Hidden delta degradation due to fluvial sediment decline and intensified marine storms

Abstract

Deltas are threatened by erosion due to climate change and reduced sediment supply, but their response to these changes remains poorly quantified. We investigate the abandoned Yellow River delta that has transitioned from rapid growth to ongoing deterioration due to a river avulsion removing the sediment supply. Integrating bathymetric data, process observations, and sediment transport modeling, we find that while the subaerial delta was stabilized by engineering measures, the subaqueous delta continued to erode due to intensified storms, losing 39% of its mass deposited before the avulsion. Long-term observations show that winter storms initiate scouring of the subaqueous delta, contributing up to 70% of seabed erosion. We then analyze 108 global deltas to assess subaqueous delta erosion risks and identify 17 deltas facing similar situations of sediment decline and storm intensification during the past 40 years. Our findings suggest that subaqueous delta erosion must be integrated into delta sustainability evaluations.

Fluvial sediment decline and intensified storms rapidly erode subaqueous deltas, revealing hidden degradation beneath surfaces.

INTRODUCTION

River deltas provide living conditions for a large population, with great ecological, social, and economic values, yet more than 340 million people living here are subjected to hazard risks (1). Situated at the critical interface of rivers and oceans, deltas are vulnerable systems threatened by human activities, land subsidence, sea-level rise, and climate change (2–4). The rapidly growing population and intensified human activities have profoundly affected deltas globally and disrupted their evolution, making them vulnerable to floods and erosion (5–7). Sea-level rise and climate change further compound the delta crisis by increasing the inundation potential of low-lying areas in subaerial deltas and causing more energetic storms that rapidly erode deltas worldwide (8, 9).

Continuous sediment deposition on delta surfaces has the potential to ameliorate the subaerial delta crisis, underscoring the importance of sustained fluvial sediment delivery for delta aggradation and long-term sustainability. However, this critical process becomes less efficient with declining fluvial sediment inputs worldwide. Over half of the world’s major deltas have experienced substantial reductions in fluvial sediment supply in recent decades, primarily due to evolving land management practices and dam construction (10). Projections suggest that this trend will persist through the 21st century, potentially outweighing the impacts of future climate change and jeopardizing the long-term sustainability of deltas (11).

Understanding the response of deltas to interactions between human-induced reductions in fluvial sediment supply and marine reworking processes is critical for predicting future change and with intervention, limiting delta loss. With the widespread applications of remote sensing analysis, recent studies have characterized the dynamics of shoreline changes and subaerial delta evolution in response to fluvial sediment decline on a global scale (10, 12, 13). Although numerous deltas around the world have achieved net land gain in recent decades (10, 14), primarily through land reclamation and shoreline protection measures, the true status of the entire subaqueous prodeltas remains unclear due to inadequate offshore observations and limited application of remote sensing techniques in turbid coastal waters (15, 16). Could the apparent stability of visible subaerial deltas be hiding submarine delta deterioration?

As the seaward extension of subaerial deltas, subaqueous prodeltas play a vital role in wave energy dissipation, biogeochemical cycling, and maintaining the overall morphological stability of deltaic systems (17–19). While subaerial deltas may appear to be stable owing to shoreline protection measures or even expand seaward with the support of offshore sediment supply, many subaqueous prodeltas, including those associated with the Mississippi and Yangtze, have experienced degradation through erosion over the past few decades due to fluvial sediment decline and climate change (20–22). The degradation of the subaqueous delta platform further exacerbates the situation by increasing wave exposure to the shoreline and steepening the nearshore seabed slope, thereby posing a potential risk of catastrophic seawall failures and ultimately undermining the stability of subaerial deltas (23, 24). The offshore environment may also provide complex feedback to the subaerial environment through rapid response to rising sea level and intensification of storms (16). The intricate dynamics of subaqueous prodeltas make it a considerable challenge to effectively monitor their morphological changes and incorporate their degradation into delta sustainability evaluations.

We investigate the evolution of the abandoned Yellow River delta in the southern Yellow Sea, which serves as an ideal study site for examining the response of the combined subaerial and subaqueous delta to fluvial sediment decline and climate change. The delta system offers (i) a complete cycle from formation to destruction over the last millennium, (ii) a history of intense anthropogenic modification of fluvial sediment supply and shoreline protection, and (iii) an analogy to present-day deltas with a substantial decline in sediment supply due to river influx change. We analyze the evolution of the delta system and identify dominant drivers for both subaerial and subaqueous delta erosion after the delta lost its fluvial sediment supply due to river avulsion. We show that the cessation of river sediment supply in 1855 led to rapid erosion of the delta. We use long-term hydrodynamic and suspended sediment data collected from its subaqueous delta, along with model simulation results of bed shear stress and sediment transport patterns, to demonstrate how intensified winter storms accelerate subaqueous delta erosion. Last, we conduct a synthesis analysis of changes in fluvial sediment flux and wave climate among 108 global deltas to identify those deltas facing similar conditions of reduced sediment influx and intensified seasonal storms.

