Introduction

Coastal wetland ecosystems are considered among the most ecologically and socioeconomically valuable resources because of their potential to provide enormous ecosystem services1. However, approximately 50% of natural coastal wetlands have been degraded or have disappeared worldwide due to intensive human activities (e.g., water projects and land reclamation) and climate change (e.g., sea level rise). This has undermined coastal resilience and resulted in rapid declines in biodiversity and coastal blue carbon storage2,3,4,5. It is urgently necessary to restore coastal wetlands and to mitigate their degradation and loss, particularly in large river deltas6. As coastal wetland restoration is usually complex and costly, identifying long-term ecological consequences is imperative to properly inform decisions on whether, how, and where to conduct wetland restoration7.

The relocation of a river flow path in a delta plain (i.e., by diverting river flow) has been proposed and implemented as a natural process-based measure for rehabilitating eroded deltas with supplemental sediments to build new land in estuarine areas8,9,10. For example, historic tributaries have been reopened to form sub-deltas in the Mississippi Delta to mitigate eroded wetlands11. The purpose of these river diversion projects in estuarine areas is basically designed to compensate for eroded coastal lands and thus reduce the maintenance dredging requirements for navigation channels by sand deposition and nourishment12,13,14. From the perspective of engineering investment, it is possible to use managed river diversions to increase the lifespan of created marshes and to decrease long-term dredging costs15. A recent research on the river diversion (i.e. Yellow River) addressed its potential societal impacts, indicating that the optimal location need consideration the co-evolution of natural fluvial processes and the balance between project costs and civil disruption caused by flooding16. There is still a lack of research on ecological feasibility for natural process-based restoration through river flow path relocation.

River-delivered sediments play a critical role in shaping delta morphology and supporting the growth of coastal wetlands17,18, but sand mining or upstream reservoirs can decrease or retain downstream sediments and stagnate the natural growth of deltas19,20. Consequently, insufficient amounts of sediment combined with sea level rise could disrupt the balance between sediment deposition and coastal erosion and result in delta sinking10,21,22. Changes in upstream hydrological and sedimentary conditions significantly shape the morphological evolution of a river delta and determine subsequent vegetation succession and distribution patterns3,23,24. In addressing natural process-based mitigation for coastal wetlands by relocating river flow paths, we need to understand how to take advantage of delta morphological dynamics in promoting ecological succession and coastal resilience of coastal wetland habitats12,25.

Relative hydro-morphodynamic modeling studies provide the possibility of predicting and quantifying the land-building process and sediment fluxes after river flow path relocation26,27. Although understanding coupled natural processes of sediment deposition, morphological dynamics and vegetation succession would facilitate prediction and assessment of ecological consequences for deltaic wetlands8, such simulation predictions covering entire evolutionary process are rare. For instance, the interplay of erosion, sedimentation and vegetation dynamics in tidal embayment has been explored28 and the ecological, geomorphic, and climatic factors influencing salt marsh evolution have been reviewed29, these studies mainly focus on individual processes and lack the integration of hydro-geomorphological and ecological processes in addressing entire eco-morphological evolution of new born coastal wetlands30,31. In addition, the importance of vegetation dynamics in promoting establishment of marsh seedlings has been targeted32,33, there is still a lack to explore feasibility of river flow path relocation employed as a natural process-based restoration by integration of geomorphological modeling and habitat suitability assessments.

Our study therefore addressed to elucidate the natural processes of new estuarine wetland formation and subsequent vegetation succession driven by possible river flow path relocation, and thereby to explore their potential ecological feasibility and mitigation effects used as a mean of ecological restoration. The Yellow River Delta (YRD) wetlands was employed as study area due to frequent diversions, continuous dynamic changes driven by its rich sand deposition in estuarine area and the need for ecological restoration to mitigate the wetland loss and degradation. In the past two decades, a total of 233 km2 of coastal wetlands have been restored in the YRD mainly from agricultural & aquacultural fields and salt pans, however this preplanned, pattern-based restoration sites accounts for only 27% of lost or degraded wetlands caused by large-scale oil extraction and intensive agricultural and aquacultural development in the past half-century34,35. On the other hand, the Yellow River brings large volumes of sediments from the Loess Plateau to the river mouth and has created one of the largest and most dynamic river deltas in the world, with only 10 ~ 12 years of periodical natural and artificial flow path relocation36,37. Siltation deposition in the current estuary has resulted in unstable development of the end-flow branch, which has shortened the lifespan of the end-flow branch and has been harmful to flood control38. Owing to the lack of replenishment of sediments and intensified sea level rise, those historically used channels (e.g., the Diaokou River channel) have experienced severe coastal erosion, with a total loss of tidal wetlands of nearly 300 km2 from 1976 to 201939.

