The sampling region (31.3 to 31.7° N, 109.9 to 110.7° E, and 398 to 3105 m above sea level) was located in the northwestern part of Hubei province, China (SI Appendix, Supplementary Figure S1). Nestled in the embrace of diverse climatic influences, The Shennongjia Forestry District (SNJFD) features a subtropical montane climate, characterized by a moderate annual mean air temperature ranging from 10 to 15°C and an annual precipitation of 800 to 1,200 mm (Zhao et al., 2005, 2018). Its unique ecological niche hosts a tapestry of vegetation types, including evergreen broad-leaved forest belt, evergreen and deciduous broad-leaved montane forest belt, deciduous broad-leaved forest belt, coniferous and broad-leaved montane forest belt, subalpine coniferous forest belt, shrub and grass, all contributing to its rich montane forest zone (MFZ) ecosystem (Zhao et al., 2005, 2018). Within the sampling area, we selected a total of seven water bodies: Xiangxi River, Nanhe River, Pingqian Reservoir, Dajiuhu Lake, Duhe River, Yandu River, and Songluo River. These seven water bodies are representative in terms of their types, hydrological conditions, and ecosystems, comprehensively encompassing all the characteristics of water bodies found in the Shennongjia Forestry District (SI Appendix, Supplementary Figure S1; Supplementary Table S1). These water bodies fall into four groups (D1: Xiangxi River. D2: Nanhe River. D3: Pingqian Reservoir, Dajiuhu Lake and Duhe River. D4: Yandu River and Songluo River). Our 53 sampling sites are distributed in seven water bodies, including 10 in Xiangxi River, 8 in Nanhe River, 2 in Pingqian Reservoir, 20 in Dajiuhu Lake, 4 in Duhe River, 5 in Yandu River, and 4 in Songluo River (SI Appendix, Supplementary Figure S1; Supplementary Table S1). We selected flood and blizzard as the main EWEs, sampling periphytic algae in four periods, include period before flood (BF, April 2023), period after flood (AF, August 2023), period before blizzard (BB, September 2023) and period after blizzard (AB, December 2023). During the period from BF to AF, the forest area experienced two floods, with the flood in August 2023 being of a larger scale. During the period from BB to AB, the forest area experienced three blizzards, with the blizzard in December 2023 being more intense.

Periphytic algae pieces were removed from the substrate using a spade and divided into four 10 cm² pieces with a knife. Each small piece was placed into a white porcelain dish and stirred evenly with 100 mL of pure water. Samples were collected with three 10 cm² algal samples serving as duplicate samples each time. These duplicate samples were stored at 4°C for species identification and cell density counting.

Each 10 cm² periphytic algal sample was individually filled into a 50 mL sample bottle and fixed by adding Lugol’s solution after being diluted with pure water to 50 mL. Each sample was divided into two equal parts: One part of the sample was treated with concentrated nitric acid (HNO₃) according to the technical guidelines for water ecological monitoring - aquatic organism monitoring and evaluation of rivers (on trial) (MEEPRC, 2024) to remove organic matter that may cause interference. The cleaned sample was then made into permanent slides for the identification and enumeration of Bacillariophyeceae. The other part of the sample was prepared into temporary slides, also following the same technical guidelines, for the identification and enumeration of Cyanobacteria, Chlorophyta, Charophyta, and Cryptista. At least 50 visual fields were surveyed for each slide with a light microscope (Olympus CX23 microscope, Tokyo, Japan) at 400× magnification, and all algae were determined to the species level. For the identification of algae, the identification monographs of Hu and Wei (Hu and Wei, 2006) were used. Additionally, AlgaeBase (Guiry and Guiry, 2024) was consulted to ensure the use of correct and modern names, taxonomy, and updated species lists in accordance with current concepts and understanding.

Water samples were collected synchronously for chemical analysis. Water samples (1 L) were collected at 0.35 m depth using a water collector and poured into sampling bottles to be stored at 4°C before test.