RESULTS AND DISCUSSION

The dominant role of human activities in delta formation

The Yellow River is well known for its enormous sediment loads and frequent floods (>1000 floods in 4000 years) and river avulsions (25). Over the past two millennia, the Yellow River has swept over its broad alluvial plain, discharging alternately into the Bohai and Yellow Seas (Fig. 1A). As the ninth superlobe of the Yellow River delta, the abandoned Yellow River delta was formed between 1128 CE and 1855 CE when the river drained into the southern Yellow Sea at Jiangsu (Fig. 1B) (26).

Fig. 1. Map of the study area and wave climate.

(A) Historical paths of the Yellow River extracted from the Atlas of the Yellow River (91). (B) The locations of the observation station and model output transect and directional distribution of waves in the abandoned Yellow River delta. (C) Wave climate at the study site. The significant wave height (H_s) data in (B) and (C) were extracted from the nearest pixel of the ECMWF 40-year Hindcast (1979–2018) at a water depth of ~21 m. Our wave climate analysis shows that waves are highest in winter, mainly driven by northeasterly winds.

The evolution of this particular superlobe was prominently driven by human activities. In 1128 CE, the government of the Song Dynasty intentionally breached the levees of the Yellow River to create a river avulsion to resist the invasion of the Jurchen army (25). This event marked the beginning of the formation of a new delta. The average annual sediment load during 1128–1855 CE was ~1 billion metric tons (Gt) per year (27), an order of magnitude higher than most major rivers today. Initially, sediment was mainly deposited on the alluvial plain because of frequent levee breaches in the lower river. In 1565, the government launched an extensive project to build artificial levees, resulting in a nearly 50% reduction in levee breach frequency and an increase in sediment delivery to the coast in the next few centuries (25). Consequently, assuming an average sediment thickness of 15 m, the progradation rate of the subaerial delta increased from 4.8 × 10⁷ m³ year⁻¹ during 1128–1560 CE to 2.3 × 10⁸ m³ year⁻¹ (28).

The enormous fluvial sediment delivery to the coast overpowered marine reworking, allowing a large subaerial delta of 7160 km² to grow into the southern Yellow Sea between 1128 CE and 1855 CE, which sequestered 140 Gt of sediment (assuming a dry bulk density of 1.3 × 10³ kg m⁻³) (29–31). The subaqueous delta (Fig. 1B) stored an additional 203 Gt of sediment (29). This progradational phase ended in 1855 CE when the Yellow River ruptured its northern levee and reversed its course back toward the Bohai Sea.

Eroding the abandoned delta

The delta system was subjected to strong reworking by marine processes and shifted into a continuous erosional phase after losing its fluvial sediment supply in 1855 CE. The shoreline retreated rapidly at >600 m year⁻¹ in the first few decades, slowing down to ~80 m year⁻¹ in the early 1900s, and to <20 m year⁻¹ after the building of seawalls in the 1950s (Fig. 2A and fig. S1). The subaqueous delta was even more dynamic and experienced seafloor erosion and sediment redistribution (Fig. 2, A and B). The subaqueous delta comprises three distinct regions based on topographic characteristics: the steep shoreface (nearshore slope) with depths less than ~15 m, a broad offshore platform ranging from 15 to 20 m in depth, and the outer offshore slope descending to a maximum depth of 60 m (Fig. 2C). Moderate erosion of the subaqueous delta occurred in the shoreface and offshore platform during 1900–1950, with an average erosion rate of 0.3 km³ year⁻¹ (Fig. 2A). Between 1950 and 1970, erosion of the offshore platform expanded to depths exceeding 20 m, driven by rising wave activity. Sediments were redistributed to areas near the 15-m contour (Fig. 2B), forming a transitional deposition layer ranging from 1 to 6 m (31). Because of sediment redistribution, the entire subaqueous delta experienced a net deposition rate of approximately 0.5 km³ year⁻¹ during this period (Fig. 2A).

Fig. 2. Morphological changes of the abandoned Yellow River delta.

(A) Changes in shoreline retreat rate, subaqueous delta erosion (negative)/deposition (positive) rate, and the long-term trend of significant wave height in the abandoned Yellow River delta since 1855 CE. (B) The continuous retreat of the 15-m isobath extracted from (31). (C) The 1979 bathymetric transect and the corresponding average combined current-wave bed shear stress during a winter storm event (21 January 2017) simulated by the Delft3D model. The shoreline retreat rate and subaqueous delta erosion/deposition rate in (A) were calculated from published data (31, 92). Time series of significant wave height were extracted from ECMWF ERA-20C Ocean Wave data from 1900 to 2010 (https://apps-dev.ecmwf.int/datasets/data/era20c-wave-daily/type=fc/), and the 10-year moving average annual significant wave height was calculated to reveal the long-term trend. The gray shaded area in (A) represents the period during which the shoreline was effectively protected by seawalls. The dashed gray line in (C) indicates the critical bed shear stress for cohesive sediment resuspension. The location of the model transect is shown in Fig. 1B.

