCA) based on Bray–Curtis dissimilarity illustrating the taxonomic structures of macroinvertebrate communities. Sites of each river were circled out with 95% confidence dotted ellipses Taxonomic structures differed significantly among rivers (perMANOVA: R² = 0.128, F = 3.144, P < 0.001; pairwise Adonis tests: adjusted P = 0.001 for all tests). Asterisks indicate significance levels for environmental variables: ∗P < 0.05; ∗∗P < 0.01. d, Mantel correlations between genus-level taxonomic structure and geographic or environmental variables, together with pairwise Spearman correlations among explanatory variables. Edge width indicates Mantel's r, and edge color indicates Spearman's correlation coefficient. H: elevation; WT: temperature; DO: dissolved oxygen; EC: electrical conductivity; S_C: sieving coefficient; J: hydraulic slope; ω: specific stream power.

Within the river basin, the deterministic process dominated the macroinvertebrate assembly in both the Nyang and the Yarlung (R² < 0.2, mean MST < 0.5); however, stochastic processes contributed slightly more to the assembly in the Parlung (R² > 0.3, mean MST = 0.53; Supplementary Fig. S5). The Mantel test was applied to explore the factors influencing the taxonomic structure of macroinvertebrates in each river. Taxonomic structure of the Nyang was significantly correlated only with hydrodynamic conditions, whereas that of the Yarlung and Parlung was significantly correlated with the spatial factor (i.e., elevation) and local environmental factors. Among the local environmental factors influencing the taxonomic structures of macroinvertebrates, grain size (D₅₀) was the only shared factor across the three rivers.

3.3. How do macroinvertebrates respond to hydrodynamic gradients?

Although biodiversity and taxonomic structure differed among rivers along the hydrodynamic intensity, their responses to the hydrodynamic intensity were consistent within each river basin (Fig. 5). Within each basin, the taxa richness (α-diversity) showed a unimodal response pattern and peaked at a relatively moderate specific stream power. By contrast, dominant taxonomic groups showed divergent responses: the relative abundance of the EPT taxa displayed a unimodal response to the specific stream power, while the chironomids exhibited the reverse pattern. This divergence was likely due to the different functional strategies among the taxa. Most chironomids are collectors, whereas the EPT taxa exhibit higher functional diversity—that is, they are collectors, shredders, and scrapers (Fig. 2a). Moderate flow intensity may increase substrate heterogeneity (Supplementary Fig. S6), promote algal and macrophyte growth, and enhance coarse particulate organic matter accumulation, thereby benefiting scraper and shredder guilds (Supplementary Fig. S7). Functional differentiation within EPT taxa may further reduce niche overlap. For example, among mayflies, Baetidae taxa are swimmers that collect organic particles, whereas Heptageniidae taxa typically attach to cobbles or boulders and scrape algae. These broader functional niches and lower niche overlap may explain why EPT taxa dominate in abundance under relatively moderate flow intensity. However, the bimodal response patterns of EPT taxa richness and relative abundance emerged across the entire middle–lower YTB, which could result from a phase shift in ecological responses among basins due to distinct taxonomic structures shaped by hydrodynamic conditions. Mayflies and stoneflies capable of resisting strong flow comprised the species pool of the Parlung, whereas that of the Nyang was comprised of caddisflies, mayflies, and chironomids better adapted to moderate-level stream power, and the Yarlung was mainly comprised of chironomids and oligochaetes adapted to low-level stream power (Supplementary Fig. S8). As these phase shifts in ecological responses were aggregated across rivers, bimodal response patterns emerged at the macroscale.

Fig. 5.

