Multi-decadal shoreline change in coastal natural world heritage sites – a global assessment

Natural World Heritage Sites (NWHS), which are of Outstanding Universal Value, are increasingly threatened by natural and anthropogenic pressures. This is especially true for coastal NWHS, which are additionally subject to erosion and flooding. This paper assesses shoreline change from 1984 to 2016 within the boundaries of 67 designated sites, providing a first global consistent assessment of its drivers. It develops a transferable methodology utilising new satellite-derived global shoreline datasets, which are classified based on linearity of change against time and compared with global datasets of geomorphology (topography, land cover, coastal type, and lithology), climate variability and sea-level change. Significant shoreline change is observed on 14% of 52 coastal NWHS shorelines that show the largest recessional and accretive trends (means of −3.4 m yr−1 and 3.5 m yr−1, respectively). These rapid shoreline changes are found in low-lying shorelines (<1 m elevation) composed of unconsolidated sediments in vegetated tidal coastal systems (means of −7.7 m yr−1 and 12.5 m yr−1), and vegetated tidal deltas at the mouth of large river systems (means of −6.9 m yr−1 and 11 m yr−1). Extreme shoreline changes occur as a result of redistribution of sediment driven by a combination of geomorphological conditions with (1) specific natural coastal morphodynamics such as opening of inlets (e.g. Río Plátano Biosphere Reserve) or gradients of alongshore sediment transport (e.g. Namib Sea) and (2) direct or indirect human interferences with natural coastal processes such as sand nourishment (e.g. Wadden Sea) and damming of river sediments upstream of a delta (e.g. Danube Delta). The most stable soft coasts are associated with the protection of coral reef ecosystems (e.g. Great Barrier Reef) which may be degraded/destroyed by climate change or human stress in the future. A positive correlation between shoreline retreat and local relative sea-level change was apparent in the Wadden Sea. However, globally, the effects of contemporary sea-level rise are not apparent for coastal NWHS, but it is a major concern for the future reinforcing the shoreline dynamics already being observed due to other drivers. Hence, future assessments of shoreline change need to account for other drivers of coastal change in addition to sea-level rise projections. In conclusion, extreme multi-decadal linear shoreline trends occur in coastal NWHS and are driven primarily by sediment redistribution. Future exacerbation of these trends may affect heritage values and coastal communities. Thus shoreline change should be considered in future management plans where necessary. This approach provides a consistent method to assess NWHS which can be repeated and help steer future management of these important sites.