Water temperature (WT), pH, turbidity (NTU), specific conductivity (SPC), Oxidation-Reduction Potential (ORP) and dissolved oxygen (DO) were monitored by YSI (proplus, Xylem, USA). The average water velocity (AV), maximum water velocity (MV) and flow rate (Q) were monitored by flowmeter (Flowatch, Switzerland). The water surface width (WW) was measured using a laser rangefinder (DDT8M, Delixi, China). The surface light intensity (SL) and underwater light intensity (UL) were measured using an underwater quantum flux meter (MQ-510, Apogee, USA). Total carbon (TC), inorganic carbon (IC), and total organic carbon (TOC) were measured using a total organic carbon analyzer (multi N/C 3100, Germany). The suspended sediment concentration (SSC) was determined using the centrifugal sedimentation method. Other environmental parameters of water sample such as total nitrogen (TN), nitrate nitrogen (NO₃^--N), ammonia nitrogen (NH₃-N), total phosphorus (TP), orthophosphate (PO₄^3--P), potassium permanganate index (COD_Mn), phytoplankton Chlorophyll-a (Chla) were determined according to APHA (APHA, 2017).

All statistical analyses were performed in the R software (V4.4.0; http://www.r-project.org/), and all figures were generated by “ggplot2” R package. Phyla relative abundances less than 1% in all samples were combined as “Others,” and phyla with abundances >5% were defined as dominant phyla. Species relative abundances less than 1% in all samples were combined as “Others,” and species with abundances >5% were defined as dominant species. Richness index (alpha diversity) was calculated using the “diversity” function (vegan package). One-way ANOVA was used to test the effects of EWEs and waterbodies on measured variables. Principal Co-ordinates Analysis (PCoA) was performed with R package Vegan (Dixon, 2003). The dissimilarity test was carried out by nonparametric multivariate statistical tests with the “xadonis” function (999 permutations) in the “vegan” R package. Algal community similarities or dissimilarities (beta diversity) were calculated by the Bray-Curtis index via the “vegdist” function in the “vegan” R package, while environmental distance was calculated by the Euclidean distance based on matrix of measured environmental variables, including ALT, WT, DO, SPC, pH, ORP, AV, MV, Q, WW, SL, UL, TN, NO₃^--N, NH₃-N, PO₄^3--P, COD_Mn, TC, IC, TOC, NTU, SSC. Altitude distance was obtained by calculating the altitude difference between sample sites. Geographic distance was obtained by calculating the direct distance between sample sites.

Co-occurrence networks were constructed for algal communities by Spearman correlations using the “corr.test” function in the “psych” R package. Spearman correlation results were filter by the thresholds r > 0.75 and false discovery rate < 0.05. Network graphs were generated by using the “igraph” R package, and network parameters were extracted, including nodes, edges, degree, eigenvector centrality, complexity (linkage density; degree/node), diameter, transitivity, and modularity. Network natural connectivity was estimated by “attacking” nodes or edges in the static network (Gravel et al., 2016; Wu et al., 2021). Importantly, robustness test was a powerful method to measure the network stability (more specifically resistance) through natural connectivity changes against node or edge removal, organized in a decreasing order of nodes’ betweenness or edges’ weight.