Wave activity in the southern Yellow Sea experienced a rapid increase between 1930 and 1950 (Fig. 2A), coinciding with a strong East Asian winter monsoon period (32, 33). Subsequently, the entire subaqueous delta entered an accelerated erosional stage at a rate of >1.5 km³ year⁻¹ (Fig. 2A), ~5 times the erosion rate during 1900–1945. The erosion rate slightly decreased from 2.0 to 1.5 km³ year⁻¹ in the 1980s, attributed to a minor reduction in wave activity caused by the weakening of the East Asian winter monsoon (34). Over this rapid erosion phase, the shoreface profile shifted from convex in the 1960s to concave after the 1980s because of continuous shoreface retreats (Fig. 2C) (31). Repeated bathymetric surveys and hydrodynamic simulations reveal that seabed erosion during this period mainly occurred in the nearshore zone within 20-m water depths where the seabed gradient steepens, and the combined current-wave shear stress surpasses a threshold for consistent erosion of the seabed (comprising very fine silt) during storms (Fig. 2, B and C).

In the past 160 years, the total mass of subaqueous delta erosion was estimated to be ~80 Gt from bathymetric surveys (29, 31, 35), a mass equal to 39% (80 Gt/203 Gt) of the sediment storage in the subaqueous delta and 23% [80 Gt/(140 + 203) Gt] of the total sediment storage in both subaerial and subaqueous deltas. Although the subaerial part of the delta appears stable after the construction of seawalls, the rapid degradation of the prodelta in the past few decades has resulted in a very steep nearshore seabed slope (0 to 20 m; Fig. 2C) and increased erosion of intertidal flats (fig. S2), undermining the stability of seawalls onshore (31, 36).

The abrupt cessation of fluvial sediment supply allowed sediment reworking to rapidly erode the subaqueous portion of the delta. As erosion occurs in response to enhanced hydrodynamics, cohesive sediment at the seabed (mean grain size, ~8 μm) is resuspended into the water column, thereby increasing the near-bottom suspended sediment concentration. Research has suggested that variations in sediment concentration in this region are controlled by seasonal wave patterns (37). During autumn and winter, average sediment concentration reached ~0.8 kg m⁻³ due to increased wave energy associated with the East Asian monsoon (Fig. 1C). To elucidate the intricate mechanisms driving subaqueous delta erosion, we conducted long-term measurements of hydrodynamic variables and unprecedented monitoring of suspended sediment at a site located approximately 19 m deep in the subaqueous delta during periods of strong wave activity from 1 November 2016 to 30 April 2017 (Fig. 3).

Fig. 3. Time series of hydrodynamics and suspended sediment concentration (SSC) from 1 November 2016 to 30 April 2017.

(A, E, and I) water depth recorded by OBS; (B, F, and J) significant wave height (H_s) calculated from the acoustic Doppler velocimeter (ADV) pressure measurements at ~19-m water depth (black line) and extracted from ECMWF dataset (orange line); (C, G, and K) bottom current velocity U (black line) and bottom wave orbital velocity U_w (red line) from ADV measurements; (D, H, and L) bottom suspended sediment concentration (SSC) from OBS turbidity sensor; and (M) net modeled residual currents averaged over the period of the storm event on 21 January [the shaded area in (F)]. The shaded areas in (A) to (L) represent the periods when extreme near-bed resuspension events occurred.

Our observations reveal that four extreme near-bed resuspension events (suspended sediment concentration > 3 kg m⁻³; the shaded areas in Fig. 3, D, H, and L) occurred during the observation period and that all of them were related to strong waves (significant wave height > 1.5 m and wave orbital velocity > 0.1 m s⁻¹; Fig. 3, B, C, F, G, J, and K). Although the near-bed current velocity was stronger than the wave orbital velocity throughout most of the observation period, high current velocity alone (>0.5 m s⁻¹) cannot trigger strong sediment resuspension from the seabed (Fig. 3). From 16 January 2017 to 14 February 2017, an Acoustic Backscatter Profiling System (ABS, AQUAscat 1000) was deployed at the observation site as a complementary measurement of the near-bed suspended sediment concentration profile and bed level change (Fig. 4). Seabed elevation changes, as determined by the maximum backscatter intensity of the profiler, indicate intense seabed erosion (~3.6 cm) during a winter storm (peak wave height > 4 m; 20 to 22 January 2017), which accounted for 70% of the total seabed elevation change over the 30-day deployment period. The seabed experienced rapid erosion during the storm’s early phase, causing the sediment concentration at 0.45 m above the bed to spike at 4 kg m⁻³ before subsiding to background levels once significant wave height dropped below 3 m (Fig. 4A). Unexpectedly, a second peak in sediment concentration occurred around January 22, even when the wave height was below 1 m. We hypothesize that this second peak was sustained by the presence of a fluid mud layer near the bed (sediment concentration > 10 kg m⁻³) (Fig. 4B). Comparable fluid mud layers have been detected through turbidity and ABS measurements in nearby modern Yellow River and Yangtze River deltas, where sediment concentrations at around 0.4 m above the seabed ranged from 2–4 kg m⁻³ (38–40).

Fig. 4. Time series of significant wave height, near-bed sediment concentration, and seabed elevation changes.

(A) SSC at 0.45 m above the seabed from OBS and significant wave height (H_s) estimated by ADV measurements. (B) Seabed elevation changes and backscatter intensity profile recorded by ABS from 16 January 2017 to 14 February 2017. The gray shading indicates the period of the observed extreme resuspension event during the storm. During this period, the seabed experienced substantial erosion (3.6 cm), and a fluid mud layer was formed at the end of the event. The values labeled in (B) represent the average bed elevations over a 24-hour period.

The sediment eroded during this winter storm was subsequently carried toward the south by strong longshore currents (Fig. 3M). A sediment budget analysis, based on model simulation results from this storm event, indicates a net sediment flux out of the subaqueous delta at 0.55 kg m⁻¹ s⁻¹. There was a sediment input flux from the north and east boundaries (refer to locations in fig. S3) at 0.18 and 0.24 kg m⁻¹ s⁻¹, respectively, countered by a sediment output flux of 0.97 kg m⁻¹ s⁻¹ to the south.

Although our long-term measurements of hydrodynamic variables in the subaqueous delta have confirmed the critical impact of winter storms on delta erosion, the interpretation of accompanying complex sediment erosion and transport phenomena remains challenging. For instance, extreme near-bed resuspension events are not solely dependent on significant wave height, and strong waves during spring tides may not necessarily cause extreme resuspension events. These complexities may arise from various processes, such as hindered settling (41), seabed fluidization (42), water-column stratification (43), the presence of fluid mud (44), and nonlinear interactions of currents and waves (45). To elucidate these complex dynamics, more consistent long-term measurements of hydrodynamic variables, sediment resuspension, and bed level changes in the subaqueous delta are needed. Nevertheless, our analysis, based on repeated bathymetric surveys and the analysis of long-term hydrodynamic measurements (Figs. 2 to 4), has identified the critical role of wind waves in subaqueous delta erosion after the cessation of fluvial sediment supply.

To further examine the impact of wind waves on long-term delta erosion, we conducted additional long-term morphological simulations of the delta using an idealized Delft3D model (see detailed model settings and parameterization in Materials and Methods). The simulation results demonstrate that continuous fluvial sediment supply during 1587–1855 promoted the formation and rapid progradation of the abandoned Yellow River delta in the southern Yellow Sea (Fig. 5, A and B). However, after losing its fluvial sediment supply in 1855, the delta system underwent strong reworking by marine processes and transitioned into erosion. Tidal currents alone were insufficient to cause substantial erosion in both the subaerial and subaqueous deltas, only resulting in minor incisions in some offshore tidal channels (Fig. 5, B and C). It was the combination of tides, winds, and waves that triggered substantial erosion of the subaqueous delta, while the implementation of a shoreline protection scheme in the model effectively reduced subaerial delta erosion (Fig. 5, D and E). Although the widespread offshore sand waves depicted by the model were not entirely realistic for this region (29, 31), these idealized morphological simulations distinctly underscore the pivotal role of wind waves in the erosion of both subaerial and subaqueous deltas. These results are consistent with our analysis based on repeated bathymetric surveys and long-term hydrodynamic measurements. Collectively, these findings suggest that, even if subaerial deltas can be stabilized through engineering measures, subaqueous delta erosion driven by wave activities is inevitable, with intensified storm waves exacerbating this process.

Fig. 5. Idealized model setup and simulation results of the abandoned Yellow River delta.

(A) Initial bathymetry in 1578. (B) Modeled bathymetry in 1855 driven by tide and river input. (C) Modeled bathymetry in 2022 driven by tide only. (D) Modeled bathymetry in 2022 driven by tide, wind, and waves. (E) Modeled bathymetry in 2022 driven by tide, wind, and waves, and with seawall protection. The red solid line in (E) shows the location of the seawall implemented in the model.

Sediment-starved deltas and enhanced storms

Our findings on the abandoned Yellow River delta demonstrate that human interventions in large river systems can dominate both delta formation and destruction, while marine processes, intensified by a warming climate, can accelerate the latter process. While engineering projects can protect shorelines from erosion, subaqueous delta erosion driven by wave activities is likely a common fate for global deltas that are experiencing rapidly declining fluvial sediment supply and intensified storm waves (46–48). Because of the proliferation of dam construction in river basins, more than 70% of global deltas have experienced greater than 50% reductions in fluvial sediment loads over recent decades (10, 49–51). However, subaqueous delta dynamics under sediment decline and increasing storm activity remain poorly characterized due to the limited application of remote sensing techniques that can penetrate turbid coastal waters and a lack of field observations during storms.

As a first-order attempt to assess the erosion risk of subaqueous deltas, we analyzed 108 representative deltas worldwide to identify deltas vulnerable to storm-induced erosion by combining long-term significant wave height (1979–2018) from the ECMWF (European Centre for Medium-Range Weather Forecasts) dataset with sediment flux data collected from published studies (see detailed descriptions about the analysis in Materials and Methods). Forty-two major deltas have experienced a notable reduction in fluvial sediment supply (>20%) compared with pre-dam periods. Among these, 36 deltas are affected by seasonal storms, with >45% of strong waves (significant wave height exceeding the 95th percentile annual wave height with a duration >12 hours) occurring in three consecutive months over an annual cycle. We applied the Mann-Kendall (MK) test to the wave height data of these deltas and identified 17 deltas with a significant increasing trend in significant wave height during the past 40 years. We postulate that these particular systems are highly susceptible to subaqueous delta erosion. These vulnerable deltas are primarily located in three regions (Mediterranean, East Asia, and Gulf of Mexico; Fig. 6). Although summer monsoons and tropical storms can also cause substantial impacts on delta erosion and sediment redistribution globally (13, 52, 53), our wave climate analysis indicates that most of the vulnerable deltas we identified are dominated by winter storms (table S1) due to their locations in subtropical and temperate regions. Consequently, we chose to concentrate our analysis on these prevalent winter storm dynamics within the three identified vulnerable regions, refraining from an exhaustive exploration of detailed seasonal storm patterns in the global context.

Fig. 6. Global distribution of deltas influenced by both reduced sediment flux and seasonal storms.

The purple shaded areas represent three critical zones (Gulf of Mexico, Mediterranean Sea, and East Asia) where deltas have experienced large reductions in fluvial sediment flux and are influenced by seasonal storms. Sediment flux data of major river deltas were compiled from (51, 89, 90).

The Mediterranean deltas have experienced >80% reductions in fluvial sediment supply due to river damming (54, 55). As a hotspot of global warming (56), the Mediterranean coast is expected to be exposed to more energetic waves and greater storminess during winter months (49), with larger bed shear stresses on the seabed, inducing strong sediment resuspension and erosion in the subaqueous deltas (23, 57). The eroded sediment can be exported from these deltas by prevailing longshore currents, resulting in severe subaqueous delta erosion (46, 58, 59). The East Asian coast has large deltas (e.g., Yellow, Yangtze, Red, and Mekong) fed by high fluvial sediment loads over the past few centuries. However, extensive human intervention, including river damming and sand mining in the last century, has caused these deltas to deviate from their previous evolution and transition toward a destruction phase (60–63). For instance, the current sediment loads carried by the Yellow and Yangtze rivers to their respective deltas are 0.17 and 0.20 Gt year⁻¹, respectively, which fall below the minimum sediment supply rates of 0.33 and 0.26 Gt year⁻¹ necessary to sustain delta growth (64, 65). Continuous erosion of the subaqueous delta has already been observed in the Yangtze and Mekong deltas in response to a modest increase in winter storms (21, 22, 61). Combined with prevailing coastal currents in these systems, these deltas are expected to experience long-term erosion in the coming decades.

The Gulf of Mexico coast hosts one megadelta (i.e., the Mississippi) and many other deltas that are profoundly affected by human activity and increasing severe storms (Fig. 6). The construction of over 50,000 dams in the Mississippi River basin and the implementation of engineering projects such as flood-protection levees along the channel have resulted in a basin-wide sediment deficit (66, 67). While subaerial delta protection and wetland restoration have become a priority for local communities, less attention has been paid to subaqueous delta erosion. The 10-m isobath of the subaqueous delta has been retreating at a rate greater than 20 m year⁻¹ since 1979, and the overall deposition rate in the delta front decreased from 317 million tons (Mt) year⁻¹ during 1984–1940 to 87 Mt year⁻¹ during 1979–2005 (20). Because of the sediment deficit of the delta and intensified winter storms over the past few decades in this region, the subaqueous Mississippi delta is highly susceptible to erosion caused by seasonal storms.

The rapid growth of human populations and intensified anthropogenic activities over the past few centuries have markedly affected global deltas and disrupted their natural evolution. Our study of the abandoned Yellow River delta illustrates how human-induced factors such as river management and levee breaches, as well as intensified marine processes, can dominate the evolution and destruction of a large river delta. The enormous fluvial sediment load of the Yellow River facilitated the growth of a large delta system in the southern Yellow Sea between 1128 CE and 1855 CE, accumulating a total sediment storage of 343 Gt. However, the system underwent substantial reworking and erosion by marine processes after the cessation of sediment supply due to a river avulsion in 1855. While sediment loss was the primary cause driving the delta’s ongoing deterioration, enhanced storm wave action over the past 70 years has notably accelerated the erosion of the subaqueous prodelta, resulting in a five-fold increase in erosion rates compared to the early erosional period (1900–1940) and a 39% loss of its sediment storage. In contrast, the rapid retreat of the subaerial delta stabilized to <20 m year⁻¹ after the 1970s owing to shoreline protection measures.

Although subaerial deltas may appear stable due to human efforts to protect their integrity with seawalls, it is important to recognize the limitations of subaerial landform remote sensing datasets in capturing delta complexity. The offshore environment can exhibit complex feedback mechanisms with the subaerial environment, which we are not yet able to fully predict, and it may respond more rapidly to rising sea levels and intensified storms compared to the subaerial changes (16). We must be cautious not to be deceived by the apparent surface stability of coastal delta plains, as erosion and degradation may be hidden underwater, posing a substantial threat to the overall morphological stability of delta systems, and potentially triggering rapid changes in the subaqueous environments. To establish a more comprehensive understanding of the long-term sustainability of deltas, it is imperative to incorporate the degradation of subaqueous deltas into delta sustainability evaluations. This approach is particularly important for those highly populated and fragile deltas in regions such as the Mediterranean, East Asia, and the Gulf of Mexico, where deltas face similar situations of diminished sediment influx and intensified seasonal storms over the past 40 years. While recent global assessments project a decrease in average winter storm wave height in these regions by the end of the century (68, 69), it is still unclear how extreme winter storms and their frequency will change in a warming climate. Because of the critical role of extreme winter storms in subaqueous delta erosion, a more comprehensive understanding of extreme winter storm dynamics under future climate change and the resultant impacts on global deltas is urgently needed. The results reported here do not imply that the impact of winter storms will increase everywhere. However, since the winter storm pattern can have varied responses to global warming in different regions (70) and, even at a fixed location, it may have complex temporal variations that are not controlled by global warming or cooling alone (71), it is likely that in the future many regions over the world will be influenced by intensified winter storms.

MATERIALS AND METHODS

Regional context

This study was conducted in the abandoned Yellow River delta in the southern Yellow Sea (Fig. 1B). This delta lobe has shifted from rapid growth to persistent erosion since the Yellow River redirected its course back to the Bohai Sea following a river avulsion in 1855. Tides in the study area are semidiurnal, with an average tidal range of 3 m (37). Tidal currents move along the coastline, flowing southeast during flood tides and northwest during ebb tides. Wind speeds in this region are influenced by the East Asian monsoon, with more frequent and stronger northerly/northeasterly winds in winter than in other seasons (37). Consequently, wave heights in this area show a distinct seasonal pattern, reaching higher levels during winter, with an average significant wave height surpassing 0.8 m (Fig. 1C).

The bed surface sediment in the subaqueous delta is predominantly cohesive sediment (very fine to fine silt; fig. S4), with an average grain size of 8 μm (31). The upper sediment layer of the subaqueous delta is characterized by relatively homogeneous fine-grained sediment delivered by the Yellow River, with an average sediment thickness ranging from 4 to 8 m (29). Because of the prevalence of fine-grained sediment in this area, sediment concentration can become relatively high, reaching a maximum during winter (averaging ~0.8 kg m⁻³ in the water column) due to increased wave energy associated with the strong East Asian winter monsoon (37).

Field observations

In situ hydrodynamic measurements were conducted in the abandoned Yellow River delta from 01 November 2016 to 30 April 2017. A bottom-mounted tripod equipped with instruments was deployed on the seabed at a depth of ~19 m (see the location in Fig. 1B). A downward-facing acoustic Doppler velocimeter (ADV Nortek Vector, 16 Hz) was installed on the tripod to measure three-dimensional current velocities at 0.45 m above the seabed. The acoustic Doppler velocimeter worked in a burst mode at 10-min intervals and recorded 8192 samples during each burst. Time series of water depth and turbidity were obtained at the same height using a nephelometer (Campbell OBS-3A, 1 Hz) every 5 min. A water sample of 50 liters was collected from the near-bottom layer on 25 November 2016, to calibrate the OBS turbidity sensor (fig. S5). Alongside the historical surface sediment grain size data from previous surveys in the subaqueous delta (124 samples; fig. S4), two more surface sediment samples were gathered at the observation site on 17 January 2017. These samples were processed in the laboratory and analyzed for grain-size distributions using a Mastersizer 2000 Particle Size Analyzer.

From 16 January 2017 to 14 February 2017, a downward-looking Acoustic Backscatter Profiling System AQUAscat1000S (ABS) was deployed at a height of 0.95 m above the seabed to measure suspended sediment concentration profiles and seabed elevation changes. The instrument had four frequency channels (0.5, 1, 2, and 4 MHz) and 100 vertical layers with a cell size of 0.02 m. It operated in burst mode every 15 min, collecting samples at a frequency of 16 Hz for a duration of 10 s during each burst. However, the lack of synchronous water samples during the observation prevented the calibration of the ABS signals at the study site, which limited the reconstruction of suspended sediment concentration profiles from the ABS recorded backscatter intensity. Nonetheless, the maximum backscatter intensity recorded by the ABS can be used to infer changes in seabed elevation for seabed erosion analysis.