Introduction
World Heritage Sites are locations of Outstanding Universal Values (OUV) selected by the United Nations Educational, Scientific and Cultural Organization (UNESCO) as having cultural, historical, scientific, or other forms of significance [1]. Of the 1 092 World Heritage Sites, 209 are classified as Natural World Heritage Sites (NWHS) [1]. NWHS have a high irreplaceability (uniqueness or rarity) factor; they are prioritised and have extraordinary biodiversity and geodiversity features compared to other protected areas [2,3]. The UNESCO World Heritage Centre established a list of 14 primary factors of deterioration of the OUV ranging from human activities (development, pollution, social and cultural use), climate change and severe weather events, to invasive species, management and institutional factors [4]. Climate change and severe weather events can affect coastal areas through flooding, inundation and increased erosion [5][6][7]. 88 NWHS intersect the coast and include sites most at risk from climate change [8]. Although they have pristine environments, their coastlines are increasingly subject to anthropogenic pressures inside and outside their boundaries such as pollution, population growth, and development including port facilities, dams and pumping stations. Following the International Union for Conservation of Nature conservation Outlook assessment conducted in 2017 [9], only 20% of coastal NWHS have a good conservation outlook, and the conservation outlooks of 39% of the sites range from significant concerns to critical. Moreover, the OUV of about two thirds of coastal NWHS are at high to very high threat from deteriorating factors. Additionally, these sites are subject to physical processes such as sea-level rise (SLR) [10][11][12][13][14] and human modifications to sediment budgets [15]. However, shoreline change is not systematically monitored or reported in many NWHS [16][17][18], so it is unclear how NWHS shorelines have or could change. As sites that have very limited internal anthropogenic disturbance, they present significant opportunities to analyse how and why shorelines change due to natural drivers and/or external pressures.
Previous assessments of shoreline change in heritage studies include local [19][20][21][22], regional [23] or global [24,25] studies. Local studies included the Sundarbans mangrove forests [20,22], the Everglades National Park [21], and the Wadden Sea [19]. A regional evaluation of 49 coastal Cultural World Heritage Sites around the coast of the Mediterranean found that 37 low-lying sites are at risk from a 100year flood event today and that 42 sites are threatened by coastal erosion [23]. Two global studies have analysed the effects of future shoreline change due to SLR. The first determined that 80% of the coastal wetlands of international importance could be affected by a 0-1 m rise in sea level [25]. The second study found that 40 to 136 cultural and mixed coastal World Heritage Sites may be affected by flooding over 2 000 years if global temperatures and sea-levels continue to rise [24]. To date, no study has explored globally past multi-decadal shoreline change and its possible drivers in NWHS in term of their geomorphology, elevation, land cover, lithology, climate variability and sea-level change.
The availability of satellite images from 1984 to present via the Google Earth Engine has allowed the creation of a global consistent shoreline change dataset that can be used to monitor coastal NWHS [26][27][28]. In this paper, global datasets of shorelines, geomorphological conditions, and relevant forcing drivers are used to evaluate historic shoreline change from 1984 to 2016 across 67 coastal NWHS (out of 88 due to data availability limitations and data cleaning). The objectives are: • To assess and classify historic shoreline change behaviour within the 67 coastal NWHS; • To evaluate the geomorphological conditions associated with different shoreline behaviours (based on their linearity against time) and shoreline trends (recessional, depositional and stable); and • To determine the impacts of historic sea-level change and climate variability on shoreline behaviour.
This paper is structured as follows. The data are introduced in section 2. The methods and results are presented in section 3 and section 4 respectively. The discussion is presented in section 5 and the conclusion in section 6.

Study sites and shoreline change time-series
Boundaries of coastal NWHS were retrieved from the World Database on Protected Areas [29]. 88 sites intersected the Global, Self-consistent, Hierarchical, High-resolution Shoreline database [30] (figure 1). Shorelines were obtained from a global assessment of derived Landsat images [26][27][28]. This provided satellite-derived shorelines (SDS) data points and their yearly positions based on transects spaced 500 m apart. SDS data points were available for 71 out of 88 coastal NWHS due to limited coverage of historic satellite imagery in offshore waters. The raw shoreline time-series data were cleaned from transects containing less than five SDS data points and having a temporal coverage shorter than seven years [26]. Approximately 1.5 million  [29], Global, Self-consistent, Hierarchical, High-resolution Shoreline database [30], and shoreline time-series data [26][27][28]). time-series data points were selected. Further conditional and outlier data cleanings were undertaken (supplementary section A.1.1, available at stacks.iop.org/ERL/15/104047/mmedia). The conditional cleaning was performed for more consistency on the assessment of shoreline trends: all transects that had at least 17 SDS data points were retained for the analysis (supplementary section A.1.1). The outliers' cleaning was performed to delete extreme SDS data points values (deviating by more than three times the standard deviation) within each transect (supplementary section A.1.1). The cleaning process (see flowchart in supplementary figure SM1) removed 3.8% of the raw SDS data points, and 67 sites remained in the analysis (figure 1).

Geomorphological conditions
Information of topography, land cover, coastal type and lithology (table 1) was obtained from global databases to analyse how depositional and recessional shoreline change rates (SCR) varied (supplementary section A.1.2). The resolution of the topography and land cover datasets (~500 m at the equator) is similar to the shoreline data. The coastal type dataset resolution is 50 km and permits the classification of sites. The resolution of lithological data varies, starting from 5 m 2 and is adequate for both transect-and site-based analysis. These datasets are suitable due to their coverage of the study area allowing for a consistent analysis; moreover, their resolutions are suitable for a global and site-based assessment of shoreline trends.