Because strong collinearity occurred among particular environmental factors, we used cluster analysis to assess the collinearity or redundancy of environmental variables by the “varclus” procedure in the “Hmisc” R package before further analyses (Wang et al., 2017). Only one variable was selected for those clustered closely (Pearson’s R² > 0.7) as the representative variable. The partial Mantel test was carried out using the “mantel.partial” function (999 permutations) in the “vegan” R package to evaluate relationships between algal community and environmental variables. The explanations of environmental factors were assessed by using partitioning of algal composition variance analysis based on canonical correlation analysis. A structural equation model was constructed by “piecewiseSEM” package (Lefcheck, 2016). The structural equation model included altitude, environmental physical factors (ALT, WT, DO, SPC, pH, ORP, AV, MV, Q, WW, SL and UL), nutrients (TN, NO₃^--N, NH₃-N, TP, PO₄^3--P, COD, Chl.a, TC, IC, TOC, NTU and SSC), community compositions (relative abundance of different phyla, mainly Bacilloriopyceae, Cyanobacteria and Chlorophyta), diversity (represented community diversities). The best model was selected according to the lowest value of AIC, a nonsignificant χ² test (0.05 < P ≤ 1.00) and the root mean squared error of approximation (RMSEA; 0 ≤ RMSEA ≤ 0.05; 0.10 < P ≤ 1.00). The altitude, physical, nutrient and diversity were represented by their PC1 scores, which explained 99.99, 78.53, 89.35 and 99.81% variance of corresponding environmental groups. Composition was represented by PCoA axis1, the first component of PCoA analysis.

To quantitatively evaluate the impact of extreme weather events, we compared environmental parameters before flood (BF), after flood (AF), before blizzard (AB) and after blizzard (AB) using pairwise Welch’s t-tests (α = 0.05).

The flood event fundamentally altered the hydrodynamic and physical structure of the water bodies. Parameters including water temperature (WT), average water velocity (AV), maximum water velocity (MV), flow rate (Q), water surface width (WW), surface light intensity (SL), underwater light intensity (UL), turbidity (NTU), and suspended sediment concentration (SSC) were significantly elevated in the AF period compared to BF. Concurrently, ammonia nitrogen (NH₃-N), total carbon (TC), and total organic carbon (TOC) also increased. In contrast, concentrations of dissolved oxygen (DO), pH, oxidation-reduction potential (ORP), nitrate nitrogen (NO₃^--N), total phosphorus (TP), orthophosphate (PO₄^3--P), potassium permanganate index (COD_Mn), phytoplankton chlorophyll-a (Chla), and inorganic carbon (IC) were significantly lower after the flood. Altitude (ALT), specific conductivity (SPC), and total nitrogen (TN) showed no significant change (Figure 1; SI Appendix; Supplementary Figure S2).

The blizzard event induced a pronounced thermal and biogeochemical shift. In the AB period, only DO was significantly higher. A suite of parameters was significantly lower, including WT, pH, ORP, TN, NO₃^--N, NH₃-N, TP, PO₄^3--P, COD_Mn, Chla, TC, IC, TOC, and NTU. Physical and hydraulic variables such as ALT, SPC, AV, MV, Q, WW, SL, UL, and SSC remained statistically unchanged (Figure 1; SI Appendix; Supplementary Figure S2).

Before and after extreme weather events (EWEs), the dominant phyla of periphytic algal communities (>5%) were similar, but there were significant differences in the abundances of some dominant phyla between them (Figure 2). From before flood (BF) to AF, the relative abundance of Bacilloriopyceae increased from 65.2% to 70.6%, and Chlorophyta increased from 27.5% to 28.7%, while Cyanobacteria decreased from 7.2% to 0.6%. From BB to AB, the relative abundance of Bacilloriopyceae increased (from 62.7% to 92.2%), while Cyanobacteria (from 14.9% to 0.1%) and Chlorophyta (from 22.4% to 7.6%) decreased. The sampled water bodies were classified into four groups: D1 (Xiangxi River), D2 (Nanhe River), D3 (Pingqian Reservoir, Dajiuhu Lake and Duhe River) and D4 (Yandu River and Songluo River). By calculating the mean relative abundance of each phylum within these four groups, it was found that the relative abundance of Bacilloriopyceae was higher in D1 (79.7%) and D2 (87.7%), Cyanobacteria in D3 (10.5%), and Chlorophyta in D4 (51.1%) (Figure 2).

In terms of species composition, during BF, Achnanthes sp. and Cymbella sp. dominated the periphytic algal community in all sites. In AF, the dominant species were Achnanthes sp., Melosira sp., Diatoma sp., Chlamydomonas sp., and Cladophora sp in all sites. When BB arrived, Achnanthes sp. and Ulothrix sp. became the dominant species in all sites. In AB, Achnanthes sp., Diatoma sp. and Navicula sp.1 were the dominant species in all sites (SI Appendix; Supplementary Figure S5; Supplementary Table S2). Within water body group 1 (D1), Navicula sp., Melosira sp. and Cladophora sp. were the dominant species. In water body group 2 (D2), Achnanthes sp. and Cladophora sp. dominated the periphytic algal community. For water body group 3 (D3), Achnanthes sp., Melosira sp., Navicula sp. and Lyngbya sp. were dominant species. Lastly, in water body group 4 (D4), Achnanthes sp., Diatoma sp. and Ulothrix sp. became the dominant species (SI Appendix; Supplementary Figure S5; Supplementary Table S3).

Regarding biomass of periphytic algae, the comparison across the four periods revealed a 90.0% decrease in the biomass of AF compared to BF, and a 95.8% decrease in the biomass of AB compared to BB, and the biomass was significantly higher in D2 and D4 than in D1 and D3 (SI Appendix; Supplementary Figures S6, S7).

Comparing the periods before and after EWEs, floods led to a 1.4% decrease in richness, a 25.5% decrease in evenness, and an 8.4% decrease in Shannon diversity of periphytic algae, while blizzards caused a 1.1% decrease in richness, a 17.5% decrease in evenness, and a 5.6% decrease in Shannon diversity, while D1 and D3 had higher richness, evenness, and Shannon diversity than D2 and D4 (SI Appendix; Supplementary Figures S8, S9). The differences in alpha diversity before and after EWEs were greater than the differences between water bodies (SI Appendix; Supplementary Figures S8, S9). Through the Adonis test (P < 0.05) and Principal Coordinates Analysis (PCoA) conducted on the community composition of the seven water bodies, it was found that these water bodies could be clustered into four distinct groups (D1-D4). After EWEs, the differences in community composition among these four groups were more pronounced (SI Appendix; Supplementary Figures S10, S11). Compared to D2 and D4, the increase in community composition differences due to EWEs was more significant in D1 and D3 (SI Appendix; Supplementary Figures S4, S11).

Decay rates of community similarities were notably higher in BF and BB than in AF and AB as environmental distance increased, based on the Euclidean distance between pairwise samples derived from the matrix of measured environmental variables (Figure 3; all P < 0.001 and SI Appendix, Supplementary Table S4). Consequently, the decline in community similarities was more pronounced per unit change in environmental distance during BF and BB. Additionally, as altitude and geographical distance increased, community similarities exhibited a decaying trend, with BF and BB showing higher decay rates compared to AF and AB (Figure 3, all P < 0.001 and SI Appendix; Supplementary Table S4).

Based on partial Mantel test, specific physical factors (average water velocity (AV), water temperature (WT) and specific conductivity (SPC)) and nutrients (total carbon (TC) and inorganic carbon (IC)) were responsible for significant decay relationships of community similarity over distances before and after EWEs (Figure 4; SI Appendix; Supplementary Figure S13). In addition, physical factors (pH and underwater light intensity (UL)), and nutrients (nitrate nitrogen (NO₃^--N), ammonium nitrogen (NH₃-N) and total phosphorus (TP)) were responsible for significant decay relationships of community similarity over distances before or after EWEs (Figure 4; SI Appendix; Supplementary Figure S13).

The observed shifts in network topology indicate that EWEs fundamentally reconfigured interspecific association patterns within the periphytic algal communities. It is important to note, however, that network stability is an emergent property arising from the interaction of multiple topological attributes, rather than a direct function of any single metric. The AF network exhibited the lowest node connectivity, centrality, and complexity (Figure 5A), which collectively suggest a loss of redundancy and architectural buffering capacity. This aligns with its most precipitous decline in natural connectivity during robustness testing (Figure 6), supporting the conclusion of severely compromised stability. In contrast, the AB network showed recovery in average node degree and centrality compared to the BB state, yet experienced a concurrent and significant drop in modularity (Figure 5B). Modular structure is known to enhance stability by compartmentalizing perturbations within semi-independent modules. This creates a potential trade-off in the AB network: recovery of local connectivity versus the loss of a higher-order structural feature that contains disturbance spread.