Macroinvertebrate community responses to hydrodynamic intensity. Generalized additive mixed models showing relationships between hydrodynamic intensity, represented by specific stream power (ω), and macroinvertebrate community metrics in the middle–lower Yarlung Tsangpo River (YTB). Grey points show biological integrity indices for all sampling sites, and colored points indicate sites from the Yarlung (blue), Nyang (green), and Parlung (red) rivers. Grey curves and shaded bands show fitted relationships and 95% confidence intervals for all sites; colored curves and dashed lines show river-specific fitted relationships and 95% confidence intervals. Taxonomic structure is represented by the relative abundance (RA) of Ephemeroptera, Plecoptera, and Trichoptera (EPT) taxa and chironomids (CHI), which were the dominant species in the middle-lower YTB (average relative abundance > 30% in the species pool). Functional structure is represented by the relative abundance of collector, shredder, scraper, and predator feeding groups.

We propose a conceptual model of community succession along the specific-stream-power gradient, with substrate type treated as a key habitat filter (Fig. 6a). Substrate type was influenced by hydrodynamic intensity (Supplementary Fig. S6) and was the only shared influential factor for variation in taxonomic structure within each river basin. Under low stream-power conditions (ω < 1 W m⁻²), the substrate type exerted primary effects on community structure. Chironomids and oligochaetes, characterized by slender morphology and low oxygen demand, prevailed on the sandy and gravel beds, while caddisflies capable of building cases outcompeted oligochaetes on the pebble beds. As the stream power increased, the influence of flow-induced disturbance on the assemblages strengthened. The local taxa richness (α-diversity) peaked under moderate hydrodynamic conditions (ω = 1–10 W m⁻²), with the EPT taxa showing the most significant increases in richness. Under high-energy conditions (ω > 10 W m⁻²), the dominant taxa included specialized stoneflies and mayflies with better adaptability to high stream power. Moreover, under ultra-high-power flow (ω > 100 W m⁻²), only a few specialized taxa capable of resisting the flow persisted, including the mayfly Epeorus, the blackflies Prosimulium and Simulium, and the chironomids Orthocladius. Taxa turnover along the gradients of the specific stream power was also pronounced at the genus level within the same families. Genus-level turnover was particularly evident within Heptageniidae along the stream-power gradient (Fig. 6b). Under moderate flow intensity, Heptagenia was dominant, with an elongated body and filamentous abdominal gills. With increasing flow intensity, the dominant taxa shifted to Rhithrogena and Epeorus, which feature shorter, flatter bodies and enlarged oval gills. These morphological traits reduced shear stress and enhanced attachment by increasing the contact area with the substrate. Under high-energy conditions (ω > 10 W m⁻²), Epeorus prevailed, likely because their robust body and gill structures conferred resistance to hydraulic abrasion and turbulence, enabling survival in torrential conditions. A similar pattern was observed among chironomids. Polypedilum prevailed under low stream-power conditions, consistent with its slender body and hemoglobin tolerance of low-oxygen fine sediments. As stream power increased, Orthocladius, which possesses dense setae along its body, prevailed in abundance due to its superior swimming performance in lotic environments.

Fig. 6.

Conceptual model of hydrodynamic filtering of macroinvertebrate communities. a, Conceptual model showing expected shifts in macroinvertebrate communities along gradients of substrate condition and flow intensity in the middle–lower Yarlung Tsangpo River (YTB). The upper left shows the names of families or genera that each taxon belongs to. b, Typical Heptageniidae taxa along the gradient of hydrodynamic intensity, with dotted lines outlining the shape of gills. Heptagenia dominated under moderate flow intensity, whereas dominance shifted towards Rhithrogena and Epeorus as flow intensity increased, with Epeorus predominating under high-flow conditions.

4. Discussion

4.1. Why does hydrodynamics shape macroinvertebrate communities?

Previous studies in temperate alpine regions have emphasized the water-temperature-driven succession of macroinvertebrate communities along meltwater contribution gradients [49,50]. In the middle–lower YTB, however, water temperature was not significantly associated with macroinvertebrate community structure (Fig. 1). This pattern contrasts with observations from temperate alpine systems, where chironomids become more dominant with increasing meltwater runoff [51]. Instead, EPT taxa became more abundant with increased meltwater runoff and so prevailed in the meltwater-dominated Parlung, while chironomids exhibited comparable abundance across the three rivers. Together, these results point to hydrodynamic conditions, rather than water temperature, as the main environmental influential factor of macroinvertebrate community structure across the YTB.

