Coastal morphological changes in the Red River Delta under increasing natural and anthropic stresses

River deltas are the best place to study intense human–earth interactions and the resultant morphological changes and sedimentary records. The coastal evolution history of the Red River Delta (RRD) is examined by time-series analysis of multiple coastline locations. We find that spatiotemporal variation in seawall locations and vegetation lines are obviously site-specific due to intense human interference, while changes in 0 m isobaths are highly dependent on external stresses. Coastal erosion and deposition patterns are determined firstly by sediment inputs from different distributaries, and secondly by sediment redistribution with tides, waves, and longshore currents. The causes of chronic erosion along the Hai Hau coast include swift distributary channels, negligible sediment supply by the regional longshore current, and continuous sediment export by local wave-generated longshore and offshore currents. The area of intertidal flats decreased significantly due to land reclamation and decelerating coastal accretion. The area of mangrove forests decreased first due to human deforestation, and then increased gradually due to artificial plantation. Poorly designed coastal infrastructures may increase risks of coastal erosion and flooding disasters. More coastal sectors in the RRD may turn into erosion due to continuous decrease in riverine sediment discharges, deserving more attention on proper coastal protection and management.


Introduction
It has been recently hotly debated whether humans have kicked off a new epoch, namely the Anthropocene (Zalasiewicz et al. 2015;Waters et al. 2016;Fan 2018). Geologists are trying to define a distinct stratigraphic signature created by human activities, clearly separating the Anthropocene from the Holocene. Whatever the outcome will be, it is not the time to celebrate this record-creation behavior, but we should introspect our rapidly deteriorating environment and explore the ways to save our planet for next generations.
Deltas are one of the best gifts from nature to humans because of their fertile lowlands for agriculture and convenient transportation for commerce. They have attracted human dwelling and farming since the Mid-Holocene Warne 1994, 1997;Fan et al. 2017), and now have developed into worldwide socioeconomic hubs. Deltas have become the most densely populated area, and continuing growth of the population is expected in the near future. Due to intense human interference, deltaic morphology and subsurface space have been greatly transformed from their natural status. Most delta plains have been diked to protect properties from riverine and coastal flooding, and numerous coastal wetlands have been reclaimed for agriculture and aquaculture. At the same time, deltaic residents have been increasingly exposed to natural hazards and disasters, which in turn may link directly or indirectly with human activities. Coastal erosion and ecological degradation are the most commonly reported issues in the deltas (Syvitski et al. 2009;Dai et al. 2014;Fan et al. 2017;Li et al. 2018). The situations are typically worsening in populous Asian mega-deltas, where the majorly sources of adverse impacts are damming rivers; sand mining in river channels; land reclamation; over-exploitation of underground water, gas and oil; and poor design of coastal infrastructures and coastal management (Li et al. 2004;Thanh et al. 2005;Saito et al. 2007;Syvitski et al. 2009;Wang et al. 2011;Anthony et al. 2015;Fan et al. 2017;Xie et al. 2017).
Monitoring coastal evolution is vital for taking appropriate action towards the sustainability of deltas (Brondizio et al. 2016). Holocene delta development is majorly determined by the interactions of fluvial and marine processes, varying over diverse time scales from a few minutes to decades and centuries (Fan et al. 2017). Various methods can be involved in studying coastal morphological changes over different spatiotemporal scales. Short-term (a few minutes to days) morphological variations can be monitored directly by repeated measurements of bed level or deduced indirectly through high-frequency measurements of sediment-dynamic parameters. Long-term (decades to millennia) coastal evolution can be studied by core data and time-series analysis of paleo-coastline marks (e.g., chenier ridges). Intermediate-term (months to decades) morphological variations can be monitored by comparison studies of different-aged satellite or aero photos or historical maps and charts (Fan 2012 and references therein).
Coastal morphological changes over the intermediate term are vital for coastal management and planning. They have been increasingly studied due to easier acquirement of higher qualified satellite images with rapid technology development. Such applications can be illustrated in a long list of works (e.g., Bi et al. (2014) in the Yellow River Delta, Chu et al. (2013) in the Yangtze Delta, Zhang et al. (2015) in the Pearl River Delta, Dien et al. (2003) in the Red River Delta (RRD), Anthony et al. (2015) and Li et al. (2017) in the Mekong Delta, and Shearman et al. (2013) in the Irrawaddy Delta and other deltas in the Asia-Pacific region). If high-quality historical maps and charts are variable, decadal variations in coastal morphology can be dated back to much earlier than the 1970s when Landsat satellite images were firstly available. For example, the earliest good-quality chart was completed in the 1930s for the RRD (Dien et al. 2003;Bui et al. 2018), and in the 1840s for the Yangtze Delta (Fan et al. 2017).