To explore ecological consequences and feasibility of natural process-based coastal wetland restoration by relocating this world’s most dynamic river flow path, it is necessary to understand the temporal‒spatial dynamics of this coastal wetland area and the ecological consequences of the various proposed river flow relocation scenarios in the YRD. In this study, we explored the ecological feasibility of natural process-based restoration in relation to coastal wetland dynamics driven by three flow path relocation scenarios (Fig. 1) using an integrated numerical simulation framework of the Delft3D and the Gaussian model, including the Diaokou River pathway (DK) scenario (historically used from 1964 to 1976 and prioritized as a backup end-flow by the Yellow River Conservancy Commission), the Qingshuigou Southeast pathway (QS) scenario (historically used from 1976 to 1996), and the Qingshuigou Northeast pathway (QN) (the existing flow path since 2007). This study aims to address the following objectives: (a) to establish a numerical model of the YRD to simulate the vegetation distribution processes following deltaic morphological evolution under different river flow path relocation; (b) to link and characterize the temporal‒spatial evolution of delta morphology, coastal wetlands and subsequent vegetation habitats across entire natural processes; and (c) to quantify and compare natural process-based restoration effects for wetland mitigation under various flow path relocation scenarios.

Fig. 1: Overview of the model setup for the Yellow River Delta.
figure 1

a Bathymetric profile and model grids with observation and tidal stations; b river flow path relocation scenarios, including current and proposed paths; and c upstream boundary conditions. QN the current existing pathway, QS the Qingshuigou Southeast pathway, DK the Diaokou River pathway. Solid lines denote the current river flow path, while dashed lines indicate the proposed relocations in this study.

Full size image

Results

Sediment deposition evolution

The deepest sedimentation occurs at the seaward front of the channels, indicating that sedimentation primarily occurs through advancing flow, spreading and accumulating along the sides of the inflow channels. In scenario QN, the main distribution area of sediment deposition extended gradually from the nearshore estuary to the offshore sea (Fig. 2a). However, in scenarios QS and DK, large amounts of sediment were deposited relatively far from the estuary in the early stages (Fig. 2b, c). The rate of increase in the deepest deposition depth under scenarios QN, QS and DK was 1.28 m yr−1, 1.17 m yr−1, 0.87 m yr−1. In all three scenarios, the rate of change in deepest sediment depth transitions from rapid to gradual, with the DK scenario initially changing more slowly than the others.

Fig. 2: Hydro-morphodynamic modeling of deltaic deposition under three river flow path scenarios.
figure 2

The figure illustrates the deposition processes of the newly formed delta under the three flow path relocation scenarios: a QN (the current existing pathway), b QS (the Qingshuigou Southeast pathway) and c DK (the Diaokou River pathway). Each two rows correspond to a scenario, with columns representing different years labeled as 2a to 10a. White areas indicate the original delta.

Full size image

The net deposition volume in each grid were calculated on the basis of sedimentary thickness (such as Supplementary Note 1). The total amount of net sediment deposition volume in the newly formed wetlands under the three flow path scenarios is relatively close (1.48~1.51 km3) (Fig. S1). In the two Qingshuigou scenarios, sediment deposition is narrowly concentrated with high intensity. In the DK scenario, it is widely dispersed with low intensity (Fig. 2).

Morphological evolution

In all three scenarios, the newly formed deltas extended seaward with increasing morphological elevation (Fig. 3). The rate of increase in the highest elevation of the wetland under scenarios QN, QS and DK was 0.070 m yr−1, 0.067 m yr−1, and 0.016 m yr−1 and the highest elevation in the 10th year of the model simulation was 1.381 m, 1.569 m and 2.513 m.

Fig. 3: Morphological evolution of a newly formed delta under three river flow path relocation scenarios.
figure 3

The figure illustrates the morphological changes in elevation dynamics (m) for the newly formed delta under three distinct flow path scenarios: a QN (the current existing pathway), b QS (the Qingshuigou Southeast pathway) and c DK (the Diaokou River pathway). Each two rows correspond to a scenario, with columns representing different years labeled as 2a to 10a. White areas indicate the original delta.

Full size image

In scenario QN, two flow paths were created in the first year of morphological evolution, but the west-side path was gradually filled by sediments at the end of the 7th year, and the delta extended seaward, which was driven only by the east-side flow path (Fig. 3a). Two stable flow paths also emerged in scenario QS and extended eastward in early stages, with the path on the north side changing direction and gradually stretching northward, leading to a large newly formed delta area on the north side of the estuary in the later stages of morphological evolution (Fig. 3b). The newly formed delta in scenario DK only developed one stable flow path throughout the whole simulation period, and the delta was an elongated shape that extended with the channel (Fig. 3c).

Evolution of newly formed wetlands

The newly formed delta area could be identified based on the relationship between the water level and geomorphic elevation during spring tides (Fig. 4). Our results reveal that intertidal wetlands initially formed in the estuary area. As the delta expanded and the elevation rose, the newly formed intertidal wetlands and channels near the estuary began to transform into supratidal wetlands (Fig. 4). Even though a larger area of intertidal wetlands formed during the early period in scenario DK, it also began to form supratidal wetlands during the middle period. Nevertheless, the time at which supratidal wetlands begin to form in scenario QS was earlier than that in the other two scenarios. Then the tidal wetlands grew and lengthened seaward around the flow paths driven by morphological evolution processes (Fig. 4a). The DK scenario created the largest amounts of supratidal (35 km2) and intertidal (148 km2) areas, followed by both (28 km2 and 91 km2) areas of the QS scenario, respectively, compared with both having the lowest values in the QN scenario (6 km2 and 50 km2, respectively) by the end of the simulation (10th yr) (Fig. 4b).