Climate variability and sea-level change
Between 1900 and 2016, global mean sea level has risen by 16-21 cm [35]. However, the effect of local SLR on the shoreline variability is poorly understood as often exceeded by climate variability, local geomorphological conditions, and/or human interventions [36]. Our study hypothesised that local trends of sea-level change [35] may have a potential observable contribution to strong linear shoreline trends within similar geomorphological categories in pristine NWHS, which should be negligibly affected by human interventions. To verify this hypothesis, local trends of sea-level change were assessed, and their effects on strong linear shoreline trends were determined within different geomorphological categories and sites. Linear available trends of local estimates of relative sea-level change [37] (measured by tide gauges) were used. These linear trends are appropriate as contemporary SLR acceleration rates are small (order of 0.1 mm 2 yr −1 ) and are often not detectable at local tide gauge sites because of the large variability present in sea level [38]. Other driving forces of regional climate variability [39] (table 2) were assessed as drivers of shoreline change. These yearly values of large-scale climate indices have been used in previous global assessments of surges and flooding [40,41] and have been shown to influence year-to-year variability in sea level [42][43][44]. The shoreline change dataset is 33 years of length, which is appropriate to capture the year-to-year variability that arises from climate forcing such as El Niño/Southern Oscillation (ENSO) or the other climate indices listed in table 2.

Methods
Three stages of analysis were undertaken, corresponding to the three study objectives.

Shoreline change time-series: linear behaviour classifications and strong linear trends
Prior to fitting a linear regression, the potential linear behaviour of SDS data points, defined by their linearity against time, was assessed using Pearson's correlation coefficient (r) (R-3.5.1 package 'psych' [60]), with the statistical significance measured using the pvalue (the closer r is to ±1 the stronger the linear relationship). Based on past qualitative description of r [60][61][62][63], shoreline change transects were divided as: To assess the contributions of the three linear categories in the long-term shoreline change, mean annual SCR for the three linear categories were assessed using an Ordinary Least Square linear regression applied to transects based SDS [64]. The linear fit is a valid option to describe and forecast long-term predictive analysis and to minimise potential random error and short-time variability [64].
For the multi-decadal period considered in the analysis, linear regressions, which assume that the relationship between shoreline change and time is linear, are not relevant for shorelines changing with weak linear or non-linear behaviours. Thus, only SCR calculated for transects with strong linear shoreline behaviour are highly probable and significant on a multi-decadal scale and were selected to analyse depositional, recessional or stable SCR between 1984 and 2016. As the SDS accuracy is within a subpixel precision for the 33 years  [48,49] SST anomalies occurring in the North Atlantic Ocean [50]. Arctic Oscillation (AO) No particular periodicity [51] Non-seasonal sea-level pressure (SLP) anomalies at the Arctic and Antarctic poles [52]. North Atlantic Oscillation (NAO) No particular periodicity [53] Atmospheric SLP between the Icelandic Low and the Azores High, which affects the westerly winds and location of storm period analysed (15 m for Landsat), SCR between −0.5 and 0.5 m yr −1 were considered stable [26]. Depositional and recessional transects were defined by SCR > 0.5 m yr −1 and <-0.5 m yr −1 respectively [26]. The mean and standard deviation of SCR were calculated for each geomorphological category and sub-category. Geomorphological categories and sub-categories with less than five transects were considered non-representative of mean shoreline change per category. Shoreline change outliers for strong linear transects were removed (<-21.16 m yr −1 for recessional transects and > 23.05 m yr −1 for depositional transects) (see supplementary figures SM10 and SM11). 6 947 transects (98%) remained within 52 sites, after outliers were removed.

Geomorphological analysis
All transects were classified by their topography, land cover, coastal type and lithology (see supplementary section A.1.2). A comparison of the different geomorphological conditions for the strong linear, weak linear and non-linear shoreline behaviours has been conducted followed by an in-depth analysis of the three transects' types of the strong linear behaviour: recessional, depositional and stable.

Climate variability and sea-level change analysis
Comparisons of SDS data points per transect against time-series of climate indexes were undertaken using Kendall τ non-parametric rank correlation [41,65]. The comparison investigated potential dependencies between shoreline change and the ten climate indices defined in section 2.3. The percentage of transects having a moderate/strong positive (τ ⩾ 0.5) or moderate/strong negative (τ ⩽ −0.5) correlation with the time-series of climate indices was assessed for each category of transects defined by Pearson's r classification. The contribution of sea-level change was assessed by fitting a linear regression between recessional and depositional strong linear SCR and local relative sea-level change for different land cover and coastal type categories. Additionally, a comparison between average shoreline evolution and relative sealevel change has been conducted for each site. Only shores with a mean elevation lower than 10 m (definition of the Low Elevation Coastal Zone [66]) were assessed.

Classification of shoreline change time-series
The first objective was to assess and classify shoreline change linear behaviours in coastal NWHS between 1984 and 2016.

Geomorphological analysis
The second objective was to evaluate the geomorphological conditions associated with different shoreline behaviours (based on their linearity against time) and shoreline trends (recessional, depositional and stable). First, a comparison of the geomorphological compositions of strong linear, weak linear and non-linear shoreline behaviours was conducted (figure 4). Transects with strong linear behaviour had a higher percentage of tidal systems (30%) and arheic systems (19%) while transects with non-linear and weak linear behaviours had a higher percentage of fjords/fjärds (14% and 9% consecutively) and islands (13% and 12% consecutively). Strong linear transects had a higher percentage of mangroves (40%) in comparison to non-linear and weak linear transects. Nonlinear and weak linear transects had a higher percentage of different rock types (such as metamorphic, acid plutonic, basic plutonic, and intermediate plutonic rocks) while transects with a strong linear behaviour had the highest percentage of unconsolidated sediments (74%). Transects with strong linear behaviour had a higher percentage of extremely low-lying (18%) and low-lying areas (61%).
Second, the geomorphological conditions associated with strong linear recessional, depositional and stable shoreline trends were evaluated. For 297 stable transects in 18 sites, 62% of the transects had their mean elevation within [1-10 m] and 29% within [10-50 m]. Stable transects consisted of 42% small deltas, 31% arheic systems and 13% tidal systems (figure 5). Within these coastal types, vegetated areas and mangroves were the prevailing land cover types (figure 5). They represented respectively 53% and 34% of the totality of stable transects. 71% of stable transects were unconsolidated sediments, 6% siliciclastic sedimentary rock and 5% acid volcanic rocks. Further analysis were not conducted for transects with stable strong linear shoreline trend as they represent only 4% of the totality of strong linear transects in 35% of the sites displaying a strong linear behaviour.
Among all elevations categories, the comparison of land cover categories shows that transects within the elevation category For all topographic categories, extremely lowelevation transects within vegetated areas had the Table 3.

Climate variability and sea-level change analysis
The third objective was to determine the impacts of historic sea-level change and climate variability on shoreline behaviour in coastal NWHS. The comparison of yearly transect-based time-series of shorelines Table 4. (within the three categories of linear shoreline change behaviour) against ten climate indices indicated no significant statistical association on a global scale (supplementary table SM18). Globally and for different geomorphological categories and sub-categories, there was no positive correlation between shoreline change and relative sea-level change for transects with strong linear recessional or depositional trend. Thus the absolute value of recessional SCR did not increase and the value of depositional SCR did not decrease with increasing relative SLR values for low lying transects (0 to 10 m) (figure 6 and supplementary figure SM14). A weak positive relationship was observed between recessional strong linear shoreline trend and relative sea-level change in vegetated tidal systems below 1 m in the Wadden Sea (figure 7). No correlation has been found between the average shoreline change rate and the average relative sea-level change for each site (supplementary figure SM15).

Discussion
This paper has presented the first global assessment of trends and drivers of shoreline change in coastal NWHS from 1984 to 2016. The data showed that both extreme erosional and accretional tendencies were apparent and one tendency did not dominate in these sites. A classification of linear behaviour with time indicated that strong linear shoreline trends have a significant contribution to the recessional (−3.4 m yr −1 , std 3.6 m yr −1 ) and depositional trends (3.5 m yr −1 , std 4.3 m yr −1 ). The prevalence of unconsolidated sediment in transects with strong linear behaviour demonstrates the potential contribution of coastal sediment processes (affected by human disturbances, waves, tides and tidal currents, wind, currents and sea-level change).
Drivers of strong linear recessional and depositional trends were assessed using geomorphological categorisation of transects, including analysis of case  [1-10 m]. This is partly explained by the lithological compositions of these low-lying environments and the presence of lagoons, sandy beaches, large rivers and large rivers under tidal influences. Río Plátano Biosphere Reserve has the highest mean shoreline recession (−11.8 m yr −1 , std 7.01 m yr −1 ) due to the 2002 opening of an inlet 12 km northwest of Iban lagoon inducing new accretive and erosive processes within the site boundaries that are influenced by Paulaya river sediment discharge and the southeastnorthwest ocean current from Honduras to Yucatan [67]. Sediment deposition, shaped by the Benguela Upwelling system, southwest of the Namib Sand Sea's Conception Bay (evaporite basin) and Sandwich harbour had induced the highest mean accretive shoreline of all coastal NWHS (13.6 m yr −1 , std 5.3 m yr −1 ) [68]. Transects with high mean rates of change (10.1 m yr −1 and −7 m yr −1 ) were found in large rivers within tidal delta situated in the vegetated shorelines of Islands and Protected Areas of the Gulf of California. This extreme trend is linked to natural forcing (wave and tides) but also to the decadal legacy of distant human alterations that interrupts completely constructive processes within the delta and creates new hydrological circulations accompanied by 'unnatural' erosive/accretive processes [69][70][71]. High sedimentary movements, found in vegetated shores (6.9 m yr −1 and −5.1 m yr −1 ) and marshes (5.4 m yr −1 and −5.7 m yr −1 ) in large river systems are due to the construction of engineered structures along the rivers and on the coasts. These extreme rates are observed in the Danube Delta that underwent a large decrease in its sediment discharge due to upstream damming projects (1970 and 1983) in parallel to the undesirable effects of extreme downdrift erosion southward of Sulina Jetties engineered in the second half of the 19th century [72][73][74][75]. Extreme rates of changes are also observed within vegetated tidal systems (8.2 m yr −1 and −6.8 m yr −1 ) and more specifically within barrier islands in the Wadden Sea. The largest unbroken system of intertidal sand and mudflats in the world is a result of dramatic morphodynamic adjustments due to land reclamation (at the boundaries of the NWHS) within the climatic environment of the Frisian coast, which supported the reduction of inlet width (and tidal prism) and thus the growth of the islands [76,77]. The mainland and some islands of the Wadden Sea are engineered (sand nourishment, breakwaters dykes, and dunes protection) and accretive transects are prevalent (supplementary figure SM19) [78][79][80][81][82]. Thus, both depositional and recessional large shoreline trends in coastal NWHS can be linked to coastlines that are highly altered by human intervention, external and internal to a site's boundaries.
Transects within small deltas and arheic systems inside 1 km geodesic buffer from coral reefs have the lowest accretive and recessional shoreline trend ((1.5 m yr −1 and −1.5 m yr −1 ) and (1.7 m yr −1 and −0.9 m yr −1 ) respectively). This trend may be explained as coral reefs provide sediments and coastal protection from waves, storms and floods and minimise the effects of coastal processes on the coastlines [83][84][85]. Most of the sites with coral reefs (such as the Great Barrier Reef (Australia), Shark Bay (Australia), and Komodo National Park (Indonesia)) are under frequent bleaching events in recent years (for instance the third bleaching event 2014-2017 was among the worst ever observed) [86,87]. Unconsolidated sediments within tidal systems protected by coral reefs show less stability than non-tidal systems with higher rates of erosion (−3.4 m y −1 ; std 1 m y −1 ) and accretion (2.1 m y −1 ; std 1.5 m y −1 ) in the Great Barrier Reef and Lagoons of New Caledonia: Reef Diversity and Associated Ecosystems (France). The reef systems within the latter coastal NWHS are among the most affected by present and projected future bleaching events [86]. Coral reefs also deteriorate through overfishing, sewage and agriculture pollution and invasive species [88,89]. Further deterioration of coral reefs would weaken their function to maintain stable coastlines, especially beaches [85,86].