RRD coast stretches over 160 km long, varying from a tide-dominated setting in the north to a wave-dominated setting in the south (Bui et al. 2018). Different coastal sectors have diverse responses to the change in sediment flux into the sea through different distributaries and other drivers. In general, most RRD coasts have been observed to undergo net progradation in terms of the vegetation line or high-water line on the accretion muddy flats in the past decades, but Hai Hau coast has been reported to exhibit persistent erosion at least since the 1930s (Imamura and To 1997;Dien et al. 2003;Thao et al. 2013). Coastal erosion can result from distributary channel shifting, longshore current enhancement, mangrove destruction, and sediment deficit (Pruszak et al. 2002;Dien et al. 2003;Thanh et al. 2004;Duc et al. 2007Duc et al. , 2012Mai et al. 2009;Hoan et al. 2010;Thao et al. 2013). In the northern RRD, coastline prograding rates were not observed to change significantly after the completion of Hoa Binh Dam (HBD) in 1988, which has been blamed for the sharp drop in yearly suspended sediment flux at Son Tay station near Hanoi from 119 Mt year −1 (1960)(1961)(1962)(1963)(1964)(1965)(1966)(1967)(1968)(1969)(1970)(1971)(1972)(1973)(1974)(1975)(1976)(1977)(1978)(1979) to 46 Mt year −1 (1989-2010) (Vinh et al. 2014). However, the subtidal flats off the northern RRD have recently been observed to undergo substantial erosion (Bui et al. 2018).
In this paper, intertidal zones between sea dikes and 0 m isobaths were investigated in detail using satellite images, historical maps, and charts. Spatiotemporal variations in coastal erosion and deposition patterns will be discussed based on changes in the locations of sea dikes, vegetation line and 0 m isobaths, and the areas of swamp and intertidal flats over different periods. We aim to better understand controlling mechanisms on the RRD coastal evolution, providing clues for coastal planning and management.

Study area
The RRD is located on the west coast of the Gulf of Tonkin (Bai Bu) in the northwestern South China Sea (Fig. 1). It is the fourth largest delta in Southeast Asia in terms of delta-plain area. The delta initiated to prograde seaward in ∼8500 cal year B.P. when the postglacial sealevel rise significantly slowed down and approached its highest level (Hori et al. 2004). Since then, a vast triangular-shaped delta-plain has been formed with an apex near Son Tray, and its current area is approximately 14 300 km 2 (Hori et al. 2004;Minh et al. 2010;Vinh et al. 2014). The delta accommodates over 17 million residents (22% of the national total), and provides 20% of the rice production of Vietnam (Thanh et al. 2004;Bui et al. 2018).
The Red River drains a basin area of ∼160 000 km 2 , with its headwater at a mean elevation of 2000 m in the mountains of the Yunnan Province in China. It receives two major tributaries, Da and Lo rivers, before entering the deltaic region. Mean annual water and suspended-sediment discharges at the Son Tay station before the HBD impoundment were reported to be 120 km 3 year −1 and 120 Mt year −1 , respectively (Vinh et al. 2014). Due to significant impact of monsoon climate, asymmetry in rainfall and the resultant riverine discharge is obviously biased to the wet season from May to October. Generally, 85%-95% of the total yearly rainfall occurs in the summer monsoon season (Dang et al. 2010;Vinh et al. 2014;Gao et al. 2015;Lu et al. 2015). Accordingly, ∼90% of the annual sediment discharge debouches into the sea during the wet season (Funabiki et al. 2007).
Below the Son Tay station, the Red River bifurcates several times to produce an intricate channel network in the delta plain, which is further complicated by irrigation canals (Fig. 1). At present, the river debouches into the Tonkin Gulf through nine distributaries. It is worth noting that no single distributary can be regarded as a predominant channel to contribute larger than one quarter of annual water runoff or suspended-sediment load. It was estimated that annually 84.8 Mt of suspended sediment could finally reach the sea over 1960-1979, of which 24.9% and 22.1% passed through the Ba Lat and Day river mouths, respectively, 13.6% through the Van Uc river mouth, 9.4% and 8.4% through the Cam and Tra Ly river mouths, respectively, 6.0%-6.1% though each of the Bach Dang, the Thai Binh, and the Lach Giang river mouths, and 3.5% through the Lach Tray river mouth (Vinh et al. 2014). Uneven distributions in water and sediment discharges through different distributaries were also reported by other studies with different estimates of their separate contribution ratios (Pruszak et al. 2002;Duc et al. 2007Duc et al. , 2012. The tide in the Gulf of Tonkin is predominantly diurnal with the spring tidal range decreasing southward roughly from 4 to 2 m along the RRD coast (Duc et al. 2012). The Hon Dau tidal gauging station gave a long-term average spring tidal range of 4.2 m (Thanh 1993;Huy 1996). The prevailing wave direction is northeast (southeast) during the dry (rainy) season. Wave heights are on average 0.7-1.3 m with a maximum of 3.5-4.5 m, but may reach over 5 m during typhoons (Duc et al. 2007(Duc et al. , 2012. The northern RRD coast is sheltered from strong wave action by Hainan Island, together with prevalent funnel-shaped estuaries, consequently dominated by tidal influence. Meanwhile the southern RRD are wave dominated due to their less wave sheltered setting and a protruding configuration . The inserted small map (a) shows the locations of the Red River drainage basin and its delta in Southeast Asia (refer to Table 1 for detailed map information). (Mathers and Zalasiewicz 1999). The Hai Hau coast between the Ba Lat mouth and the Lach Giang mouth, stretching relatively straight over 30 km long in a northeasterly direction, has undergone continuous erosion for several decades with the predominance of longshore sediment transport (Pruszak et al. 2002;Duc et al. 2007Duc et al. , 2012Hoan et al. 2010).