Fig. 4: Dynamic evolution of newly formed delta wetlands under three river flow path scenarios.
figure 4

a The spatial distribution of intertidal (green) and supratidal (orange) zones. The sequence of maps for each scenario shows the expansion and transformation of these areas over time. White areas indicate the original delta. b Two graphs depict the temporal changes of the supratidal area (left) and the intertidal area (right) in wetland area. QN the current existing pathway, QS the Qingshuigou Southeast pathway, DK the Diaokou River pathway; a: year.

Full size image

The intertidal wetlands under scenario DK that increased from south to north, expanded most rapidly in the early period (1st year-3rd year), and accumulated the largest area throughout the morphological evolution. Although supratidal wetlands occurred late under scenario DK, they expanded rapidly in the later stage and finally created the largest area of tidal wetlands (Fig. 4). In comparison, the newly emerged delta under scenario QS expanded with the growth of both intertidal and supratidal wetlands in the early stage (Fig. 4a); however, the intertidal wetlands tended to stagnate with accelerated growth in supratidal areas during the 2nd year~4.5th year period, indicating trade-offs between the growth of intertidal and supratidal wetlands with limited sediment budgets. The growth rates of the delta were the lowest in scenario QN, with the smallest values of both intertidal and supratidal wetlands (Fig. 4b).

Evolution of vegetation habitats

A suitable growth habitat mainly emerged in the newly formed wetlands and had an across-shore distribution along the shoreline throughout the process of morphological evolution (Fig. 5a). Under scenario QN, suitable growth habitat mainly formed in the western part of the estuarine area and created a large area of zonal intertidal wetlands around the estuarine area in the later stage of morphological evolution (Fig. 5a). The suitable growth habitat in scenario QS formed into a strip shape during early morphological evolution and slowly expanded seaward following evolution processes (Fig. 5a). The suitable growth habitat also formed an across-shore distribution under scenario DK (Fig. 5a), and a strip shape developed until the middle stage of morphological evolution because of the lack of newly formed wetlands during early evolution. The spatial and temporal patterns of suitable growth habitats were largely determined by the growth rates and patterns of the newly formed wetlands under each scenario. For example, the loss of tidal wetlands may lead to a decrease in suitable growth habitats under the QS scenario during early and midterm morphological evolution. In all three scenarios, the total values of the optimum and suitable growth habitat areas under the DK scenario were the largest, whereas those under the other two scenarios were similar. In addition, the areas of optimum growth habitat under scenarios QS and QN were much more similar than the amounts of suitable growth habitats. In particular, the areas of optimum and suitable growth habitats were almost equal under scenario QS, indicating their relatively high habitat suitability for pioneer tidal vegetation (Fig. 5b).

Fig. 5: The evolution of the Suaeda salsa habitat suitability under three flow path scenarios. under the three flow path relocation scenarios.
figure 5

a The evolution of the Suaeda salsa vegetation habitat suitability. b Two line graphs depict the temporal changes in the optimal and suitable areas for Suaeda salsa growth. QN the current existing pathway, QS the Qingshuigou Southeast pathway, DK the Diaokou River pathway; a year.

Full size image

Discussion

In this study, we simulated the distribution patterns of newborn wetlands and vegetation in the YRD estuary within 10 years after river diversion under stable water and sediment transport conditions after the implementation of the water‒sediment regulation scheme project. The differences between the model results and the actual subaerial geomorphological evolution and vegetation succession have been verified. Based on the Yellow River Water Resources Bulletin, in the decade following artificial diversion to the QN in 1996, the sediment discharge experienced a dramatic change from 0.02 Gt yr−1 to nearly 0.44 Gt yr−1. In 1997, the period of no flow lasted 226 days at the Lijin hydrological station, and the river channel with no flow extended 700 km upstream from the river mouth40. The active YRD lobe decreased considerably in area from approximately 760 km² in 1996 to nearly 735 km² in 2002, with a mean erosion rate of 4.28 km² yr−141. However, this erosion scenario did not align with the model results, which was due to the setting of the discharged sediment concentration in the model. After the initiation of the water‒sediment regulation scheme project in 2002, the actual sediment‒water ratio matched the model’s simulated scenario, and the average accretion rate was approximately 6.56 km² yr−1 between 2003 and 201141. In our model simulation results, the QN scenario added 56 km² of new wetland area over 10 years, with an average growth rate of 5.6 km² yr−1.

Furthermore, the QS promontory generally grew during the period from 1976–2000 following artificial diversion to QS, and its total area increased by 384.16 km², with a mean net accretion rate of 16.4 km² yr−141. In our model simulation results, the QS scenario added 123 km² of new wetland area over 10 years, with an average growth rate of 12.3 km² yr−1. Considering that the cumulative area of increase over the entire subaerial delta and the cumulative Yellow River sediment discharge at Lijin were calculated and that the results show that they are closely related42, the increase in the decade after 1976 was significantly greater than that in the decade after the implementation of the water‒sediment regulation scheme project. Additionally, previous studies revealed that the interdecadal growth in sediment accumulation rates in the QS abandoned river mouth was 5.77 × 10⁸ m³ yr−1 from 1976–198543, which was greater than that of the river delivered by the QN after 199644. Therefore, it is acceptable that the results of this study are lower than the historical values.

Vegetation patterns have been well documented in the YRD, S. salsa is a major plant specie in the YRD, forming belts along the coast45,46. In particular, the vegetation also demonstrated the same distribution pattern around the Qingshuigou course47. This vegetation pattern is notably consistent with our modeling results (Fig. 5). Concomitant with the expansion of the new wetland, the zonal distribution of S. salsa in the new estuary has also gradually shifted seaward, which aligns with the outcomes of our model (Fig. 5).

The YRD is a typical river-dominated delta where the marine energy is weak and the sediment is mainly composed of cohesive silt48,49. Our results indicate that the deltas formed with few stable narrow channels and that the delta shape was elongated with the channel (Fig. 3), which is consistent with previous field observations and numerical studies50. The fine-grained, cohesive sediment inputs have grain sizes with high values of critical shear stress for erosion and slow values of settling velocity, which stabilize the channels and lead to fine-grained sediments being transported far into the basin, resulting in the delta being dominated by the process of levee progradation (Fig.2), thus creating a smooth delta front and elongated delta shape50.

As a river-dominated delta, the YRD is usually influenced by runoff and sediment supply which were identical settings across all scenarios. Here, we suggest that the primary causes of the variation were the initial bathymetry. We illustrate the initial bathymetry at the estuary in each scenario (Fig. 1), and it is clear that scenario QN had a far steeper initial bathymetry than scenarios QS and DK. Owing to the delta’s steeper initial bathymetry (as in the QN scenario), which requires more sediment deposition for delta growth48, leading to a higher rate of deepest deposition depth (1.28 m yr−1), a smaller deposition area (Fig. 2), and slower initial wetland formation (5.6 km2 yr−1). In contrast, a more gradual shelf slope (as in the DK scenario) allows for more rapid delta growth (18.3 km2 yr−1) and a lower rate of deepest deposition depth (0.87 m yr−1). The formation of two stable channels in some cases (e.g., QS scenario and early stage in the QN scenario) can be attributed to the initial bathymetric conditions and the hydrodynamic processes. Compared to the QN scenario, the QS scenario’s gentler initial bathymetry accompanied slower sediment deposition (Fig. 2), enabling the possibility of dual-channel operation. Conversely, in contrast to those in the DK scenario, the two stable channels in the QS scenario may have enhanced sediment deposition and accelerated delta growth, causing supratidal wetlands to form earlier.

The interaction between sediment deposition and coastal erosion governs how the delta progrades51. As a result, we further computed the overall volume of sediment deposition and erosion for each scenario (Fig. S1). Although their total net deposition volumes were similar, the differences in the newly formed wetland areas could be explained by the variations in the magnitudes of coastal erosion across the scenarios (Fig. S1). In particular, scenario DK had the least amount of erosion, whereas the QS and QN flow routes experience the most erosion (Fig. S1).

The comparison of potential ecological benefits among the scenarios could provide valuable decision-making information concerning flow path relocation. Obvious advantages can be observed for all evaluation elements (newly formed wetlands, S. salsa growth area and net deposition volume) in scenario DK in the 10th year (Figs. 4, 5, S1). The ecological benefits achieved in scenario QS were less than those in scenario DK but much greater than those in scenario QN (the existing entrance to the Bohai Sea). Our results reveal that the sedimentary deposition process around estuarine areas would slow down in later stages of morphological evolution, resulting in the tardy growth of coastal wetlands. This was verified by the performance of the flow path of the QN scenario, which has already shown stagnant morphological development since the artificial river diversion project to facilitate oil exploitation in 1996. In conclusion, these ecological benefits across the scenarios indicate that the least ecological benefits are likely to occur if the use of the current existing flow path is maintained, whereas the best ecological benefits could be achieved by restoring and reusing the DK flow path.

Under the three scenarios, 57.5~183 km2 of new coastal wetlands could be created (Fig. 4). Among them, scenario QN would form the least tidal wetlands. Although the restoration effects on the tidal area across the scenarios are limited compared with the loss of wetlands mentioned in the introduction, multiple ecological services could be achieved and amplified with pioneer vegetation (e.g., S. salsa) in newly emerged tidal wetlands. For example, newly formed salt marshes could enhance habitat suitability for shorebirds by 76% and increase carbon storage by 0.18 kg C m−2 and 0.55 kg C m−2 compared with seawater and tidal flats, respectively52. The nitrogen and phosphorus removal efficiency of newly formed salt marshes dominated by S. salsa could be improved by 0.09% ~ 0.2% and 0.31% ~ 0.45%, respectively, relative to that of tidal flats and seawater52. The areas of newly formed wetlands under scenario DK were 1.5 and 2.1 times greater than those under the QN and QS scenarios, indicating that the increase in ecosystem services is greater. This demonstrates and justifies the rationale for the DK scenario to be implemented given its best ecological benefits. In general, by linking multiple processes of sedimentary deposition, morphologic evolution, and habitat succession, our research demonstrates that the optimized relocation of flow paths in river deltas (e.g., the DK scenario) can facilitate the growth of new estuarine wetlands.

We proposed a general framework for natural process-based restoration schemes by relocating river flow paths in a river-dominated delta (Fig. 6). Optimizing flow paths in river deltas can boost sediment deposition, thereby fostering the growth of coastal wetlands and the development of typical tidal salt marshes. This strategy can effectively mitigate the loss of tidal coastal wetlands caused by intensified coastal erosion — as seen in the northern abandoned delta lobe, which has suffered significant erosion since 1976 due to the lack of sediment supply from the Yellow River42 — and by land reclamation (e.g., oil and aquacultural development) (Fig. 6). In addition, our research also demonstrates the necessity of combining innovative natural process-based strategies with traditional pattern-based restoration strategies to achieve adequate and efficient mitigation for the loss of coastal wetlands (e.g., the scrolling development mode) (Fig. 6).

Fig. 6: General natural process-based restoration schemes and implications by relocating river flow paths in a river-dominated delta for sustainable resource use under sea level rise (SLR) and land reclamations.
figure 6

The figure illustrates a universal four-step process of river flow path relocation and its implications for ecological restoration and resource management in a river-dominated delta environment. Each panel represents a different stage of the restoration process. The arrows indicate the direction of change and the processes involved. The color coding (e.g., blue for water, brown for land, green for vegetation, brown and black for land reclamations) helps differentiate the various elements of the delta ecosystem.

Full size image

Pattern-based and process-based restoration strategies are two major approaches to restore degraded ecosystems. Pattern-based restoration focuses on restoring the physical and biological structure of an ecosystem, including its landscape patterns53, species composition54, and habitat structure55. This approach emphasizes the importance of achieving specific, predetermined ecological targets or goals (e.g., returning farmland to wetlands and removing some fish farms)56 and is more appropriate for restoring highly degraded or simplified ecosystems57. Process-based ecological restoration, on the other hand, focuses on restoring the ecological processes that underlie ecological functions58, such as nutrient cycling, hydrological and energy flow, and biotic interactions59. This approach emphasizes the importance of restoring the dynamic, adaptive capacity of ecosystems to respond to environmental change and anthropogenic disturbances60 and is more appropriate for restoring complex or dynamic ecosystems to enhance ecological processes and self-sustaining ecosystems60,61,62. The choice between pattern- and process-based ecological restoration strategies depends on the specific goals of the restoration project and the characteristics of the degraded ecosystem being restored60.

Our research indicates that a large area of intact wetlands could be generated by natural process-based restoration (i.e., flow path relocation) in undisturbed lower tidal areas (Fig. 4), in contrast to prevailing pattern-based restoration for coastal wetlands through the return of agricultural and aquacultural fields. Long-term resilient, and innovative solutions for coastal management need to work with the natural processes inherent in coastal ecosystems63. However, our research revealed that only 22% of the human-induced loss of coastal wetlands could be compensated for by natural process-based measures (i.e., flow path relocation) under the DK scenario34 (Fig. 4). This may demonstrate that ecological restoration projects may need a combination of both pattern-based and process-based approaches to obtain the best ecological benefits64.

In addition, successful ecological restoration also requires careful consideration of the socioeconomic context to ensure feasible and achievable goals for resilient coastal restoration58. These goals should incorporate not only ecological benefits but also consider socioeconomic demands such as flood control, coastal defense, and ecotourism and thus mitigate the conflicts between ecological benefits and socioeconomic development. Furthermore, our natural process-based restoration schemes provide a method to identify the appropriate magnitude for the “scrolling” sustainable resource use of coastal wetlands (Fig. 6) that addresses unavoidable socioeconomic development (e.g., land reclamation) scales, which should be controlled under predicted rates of tidal wetland growth to maintain “no net loss” of tidal wetlands65.

In this study, all the scenarios studied were defined based on the “single-flow path” framework, as our test simulation indicated the invalidation and crumbling of the proposed “dual flow paths (DK & QS)” due to severe flow path blockage by river-delivered sand deposition. The sediment deposition process would have been considerably accelerated because of the divided water flow allocated to the two flow paths into the Bohai Sea. In addition, none of the historical records concerning long-term stable dual or multiple flow paths in the YRD have demonstrated the feasibility of using multiple flow paths66. Future research is suggested to explore other possibilities of river flow relocation, such as the Beicha channel. This is another alternative mouth channel located at the QN active river mouth, as the Beicha channel presents an alternative flow path with potentially lower channel migration costs. We acknowledge that while our model provides valuable insights into the distribution of S. salsa, a comprehensive assessment of ecological benefits would require a more holistic approach that considers multiple species and additional evaluation elements. This study is a relatively short-term simulation of 10 years that ignored mutual feedback and focused more on the influence of topography on vegetation patterns. In addition, the feedback of S. salsa to topography is weak67 because of its small vegetation size. The simulation of vegetation distribution types, including P. australis and S. alterniflora, should be expanded, and the mutual feedback between geomorphology and vegetation should be considered. Recent research has revealed that the cohesive and sandy sediment composition in the YRD varies spatially, with a D50 ranging from 6 to 120 μm68. This finding underscores the importance of incorporating a diverse range of sediment grain sizes in future geomorphological simulations to more accurately reflect the nearshore conditions of the delta. We are committed to addressing these limitations in our ongoing and future research.

Although this work attempted to address natural-based wetland restoration by flow path relocation in the YRD, we suggest that our theoretical and methodological framework can be replicable for silt-rich river-dominated deltas where natural coastal wetlands have been threatened by socioeconomic development and are vulnerable to climate change (e.g., sea level rise).

Conclusions

River flow relocation historically was performed primarily for socioeconomic purposes, such as flood control, navigation enhancement, and land resource creation within river deltas, and its potential application as a natural process-based restoration measure has rarely been targeted. To address this gap, as a novel natural process-based restoration method to the best of our knowledge, we developed an integrated methodological framework for spatially explicitly projecting the ecological outcomes driven by river flow relocation. This research predicted the evolution of delta morphology, wetlands and vegetation habitats based on three flow path scenarios in the YRD during the cumulative simulation time. The results reveal that intertidal wetlands formed first at the beginning of evolution and then transitioned into supratidal wetlands when the deltas evolved. Obvious advantages can be observed for all evaluation elements (newly formed wetlands, S. salsa growth area and deposition volume) in scenario DK in the 10th year. Nevertheless, only 22% of the human-induced loss of coastal wetlands could be compensated for by natural process-based measures under the DK scenario, which demonstrates that better ecological restoration projects may need to be a combination using pattern-based approaches. Finally, we propose a natural process-based restoration scheme based on river flow path relocation and implications for sustainable resource use in a river-dominated delta. In conclusion, our research coupled multiple natural processes and explored the long-term ecological consequences of relocating river flow paths in a typical river-dominated delta, which is particularly replicable for silt-rich and dynamically migratory rivers worldwide.

Methods

The variation in the Yellow River flow path into the sea and modeling scenarios

In the past half century, the flow path of the Yellow River first utilized the DK and drained into the Bohai Sea as a consequence of river relocation in 1964 and then it was artificially modified and shifted southeastward to the QS in 1976 to protect the northern part of the Shengli oil field (the third largest national oil field) from river flooding. To reclaim land from the sea, the flow path was rediverted northeastward to the QN in 1996, creating a large amount of new land to facilitate offshore oil exploitation36. Although the existing Qingshuigou flow path (i.e., QN) had already stably operated for four decades and had produced a large area of tidal wetlands, it has been considerably encroached upon and polluted by intensive oil exploitation. In 2013, the “Comprehensive Plan for the Yellow River Basin (2012-2030)” was officially approved by the State Council of China to primarily use the Qingshuigou River, maintain the stable flow of this channel, and then use the Diaokou River as a priority backup flow option. There is a need to explore alternative flow paths that could be used to prevent human impacts on tidal wetlands, particularly considering the insignificant role of the existing flow path for serving oil development due to the limitation of exploitable oil resources.

Considering historical river migration and the possibility of flow path changes in the YRD, we assumed three single river flow relocation scenarios (Fig. 1), including the DK and QS, which were both historically used as the end-flow paths during the years of 1964-1976 and 1976-1996, respectively. The QN, the currently used river flow pathway, was employed as the baseline. In the model simulation, the Yellow River chooses one path from the upstream boundary as its outlet flow path to the sea, with other paths being normal terrain without river flow pathway (Fig. 1). The three hypothetical scenarios could be realized through future artificial diversions.

Hydro-morphodynamic model

As a process-based hydro-morphodynamic model, the Delft3D model has been increasingly used in simulating estuarine-deltaic morphological evolution17. In this study, we employed a two-dimensional depth-averaged hydro-morphodynamic model to simulate a hypothetical river-dominated delta69. This type of 2D model is commonly used for modeling delta morphology and has been validated by comparing discharge ratios between bifurcating channels, bed heights of bifurcating arms, and flood maps with field observations47,48,50,70. The Delft3D model was used to characterize delta morphological evolution in the YRD considering the hydrological processes of river and tidal flows, estuarine geomorphology and sediment deposition.

Modeling domain and boundary

Owing to a lack of tidal data, the modeling domain was set greater than the YRD to incorporate those coastal cities (e.g., Lushunkou and Penglai), where more tidal data along the Bohai Sea could be obtained from their tidal stations (Fig. 1). The upstream open boundary of the model is set at the Yellow River estuary where it enters the Bohai Sea. The model has one river input boundary at the estuary and the consistent open sea boundary in each scenario (Fig. 1). To compromise between the modeling resolution and computational capacity, a total of 706,446 grids with sizes ranging from 50 m*50 m~500 m*500 m were created to ensure higher nearshore resolution and coarser offshore resolution (Fig. 1).

Modeling parameters

As the wave magnitude and strength in the YRD are relatively weak in the nearly enclosed Bohai Bay, the effect of waves on estuarine geomorphological development is insignificant, and thus it was excluded to simplify the modeling71. The bathymetric data of the Bohai Sea were obtained and digitized from the 1:250,000 Sea Charts of the Bohai Sea, which were published by the People’s Liberation Army Navy in 2012 and 2013. The elevation data used in the hydro-morphodynamic simulations are referenced to a geodetic datum (GPS reference level), which were consistent with the Sea Charts, to ensure accurate topographic and bathymetric representation. Given the absence of tide gauges and long-term tidal data near the YRD estuary, and the need for accurate tidal data as boundary conditions for the hydro-morphodynamic model of the estuary, the sea boundary tidal conditions are controlled by the tidal harmonic constants K1, O1, M2, and S2 of nine tidal stations in the Bohai Sea, including Dongying Port (Fig. 1). Since runoff and sediment loads have become relatively stable after the water‒sediment regulation scheme project was initiated upstream in 2002, we averaged river discharge and sediment concentration data obtained from the Lijin hydrological station (Fig. 1), which is the last hydrological station on the Yellow River, to define river discharge. As empirical coefficient, the discharged sediment concentration was calculated as 70% of the multiyear average data72. The sediments supplied to the YRD are mostly fine-grained cohesive silts with a median grain size(D50) of approximately 0.03 mm49. The erosion and deposition processes of cohesive sediment were considered following the Krone-Mehta-Partheniades formulations73, with the critical shear stress for both processes set to be the same as the previous defined50. The cohesive sediment critical shear stress for erosion and deposition are set at 0.7 N m−² and 0.1 N m−².

The time step was set as 0.02 min, with the cumulative simulation times were set at 10 years. The morphological scale factor was set as 60 to accelerate the morphological evolution. The hydrodynamic field and sediment deposition calculation were initiated using a constant flow cold start method. After 12 hours of stable computation, the morphological evolution processes were then calculated by the hot start calculation pattern based on the results obtained.

Model validation

Near the YRD, tidal currents are predominantly irregular semidiurnal. The spring tide range in the study area is about 1 m, while the neap tide range is approximately 0.5 m68. Based on the Yellow River Water Resources Bulletin, the average measured river discharge and sediment concentration at the Lijin hydrological station of the Yellow River from 2005 to 2022 were 667.91 m³ s−1 and 8.97 kg m−³, respectively.

The simulated versus observed data were compared to verify the hydrodynamic geomorphic modeling results (Fig. 7). The observed water level, depth-averaged flow velocity and flow direction data were obtained from a self-built observation station on the KD47 platform (119.18°E, 37.94°N)71 in 2012.6. The depth-averaged suspended sediment concentration data were collected from the observation station (119.02°E, 38.16°N) established by the First Institute of Oceanography of the State Oceanic Administration in the waters north of the YRD74. This station uses a tethered setup to suspend turbidity meters (OBS-3A) in the water. The measured data were calibrated in the laboratory using sediment from the study area, and the NTU values (dimensionless parameters that reflect turbidity) obtained from the OBS were converted into a time series of the suspended sediment concentration in 2013.4.

Fig. 7: Hydro-morphodynamic model validation against observed data.
figure 7

This figure presents the time series comparison between observed data (solid lines) and simulated results (dotted lines) for hydrodynamic parameters. a depth-averaged water level, b flow velocity, c flow direction and d suspended sediment concentration.

Full size image

The model accuracy is analyzed using the root-mean-square error (RMSE), the correlation coefficients (CC), and the skill score (SS), with formulas as follows:

$${{{\rm{RMSE}}}}=\sqrt{\frac{\sum {\,({X}_{m}-{X}_{o})}^{2}}{N}}$$
(1)
$${{{\rm{CC}}}}=\frac{\sum \,({X}_{m}-\overline{{X}_{m}})\,({X}_{o}-\overline{{X}_{o}})}{{\left[\sum {\,({X}_{m}-\overline{{X}_{m}})}^{2}\sum {\,({X}_{o}-\overline{{X}_{o}})}^{2}\right]}^{1/2}}$$
(2)
$${{{\rm{SS}}}}=1-\frac{\sum {\,({X}_{m}-{X}_{o})}^{2}}{\sum {\,({X}_{m}-\overline{{X}_{o}})}^{2}}$$
(3)

Where \({X}_{m}\) and \({X}_{o}\) denote model simulations and observations, respectively. \(N\) is the number of data points, and the overbar − indicates the arithmetic mean. When 0.2 < SS < 0.5, it indicates good simulation results. When SS > 0.5, it indicates excellent simulation results75,76.

As shown in Fig. 7, as the simulation progressed, the model results stabilized corresponding observed data start from noon on June 18th, and the water level, velocity, and flow direction exhibited insignificant discrepancies and temporal trends similar to those of the measured data. During the steady-state period of the model simulation, the RMSE values are 0.30 m, 0.10 m s−1, 69 Degree, and the CC are 78%, 72%, 70%, the SS values are 0.53, 0.51, 0.91 for water level, flow velocity, and flow direction (Fig. 7a–c). The verification of the sediment concentration reveals that the simulated values are of the same order of magnitude as the measured data. As shown in the Fig. 7d, the RMSE is 0.09 kg m−3, the CC is 69%, and SS is 0.78. For sediment transport models of other delta evolution studies, SS of 0.6–0.8 was reported for the Mississippi River Delta26, indicating good performance, while SS above 0.75 was achieved for the YRD70, reflecting high accuracy. Thus, the model and parameters developed in this study accurately represent the actual hydrodynamic and sediment conditions in the Yellow River estuary, indicating high and acceptable modeling results.

Identification of newly formed delta wetland types

The newly formed wetlands can be differentiated into supratidal and intertidal wetlands according to the relationship between the water level and geomorphic elevation from the hydro-morphodynamic simulation results of the Delft3D model at every time step in the estuary area. The newly formed wetland type was identified by the flooding conditions during the spring tidal period. Supratidal zones are defined as elevated uplands that remain unaffected by spring tides and are typically dry, identified as newly emerged grids that become wet only under extreme conditions. In comparison, intertidal zones are only periodically submerged by tides at intermediate elevations and therefore can be identified as emerging grids that experience a recurrent alternation of wetting and drying.

Simulation of vegetation habitats

The Gaussian model is commonly used to predict pioneer vegetation dynamics (e.g., biomass) driven by habitat environments77. For coastal wetlands, the impacts of ocean tides along wetland elevation gradients cause drastic changes in habitat environments and create vegetation patterns. In particular, the tidal regime determines water depth and flood frequency and fundamentally shapes the distribution pattern of tidal vegetation78. Functioning as the dominant tidal vegetation and the indicator habitat for shorebirds in the succession and biodiversity of ecosystems as highly salt-tolerant, Suaeda salsa (S. salsa) was employed as the indicative vegetation to assess the habitat quality of tidal wetlands in the YRD47,79. The expression of the Gaussian curve80 is:

$$y={c\exp }^{-\,\frac{1}{2}\, \cdot \, \frac{{(x-\mu )}^{2}}{{t}^{2}}}$$
(4)

Where \(y\) represents biological indicators such as abundance, density, or biomass. \(c\) is the maximum value of the corresponding indicator, \(\mu\) is the optimal value for a species regarding a specific environmental factor, and \(t\) is the tolerance level of the species, indicating its ecological threshold. The suitable habitat is [\(\mu -2t\), \(\mu +2t\)], with the optimum range being [\(\mu -t\), \(\mu +t\)].

Our research group’s previous studies revealed that the habitat suitability of S. salsa is generally determined by water depth, and the correlations between its probability of biomass and water depth fit well with Gaussian curves78. Previous research was conducted to determine the suitable water table depth for S. salsa. Based on an extensive investigation of S. salsa habitats in the YRD, we surveyed 69 sampling sites across three districts. At each site, five 50 cm × 50 cm quadrats were placed to collect aboveground and underground biomass samples. The samples were dried to constant weight at 80°C and weighed to obtain dry biomass. Using the Gaussian Model, we explored and analyzed the ecological thresholds of S. salsa in response to the water depth gradient in the YRD. Accordingly, we constructed Gaussian fitting curves and identified the water depths for suitable (−0.92~0.08 m) and optimum (−0.67~−0.17 m) habitats (Fig. 8). We used Gaussian curves to calculate and identify the biomass in each suitable habitat grid. Thus, we retrieved water-level and bed-level elevation data from the Delft3D model at the end of each time step morphological run to simulate the average water depth over a spring tide cycle and calculated the biomass of S. salsa based on Fig.8. After identifying the maximum biomass value and the biomass values of suitable habitats, we used min–max normalization to standardize each biomass quantity to a value between 0~1 to describe habitat suitability.

Fig. 8: Gaussian model fit for Suaeda salsa biomass (g m−2) with water depth (m).
figure 8

The water depth used in this study to simulate vegetation habitat refers to the average water depth over a spring tide cycle. The green solid curve shows the fitted Gaussian function. Short arrow area indicates the optimal habitat range (±0.25 around the peak), while long arrow area extends to ±2 × 0.25, representing the suitable habitat range. The vertical dashed line at 0.42 m marks the depth of maximum biomass.

Full size image

This study did not account for dynamic vegetation feedback, primarily for the following reasons: Firstly, S. salsa, characterized by its much smaller stature and sparse distribution, exerts an insignificant influence on landform dynamics compared with those ecosystem-engineering species such as Spartina alterniflora67. Secondly, our research is centered on the newly formed tidal wetlands, rather than on the morphological alterations within intertidal salt marshes. Consequently, our emphasis lies in comprehending the dynamic environmental conditions that shape habitat suitability of newly emerged wetlands for germination and growth of indicator vegetation (i.e., S. salsa) these newly emerged wetland habitats. In contrast, vegetation feedback mechanisms predominantly influence micro-topographical features within vegetated areas, such as the formation and evolution of tidal channels31,33. Given that our study aims to elucidate the broader environmental dynamics rather than the localized effects of vegetation on micro-topography, the incorporation of dynamic vegetation feedback was deemed less pertinent to our specific research objectives.

Author contributions

X.L., J.B. and B.C. conceived the study. L.Z., X.L., and X.X. designed the model scenarios and performed the simulations. L.Z. curated the data and conducted all post-processing analyses. L.Z. drafted the manuscript and prepared all figures. X.L. and D.S. supervised the project and secured funding. All authors discussed the results and contributed to the final manuscript.