We attributed this discrepancy in terms of the temperate patterns to regional differences in intrinsic properties. The YTB receives more intense solar radiation [52] and experiences higher baseline temperatures [53], which may buffer the thermal effects of meltwater on the river ecosystem. In addition, the middle–lower YTB lies in a geomorphic transition zone with abundant water resources and strong hydrodynamic forcing. The stability of habitat conditions, which is essential for habitat suitability, is primarily governed by hydrodynamic processes in river ecosystems. Our findings revealed that while the α- and β-diversity in the Parlung were comparable to those in the Nyang, the γ-diversity of the Parlung was significantly lower. This suggests that intense hydrodynamic processes have destabilized certain habitats in the Parlung, rendering them unsuitable for supporting diverse macroinvertebrates.

Strong hydrodynamic processes can affect community assembly indirectly by altering flow intensity and river topography, as shown in the Parlung. The assembly process was slightly more stochastic in the Parlung, while pairwise dissimilarity in taxonomic structure was more strongly correlated with elevation, indicating that dispersal limitation hinders species from tracking environmental gradients longitudinally. The steep elevation gradient along the Parlung was largely attributed to the dense formation of knickpoints in the river profile [26], reflecting the river's self-regulation in response to the strong hydrodynamic processes. These conditions may amplify the direct and indirect effects of hydrodynamics on macroinvertebrate community structure.

Food-source availability may further mediate the effects of hydrodynamic conditions on macroinvertebrate communities. Biomass and diversity are often co-limited by food availability [54,55], and the effects of food resources on aquatic biota are influenced by hydrodynamic conditions in river ecosystems [56]. Primary productivity of the YTB is low due to the cold, oligotrophic conditions, and the dominance of detritivores among functional feeding groups suggests that detritus serves as the primary food source. Competition for a limited food source limits the community size. Moderate flow can erode soil from riverbanks and transport it into the river, increasing nutrient inputs for algal and macrophyte growth, enhancing the diversity and biomass of non-detritivores, and increasing food web complexity [57]. As flow intensity increases further, however, greater nutrient inputs may be offset by vegetation damage [58] and reduced retention of organic particles, ultimately reducing the food supply for macroinvertebrates. These patterns suggest that, in oligotrophic rivers with limited exogenous nutrient inputs, hydrodynamic conditions significantly affect food-source availability [59]. Moderately intense hydrodynamic conditions can improve the river nutrient conditions and enhance the habitat suitability. Overall, hydrodynamic processes influence habitat stability, availability, and suitability, leading to pronounced effects on macroinvertebrate communities in the YTB.

4.2. Potential implications

Our findings have two major implications. First, we identified hydrodynamic conditions, rather than water temperature, as the primary factors shaping macroinvertebrate communities in the middle–lower YTB. This finding may offer additional insight into the ecological responses in other high-altitude glacier-fed rivers. For instance, in the Andes’ glacier-fed rivers, where baseline water temperatures are also higher than those in temperate regions [20], macroinvertebrate communities are highly sensitive to variations in discharge associated with shifts in the meltwater runoff [60], whereas the positive relation between taxa richness and water temperature did not exhibit statistical significance (for example, Peru and Ecuador in Fig. 1). Jacobsen et al. [20] also found that Chironomidae composition differed between glacier-fed rivers in the Ecuadorian Andes and rivers in temperate regions. Our findings suggest that hydrodynamic conditions may be the primary factor influencing macroinvertebrate communities when the effects of water temperature are buffered. This provides an insight into assessing the state of glacier-fed river ecosystems as meltwater runoff changes in high-altitude regions. In many high-altitude glacier-fed rivers, peak meltwater runoff has not yet occurred, suggesting that the hydrodynamic conditions in these rivers may continue to intensify before 2050 [61]. When attempting to mitigate global warming and delay the peak of meltwater runoff, optimizing flow regimes may be a more effective strategy for addressing the ecological impacts of these changes in the coming years.