Data sources and methodology
Five nautical charts and four topographic maps were collected in the time period from 1930 to 2008 with different spatial scales (Table 1). Two scenes of satellite images (Landsat 7 ETM) with their acquired times on 1 April 2008 and 8 August 2017 were freely downloaded from the USGS website (http://earthexplor.usgs.gov).
The maps were established using different coordinate systems, making direct overlaid comparison study impossible. WGS-84 proposed by the World Geodetic System is currently used by the Global positioning system as the worldwide reference system. VN-2000 is the current national geodetic system in Vietnam. However, other coordinate systems were used before 2000 (Table 1). Therefore, different formulas should be used to transform the vector maps into the same coordinate system. We performed the transformation based on the official guidebook "Guidelines for the Application of VN-2000 System" (Duc et al. 2012), using ENVI software package. The 1930 map was separately analyzed because of unknown coordinate system. The geometric correction was carried out using true ground control points (GCPs) of main road intersections and streets identified in the image. For each image at least 15 GCPs have been fitted to a first-order polynomial function and the resulting RMS errors are less than half of one-pixel size (White and El Asmar 1999). For the older maps, the resultant errors can be much larger because of their relatively lower resolutions. After registering all maps and satellite images to the VN-2000 coordinate system, locations of sea dikes, vegetation lines, and 0 m isobaths were digitized using ArcGIS Desktop 10.1 software. Coastal erosion and accretion patterns were shown by comparing the locations of three featured coastlines between two characteristic years .

Results
The study coast stretches roughly 140 km from the Don Son peninsula to the Day river mouth. It is further divided into four sub-regions on the basis of their general accretion and erosion patterns in the past few decades (Fig. 1). Sub-region A in the north, including three active river mouths (Van Uc, Thai Binh, and Tra Ly), features moderate accretion; sub-region B is the modern Red river delta (we use the lowercase river for the distributaries of the Red River system in the delta, so the Red river delta is just a lobe produced by the Red River in the RRD), characterized by rapidly seaward progradation; sub-region C between the Ba Lat mouth and Lach Giang mouth, features persistently severe erosion; sub-region D in the south, including two active river mouths (Lach Giang and Day), is also characterized by rapidly seaward progradation. Coastal evolution was analyzed in detail by the comparison of different-aged locations of sea dikes, vegetation lines, and 0 m isobaths in the four sub-regions (Figs. 2-5). The variations were further calculated by the mean changing rates of featured lines over accretional and erosional sectors, respectively; and the net accretion-erosion (positive-negative) areas were summarized over different enclosure areas within different-aged featured lines (Table 2). Therefore, land gain or loss are indicated by seaward or landward relocation of sea dikes, mangrove wetland expansion or reduction are indexed by seaward or landward shifts of vegetation lines, and coastal erosion or accretion are represented by seaward or landward migration of 0 m isobaths.

Sub-region A
Sub-region A was characterized by the rapid coastal accretion in terms of 0 m isobaths over the period 1965-1995, alternating with two minor erosion periods before 1965 and after 1995 ( Fig. 2; Table 2). Except for small erosion in the estuaries produced by channel shifting, tidal flats prograded seaward with a mean rate of 17.64 m year −1 , resulting in a net increased area of 98.13 km 2 in the period 1965-1995. At the same time, rapid seaward sea-dike relocations and mangrove expansions were observed with net increased areas of 48.13 and 56.73 km 2 , respectively. Mangrove expansion majorly occurred between the Tray Ly and the Thai Binh estuaries where only a narrow bank of mangrove was originally observed in the 1965 map. Minor wetland reduction was only observed in the inner part of Van Uc estuary, produced by river bank erosion. However, severe land loss in terms of seadike landward relocations occurred entirely along the Kien Thuy coast and some sectors along the Thai Thuy coast, producing a mean erosion rate of 4.96 m year −1 ( Table 2).
In the period 1930-1965, except that some small bars outcropped above the 0 m isobaths at the mouth of Van Uc and Thai Binh estuaries, most tidal flats underwent erosion. A big mouth bar off the Van Uc estuary was observed to migrate landward (westward) entirely, and the river channels of Tray Ly and Diem Dien were also found to shift westward slightly. A net erosion area of 13.33 km 2 was cumulatively produced by the change of 0 m isobaths in the period. However, it was concurrent with a net land-gain and wetland-expansion scenario with the areas of 11.31 and 20.80 km 2 , respectively (Table 2). Mangrove expansions mainly occurred at the estuaries of Van Uc and Thai Binh, but minor reduction occurred in between these two estuaries and around the Tray Ly estuary (Fig. 2). Land gain mostly In the period 1995-2008, minor coastal erosion and accretion occurred on the Kien Thuy and Tien Hai tidal flats, respectively. Other tidal flats remained stable or underwent slight erosion. This was summed up with a net coastal erosion area of 3.40 km 2 (Table 2). Wetland reduction was inferred by a minus statistic of 3.11 km 2 , but minor land gain was produced by sporicidal aquaculture pond constructions or land reclamation projects with a total area of 4.12 km 2 .
