Predicting possible bank erosion and bed incision in the Zhengzhou Reach of the Yellow River with hydrodynamic modeling

BL

Bed load

D_T

Decision attribute

HF

High flood

ICUS

Iso Cluster Unsupervised classification

NIR

Visible and near-infrared

RST

Rough set theory

rRMSE

Maximum relative root mean square error

SSL

Suspended sediment load

WLs

Water levels

Ke

Erodibility factor

τ_c

Critical shear stress

C_T

Conditioned attribute

FSD

Flow–sediment transfer

HYK

Huayuankou

LF

Low flows

NSEs

Nash–Sutcliffe efficiency coefficients

RMSEs

Root mean square error

SSC

Suspended sediment concentrations

SC

Sediment concentration

XLD

Xiaolangdi dam

X-SP

Cross-sectional profile

ZRYR

Zhengzhou Reach of Yellow River

Morpho-sedimentological alterations in river channels are pivotal for understanding erosion dynamics, particularly concerning riverbank erosion and riverbed incision, which have profound economic and environmental consequences (Gava 2023). These processes are crucial, leading to land loss, infrastructure damage, and sediment management challenges (Chassiot et al. 2020). Asian rivers like the Indus, Yangtze, Brahmaputra, and Jamuna have undergone significant morphological changes due to factors such as sediment accumulation, dam construction, and channelization (Boota et al. 2022; Rashid & Habib 2024). These rivers, including the Yellow River (YR), experience dynamic changes that have been widely studied due to their critical implications for local communities and ecosystems (Yan et al. 2023a, 2023b).

The Zhengzhou Reach of Yellow River (ZRYR) experiences continuous variations in its banks and beds at various spatiotemporal scales due to recurring high flood (HF) events and sediment accumulation (Li et al. 2022). Engineering efforts have focused on managing the risks associated with unpredictable changes in channels and devastating flooding (Dharmarathne et al. 2024). In the Lower Yellow River (LYR), the processes of bank erosion and bed incision have led to vegetation loss, increased sedimentation, variations in channel width, impacts on aquatic ecosystems, and significant agricultural land loss (Scopetani et al. 2024). One effective strategy for mitigating these issues is the use of predictive hydrodynamic modeling tools. By simulating potential erosion scenarios and river channel dynamics, these tools can help inform the design of embankments and other protective infrastructure, ultimately reducing the impact of erosion on riverbanks (Wei et al. 2024). Furthermore, employing such modeling approaches can streamline the development of cost-effective training programs for local communities living near or within river infrastructure, enhancing their capacity to manage and respond to river dynamics effectively.

Research on the LYR has primarily focused on understanding the causes and effects of these morphological changes. Previous studies have examined the influence of sediment load, human activities, and natural processes on river channel morphology (Zhang et al. 2024). However, a comprehensive understanding of the specific interactions between riverbank erosion and bed incision remains underexplored. While numerous studies have investigated these phenomena separately, the combined impact and interrelationship between these two processes, particularly in the context of the LYR, require further investigation.

This study aims to address this gap by focusing on the combined effects of riverbank erosion and bed incision in the LYR. Using rough set theory (RST) and the Hydrologic Engineering Centre–River Analysis System (HEC-RAS) 2D model as analytical tools, this research seeks to identify and analyze the primary factors influencing these morphological changes. RST serves as a practical analytical approach for discerning critical factors impacting river channel morphology (Yan et al. 2023a, 2023b). RST is adept at managing uncertainty and imprecision in data, rendering it an optimal instrument for evaluating intricate environmental systems with interrelated variables (Pawlak 1982). By utilizing RST, one can identify the primary determinants from an extensive array of variables, thereby enhancing the comprehension of the causes of riverbank erosion and bed incision. This proficiency is essential for formulating efficient management strategies and forecasting models. The initial aim of this study is to utilize RST to identify key factors affecting channel morphology. The following aims involve creating a prognostic framework that combines these insights with the HEC-RAS 2D model for analyzing riverbank erosion and bed incision, as well as evaluating the socio-economic and environmental repercussions of these phenomena on local communities, especially regarding land loss.

However, it is essential to clarify that the novelty of this research does not lie in the use of these models themselves. Instead, this study's contribution stems from the new insights it provides into the complex interactions between riverbank erosion and bed incision, particularly in how these processes influence each other and the resulting morphological changes in the river channel. To establish the significance of this research, a thorough review of the existing literature on riverbank erosion and bed incision in the YR and similar contexts worldwide has been conducted. Previous studies, such as those by Yan et al. (2023a, 2023b) and Scopetani et al. (2024), have provided a foundation for understanding the individual processes of erosion and incision. However, this study aims to extend this understanding by exploring how these processes interact, leading to the observed changes in river channel morphology.

The specific objectives of this study are: (1) to identify the key factors influencing the combined effects of riverbank erosion and bed incision in the LYR, (2) to develop a comprehensive understanding of how these processes interact and contribute to morphological changes in the river channel, and (3) to assess the socio-economic and environmental impacts of these changes on local communities, particularly in terms of land loss. By focusing on these objectives, this research seeks to provide new contributions to the field, offering a deeper understanding of the dynamics of riverbank erosion and bed incision and their implications for river management in the LYR and similar environments worldwide. The findings of this study will contribute to the broader understanding of fluvial dynamics, providing valuable insights for managing river systems subject to significant morphological changes. This research aims to inform strategies for mitigating the adverse effects of riverbank erosion and bed incision, ultimately contributing to more effective and sustainable river management practices.

The YR is the second largest river in China, known for its sandy material, frequent flooding episodes, distinctive channel patterns in the lower parts (the channel bed is higher than the land beyond the banks), and presently overutilized water resources (Yao et al. 2013). The YR has a total length of 5,464 km and a watershed drainage area of 795,000 km² (Wang & Mei 2016). The ZRYR and its reaches constantly change, and 26 significant channel shifts have occurred since 602 BC. These channel variations in the YR should be predicted in the context of bank and bed changes (Xu 2002). The LYR, which stretches from Mengjin (Henan province) to Linjin (Shandong province), has been split into three parts based on geomorphological landforms such as braided reach, transitional reach, and meandering reach. Significant soil erosion on the Loess Plateau has resulted in severe sediment load in the LYR, with overall sediment deposition observed to be approximately 5.52 × 10⁹ m³ (1950–1999) based on actual data. About 60% of this deposition occurred in the braided reach (Yang et al. 2024). The greatest effect of the huge sediment load in the LYR has been in the evident shrinkage of the main channel and, consequently, a sudden decrease in runoff capacity, which has culminated in the phenomena of secondary suspended rivers in specific reaches and poses a threat to local community and channel management.

The ZRYR displays a complex and dynamic geological structure characterized by merging diverse geological formations and landform attributes. The ZRYR flows through the northern region of Henan Province, where it intersects the southern boundary of the North China Craton. The foundational geology predominantly comprises ancient crystalline formations covered by thick layers of Quaternary alluvium and loess. The Quaternary sediments, accumulated during alternating glacial and interglacial epochs, have substantially shaped the river's geomorphology, forming extensive floodplains, terraces, and levees (Yan et al. 2023a, 2023b). The loess, derived from eolian deposits dating back to the Pleistocene era, plays a significant role in the channel's elevated sediment concentration, affecting its hydrodynamics and conveyance mechanisms. Geological processes, such as the Taihang Mountains' uplift and the North China Plain's subsidence, significantly impact the channel's longitudinal profile and morphology.

River bank erosion was computed using image analysis with geographic information system (GIS) software, a method extensively employed by numerous researchers (Hasanuzzaman et al. 2023; Mallick et al. 2023). Landsat images (30 m resolution) were used for the computation of river bank erosion, and the selected images were obtained before and after the peak flows. Wang et al. (2014) used the NIR images, which are helpful for identifying vegetation boundaries to recognize the outer river bank lines. According to Ashraf & Shakir (2018), the ICUS was used to detect channel characteristics such as channels, islands, and sandbars in the ZRYR.

The data pertaining to runoff, water levels, sediment composition, and the bathymetric surveys utilized in this research were acquired from diverse sources within a uniform temporal framework to guarantee the dependability and precision of the hydrodynamic model. More precisely, the data encompasses the duration from 2016 to 2018, with comprehensive documentation outlined as follows. First, runoff and water-level data were sourced from hydrometric observations at the HYK station. Measurements spanned from 2016 to 2018, emphasizing HF events during flood seasons. Water-level data were essential for defining model boundary conditions and validating predictions. This consistency over several years addresses annual hydrological variations. Second, sediment data, comprising sediment concentrations (SC) in kilograms per cubic metre (kg/m³), were collected at the HYK station from 2016 to 2018. This dataset encompasses both suspended sediment load (SSL) and bed load (BL), critical for modeling sediment transport and deposition dynamics. Sediment concentration data was crucial for evaluating sediment dynamics under varying flow conditions. Standardized measurement methods were utilized over the years to guarantee comparability and precision. Lastly, bathymetric surveys from 2016 to 2018 mapped riverbed topography and provided cross-sectional profiles. The data encompassed bed elevations, bank positions, and topographic features utilized to generate the DEM for the study area. The bathymetric data were acquired through standardized methods, guaranteeing consistency and accuracy throughout the study duration. This data were essential for establishing model geometry and simulating morphological transformations in the riverbed and its banks. In summary, the data utilized in this research – covering the years 2016–2018 – were uniformly gathered and implemented across various model components, such as runoff, water levels, SCs, and bathymetric assessments. This uniformity guarantees that the model is founded on dependable and comparable data, thereby augmenting the precision and legitimacy of the simulation outcomes.

The morphological dataset used for modeling was obtained from the Yellow River Conservancy Commission through a channel reach survey conducted between 2016 and 2018. This dataset was crucial for precisely delineating the morphology, dimensions, and configuration of the river channel, encompassing the floodplain, within the model. Cross-sections define the shape, size, and layout of the river channel and floodplain in the model (Figure S3, Supplementary Information). They include detailed information about the channel bed elevation, bank positions, and floodplain topography. Accurate cross-sections are essential for the model to simulate water flow and water surface elevations correctly. The X-SP data obtained through this survey provided comprehensive insights into channel bed elevation, bank configurations, and floodplain geomorphology. The data were collected across the full breadth of the channel, extending from the right bank to the left bank, and included multiple points along each cross-section, not just at the banks and bed. Specifically, each cross-section included a series of elevation points at regular intervals across the riverbed and floodplain, ensuring that the entire profile of the cross-section was captured in detail. This means that for every cross-section, the dataset contains the coordinates (latitude, longitude, and elevation) for multiple points across the riverbed and floodplain. To generate the 2D geometry data required for the HEC-RAS model, these detailed cross-sectional points were used to create a continuous surface representation of the channel and floodplain. The inverse distance weighting (IDW) method was utilized to generate a continuous surface representation from the discrete cross-sectional data. This interpolation methodology was selected due to its capacity to approximate values among known data points, contingent upon the spatial distance separating them, which proves particularly advantageous in illustrating the diverse elevations throughout the river channel and floodplain. The application of the IDW method enabled the conversion of the cross-sectional data into a high-resolution digital elevation model (DEM) that accurately represented the river's topographical features. This high-resolution DEM was then geo-referenced using a GIS platform and converted into GeoTIFF format through HEC-RAS software. The GeoTIFF file, representing the DEM of the surveyed cross-sections, was subsequently integrated into an existing open-source 30 m resolution DEM. This integration ensured that the model included both the precise surveyed data and a broader context, necessary for accurately simulating water flow and surface elevations in the 2D HEC-RAS model. The resulting DEM was used to construct the mesh necessary for the 2D HEC-RAS model. This mesh delineated both the main channels and the floodplains with high precision, allowing the model to simulate water flow dynamics and surface elevations across the heterogeneous topography of the study area accurately.

The delineation of boundary conditions for the hydrodynamic model was meticulously formulated to guarantee precise simulations of the flow–sediment interactions within the ZRYR. These conditions encompassed both upstream and downstream inputs, which were essential for facilitating the model's computational processes. The upstream boundary condition was delineated utilizing hydrograph data that encapsulates river discharge (Q, quantified in m³/s) alongside graded SCs (quantified in kg/m³). This dataset was procured from empirical observations at the HYK monitoring station and employed to replicate various flooding scenarios through the modulation of both the magnitude and duration of the discharge. The downstream boundary condition was formulated utilizing a rating curve derived from empirical observations of water levels and discharge metrics at the ZRYR. The preliminary water levels were meticulously established, as inadequate initial water levels may result in suboptimal model performance, encompassing the manifestation of dry nodes within the simulation. The model additionally utilized open boundary conditions at the outlet, thereby facilitating the calculation of water levels through the application of the kinematic wave condition. Furthermore, the Strickler formula was employed to ascertain the bed roughness for the semi-stable sandy bars and the riverine channel, thereby enhancing the overall accuracy of the model.

The morphology of the LYR reach is determined by a multitude of variables, including hydraulic parameters (velocity, shear stress, discharge, and precipitation), SSL, BL, channel attributes (slope, shape, and roughness), as well as the physical characteristics and composition of the bed and bank materials. Scholars, including Morisawa (1985) and Pawlak (2002) have examined these variables extensively. The RST methodology was formulated to ascertain the predominant factors that affect channel morphology. The RST approach was devised to identify the dominant factors influencing the channel/estuary morphology. The data are processed using indiscernibility in the RST (Yan et al. 2019). The decision-table information is expressed as D = (Ú, ą, V, f); where Ú represents the non-empty set, ą represents the non-empty attribute finite set, V is the range of ă, ă: Ú⇒ V is a single mapping function, f: Ú׹ ⇒ V, for each ă ɛ ą, ϰ ɛ Ú, then f (ϰ, ă) ɛ V. If the attributed set (ą) is based on C_T (condition attribute) and D_T (decision attribute) then C_T⋒ D_T=Ҩ, and S = (Ú, C_T⋒ D_T), where D_Tɛ C_T is the decision table. We used the Rosetta software for attribute reduction [C_T1, C_T2, C_T3, C_T4, C_T5, C_T6, C_T7, C_T8], ensuring the necessary attributes were included (Yan et al. 2019, 2022). These input parameters were used in the hydrodynamic model to simulate river bank erosion and riverbed incision (Section S1, Figure S2; Supplementary Information).

In HEC-RAS 2D modeling, geometrical cross-section profiles were essential inputs for simulating bank erosion and bed incision. The model necessitated precise boundary conditions at both ends to effectively emulate the river's hydrodynamics. The cross-sectional profiles delineated the river's geometric characteristics, encompassing channel and floodplain morphology. These profiles furnished critical data on bed elevation, bank locations, and topography, vital for accurate modeling of hydrodynamics, sediment dynamics, and morphological transformations. The upstream boundary employed hydrographs of runoff and SCs. These hydrographs supplied dynamic data on flow rate and sediment load for the model domain. This input was essential for modeling hydrodynamic processes, enabling simulations of flow variations and sediment influx effects on bank erosion and bed incision. The downstream boundary was characterized by water-level hydrographs. These hydrographs illustrated temporal water-surface elevation, crucial for determining model outflow conditions. The water-level data validated the model's representation of tailwater conditions, impacting the hydraulic gradient and flow velocity, subsequently influencing erosion and sedimentation dynamics. Through the integration of these variables – geometrical cross-section profiles to delineate the physical configuration of the river and hydrographs representing the upstream and downstream limits – the HEC-RAS 2D model successfully simulated the dynamic phenomena of bank erosion and bed incision with a significant level of precision. This method facilitated the model's ability to account for spatiotemporal variations in channel geometry and flow conditions, resulting in enhanced predictive accuracy regarding morphological alterations in the ZRYR.

The decision to choose the HEC-RAS model was primarily driven by its capability and simplicity in being incorporated with the RST model. HEC-RAS offers a robust framework that enables seamless integration with RST, thereby enhancing the effectiveness of data transfer and model functioning. The integration of various factors is crucial in accurately forecasting the processes of river bank erosion and riverbed incision. First, while TELEMAC and other models provide high-resolution subgrid capabilities, HEC-RAS has exhibited a well-documented history of success in comparable research endeavors, yielding dependable outcomes (Czuba et al. 2019). The ability of the HEC-RAS 2D model to establish geometric and hydraulic characteristics for cells and cell boundaries according to the terrain beneath guarantees a high level of precision when simulating intricate river networks such as the ZRYR. Second, HEC-RAS is commonly employed in hydraulic modeling for its user-friendly interface and comprehensive documentation. Third, the distinctive geomorphological and hydrological features of the ZRYR require a model capable of effectively managing intricate geometrical cross-section profiles, connections within the stream network, and varying hydraulic properties. HEC-RAS has been selected based on its capability to adequately address these criteria, thereby guaranteeing accurate simulation of runoff and changes in slope, geometry, and surface roughness at key locations. Lastly, HEC-RAS provides a level of flexibility in the delineation of upstream and downstream boundary conditions, a critical aspect in ensuring the precise portrayal of runoff and fluctuations in water depth within the ZRYR. The capacity for adaptability, in conjunction with the ability to introduce transverse inflows, facilitates a hydrodynamic simulation that is both thorough and precise.

The HEC-RAS 2D model was utilized to model bank erosion and bed incision in the ZRYR, referencing its use in analogous regional investigations. The model employed a heterogeneous mesh that integrates both structured and unstructured cells to represent the intricate geometry of the river channel and floodplain. The mesh dimensions ranged from 30 m in crucial regions to 50 m in less-critical areas, optimizing computational efficiency alongside model precision (Maingi & Marsh 2002; Benito et al. 2003; Chevalier et al. 2021). The boundary conditions of the model were established via hydrographs. The upstream boundary utilized runoff and SC to deliver dynamic flow and sediment load inputs. The downstream boundary employed WL hydrographs to accurately depict hydraulic gradients and flow behavior. Key parameters from RST were integrated into the model, such as flow velocity, shear stress, channel slope, width-to-depth ratio, and sediment load. These elements were essential for modeling erosion, sediment transport, and morphological alterations in the channel. The computational methodology entailed iterative computations at each temporal interval, adjusting shear stress, sediment transport rates, and modifications to the bed and bank per the model's governing equations. Upon the identification of bank degradation, the bank failure function was instigated. The model perpetually revised the channel morphology to capture dynamic alterations.

The aforementioned boundary conditions were incorporated into the HEC-RAS model to create a comprehensive simulation framework. In this configuration, the mesh derived from the morphological data functioned as the basis for modeling the hydrodynamic and sediment transport phenomena. To enhance the elucidation of the model configuration, it is imperative to incorporate a diagram that delineates the boundary conditions. This diagram would depict the upstream and downstream boundary conditions, illustrating the placement of input parameters such as discharge hydrographs, SCs, and water levels. Furthermore, it would signify the manner in which these boundary conditions were assimilated into the model, thereby offering a visual representation of the model's comprehensive architecture.

Discharge is simulated regularly for every grid cell of the contributing area using a seven-parameter model of the surface water budget. The seven-parameter model input data are rainfall and evapotranspiration on a spatial grid comprising a water balance storage basin, a separating basin, and two transfer storage basins. The daily output results for each grid are actual evapotranspiration, seepage, and runoff. Basin discharge is then routed daily to the channel cells by a module based on isochronism at 24h time-steps. The model was run on a heterogeneous mesh with structured and unstructured cells. The weighted mean elevation and roughness values across all intersecting cells of the 30 m roughness grid were incorporated into the model regardless of cell size. The two-dimensional diffusion wave solver was selected to provide more reliable numerical solutions to minimize computation time for the current scenario, and determine the prevailing barotropic and bottom-friction terms (Shustikova et al. 2019).

The computational process at each time-increment encompasses the subsequent steps: (1) incorporate the RST-based dominating factors as input parameters (flow velocity, shear stress, channel slope, width-to-depth ratio, and sediment load) into the model; (2) determine the shear stresses exerted by both the bed and the bank; (3) calculate the sediment transport rate and the capacity of each cell; (4) the governing calculations must be assessed to revise the sediment concentration and hydraulic variables at the subsequent time increment; (5) modification of the bed elevations and erosion rate of the banks should be determined according to τ_c; (6) in the event of bank degradation, it is recommended to activate the bank failure function; (7) keep the channel morphological changes updated; (8) restart the computation from step 1 at an alternative time instance; (9) repeat steps 1–8 iteratively until the simulation is concluded, as shown in Figure 2.

The upstream boundary condition was established based on runoff values over time, while the downstream boundary condition was developed using the water-level values over time. The measured values of water levels at certain X-SPs on the main reach of the YR and three X-SPs within the study range are all from the observations. Two sets of water-level observations were used for model validation. As previously stated, boundary conditions were set using known runoff values over time for the upstream boundary and water-level values for the downstream boundary. This was done for model calibration and verification across all studied scenarios. Model calibration involved varying the coefficient values to achieve good agreement among observed and estimated water levels.

The hydrodynamic model was calibrated by adjusting essential parameters, notably Manning's roughness coefficient, to align model outputs with empirical data. Manning's roughness coefficient is vital in hydrodynamic models, indicating water flow resistance from surface roughness. The coefficient n is surface-dependent; smoother surfaces yield lower n values, whereas rough, vegetated surfaces result in higher values. This research systematically modified Manning's n values during calibration to align computed water levels and SCs with observed data. The calibration process entailed systematic coefficient modifications to ensure the model aligned with observed hydrological conditions. For the ZRYR, Manning's coefficient for the main channel was established at 0.0344 m^−1/3 s, while for the floodplain, it was determined to be 0.059 m^−1/3 s. The values were derived to optimize the alignment of computed water levels with observed data from multiple cross-sectional profiles along the river, especially at the HYK station. Calibration was carried out by replicating the time interval 2018, which corresponds to the observed water surface elevation data period. For brevity, a detailed discussion about the time step, model stability, and model reliability is described in the Supplementary Information (Section S1).

We used Rosetta software to determine the dominating factors in the ZRYR, with the test period being 2018. The conditioned attributes were C_T1 (discharge), C_T2 (shear stress), C_T3 (velocity), C_T4 (roughness), C_T5 (gradient), C_T6 (precipitation), C_T7 (suspended sediment and BL), and C_T8 (width/depth ratio), with the decision attribute (D_T1) being erosion and deposition (Table 1). Using the naÏve algorithm, Rosetta software reduced the condition attributes to five attributes: C_T2 (shear stress), C_T3 (velocity), C_T5 (gradient), C_T7 (suspended sediment and BL), and C_T8 (width/depth ratio) (Table 2). Based on the minimum set rule, the forecasting pattern was achieved by analyzing the significance and frequency of rules. The results obtained from the forecast model were highly accurate, precisely representing the interrelationship between predicted factors and objects. For brevity, an in-depth discussion is not provided here but it is discussed by Pawlak (2002). The results based on RST could refine the sample structure and improve the classification of dominating factors. Therefore, RST effectively predicted the dominating channel factors based on available data. It is suggested that more channel morphological factors be included in future studies. These dominating parameters were input in the hydrodynamic model to predict bank erosion and bed incision in the ZRYR. The results illustrate that reducing unnecessary attributes in the RST can enhance the sample's structure and improve its accuracy in identifying dominant factors in the ZRYR morphology. Therefore, the RST technique is a pragmatic approach that effectively anticipates the primary influencers by analyzing the existing dataset. The discrepancy between the predicted and actual outcomes after implementing the rule-matching algorithm indicates a comprehensive prediction accuracy of 87.98%. The forecast's precision signifies the approach's viability and enhanced capability to identify the primary factors influencing changes in the river's morphology. The outcomes from RST analysis can improve the sample configuration and offer greater precision in categorizing the key factors that dominate channels. The summary explains that RST provides a wide range of opportunities to address the selection of influential factors impacting channel morphology. The aforementioned contemporary technique establishes various distinct approaches. Hence, leveraging existing data, RST made astute predictions about the prominent channel factors. It is recommended that additional morphological channel factors be incorporated into future research.

It should be noted that the accessibility of outer region flow data for runoff allowed for many simulations, yielding total, form, and skin drag components of shear stress across a different range of flow runoffs. The shear-stress components were linked to runoff, and the channel bank shear-stress value was observed to be smaller (≪10 Pa), even if runoff exceeds 25,000 m³/s. The findings are parallel with the lower channel gradients noticed on this large river. In addition, the importance of the form drag component in the shear-stress partitioning is evident, accounting for 61% of the total stress imparted on both banks. The right and left beds were observed to be incised to the elevation of 81.50 and 80.50 m, respectively, on the island, which separates the flow into right and left parts. All indicated dredging areas are 30 m wide, and the river bank and bed study are based on three scenarios.

Flood simulation using two-dimensional hydrodynamic models is influenced by various uncertainties arising from model configuration, boundary conditions, and parameterization. Sensitivity analysis, a widely employed technique in modeling, serves multiple purposes, including identifying and ranking influential parameters. In this study, sensitivity analysis was performed by varying several model-configuration factors, such as floodplain and channel roughness coefficients, terrain and mesh size, and river boundary conditions. Key parameters in hydrodynamic modeling, particularly the river section and roughness coefficient, play a crucial role in model accuracy. Roughness coefficients, which represent the resistance to flow within the river channel and floodplain, significantly affect water-level predictions. The terrain and mesh size determine the resolution of the model and can impact the accuracy of flow representation and the capture of critical topographical features (Sanders 2007). The sensitivity analysis revealed that as the flow rate increases, the roughness coefficient's sensitivity in determining the water level in the river reach also rises. This finding aligns with other studies highlighting the importance of accurately calibrating roughness coefficients to improve model reliability under varying flow conditions.

Numerous models are available for simulating river flow, such as HEC-RAS, CCHE2D, and MIKE (Dhakal et al. 2021). Developed by the US Army Corps of Engineers, HEC-RAS is one of the most renowned, extensively studied, and widely applied models for flood mapping in both scientific research and practical scenarios. Comparative studies have illustrated the strengths and weaknesses of HEC-RAS 2D compared with other models Muñoz et al. (2022); Orozco et al. 2024). First, HEC-RAS is generally precise for river hydraulic modeling, including complex flow patterns and sediment scenarios. Second, renowned for its equilibrium between computational effectiveness and precision, HEC-RAS 2D proves to be appropriate for a diverse array of applications, all without necessitating an abundance of computational resources. Third, HEC-RAS 2D is recognized as a highly intuitive modeling tool characterized by its comprehensive documentation, thus enabling its utilization by a diverse user base, even those lacking an extensive modeling background.

Morphological alterations, such as bank erosion and bed incision, have substantial ecological consequences for river ecosystems. In the realm of ZRYR, the ramifications can be far reaching and diverse, influencing both the aquatic ecology and the adjacent terrestrial habitats. Prior to the construction of the Xiaolangdi operation, water flowed through the active creeks in the LYR braided flood plain, which provided sufficient water for almost every morphological unit. During HFs, most morphological units are inundated with flood water and receive a fertile soil layer conducive to a sustainable terrestrial ecosystem. Water was also available in the main creeks filled from the previous HF period during low flows (LF). This water was pumped for use in agricultural practices within the accessible area and for maintaining the aquatic ecosystem. After the operation of Xiaolangdi, the discharge has been diverted to the power channel for electricity generation. However, despite the release of environmental flow, the following impacts have been induced after the morphological changes in the ZRYR:-

• (i) Habitat degradation/loss and negative impacts on migratory birds: Bank erosion and bed incision in the ZRYR may result in the degradation of aquatic habitats. The erosion of banks and beds has the potential to eliminate benthic habitats essential for macroinvertebrates and other aquatic organisms, thus causing a reduction in biodiversity. Erosion has the potential to compromise the stability of riparian areas, which play a crucial role in upholding the ecological coherence of river ecosystems. These regions provide essential habitats for a diverse array of flora and fauna while serving as protective barriers that strain out contaminants and fine particles.

• (ii) Altered hydrology: Bed incision has the potential to modify the flow patterns within a river through adjustments in water flow depth and velocity. Such alterations can influence the biological processes of aquatic species that have evolved to thrive under particular flow circumstances and can also intensify the consequences of both droughts and floods. Changes in the morphology of the channel bed may have an impact on the relationship between surface water and groundwater. The process of bed incision has the potential to decrease the water table, which in turn diminishes the access to groundwater for riparian vegetation and could have implications for the broader regional hydrological system.

• (iii) Reduction of fisheries development and reduced agricultural production: Erosion and sedimentation can potentially harm fish-spawning habitats. Fine sediments can obstruct the gravel beds frequently utilized by various fish species for spawning, consequently diminishing their reproductive efficacy. Bed incisions may lead to the formation of impediments to the movement of fish, causing the fragmentation of populations and limiting their ability to reach crucial habitats necessary for activities such as feeding, breeding, and seeking shelter.

• (iv) Deterioration of river water quality and depletion of groundwater table: Increased bank erosion leads to elevated sediment loads in the river, consequently diminishing water clarity and impacting photosynthesis in aquatic vegetation. Elevated sedimentation rates have the potential to suffocate benthic organisms and their habitats, ultimately resulting in a depletion of aquatic biodiversity. Sediments frequently transport associated nutrients and contaminants. Consequently, erosion can escalate the concentrations of these compounds in the aquatic environment, potentially resulting in eutrophication and the deterioration of water quality.

• (v) Negative impacts on flora and terrestrial wildlife: Bank erosion may result in the depletion of riparian vegetation, a critical element in supplying shelter, habitat, and sustenance for numerous species. The disappearance of trees and plants could also diminish bank integrity, consequently exacerbating erosion. Terrestrial species depend on riparian zones for both habitat and food sources. The deterioration of these areas has the potential to result in a reduction in terrestrial biodiversity and the disturbance of crucial ecosystem services like pollination and seed dispersal.

These environmental and social impacts have significant influences on the lives of the local population. The population has soared, which has increased the requirement for drinkable water, livestock, trees, and agriculture, ultimately dictating the review of the present environment flows downstream of XLD.

This study proposes a two-dimensional hydrodynamic model coupled with RST findings for both bank erosion and bed incision in the ZRYR, considering interactions between three separate processes. Five channel morphological dominating factors – velocity, stress, slope, channel width/depth ratio, and sediment load – were computed based on RST. The model includes the simulation of three modules: bank erosion, bed incision, and sediment transport based on hydrologic and cross-sectional data. The maximum NSE value for measured and computed runoff, sediment load, and water levels at HYK was 0.94, and the RMSE concerning runoff was 0.10 m, pointing to significant computational accuracy for the proposed model. The model can also precisely simulate the bank deformation, with computed and measured bank erosion/bed incision values at specific cross-sectional widths in close agreement. The highest computed post-flood reach scale width was 289 m/day, which is only 16 m less than measured values, and bed incision was observed to be 1 m/day or more than that at certain sections.

The scale and extent of bank erosion were relatively more than bed incision, which can cause a significant impact on longitudinal channel deformation. The high sediment load and LF velocity at certain cross-sections can cause large-scale bank erosion, while changes in sediment runoff and flow velocity have the opposite effect, leading to large-scale deposition in the study reach. The total volume driven by bank deformation accounted for a significant portion of the volume of river deformation volume, indicating that the bank erosion/deposition module is crucial in the computation of runoff, sediment transport, and channel deformation in the Zhengzhou reach of the YR. The HEC-RAS 2D model, coupled with RST findings, can simulate the bank erosion, bed incision and flow-sediment transport in a long channel reach. However, future modeling should consider the system-changing thresholds in natural and human-induced contexts.

Authors are very grateful to Chaode Yan for providing all resources and facilities during the research.

M. W. B., S.-e-h. S., S. S. A., H. X., and Y. Q. conceptualized the whole work, developed the methodology, wrote the original draft, wrote the review, and edited the article. C. Y. rendered support in funding acquisition, project administration, developed the resources, and supervised the study. M. D., X. Y., and L. W. wrote the review and edited the article, and visualized the process. H. X., Y. Q., and S. S. A. rendered support in literature search, data curation, and analysis of data.

This work was supported by the Science Foundations of Zhengzhou University, with grant (i) 32340370/22 Academician Team Research Launch Project (No. 13432340370), and (ii) First-class Project Special Funding of Yellow River Laboratory (No. YRL22IR13).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.