We further observed a universal unimodal response pattern in taxa richness and functional diversity to specific stream power across river basins, which could inform sustainable river management. Moderate specific stream power corresponds to intermediate flow intensity and disturbance of the riverbed. Such conditions can regulate interspecific competition within the same functional feeding group, reducing overcompetition and facilitating coexistence among taxa. Similar unimodal responses to the hydrodynamic intensity have been reported for other biological groups in the YTB. For example, suitable habitat for the endemic fish Schizopygopsis younghusbandi peaks under moderate flow intensity [62]. This suggests that modifying hydrodynamic intensity relative to the basin's moderate level may be an effective strategy for conserving the QXP river ecology. Rivers along the southeastern margin of the QXP are projected to experience drastic hydrological shifts in the coming decades [8,63], potentially simplifying the trophic structure [64]. In the YTB, intensified precipitation and glacier mass loss are expected to alter runoff composition and increase discharge [8,65]. These changes may increase the frequency of natural hazards, such as floods and glacial lake outburst floods [6,66], placing pressure on both ecosystems and human society. From the perspective of optimizing the hydrodynamic conditions, the hydropower development in the middle–lower YTB over the coming decades [67], in addition to contributing to carbon-neutrality goals [68], could temporarily restore the water during floods, dissipate the flow energy, modify the flow toward moderate intensity, and benefit ecological restoration by increasing the complexity of the community structure and improving the habitat suitability for aquatic biotas [62,69].

This study has several limitations. First, we did not disentangle the ecological responses of multiple biotic communities and their biotic interactions from hydrodynamic conditions, which can strongly influence the complexity of an abiotically driven ecosystem [70]. Second, due to data limitations, we were unable to evaluate the temporal consistency of hydrodynamic conditions as the primary driver of macroinvertebrate communities, relative to water temperature. Future studies should apply interdisciplinary approaches to examine how biotic interactions respond to hydrodynamic variability and conduct long-term investigations of the QXP to better characterize ecological responses to key environmental changes over time. Such efforts would provide a more integrated perspective for understanding how alpine river ecosystems respond to environmental changes and, therefore, aid in the protection and sustainable management of these rivers in fluctuating environments.

5. Conclusion

This study shows that hydrodynamic conditions, rather than water temperature, were the primary factor shaping macroinvertebrate communities in the middle–lower YTB. Taxa richness and functional diversity consistently exhibited a unimodal response to hydrodynamic intensity, as indicated by specific stream power. Moderate hydrodynamic intensity might enhance the stability, availability, and suitability of the habitats for diverse biotas. This study elucidates a clear conservation strategy for glacier-fed rivers in high-altitude regions and high-energy mountain rivers. Rather than focusing enormous effort on controlling temperature increases, locally modifying hydrodynamic processes may offer a more feasible and effective pathway to sustain ecosystem function in fluctuating environments. Moreover, our findings provide a scientific basis for ecological protection and sustainable river use in the YTB and can be extended to other alpine rivers affected by climate change and anthropogenic activities.

CRediT authorship contribution statement

Yaqi Luo: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Conceptualization. Mengzhen Xu: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization. Xiongdong Zhou: Writing – review & editing, Investigation, Formal analysis. Dongfeng Li: Writing – review & editing. Xudong Fu: Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Dr. Mengzhen Xu is an Associate Editor for this journal and was not involved in the editorial review or the decision to publish this article.

Appendix A. Supplementary data

The following is the Supplementary data to this article: