Yinghu Lake, located in the upper reaches of the Han River, the largest tributary of the Yangtze River, is the largest freshwater lake in northwest China. It has a controlled catchment area of 35,700 km², a total storage capacity of 3.2 billion m³, a maximum flood storage capacity of 980 million m³, a normal water level of 330 m, an average depth of 48.5 m, and a maximum depth of 80 m. The reservoir supplies 75% of the water for the central route of the South-to-North Water Diversion Project (Liu et al., 2019). As a large multipurpose reservoir integrating power generation, flood control, water supply, navigation, and tourism, Yinghu Lake was included in China’s first batch of 15 key lake ecological and environmental protection projects in 2014. The reservoir area encompasses 34 administrative villages, where untreated domestic sewage from local communities is discharged directly into the lake. Additionally, water quality is further affected by tourism-related waste, including increased domestic sewage from tourist facilities, trash left by visitors, pollutants from recreational activities, as well as feed and waste from aquaculture and livestock farming, and agricultural runoff, all of which exacerbate nutrient enrichment. Furthermore, Yinghu Lake is located within the subtropical monsoon climate zone, characterized by significant seasonal hydrological variations (Liu et al., 2019). During summer (July-September), rainfall accounts for approximately 50% of the annual precipitation in the watershed. The reservoir manages inflowing flood peaks to adapt to variations in rainfall. These anthropogenic activities and natural changes collectively influence water quality, nutrient levels, and macrobenthic communities.

We established twelve sampling sites along the riverine sections of Yinghu Lake, evenly distributed from upstream to downstream, and covering the main tributaries. And we sampled macrobenthos and their ecological variables at these sites ( Figure 1 ). We obtained land use data from the Zenodo website (https://zenodo.org/), which contains the 2022 Yinghu Lake watershed surface cover types at a 30 m resolution. We used ArcGIS (ver. 10.8) to calculate land use within a 1 km radius around each sampling point (Nelson Mwaijengo et al., 2020) and determined the percentage coverage of land use types in the buffer zone. We categorized the sampling sites into developed sites (DS) including S1, S4, S5, S6, S7, S8, S10, and S11, and undeveloped sites (UDS) including S2, S3, S9, and S12 based on the presence or absence of impervious surfaces. The actual land use conditions of each sampling site are presented in Supplementary Table A.1 .

We conducted sampling in 2023 during February (winter), May (spring), August (summer), and November (autumn). The winter data were excluded from the study because the plastic storage bags for sediment samples at multiple sites ruptured, leading to the mixing of sediment samples from different sites and compromising subsequent data analysis. To prevent such issues, samples from each site were stored separately during the other seasons, ensuring data integrity for later measurements and correlation analyses with macrobenthic organisms. Water samples were collected in layers (0 m, 4 m, 8 m, 12 m, 16 m and 20 m) using a 5L acrylic water sampler (GWS-117, CN).

Total phosphorus (TP), soluble reactive phosphorus (SRP), total nitrogen (TN), ammoniacal nitrogen (NH₄ ⁺-N), nitrate nitrogen (NO₃ ^–N), nitrite nitrogen (NO₂ ^–N) and the permanganate index (COD_Mn) were measured according to Standard Methods (Rice et al., 2012). We measured water temperature (hereafter Temperature), dissolved oxygen (DO), conductivity (EC), total dissolved solids (TDS), pH, and turbidity (hereafter Turb) using a multi-parameter water quality analyzer (YSI ProDSS, USA), and calculated the mean values for each water layer for subsequent analyses. We determined chlorophyll a (Chl-a) using the BBE Algae Analyzer (FluoroProbe, GER), Secchi depth (SD) using a Secchi disk, and water depth (Depth) using a sonar depth gauge (SM-5A, USA).

We collected macrobenthos in the study area using a 1/16 m² grab sampler (PBS-411, CN), taking two collections per sample site and mixing them into one sample. At each sample point, we cleaned the samples by filtering them through a nylon mesh with a pore size of 450 μm. We placed the residue into 250 ml plastic bottles and immediately fixed it in 10% formalin. Subsequently, we rinsed off the formalin, picked out the benthos, and collected them in petri dishes. We counted all organisms at each site and identified them to the lowest possible taxonomic unit (Thorp and Covich, 2010). We classified FFG based on their taxonomic characteristics and feeding types (Barbour et al., 1996), with major FFG including predators, scrapers, filtering collectors, and gathering collectors (Pratiwi et al., 2024).

During the collection of macrobenthos samples, we simultaneously collected small sediment samples and transported them to the laboratory under refrigerated storage at 4 °C. In the laboratory, we naturally air-dried the sediment samples, ground them, and sieved them through a 100 - mesh sieve. We used the sieved powder samples to measure the mechanical composition using the densitometer method and extracted the proportion of 0.002 - 0.02 mm fines. We measured the total carbon of sediment (TC_s), organic carbon of sediment (OC_s), and total nitrogen of sediment (TN_s) using an elemental analyzer (Vario Macro-CHNS, GER). We measured total phosphorus of sediment (TP_s) using the molybdenum-antimony colorimetric method (Rice et al., 2012).

We determined the dominant species based on the dominance degree Y ≥ 0.02. The formula for calculating Y statistics was as follows:

where n_i is the number of individuals of a species in the sample; f_i is the frequency of occurrence of the species; and N is the total number of individuals of all species in the sample.

We used the Richness of macrobenthic species, Abundance, Margalef index of macrobenthic fauna (D), and Margalef index of FFG (FFG-D) to describe the diversity of macrobenthos at different sampling sites (Yi et al., 2018).The formula general for Margalef index (D) was as follows:

where S is species number. When calculating the Margalef index of FFG (FFG-D), the formula remains unchanged, but S is replaced by the number of FFG, and N representing the total number of individuals across all FFG in the sample.

We used the ‘DescTools’ and ‘rstatix’ packages in R (ver. 4.3.3) to test the significance of spatial and temporal differences in environmental factors and macrobenthic bioindicators using a two-way ANOVA (Analysis of Variance). We applied a T-test to analyze the significance of differences in macrobenthic communities across different land use classes, considering P < 0.05 as significant (Tomanova et al., 2006). We organized the macrobenthos abundance data from the three seasons using the ‘vegan’ package in R for Non-metric Multidimensional Scaling (NMDS) analysis, aiming to examine differences in macrobenthos ecosystems across the three seasons. We performed a Hellinger transformation of abundance before the analyses. We used PERMANOVA to assess the significance of differences in macrobenthic communities across different spatial and temporal scales. We identified species with the most significant contribution to differences across spatial and temporal scales using similarity percentage analysis (SIMPER). We conducted Detrended Correspondence Analysis (DCA) and Redundancy Analysis (RDA) using CANOCO (version 5) to explore the relationship between environmental factors and macrobenthic community structure (including FFG). Prior to analyses, we transformed data on environmental variables of percentage type (e.g., substrate composition) by an inverse chord square root transformation, and we applied a log (x+1) transformation to species abundance and other physico-chemical environmental variables (except pH).

We used a Generalized Additive Model (GAM) to fit the association between the response variables (diversity indices and abundance of FFG) and environmental conditions. We included the explanatory variables using the “mgcv” package in R. The formula general for GAM was as follows (Politou et al., 2008):

where g(Y) is the connection function, Y = lg (y+1), where y is the response variable; ɑ is the constant functional intercept; s_i(x_i) is the spline smoothing function for each environmental factor; and θ is the random error.

We first examined the relationship between each environmental factor and the Margalef index to test whether the effect of individual environmental factors on the community Margalef index was significant, with a significance level of 0.1. We then applied the forward selection method for adding variables. To avoid the effect of covariance, we excluded environmental variables with absolute values of correlation coefficients (R) in Pearson exceeding 0.4 from the same model. We comprehensively evaluated the impact of the resulting models based on the corrected coefficient of determination (Adj-R²), Deviance Explained (DE), and Akaike Information Criterion (AIC), and then fitted the best model (Maravelias et al., 2007; Martínez-Rincón et al., 2012).

Our results indicated that NO₃ ^–N, NO₂ ^–N, Temperature, DO, pH, Chl-a, Secchi depth and TLI had varied among seasons (P < 0.05). DO, pH, Secchi depth, and NO₃ ^–N were significantly lower during the summer compared to the spring and autumn seasons, Temperature, TLI, and Chl-a were highest during the summer, and NO₂ ^–N was highest during the spring season. The rest of the environmental factors did not vary among seasons (P > 0.05). We observed differences between land use development classes in TN, TP, pH, and TLI (P < 0.05). TN, TP, and TLI significantly higher in DS than in UDS (P < 0.05), and pH significantly higher in UDS than in DS (P < 0.05), while the rest of the environmental factors were similar among land use classes. Secchi depth was influenced by a season-by-land use class interaction, and the rest of the environmental factors were insignificant (P > 0.05, Table 1 ). Environmental factors data for each sampling site during the three seasons is presented in ( Supplementary Tables A.2 – A.4 ).

We performed a detrended correspondence analysis (DCA) on the macrobenthic community data, yielding principal axis lengths of 2.1 and 1.8, respectively. We then conducted a redundancy analysis (RDA) for the macrobenthic communities and FFG. The redundancy analysis results between macrobenthic communities and environmental factors showed that the first two axes explained a total of 24.76% of the variation ( Figure 5A ). Among the environmental variables, Chl-a (F=4.3, P < 0.01) was the most closely related to macrobenthos with a contribution of 9.9%, followed by TP_s (F=3.3, P < 0.01) with a contribution of 8.8% and COD_Mn (F=3.3, P < 0.01) with a contribution of 8.3%. From the results, it can be seen that L. hoffmeisteri, L. grandisetosus and B. sowerbyi were positively correlated with COD_Mn. C. fluminea were positively correlated with TP_s. Macrobenthic communities were negatively correlated with Chl-a. Redundancy analysis of FFG with environmental factors showed that the first two axes explained a total of 24.35% of the variation ( Figure 5B ). Among the environmental variables, TP_s (F=4.1, P < 0.01), total dissolved solids (F=3.1, P < 0.05) and Chl-a (F=2.6, P < 0.05) were the environmental factors most closely related to FFG, and their contributions were 16.3%, 11.4% and 9.5%, respectively. From the results, it can be seen that gathering collectors was positively correlated with total dissolved solids, while predators, scrapers and filtering collectors were positively correlated with TP_s, and FFG was negatively correlated with Chl-a. Compared with UDS, DS were mainly located in the lower left part of the ordination diagram and had a stronger relationship with total dissolved solids.

As the Margalef diversity index performs well in distinguishing differences between communities (Yi et al., 2018), the Margalef index was selected to reflect changes in macrobenthic diversity. We then examined the influence of environmental factors on both the D and FFG-D indices. The results indicated that SRP, Temperature, total dissolved solids, pH, Secchi depth, OC_s, and TP_s significantly influenced D (P < 0.1); SRP, Temperature, DO, EC, total dissolved solids, Depth, and TP_s were the environmental factors that significantly influenced FFG-D (P < 0.1). Since the correlation coefficients of DO with Temperature and EC, as well as EC and total dissolved solids were > 0.4, these two variables could not be added to the model simultaneously. We selected Temperature and total dissolved solids for the following analysis and performed forward stepwise regression to screen the optimal model for each. The Temperature with the lowest AIC was first added to the model of D, and the addition of total dissolved solids and OC_s on top of this significantly improved the model performance (P < 0.05). SRP, pH, and TP_s were removed because adding these environmental factors did not improve model performance, and the effect of this environmental factor on D was not significant after addition (P > 0.1). The final model was optimal with the addition of Temperature, total dissolved solids, and OC_s, and the AIC was minimal, with a cumulative model deviation of 39.88%. D was linearly negatively correlated with Temperature, total dissolved solids, and OC_s ( Figure 6A ). We first added SRP with the minimum AIC to the FFG-D model, and the model’s performance significantly improved after adding Depth, Temperature, and TP_s. We then removed total dissolved solids, as it did not improve the model’s performance. The final optimal model, which included SRP, Depth, Temperature, and TP_s with the minimum AIC, explained 45.52% of the cumulative deviation of the model ( Table 5 ). FFG-D was linearly negatively correlated with SRP and Temperature, and linearly positively correlated with TP_s. FFG-D exhibited the highest index values at a depth of approximately 20 m, followed by a gradual decline, and then a slight increase after reaching about 60 meters ( Figure 6B ).

Urban areas characterized by non-permeable pavements facilitate the rapid transport of pollutants into water bodies via surface runoff. In aerobic conditions, microorganisms break down organic matter, resulting in the production of CO₂. Simultaneously, certain nitrogenous compounds undergo oxidation to nitrate through nitrification in the water column, which generates hydrogen ions, leading to a decreased pH at the DS compared to the UDS (Gałuszka and Migaszewski, 2020). Furthermore, the direct or indirect discharge of domestic sewage and aquaculture waste into the reservoir area at the DS frequently causes the enrichment of organic constituents, resulting in water quality degradation and hydrological disruption in the reservoir (Dudgeon et al., 2006). Furthermore, due to the impervious surfaces of developed areas, the increased rainfall and elevated temperatures during summer facilitate the transport of pollutants via surface runoff, significantly increasing the likelihood of nutrient pollution at the DS (Yang et al., 2020). The mean TLI value at the DS throughout the summer was the highest for the whole year (44.68), with the most pronounced difference in pollutant concentrations (TN and TP) between the two regions. However, during the hydrologically stable spring and autumn seasons, these differences were relatively smaller. Additionally, the TLI values at the UDS in spring were higher than those at the DS, likely due to sampling conducted after rainfall, resulting in poorer water quality. Sampling points S2 and S3, located upstream, were directly influenced by non-point source pollution, whereas the S9 and S12 sites, situated in the lower reaches of the mainstem and tributaries, experienced slower flow rates, potentially facilitating nutrient retention (Wang et al., 2020, 2021).

The main groups of macrobenthos in Yinghu Lake are Oligochaeta and Insecta, of which, L. hoffmeisteri, L. grandisetosus and P. choreus are the representative species of these two orders, and they are the dominant species in Yinghu Lake. The distinct structural and functional characteristics of chironomid larvae and some oligochaetes enable these r-selected organisms to recover more rapidly from environmental disturbances and to thrive in deep-water habitats (Hayford et al., 2015). Furthermore, chironomid larvae eventually metamorphose into adults, which possess strong dispersal capabilities (Cummins and Klug, 1979), which may contribute to their apparent prevalence in deep-water reservoirs.

Impervious surfaces significantly impact the ecosystem, resulting in distinct habitat environments between DS and UDS. Additionally, Yinghu Reservoir is located in a monsoon climate zone, where temperature, rainfall, and wind exhibit significant seasonal variations (Chi et al., 2017; Sun et al., 2024). Substrate diversity, nutrient concentrations, and hydromorphological characteristics largely determine the community structure of macrobenthic fauna (Mathooko and Mavuti, 1992). As a result, the changes in macrobenthic community structure across different land use vary significantly under different seasons. At the DS, domestic sewage discharge and industrial production directly degrade water quality, leading to a decline in pollution-sensitive macrobenthic fauna due to reduced food availability and potential toxicity or harm (Pratiwi et al., 2024). Differences between the two regions were primarily reflected in the proportion of pollution-tolerant individuals. L. hoffmeisteri, a widely known pollution-tolerant species, exhibited significantly higher abundance at the DS with higher trophic state indices compared to the UDS. In contrast, less pollution-tolerant mollusks were predominantly found at the UDS, where the abundance of C. fluminea was significantly higher than at the DS and served as a dominant species at the UDS, which indicate that the spatial distribution of macrobenthic communities in Yinghu Lake exhibited significant differences. When habitats are degraded, the abundance of semi-tolerant taxa decreases, and habitat-sensitive species like C. fluminea may serve as effective indicators of habitat degradation. Additionally, under the influence of climate variability, differences in community structure between the DS and UDS were most pronounced during the summer, compared to the other two seasons. Given that the DS dominate the study area, the community composition was primarily characterized by Oligochaeta. Furthermore, the sediment properties did not show significant seasonal variation, and the habitat uniformity influenced the distribution patterns of macrobenthic fauna (Hayford et al., 2015), leading to a non-significant difference in the macrobenthic community structure of Yinghu Lake across seasons.

Typically, changes in trophic structure are a community’s response to specific changes in its food sources or disturbance patterns (Cummins and Klug, 1979; Mendoza and Araújo, 2019), and FFG more accurately represents the vulnerability of macrobenthos to environmental stressors. In Yinghu Lake, the FFG exhibited significant spatial variation. During the summer, with increased rainfall and rising temperatures, the FFG composition between the two regions showed a significant difference. There were no significant differences between the two regions for predators and scrapers, although higher abundance was observed at the UDS than at the DS. Collectors are typically organisms that gather organic particles or microorganisms (Pratiwi et al., 2024). The increased organic matter at the DS offers abundant nourishment for gathering collectors, which thrive in such environments due to their high contamination tolerance, diminutive size, brief lifecycle, and incessant reproduction, facilitating swift community recovery and sustained survival. Nevertheless, filtering collectors, being collectors as well, exhibit greater environmental demands and face disadvantages in interspecific competition with gathering collectors, which are primarily composed of smaller, pollution-tolerant organisms (Wang et al., 2021a), leading to an inverse distribution pattern between filtering collectors and gathering collectors. Temporally, gathering collectors and filtering collectors abundance peaked in the autumn when hydrology was stable. Pratiwi et al. (2024) observed that the abundance of predators correlates with the species present in the habitat. The water pollution has led to a decline in biodiversity at the DS (Wang et al., 2012), thereby diminishing the availability of food sources for predators and subsequently reducing their abundance. Scrapers are species that consumes sedimentary organic matter and algae, is preferentially supported during steady hydrological conditions (Rahman et al., 2022), paralleling the fluctuations in the relative abundance of scrapers in Yinghu Lake. The stable hydrological conditions during autumn facilitated the deposition of organic matter. Following the autumnal deposition, phytoplankton and supplementary organic matter, including decomposed leaves, accumulate at the bottom of reservoirs, resulting in a peak abundance of scrapers during the autumn. Spatially, scrapers exhibited reduced abundance at DS compared to UDS due to its lower tolerance to pollution.

Richness, D and FFG-D were all significantly lower during the summer compared to the autumn. This result is similar to conclusions drawn from studies on changes in macrobenthic communities due to variations in rainfall or changes in reservoir water level regulation (Milner et al., 2013; Quadroni et al., 2016; Sun et al., 2024), indicating that variations in seasons may be the predominant factor affecting the diversity of macrobenthic communities. Variations in macrobenthic diversity indices are contingent upon environmental stability (Duan et al., 2009). Compared to the summer, the spring and autumn seasons exhibited more stable hydromorphological characteristics, resulting in higher diversity indices during these seasons. Additionally, spring sampling conducted after rainfall revealed that nutrient concentrations in the water were influenced, leading to lower water quality in the spring compared to the autumn. This may explain why structural indices in the spring were lower than in the autumn. The r-selected macrobenthos, which dominate the abundance in Yinghu Lake, are less affected by environmental disturbances, resulting in no significant differences in abundance across the three seasons. Spatially, there were significant differences in the FFG-D between the two regions, indicating that changes in land use patterns and human activities have altered habitat conditions, thereby constraining the survival of various organism groups, which in turn has led to spatial differences in species diversity (Wang et al., 2021b). Due to environmental change, richness, D, and FFG-D at DS were lower than those at UDS across all three seasons.

Environmental factors directly influence the composition, life cycle and distribution of macrobenthic communities (Cooper et al., 2007). Evaluations of the correlation between environmental variables and macrobenthos can accurately determine the ecological and biological condition of sampling locations (Poikane et al., 2016; Yang et al., 2020). Previous studies have shown that nutrients have a strong influence on macrobenthic community structure (Poikane et al., 2016). COD_Mn serves as a crucial metric for assessing the organic matter concentration in water, mostly utilized in evaluating river pollution and analyzing the characteristics of industrial effluent (Kawabe and Kawabe, 1997). Pollution-tolerant species among the principal macrobenthic communities (L. hoffmeisteri, B. sowerbyi, and P. choreus of oligochaetes) had a positive correlation with COD_Mn. Moreover, the discharge of municipal and industrial wastewater elevates the total dissolved solids in water (Chen et al., 2022), and overly high total dissolved solids adversely affects aquatic species and water quality (Pratiwi et al., 2024). The DS was primarily composed of gathering collectors, and total dissolved solids was positively correlated with DS, aligning with prior research findings. Generally, COD_Mn and total dissolved solids concentrations in DS are relatively high, providing abundant food sources for the more pollution-tolerant gathering collectors and oligochaetes, while inhibiting the growth of the less pollution-tolerant filtering collectors. UDS were mainly dominated by filtering collectors, scrapers and predators and were somewhat negatively correlated with gathering collectors abundance.

Prior research has demonstrated that elevated TP_s and TN concentrations led to diminished DO levels in the benthic environment and increased toxicity in both sediment and water column, potentially resulting in the decline of macrobenthic fauna (Jinlin et al., 2003). This study found a positive correlation between scrapers, represented by the B. purificata, and filtering collectors, dominated by the C. fluminea, with TP_s. Phosphorus, as a vital ingredient for living organisms, enhances the stability and complexity of benthic food web structures, ensuring a sufficient food supply for predators while supporting increased mollusk biomass. The better quality of the aquatic environment at the UDS encouraged the growth of filtering collectors that have specific environmental requirements, leading to competition for food resources between filtering collectors and gathering collectors. This competition subsequently inhibited the growth and reproduction of gathering collectors, which occupy lower ecological niches. Theoretically, Chl-a promotes photosynthesis in algal organisms, increasing primary productivity and providing a food source for filter-feeding benthos. However, excessive Chl-a concentrations are commonly associated with eutrophication of aquatic systems, resulting in a decline in macrobenthic abundance due to environmental stresses (Carvalho et al., 2011).

The key environmental factors selected based on GAM are similar to those in previous studies of macrobenthic community diversity as affected by anthropogenic disturbance as well as changes in the natural environment (Alvarez-Cabria et al., 2011; Fu et al., 2015, 2022; Yi et al., 2018). Both the Margalef index of macrobenthos and the FFG Margalef index showed a negative correlation with Temperature, suggesting that seasonal changes significantly affected macrobenthic diversity. During the summer with the highest temperature, alterations in the aquatic environment due to precipitation and anthropogenic gate operations subsequently affected biodiversity (Milner et al., 2013). During the autumn with the lowest temperature, a period characterized by stable hydrological conditions and abundant food resources, which favored the growth of macrobenthic organisms. Environmental factors such as OC_s, SRP, and total dissolved solids, influenced by anthropogenic activities (Cooper et al., 2007; Yuan et al., 2019), are marked by a decline in water quality resulting from heightened stressor intensity, which affects the diversity of macrobenthos (Firmiano et al., 2021). However, TP_s, also subject to anthropogenic disturbance, showed a positive correlation with FFG diversity. TP_s successfully sustained a significant diversity of FFG throughout the monitoring range, likely due to the concentration of TP_s in Yinghu Lake not reaching a detrimental threshold for aquatic life. Consequently, the scope of environmental component sampling may be broadened to improve the reliability of the model. In addition, depth was identified as a significant environmental factor limiting the diversity of FFG in Yinghu Lake, aligning with the conclusions of Petridis and Sinis (1993). Water depth influences the physical characteristics of macrobenthic habitats, DO, and food supply, significantly altering the structure of benthic communities (Cui et al., 2008; Ohtaka et al., 2006). It is noteworthy that the diversity index was slightly greater when the water depth exceeded a certain threshold. They may be due to deeper waters having less variation in environmental parameters (Nalepa, 1989), and certain oligochaetes have a competitive advantage under these conditions, resulting in a slight increase in their diversity.

Variations in natural conditions and the intensity of human disturbances lead to changes in water environmental factors and food resources for macrobenthic fauna in deep-water reservoirs. These are the primary factors influencing changes in the community structure and trophic composition of macrobenthic fauna. Notably, environmental factors with significant differences between DS and UDS (e.g., TN and TP) may be important drivers of macrobenthic community structure, but did not show significant correlations in this study. This may be due to the complex ecological roles of these factors or potential lag effects in the response of macrobenthic fauna. Additionally, undetected pollutants in the current dataset (e.g., heavy metals and organic pollutants) may exhibit significant differences between the two regions and directly or indirectly influence macrobenthic community composition (Bendary et al., 2023). Therefore, it is essential to expand the range of detected indicators and implement continuous monitoring measures to maintain macrobenthic biodiversity and prevent further ecological degradation.