In the period 2008-2017, there was no data for 0 m isobath change. Sea dikes almost remained the same except for two new aquaculture ponds constructed at the southern Tien Lang coast, and one land reclamation project along the Tien Hai coast, resulting in a land-gain area of 2.67 km 2 . At the same time, mangrove distribution patterns changed a lot in that the former wide wetland zones became slightly narrower, and vice versa, some narrow zones expanded a lot (Fig. 2), resulting in a net wetland increasing area of 12.07 km 2 (Table 2).

Sub-region B
Sub-region B featured the continuous growth of barrier ridges and intertidal bars over the extensive intertidal zones of the Red river delta, changing its original triangular shape into a trapezoid configuration after 1930 (Fig. 3a). In the period 1930-1965, the central part of Ba Lat mouth underwent severe erosion with a mean retreating rate of 10.21 m year −1 . This was compensated for by rapid accretion at both proximal flanks with a mean progradational rate of 14.53 m year −1 . Although the distal flanks were also observed to undergo slight erosion, net coastal accretion was produced with an area of 48.13 km 2 (Table 2). At the same time, rapid seaward sea-dike relocations and mangrove expansions were observed with net increased areas of 21.48 and 39.91 km 2 , respectively.  13.14 ---Note: "-" means no data available.
In the period 1965-1995, tidal flats continued large-scale adjustments with obvious erosion over the previous accretion sections, almost balanced by rapid accretion over the previous erosion parts ( Fig. 3b; Table 2). However, land gain totalled 38.54 km 2 , resulting from large-scale land reclamation along the mainland margin and Ngan and Vanh islands. Wetland expansion was also observed significantly with a net increasing area of 20.32 km 2 .
The tidal-flat adjustment continued in the period 1995-2008 but with minor erosion and accretion magnitudes, resulting in a net increasing area of 1.75 km 2 . At the same time, sea dikes were slightly relocated along the mainland coast with a net land loss of 0.46 km 2 . However, moderate wetland expansion was indicated by a net increasing area of 5.23 km 2 ( Table 2).
In the period 2008-2017, sea dikes continued slight relocation by the aquaculture practice with a net increasing area of 0.60 km 2 , mainly resulting from new land reclamation at the NE side of Vanh Island ( Fig. 3b; Table 2). Significant wetland expansion was concurrently observed with a net increasing area of 12.93 km 2 .

Sub-region C
Sub-region C was characterize by large-scale persistent coastal erosion, producing a narrow intertidal zone along the straight coast of Hai Hau district (Fig. 4a). Natural irregular coastlines were populous in 1930 with extensive intertidal zone and broad mangrove wetland at the northern SRC part. But they were almost replaced by artificial sea dikes in 1965 with relatively straight coastline. Land gain toalled 19.65 km 2 because of sea-dike construction in the period 1930-1965, although coastal erosion occurred concurrently along the whole coast in terms of 0 m isobaths with a net erosion area of 6.50 km 2 ( Fig. 4b; Table 2).
Coastal erosion continued in the period 1965-1995 with a greater magnitude, resulting in a net erosion area of 8.62 km 2 . Land reclamation also continued with a net land-gain area of 10.08 km 2 . Their combined effects led to a sharp decrease in the intertidal width, with some sections where high tides almost reached the foot of sea dikes in 1995. At the same time, mangrove area almost disappeared due to coastal erosion and land reclamation. However, coastal residents were forced to retreat and construct new sea dikes about 200-300 m behind the previous ones because of severe breakdowns by increasing erosions and storm waves. The landward relocation of sea dikes resulted in a net land-loss area of 11.34 km 2 in the period 1995-2008 ( Fig. 4b; Table 2). Concurrently, the 0 m isobaths underwent adjustment with tiny movement.
In the period 2008-2017, coastline remained almost the same except for the construction of three aquaculture ponds near the So estuary (Fig. 4b). Mangroves were found to grow again at the high tidal zone, typically north of the So estuary, resulting in a net wetland expansion area of 2.22 km 2 (Table 2).

Sub-region D
Sub-region D featured large-scale land reclamation over the extensive intertidal flats in the period 1930-1995 ( Fig. 5; Table 2). Intertidal flats extended over 10 km wide with welldeveloped mangrove wetlands in 1930 at the Day estuary. An approximate area of 86.5 km 2 was reclaimed from tidal flats in the period 1930-1965, and the wetland expanded seaward rapidly with a net increasing area of 72.74 km 2 , but tidal flats concurrently underwent minor adjustment with a net erosion area of 2.36 km 2 .
In the period 1965-1995, tidal flats underwent rapid accretion with a net increasing area of 32.83 km 2 . Wetland expansion was also observed significantly with a net increasing area of 71.57 km 2 . At the same time, large-scale land reclamation continued with a land-gain area of 56.06 km 2 (Table 2).
In the period 1995-2008, tidal flats underwent slight adjustment with a net increasing area of 0.97 km 2 . Severe wetland reduction was concurrently observed with a net decreased area of 10.61 km 2 . Sea dikes underwent moderate relocations in that land reclamation continued on the coast west of the Day estuary, but sea dikes were forced to relocate landward along the coast of Lach Giang estuary and the inner part of Day estuary (Fig. 5). Their combined effects produced a small land-gain area of 1.29 km 2 over the period 1995-2008.
In the period 2008-2017, major land reclamation occurred along the Kim Hai coast and the Nga Tan coast, resulting in a land-gain area of 10.74 km 2 . At the same time, significant wetland expansion was observed in the Day river estuary with a net area increase of 31.87 km 2 ( Fig. 5b; Table 2).

Toward a standardized workflow of coastline change analysis
Coastline change is the principal content of coastal morphodynamic study. Evidence of coastline changes can be obtained from comparisons of historical maps, charts, aero photos, and satellite images (Bird 2005;Fan 2012; and references therein). Historical maps and charts may have implemented different coordinate systems when they were produced, so they should be registered into the same coordinate system before carrying out overlaying comparison studies of different-aged coastline positions. In addition, tidal datums used for coastline mapping may be different between different countries over different ages (Bird 2005;Hoan et al. 2010;Fan et al. 2017). Digital data should be adjusted to the present regional datums using some transformation methods before comparison.
After the above-mentioned assimilations of different sourced data, selecting a single representative coastline or multiple coastlines is vital for comparisons, because coastline positions vary significantly with tidal cycles and wave processes by themselves. The most used coastline positions include the intersections of beaches or tidal flats with mean high-water level (MHW), mean sea level (MSL), and mean low-water level (MLW). There are also means for spring high and spring low tides, as well as neap high and neap low tides (Oertel 2005;Fan 2012). Because of their different interests, geologists and oceanographers do not agree on the placement of coastlines on maps and charts (Oertel 2005). The MHW-coastline is generally used as the datum for topographical maps, while the mean lowest low water is extensively chosen as the datum for nautical charts. Furthermore, as remote sensing data have been increasingly used to monitor coastline variations, additional coastlines have been defined based on the wet-dry boundary or the vegetation-bare-flat boundary on satellite images. However, these two boundaries are difficult to assign to some specific tidal elevations because of irregular vegetation fronts, and changing water levels when each satellite image is shot at a different tidal stage.
Coastline comparison studies may involve using nautical charts, topographical maps, and satellite images, adding further complexity to coastline selection. For example, coastline identifications on satellite images were usually put at the foot of erosion cliffs or sea dikes (near the MHW-coastline) along beach sectors, and at the mangrove forest front boundary (near the MSL-coastline) along muddy coast sectors in the RRD (Dien et al. 2003;Hoan et al. 2010;Thao et al. 2013). This inconsistent coastline identification makes it difficult to do large-scale coastal evolution comparisons. Coastal changing rates deduced from differentsourced coastline data differ substantially from each other, as discussed by Hoan et al. (2010). Furthermore, a single coastline is generally insufficient to study complex coastal morphodynamic responses to storm events and increasing human disturbances. For example, the MHW-coastline may shift landward due to storm erosion, but the MLW-coastline concurrently shifts seaward due to deposition of eroded materials from the upper part of the shore or tidal flats (Oertel 2005;Fan et al. 2006). The opposite changes of the MHW-and MLW-coastlines were observed along the Kien Thuy coast over the period 1965-1995 (Fig. 2). Due to increasing human interference, changes in the MHW-coastline are generally linked with the construction of sea dikes for agriculture and aquaculture needs, and changes in the vegetation line are highly influenced by wetland destruction and rehabilitation activities over different decades. Therefore, we propose using multiple coastlines to study coastal evolution, typically over larger spatiotemporal scales along the mega-river deltas.

Spatiotemporal variations in featured coastlines along the RRD coast
We selected the positions of sea dikes near the spring MHW-coastline, vegetation lines near the MSL-coastline, and 0 m isobaths near the spring MLW-coastline to study coastal evolution in the RRD over the past few decades (Figs. 2-5). Sea dikes have been constructed to protect from sea flooding in the RRD for more than one thousand years (Thanh et al. 1997), and now the RRD coasts have almost been completely diked by different standard seawalls (Fig. 6). These include earth dikes with crude rock revetments along the accretion muddy coasts (Figs. 6a and 6P1), and elevated dikes with consolidated revetments using stones and mortar along the erosion beach coasts (Figs. 6c, 6d, and 6P2-6P5). The latter may be further reinforced by constructing groins perpendicular to or piling tetrapod armors as wave breakers parallel to the seawalls. Sea dikes are easily identified on satellite images, and are clearly marked in both topographic maps and nautical charts. Therefore, their positions can be verified for accuracy, and time-series analysis of seawall positions and architectures can effectively disclose land gain or loss induced by natural processes and human activities.
Rapid seaward relocations of sea dikes along most RRD coasts in the periods 1930-1965 and 1965-1995 were coincident with large-scale land reclamation for agriculture in 1957-1962and 1968-1971(Thanh et al. 1997. Small-scale seaward relocations in the past two decades were majorly linked with the construction of aquaculture ponds. Landward relocations of sea dikes were also frequently observed along the muddy coasts because they were generally constructed by compacted soils with crude rock revetments and were easily broken by storm impacts or channel shifting near the river banks (Figs. 2, 3, and 5). Along the Hai Hau beach coast (Fig. 4), simple dikes of compacted soils with crude revetments were common before the 1990s, so dike damage with landward relocations occurred frequently with severe coastal flooding disasters. Severe dike collapse events occurred in the period 1995-2008 to produce huge economic losses and rapid landward relocation of seawalls because of concurrent active typhoon impacts and unreasonable land reclamation along the severe erosion coast in the previous stages. These dikes have been continuously reinforced since the mid-1990s (Imamura and To 1997;Mai et al. 2009;Duc et al. 2012), and no more landward relocation occurred in the last decade (Fig. 4). Therefore, studies on seawall construction and relocation histories should shed light on future coastal protection.
River estuaries and muddy coasts are generally fringed with mangrove forest in the RRD (Figs. 6P1 and 6P3). A hundred metres of mature mangrove could reduce 0.1 m of wave height, as reported by Quartel et al. (2007), so mangrove forest is vital for not only biodiversity conservation but also coastal protection. However, mangrove forests have been obviously influenced by human activities in the past few decades. They have been extensively exploited for agriculture and aquaculture before the 1990s . Severe coastal erosion may also threaten mangrove growth. For example, mangrove was almost eliminated on the narrow erosional beach along the Hai Hau coast over the period 1995-2008 (Fig. 4). However, mangrove plantation became populous for wetland rehabilitation and coastal protection measures in the last two decades, resulting in rapidly seaward expansion of mangrove front boundaries along several coastal sectors .
In comparison with sea dikes and vegetation lines, 0 m isobaths are much less influenced by direct human interference, potentially served as a best recorder of the study coast in response to external stresses. It is interesting to find that most RRD coasts have become much stabilized since 1995, in sharp contrast to intense erosion or accretion during the previous two periods (Figs. 2-5). Sub-regions A and D were quite similar in terms of the changing patterns of 0 m isobaths, in that net minor erosion was indicated in the period 1935-1965, followed by intense accretion in the period 1965-1995. The latter may be linked with increased sediment flux by concurrently severe deforestation in the drainage basin (Thanh et al. 1997). Sub-region B featured large-scale morphological adjustment from triangular to trapezoid shapes with significant longshore sediment transport (Fig. 3), and the reason for such a magnificent morphological change was not well understood. In the sub-region C, coastal erosion occurred persistently from 1935 to 1995, but the erosion magnitude was much lower than that before 1935 (Pruszak et al. 2002;Duc et al. 2012;Thao et al. 2013). The occurrence of severe erosion along the Hai Hau coast in the early 20th century could be ascribed to the abandon of a major distributary debouched into the sea through the Ha Lan river course or nearby (Duc et al. 2012). The reduced rate of erosion was conceived to majorly result from the long-term coastal self-regulation towards a new morphodynamic equilibrium (Pruszak et al. 2002). However, the reduced accretion rate in the other three sub-regions after the mid-1990s should be linked with a sharp decrease in sediment discharges by the operation of HBD reservoir in 1988 (Vinh et al. 2014). Additional discussion on erosion causes will be given in Sect. 5.4.

Changes in total areas of mangrove forests and intertidal flats
We also calculated areal areas of swamps (mostly mangrove forests off the sea dikes) and intertidal flats (including swamps and bare flats above 0 m isobaths), and their changing rates over different periods (Table 3). In total, the area of intertidal flats decreased from 578.01 km 2 in 1930 to 394.09 km 2 in 1995, and then was followed by a slight increase towards 425.82 km 2 in 2008 ( Fig. 7b; Table 3). The intertidal area reduction in the period 1930-1965 was mostly induced by intense land reclamation at the sub-regions A, C, and D where slight to moderate coastal erosion occurred concurrently, whereas areal loss by land reclamation (21.48 km 2 ) at the sub-region B was almost balanced by areal gain by coastal accretion (20.25 km 2 ) ( Table 2). In the period 1965-1995, the reduction in intertidal area was largely sourced from land reclamation at the sub-regions B and C where the coasts remained stable or underwent moderate erosion, and land reclamation area (56.06 km 2 ) much larger than coastal accretion (32.83 km 2 ) at the sub-region D, although coastal accretion area (98.13 km 2 ) was much larger than land reclamation (48.13 km 2 ) at the sub-region A. A slight increase in intertidal area from 1995 to 2008 produced slight coastal accretion and significant seawall landward relocations due to storm damages at the sub-region C.
Mangrove forests can grow seaward to increase swamp area along the accretional muddy coasts. They are less impacted by coastal erosion except those along narrow river banks or erosional coasts like the Hai Hau beach, where mangrove forests were almost eliminated by severe erosion during the period 1965-1995 (Fig. 4). Furthermore, mangrove forests have been highly influenced by human activities in recent decades, in that land reclamation may remove large-scale mangrove forests, and vice versa, mangrove plantation increases swamp area. Consequently, changes in swamp area are determined by the combined effects of natural and human stresses on mangrove forests over a certain period and domain. The total swamp area increased substantially from 1930 to 1965, majorly resulting from the mangrove growing seaward rates being much higher than the reduction rates by land reclamation in sub-regions A and B (Fig. 7a). In the period 1965-1995, the mangrove growing rates were substantially lower than the reduction rates, resulting in a slight increase in total swamp area. Since the mid-1990s, mangrove planting has become popular as a countermeasure for coastal protection and biodiversity conservation in the RRD, leading to gradual increase in total swamp area from 1995 to the present.  1930-1965 1965-1995 1995-2008 2008 Minus change rates denote decrease in the statistic areas, and "-" means no data available.

Controlling factors on the RRD coastal evolution
Deltaic coasts are highly dynamic due to intense interactions between fluvial and marine processes. Their development is further complicated by increasing impacts of global changes and human activities (Syvitski et al. 2009;Anthony et al. 2015;Brondizio et al. 2016;Fan et al. 2017;Li et al. 2017). The most frequently cited natural factors include rising sea level (RSL) and more active tropical cyclones in terms of strength and frequency (Fan and Liu 2008;and references therein). Long-term tidal gauged data show the mean RSL rate of 2.24 mm year −1 over the period 1950s-1990s in the RRD, a few higher than the contemporary global average rate of 1.8 mm year −1 (Thanh et al. 2005;Thao et al. 2013;and references therein). However, the RSL has been overwhelmingly compensated for by rapid deposition at the active distributary mouths and their adjacent coasts in the RRD, producing fast coastal progradation instead of erosion (Figs. 2,3,and 5). The RSL may have a substantial contribution to coastal erosion along the Hai Hau beach between the two active river mouths of Ba Lat and Lach Giang (Fig. 4), but it is difficult to estimate its accurate contribution on the erosion rates. Using the simple formula based on Bruun's rule, it was estimated that the RSL contributed 34% to the coastal erosion rate during the period 1965-1995, and 12% in the period 1995(Duc et al. 2012. On average, five to six typhoons annually affect the Vietnamese coast, and most of them influence the north and central part of the coast Larson et al. 2014). According to the Typhoon Best Track Data of the Joint Typhoon Warning Center, over 200 typhoons landed on the Vietnam coast in the period 1951-2010 (Takagi et al. 2014). According to these data, yearly and decadal variations in typhoon activities were obvious, but the long-term trend due to global warming was not significantly detected (Takagi et al. 2014). The statistics on typhoon yearly records of the Vietnam National Centre for Meteorology and Hydrology show that 86 typhoons directly hit the coast of the RRD over 1962(Duc et al. 2012. The magnitude of coastal erosion and resultant disasters is not only determined by typhoon intensity, but also related to tidal states during the typhoon landing moments. Storm waves induced by typhoons may induce dike bleaching and severe erosion at the upper part of beach or tidal flats, while they have potential to produce concurrent deposition at the lower part as discussed previously. Typhoon-induced coastal erosion could be recovered soon after the storms along the muddy coasts as shown at the Kien Thuy coast (Fig. 2). However, the recovery may not happen along the Hai Hau beach because of sediment deficit. Consequently, more attention should be paid to storm impacts on persistent erosion sandy coasts, and take appropriate measurements to reduce coastal erosion and flooding risks .
The RRD coasts, stretching over 160 km long, are now unevenly dissected by nine active distributary estuaries. Before the HBD completeness, roughly 120 Mt suspended sediments were annually debouched into the Tonkin Gulf through these nine distributaries. Consequently, their estuaries and adjacent coasts were all observed to undergo accretion in the past decades as partly shown in Figs. 2, 3, and 5. The similar coastal accretion pattern in the north RRD, including the Bach Dang, Cam, and Lach Tray mouths, has been well delineated by Bui et al. (2018). However, the Hai Hau beach coast has been reported to have undergone chronic erosion since 1905 (Imamura and To 1997;Pruszak et al. 2002;Dien et al. 2003;Thao et al. 2013). This coastal erosion is principally induced by sediment deficit due to gradual siltation and final abandon of a major distributary channel that previously debouched into the sea through the Ha Lan river course or a little further south (Duc et al. 2012). The evidence was also obvious from the gradual decrease in erosion rates since severe coastal erosion was observed at the beginning of the 20th century, because of coastal self-adjustment towards a new morphodynamic equilibrium (Pruszak et al. 2002).
The accretion rates at the active distributary mouths have been observed to slow down obviously since the mid-1990s (Figs. 2, 3, and 5). This reduction can be attributed to the sharp decrease in sediment discharge by the operation of HBD reservoir since 1988, which has been accused of reducing suspended sediment flux at the Son Tay station from 119 Mt year −1 (1960)(1961)(1962)(1963)(1964)(1965)(1966)(1967)(1968)(1969)(1970)(1971)(1972)(1973)(1974)(1975)(1976)(1977)(1978)(1979) to 46 Mt year −1 (1989-2010) (Dang et al. 2010;Wang et al. 2011;Vinh et al. 2014). Recently, two more dams have been built in the lower Red River system, including the Tuyen Quang Dam completed in 2009 on the Lo River with a reservoir surface area of 42 km 2 and an effective storage capacity of 2.2 km 3 , and the Sonla Dam completed in 2010 on the Da River with a reservoir surface area of 440 km 2 and an effective storage capacity of 12.5 km 3 (Duc et al. 2012;Lu et al. 2015). In the near future, more dams are expected to be constructed in the drainage basin, and the decrease in suspended sediment flux from the river will also continue leading to a much lower suspended sediment rate level than the present. Consequently, the reduction in accretion rates along the muddy RRD coasts will deteriorate, and some coastal stretches may shift from accretion into erosion soon, deserving more attention on the coastal planning and protection.
Changes in longshore current directions and strengths in response to monsoon climate and large-scale coastal configuration changes have the potential to induce coastal erosion. For example, the modern Red river delta has been greatly reshaped by longshore currents from a seaward-extending triangular form to a shoreward-attaching trapezoid configuration since the 1930s (Fig. 3). Because of the protrusion of modern Red river delta, the downstream longshore current has been detached offshore and run southwestward between 10 and 30 m in depth, so no sediments can be supplied from the delta to the Hai Hau Beach (Duc et al. 2007;Hung and Larson 2014). The longshore current generated by local wave shallowing runs parallel to the coast within the depths of <5 m, and transports large amounts of resuspended sediments from the Hai Hau Beach towards the Lach Giang mouth, leading to erosion at the beach and deposition at the river mouth (Duc et al. 2007;Hung and Larson 2014).
In addition, large-scale land reclamation projects squeeze the areal areas of mangrove forests and intertidal zones, increasing the risk of coastal erosion by the less damped waves. Poorly designed coastal infrastructure may not protect the coast as expected, but may increase the risk of erosion at the foot of seawalls, potentially leading to seawall collapse and coastal retreat. Other human activities, including coastal mining practices and mangrove destruction and plantation, are also frequently reported to exert great impacts on coastal morphodynamic changes (Thanh et al. 1997;Thao et al. 2013).

Conclusions
The coasts of the RRD are composed of ∼130 km muddy flats and ∼30 km sandy beaches. Most muddy flats have undergone continuous accretion since 1930, while sandy beaches have suffered chronic erosion in the last century. This accretion-erosion pattern is dependent firstly on sediment fluxes of the different distributaries, and secondly on sediment redistribution by tides and longshore currents. Therefore, coastal accretion rates were usually highest at the front of distributary mouths, and decreased away from there. However, due to the distributary channel swift event, the Hai Hau coast located between the Ba Lat and Lach Giang mouths has been cut off from direct riverine sediment supply at least since the beginning of the 20th century. Moreover, the coast received negligible longshore sediment supply from the upstream river mouth because the regional southwestward longshore current has been detached offshore by the protruding deltaic lobe at the Ba Lat mouth. But wave-generated longshore and offshore currents continued to export sediments from the Hai Hau beach, producing severe erosion thereof. The combined effects prolonged coastal erosion over a century on the Hai Hau beach.
Spatiotemporal coastal evolution was monitored in detail by examining changes in the locations of both sea dikes near the spring MHW-coastline, vegetation lines near the MSL-coastline, and 0 m isobaths near the spring MLW-coastline. We found that 0 m isobaths are less influenced by local human interference but highly sensitive to external stresses, including storm waves and river discharges. The recently decelerating accretion rates on the muddy coasts were linked with the decrease in sediment discharge from the river after the completion of HBD in 1988. However, gradually decreasing erosion rates on the Hai Hau beach resulted from the chronic coastal self-regulation towards a new morphodynamic equilibrium, together with recently strengthening coastal embankments. Both sea dikes and mangrove forests were highly influenced by human activities but less by external forcing. However, large-scale mangrove destruction and poorly designed coastal infrastructures had potential to increase risks of coastal erosion and flooding disasters. A general decreasing trend of intertidal flats was observed on the RRD coast because of land reclamation for agriculture and aquaculture to satisfy increasing socioeconomic development. Because of increasing awareness of their biodiversity and coastal protection values, the area of mangrove forests has been observed to increase persistently after a reduction period in 1965-1995. In the near future, more attention should be paid to increasing hazards along the erosional muddy coasts in the RRD, because local residents are not effectively protected by coastal infrastructures and insufficiently aware of potential risks.