Variance partitioning analysis revealed that environmental variables explained 46.0%, 21.0%, 60.0%, and 52.0% of the variation in community composition before and after EWEs (SI Appendix; Supplementary Figure S16). Specifically, physical factors explained 10.0% to 46.0%, nutrients explained 20.0% to 39.0%, and altitude explained 8.0% to 17.0% of the variation (SI Appendix; Supplementary Figure S16).

Structural equation modeling (SEM, all P > 0.05, AIC < 200; Figure 7) showed that changes in community diversity were directly affected by species composition and physical factors, and indirectly affected by altitude and nutrients across four periods. Before and after floods or blizzards, community diversity was influenced differently by these factors, with direct and indirect effects varying across periods (Figure 7).

After flood, altitude and water temperature positively correlated with community richness, evenness, and Shannon index; maximum water velocity negatively correlated (SI Appendix, Supplementary Figure S19). A 10% increase in max water velocity led to decreases in richness (7.5%), evenness (2.8%), and Shannon diversity (3.1%). After blizzard, altitude and max water velocity negatively correlated, while water temperature positively correlated with these metrics (SI Appendix, Supplementary Figure S20). The analysis quantified the effect of unit environmental changes: a 10% decrease in water temperature corresponded to reductions in richness (9.4%), evenness (2.2%), and Shannon diversity (3.5%) (SI Appendix, Supplementary Figure S21). Critically, given that actual temperature shifts during these EWEs were far greater, the realized ecological impact was substantial. Similarly, a 10% increase in altitude led to a 1.1-1.2% increase in community similarity, while a 10% increase in maximum water velocity led to a 1.5-2.1% decrease (SI Appendix, Supplementary Figure S21). As the floods involved water velocity increases often exceeding 100%, the consequent decline in community similarity was profound.

Previous studies indicated that Extreme Weather Events (EWEs) have reduced the diversity of periphytic algae (Smith, 2011; Zhao et al., 2024) and similar results were obtained in this study. Floods significantly increased water velocity, which exerted multiple impacts on periphytic algae. Firstly, the elevated hydraulic scour directly detached and removed algal biofilms from the substrates, leading to an immediate reduction in biomass and a decrease in species richness (Musselman et al., 2018; Chen et al., 2022). Concurrently, the high velocity disturbed riverbed sediments, increasing water turbidity and inducing a shading effect that limited photosynthesis and suppressed algal growth (Zhong et al., 2019; Medeiros et al., 2020). Beyond direct physical removal and light limitation, the high-flow conditions also diluted nutrient concentrations and altered the competitive landscape, favoring a few fast-colonizing species over the majority of other species (Shanafield et al., 2020; Liu et al., 2024). The observed changes in biomass and species composition in this study are consistent with predictions from previous research, demonstrating that floods reduce periphytic algal diversity through multiple direct and indirect pathways (Ren et al., 2021; De Gallardo et al., 2023). Blizzard significantly reduced water temperature, and low temperatures were unfavorable for the survival of periphytic algae (Battin et al., 2016; Osório et al., 2019). The excessively low temperatures and snowfall could cause water bodies to freeze, preventing water flow and reducing nutrient transport, further killing periphytic algae (Dunck et al., 2016; Hampton et al., 2022), which resulted in the death of algal cells. The results were obtained in this study indicated that blizzard directly reduced the diversity of periphytic algae by rapid temperature changing. EWEs could make some algae species disappeared by rapidly altering water velocity, light, and temperature, which is the direct mechanism for reducing the diversity of periphytic algae.

Extreme weather events (EWEs) can also reduce periphytic algal diversity indirectly by diminishing habitat heterogeneity (Bondar-Kunze et al., 2021; Lukacs et al., 2021). Previous studies have established that intense scouring from floods and sharp reductions in water temperature from blizzards can lead to habitat homogenization, which is frequently accompanied by declines in community diversity (Dillon and Conway, 2021; Liu et al., 2022). Floods significantly increased water velocity, and the blizzard significantly reduced water temperature, creating a synchronous pattern of habitat homogenization and diversity loss (Balke et al., 2014; Musselman et al., 2018). Critically, habitat homogenization not only directly reduces diversity but also hinders its recovery. Due to their sessile habitat, periphytic algae cannot readily recolonize homogenized habitats from external sources, which prolongs the post-disturbance low-diversity state (Nascimento et al., 2024; Tang et al., 2025). As EWEs intensify, periphytic algal diversity is projected to decline further, making this indirect pathway of impact increasingly pronounced (Loarie et al., 2009; FAO, 2020; CMA, 2024). Therefore, the indirect mechanism for sustained diversity loss involves the direct homogenization of habitat by EWEs, compounded by the sessile habitat of periphytic algae which severely limits post-disturbance recovery.

Studies have found that increased node connectivity, centrality, and complexity are associated with decreased network stability, aligning with the changes observed in community networks after blizzards, which suggest a decline in periphytic algal network stability (Krug et al., 2020; Polazzo et al., 2023). Although floods slightly increased network stability based on node attributes, the lowest biomass and diversity during the flood period indicated that floods did not enhance stability but rather reduced the number of network nodes and weakened node connectivity, resulting in opposite trends between network attributes and stability (Garcia et al., 2014; Zhu et al., 2022). Robustness tests further showed that EWEs significantly reduced the resistance of periphytic algal networks to disturbances (Fan et al., 2018; Morrissey et al., 2021). Taken together, these findings demonstrated that weakening network stability was a crucial mechanism by which extreme weather events reduce the stability of periphytic algal communities.

Our network analysis elucidates a complex mechanism through which EWEs undermine community stability by altering inter-specific interaction patterns. Crucially, the contribution of individual network properties to stability is not additive but involves interactions and trade-offs (Carpentier et al., 2021; Meena et al., 2023). The AF period presented a relatively direct case: a simultaneous decline in node degree, eigenvector centrality, and linkage density led to a simplified architecture with diminished redundancy (Figure 5A). This structural simplification directly translated to the weakest resistance against node or edge removal in robustness tests (Figure 6), consistent with classical theory linking network complexity to stability (Meena et al., 2023; Wang et al., 2024b). The impact of the blizzard, however, revealed a more nuanced scenario. While the AB network showed recovered average connectivity and centrality, it exhibited a significant loss of modularity (Figure 5B). High modularity is a key stabilizing structure that limits the propagation of shocks by isolating them within modules (Meghanathan, 2021; Engsig et al., 2024). Therefore, the AB network likely existed in a state where recovered local connectedness was counterbalanced by the erosion of this protective, compartmentalized structure. This configuration—potentially more “brittle” or “globally connected”—might explain why its robustness, though better than the AF network, remained substantially lower than the BB network (Figure 6). Consequently, EWEs weaken stability not only by altering individual network metrics but, more profoundly, by disrupting the strategic balance between different topological properties, such as the trade-off between connectivity and modularity. This shift in the network’s structural paradigm may render the community more vulnerable to future perturbations, even if some metrics appear to recover.

Previous studies have established that higher turnover rates in the distance-decay relationship between community similarity and environmental distance indicated high environmental sensitivity and low community stability (Liang et al., 2015; Peguero et al., 2022). In this study, EWEs led to a decrease in community stability, with the distance-decay relationship exhibiting a lower turnover rate. It was attributed to the environmental perturbations caused by EWEs greatly exceeding previous thresholds, leading to the reconstruction of algal communities with reduced environmental sensitivity (Battin et al., 2016; Osório et al., 2019). The low total β-diversity but high turnover in communities after EWEs indicated that the decline in stability was due to rapid species elimination and replacement during community reconstruction (Wang et al., 2017; Wu et al., 2020). Overall, these findings suggested that dramatic environmental perturbations disrupting community structures constituted a significant force by which extreme weather events reduced the stability of periphytic algal communities under disturbances.

Studies have shown that periphytic algae inhabiting harsh environments often had weak environmental adaptability, making communities vulnerable to environmental impacts (Swanson et al., 2015; Osório et al., 2019). Our results indicated that harsh environments drove a decline in community stability, implying that the reduction of community stability by EWEs was related to the weak adaptability of periphytic algae (Garcia et al., 2014; Krug et al., 2020). Specifically, extreme perturbations during EWEs could inhibit or even kill some algae in the original community by altering metabolic activities and damaging physiological functions (Hernandez et al., 2021). Additionally, specific conductance significantly impacted periphytic algal communities, as the influx of dissolved solids during floods might disrupt the osmotic balance inside and outside algal cells (Waadt et al., 2022). The decrease in some nutrient levels during EWEs could further exacerbate living conditions for periphytic algae (Hao et al., 2020). Moreover, the decline in periphytic algal community richness, evenness, and Shannon diversity during EWEs, which was substantial when scaled to the major changes in water temperature and velocity, likely contributed to the decrease in community stability. This aligns with the established principle that lower diversity generally weakens a community’s resistance to disturbances (Tsang et al., 2023; Guo et al., 2024). In summary, the fragile resistance of periphytic algae to harsh environmental conditions represented a significant vulnerability in community stability and serves as a key target for extreme weather events to significantly reduce community stability.

Changes in periphytic algal community composition with EWEs could potentially lead to stability reductions due to nonlinear relationships between community composition and stability and cascading impacts of these property changes (Leps et al., 2018; Wang et al., 2021). Studies on the colonization process of periphytic algal communities have shown that different algae had specific habitat requirements when joining a community (Zhu et al., 2022). When the degree of community structure destruction and diversity decline reached a threshold, the stability of periphytic algal communities might undergo a cliff-like decline, requiring longer recovery times after EWEs, posing potential risks to ecosystem biological stability (Leps et al., 2018; Wang et al., 2021). Our structural equation modeling (SEM) analysis revealed that the decline in periphytic algal community diversity during EWEs had a direct and close correlation with community composition. Combined with disturbance and community correlation studies, the intense disturbances from extreme weather were important for the decline and difficult recovery of community diversity (Coleine et al., 2024; Doolittle and LaManna, 2025). In a word, it was found that physical indicators significantly contributed to the indirect impact on community structure during EWEs. Additionally, the relationship between EWEs and the reduction of community diversity was quantified. The mechanism by which EWEs weaken the stability of periphytic algae was also demonstrated, which will be helpful for estimating the impacts of intensified EWEs in the future.

This study employed a before-after sampling design to capture the net ecological impact of discrete EWEs. While this approach effectively revealed significant shifts in periphyton communities, it inherently lacks the temporal resolution to trace the transient dynamics and short-term recovery processes during and immediately after the events. Furthermore, while we attribute the observed consistent changes across multiple water bodies to EWEs, the potential influence of unmeasured micro-scale factors cannot be entirely ruled out. To mitigate these concerns, our conclusions are strengthened by the strong spatial replication across 53 sites and the use of statistical controls for co-varying environmental factors. Future research would benefit greatly from high-frequency, continuous monitoring throughout the duration of EWEs. Coupling such detailed time-series data with advanced techniques like environmental DNA (eDNA) meta-barcoding and transcriptomics could precisely track community turnover and elucidate the real-time physiological responses of algae, thereby solidifying causal links and deepening our understanding of ecosystem resilience.