In addition to the tripod deployment, hourly hydrodynamic profile measurements were conducted simultaneously at six sites in the study area (see locations in fig. S3) from 11:00 on 30 December 2016 to 13:00 on 31 December 2016. Six downward-looking Acoustic Doppler Current Profilers (TRDI ADCP, 1200 kHz) were deployed on anchored boats to measure water depth and velocity profiles. These measurements were used to validate the Delft3D model under calm weather conditions (see Supplementary Text for detailed model validation).

Calculation of wave parameter based on ADV measurement

The wave height, wave period, and spectra estimates for every burst-averaged (10 min) ADV data are determined using the PUV processing method (72). The wave orbital velocity u_w is calculated by integrating the contributions of each frequency component of S_uv,i from the variance of the wave-induced near-bed velocity (73)

$$u_w=\sqrt{2∆f{\sum }_i{S}_{\mathit{uv},i}}$$ (1)

Here, S_uv = S_uu + S_vv is the combined spectrum of the eastern (S_uu) and northern velocity (S_vv).

Hydrodynamic and sediment transport simulation

The three-dimensional physics-based fluid dynamics modeling package Delft3D was used to investigate the flow and bed shear stress patterns in the abandoned Yellow River delta. Delft3D has been widely used for hydrodynamic and sediment transport research in various coastal environments (74). Our model covered the coastal areas from Haizhou Bay to Hangzhou Bay (fig. S3) and included a total of 81,741 grid cells. The southern and eastern boundaries were open ocean boundaries that were forced with principal tidal constituents (M2, S2, K1, and O1) extracted from satellite altimeters and tidal gauge records (75). The resolution of model grid cells was ~900 m near the coast and up to ~4000 m near the open boundaries. Bathymetry data were extracted from detailed marine charts (1:250,000 scale) in 1979 from the Maritime Safety Administration of China and adjusted to mean sea level.

The hydrodynamic model was implemented in depth-averaged mode from 25 December 2016 to 26 January 2017, with a time step of 5 min. The model was driven by tides, winds, and waves. Manning’s formula for parameterizing a depth-dependent bottom friction coefficient was used (76), and the upper limit of the Manning coefficient was set at 0.023, which is the empirical maximum value in the study area (77). The wind field was extracted from the ECMWF dataset (https://apps.ecmwf.int/archive-catalogue/?type=an&class=ei&stream=oper&expver=1:) with a spatial resolution of 0.125° (~11.5 km) and a time interval of 3 hours. The SWAN wave model (78) was coupled with the FLOW module every 60 min. To determine the net water transport flux in the study area, the average residual currents were calculated by applying a low-pass Butterworth filter to the modeled velocities during the simulation period (79).

In addition to depth-averaged hydrodynamic simulations, we conducted three-dimensional sediment transport modeling during the winter storm event from 17 to 26 January 2017, to assess sediment transport fluxes in the study area during the storm event. The model used a σ vertical coordinate system with 12 layers. Each of the top six layers had a thickness of 12% of the water column, while the lower layers gradually decreased in thickness, reaching a bottom layer thickness of 2.5% of the water column (about 0.45 m above the seabed). This configuration enabled a direct comparison between the model results and the measured suspended sediment concentration from OBS at the same height.

The layering of bottom sediment and size characteristics were based on data from previous field surveys (fig. S4). Four sediment classes were specified in the model: cohesive mud with a settling velocity of 0.75 mm s⁻¹, cohesive fine silt with a settling velocity of 2 mm s⁻¹, cohesive coarse silt with a settling velocity of 3.6 mm s⁻¹, and non-cohesive sand with a representative median grain size of 200 μm. The model used the Partheniades-Krone formulation (80) to calculate cohesive sediment erosion and deposition fluxes and the Van Rijn et al. (81) approach to estimating non-cohesive sediment transport rates. Initial sediment thickness distributions were derived from previous model studies in this region (77). Sediment grain density and dry bed density were set to 2650 and 1600 kg/m³, respectively, for all sediment classes. The critical bed shear stress for cohesive sediment erosion and the erosion parameter were set to 0.25 N m⁻² and 0.0002 kg m⁻² s⁻¹, respectively (37, 77). Three monitoring transects (North, East, and South; see locations in fig. S3) were constructed in the model to quantify depth-averaged sediment fluxes into and out of the subaqueous delta during the simulation period.

The model performance was evaluated for two time periods. The modeled water levels and velocities were compared with observations for one complete tidal cycle under calm weather conditions (30 to 31 December 2016), while the modeled significant wave height and near-bed suspended sediment concentration were compared with measurements during a storm period (17 January 2017 to 28 February 2017). Overall, the model can produce reasonable simulations of hydrodynamics and sediment resuspension in our study area in good agreement with in situ measurements (see detailed model validation results in figs. S6 to S8 and table S2).

Idealized long-term morphological simulations of the delta

To further explore the impact of wind waves on long-term delta erosion after the cessation of river sediment supply, we conducted morphological simulations of the delta spanning from 1587, when sediment delivery to the coast substantially increased in response to extensive levee constructions along the river, to 2022 using an idealized Delft3D model adapted from previous studies in the area (82). This idealized morphological model covers a smaller domain than the hydrodynamic model, ranging from Haizhou Bay in the north to the central Jiangsu coast (Fig. 5). It maintains a spatial resolution of approximately 200 m near the abandoned Yellow River delta. Open ocean boundaries along the north, east, and south are forced with tidal conditions extracted from the OSU TPXO Global Tide Models (83). The initial bathymetry and shoreline were adapted from (84) based on historical maps, and a uniform bed slope of 0.177‰ was applied in the delta region (Fig. 5A). To enhance computational efficiency, the morphological acceleration factor was used as a scalar multiplier for the sediment continuity equation to reduce computational time, which is a common strategy in long-term Delft3D morphodynamic simulations (85). A morphological acceleration factor of 108 was determined through sensitivity tests (82). Furthermore, a 5-m spatially uniform non-cohesive sand layer with a median grain size of 200 μm was initialized in the model to represent the predominant sand composition before the Yellow River’s discharge into the Yellow Sea (29).

Between 1587 and 1855, a consistent river discharge of 2,550 m³ s⁻¹ and sediment concentration of 3 kg m⁻³ were imposed at the Yellow River mouth, mimicking the average water and sediment fluxes of the river during that period (84). The suspended sediment carried by the river is characterized by fine silt with a median grain size of around 15 μm (86). Consequently, river-borne sediment was represented as cohesive sediment in the model with a settling velocity of 2 mm s⁻¹ and a critical bed shear stress for erosion of 0.25 N m⁻². After 1855, after the river was redirected toward the Bohai Sea, both water discharge and sediment flux from the river were turned off in the model.

For the period from 1855 to 2022, three model scenarios were designed to assess the impact of different factors on delta erosion: tide only (T), tide coupled with wind waves (TW), and tide and wind wave coupling integrated with shoreline protection (TWP). In wave simulations, we applied a synthetic hourly wind input based on the 1900–2012 ERA reanalysis wind data in the subaqueous delta (table S3) using the input reduction tool from the Delft3D Open Earth Tool (87). In addition, a thin dam feature in Delft3D was incorporated in the TWP model run along the coastline at an elevation of 0 m. This alteration effectively prevented water and sediment exchange between the subaerial and subaqueous deltas.

Evaluating the erosion risk of global subaqueous deltas

To assess the erosion risk of subaqueous deltas, we selected 108 representative deltas of various sizes and locations worldwide and conducted the following analyses. (i) We identified major deltas with an average river discharge >500 m³ s⁻¹ from the dataset of Nienhuis et al. (14) and verified delta locations with Google Earth images. (ii) We expanded the delta dataset by including an additional 26 known deltas from previous studies (88, 89). (iii) We calculated the sediment fluxes relative to pre-dam periods for all deltas in the integrated dataset and identified the deltas with >20% reductions in sediment flux compared to pre-dam periods, which we marked as sediment-starved deltas (42 deltas). Sediment flux data for large densely populated deltas were extracted from (51, 89, 90), and for other deltas where direct sediment flux observations were unavailable, estimates were derived from the Nienhuis et al. (14) dataset. (iv) We calculated wave parameters and statistics for the sediment-starved deltas across various delta types (river-, tide-, and wave-dominated deltas) using the 40-year wave records (1979–2018) extracted from the ECMWF dataset (https://apps.ecmwf.int/archive-catalogue/?type=an&class=ei&stream=wave&expver=1), which has a spatial resolution of 0.125° and a temporal resolution of 6 hours. The wave data closest to the selected deltas were used to represent the prevailing wave conditions at the site. (v) We determined whether these sediment-starved deltas are controlled by seasonal storms by selecting deltas with >45% of large waves (with significant wave height exceeding the 95th percentile and a duration of more than 12 hours) occurring in any three consecutive months. These parameters were chosen after a series of sensitivity tests as the most effective criteria to identify seasonal storm patterns in global deltas (see Supplementary Text for detailed evaluations of seasonal storm pattern identification; tables S4 and S5). (vi) For the deltas affected by seasonal storms (36 deltas), we determined their significant wave height trend during the past 40 years using the MK test and identified those deltas with a significant increasing trend of significant wave height as most vulnerable to subaqueous delta erosion (17 deltas). (vii) We identified three critical zones based on the spatial distribution of the most vulnerable deltas to subaqueous delta erosion. Because of a lack of available wave data, most of the inland deltas and high-latitude deltas (26 in total) were not included in our analysis. It is important to note that our global delta analysis primarily focuses on identifying delta vulnerability to erosion induced by reduced sediment flux and seasonal storms. Therefore, the deltas we identified as high subaqueous delta erosion risks might not encompass those vulnerable to other erosion factors such as river flow and tides. The complete list of the selected deltas in our analysis is provided in table S6.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S8

Tables S1 to S6

References