While the shoreline change dataset describes well the changes for continental unconsolidated sediments or sedimentary rocks, it does not demonstrate well shoreline change for coastal transects situated within complex narrow bodies of water as fjords (such as Te Wahipounamu, and the West Norwegian Fjords) or remote rocky cliffs (such as the Galápagos Islands (Ecuador)). A visual verification using Google Timelapse does not show the extreme linear shoreline trend captured by the SDS for these natural systems and informs on the limitation of shoreline detection methodology using satellite images. These errors may occur during (1) image detection: geometric distortion and radiometric errors [90] or (2) image processing: geo-rectification, ortho-rectification [91] and shoreline extraction.
Overall, there are no statistically significant correlations between transect-based shoreline change and the climatic indices of sea surface temperature and pressure anomalies. This may be explained by the limited spatial and temporal resolution of the climatic data and the underlying satellites images used to assess shoreline trends. In Low Elevation Coastal Zones, the analysis of shoreline trends demonstrate that no major historic role of relative sea-level change in accretional or recessional shoreline trend can be identified. One issue is that SLR shows limited variability in time and space over the study period. Further, the high variability at many sites emphasises that other processes, in addition to SLR, are operating. This may be due to different responses of sites to sea-level change, the lack of observations on coastal dynamics and their driving processes and that even in rapidly subsiding coasts other processes (i.e. storms, wave action, human activities) may dominate the shoreline trend [36,92]. However, for transects below 1 m in the vegetated tidal sedimentary systems and marshes of the Wadden Sea, a weak correlation between increasing relative sea-level and shoreline strong linear retreat was detected. This may be explained by rising sea-levels resulting in more inundation but also coastal erosion in low-lying areas [93,94]. The detection of this weak correlation may be related to the better quality of tide gauge data available in the Wadden Sea and to the site's highly dynamic tidally influenced inlets that experience one of the highest mean recession (−8.1 m yr −1 , std 5.2 m yr −1 ) in NWHS worldwide [76,95]. This finding is supported due to the accuracy of shoreline detection methods (0.5 m yr −1 ) allowing observation of increased shoreline change as a result of SLR. For instance, following the Bruun rule [96], 1 mm yr −1 of SLR could induce at least an incremental horizontal change of 1.65 m in a beach slope of 1:50 over 33 years. Detection of climate variability and sea-level change effects on shoreline behaviour could be improved by using higher satellites image resolution (e.g. 1 m), developing monthly time-series of shoreline change (instead of annual time-series) and improving the spatial and temporal resolution of sea level and climatic data especially in remote areas.
The intensification of human interferences, climate change, SLR and wave climate change will affect coastal processes inducing variations in sedimentbudgets [97]. Future SLR may become the main driver of recession [97] effecting geomorphological responses. Eroding low-lying shorelines within tidal systems, large rivers and large rivers under tidal influences, altered by human interferences to coastal processes, may become the most affected coastal NWHS by future SLR and its related changes in sediment dynamics. In the Wadden Sea while contemporary slow sea-level change has expressed itself in losses of beaches or island displacements [98][99][100], future acceleration of SLR may induce back-barrier erosion and sediment deficit in the tidal basin and result in the transformation of the inter-tidal system to a lagoon system [19,101]. The mapping of shoreline linear behaviour and depositional/recessional trends distinguishing abrupt and gradual changes at the transect level, coupled with socio-economic and ecologic indicators, can be used by coastal managers as a preliminary classification of shorelines in term of the importance and urgency of their management, supporting NWHS conservation triage (process of prioritising actions) [102,103]. The enhanced predictive capacity of strong linear shoreline behaviour and the improved understanding of the factors causing this strong linear changes need to be followed by more appropriate management actions, monitoring and planning of coastal NWHS evolving shorelines (when required and to the extent possible). Unconsolidated sediment shorelines in coastal NWHS, not affected by external human interferences, which exhibit a strong linear behaviour of shoreline change, may become primary observatories to assess SLR impacts on natural coastal processes such as in Río Plátano Biosphere Reserve and the Namib Sea. Thus, this study contributes to informing coastal management plans and decisions of coastal and marine protected areas by providing a quantitative evaluation of shoreline behaviour that could improve the guide for Planners and Managers for Marine and Coastal Protected Areas (developed by Salm & Clark [104]).

Conclusions
Despite the high local and international values of coastal NWHS, shoreline change has not been systematically monitored or reported to date. Therefore, it was unclear how NWHS coasts have been changing across the world. This study comprises the first global assessment of multi-decadal shoreline change from 1984 to 2016 within coastal NWHS asking: 'how are coastal NWHS shorelines changing around the world and why?' .
Based on newly available open-access datasets, shoreline change was analysed for 67 NWHS worldwide, in terms of linear behaviour, recessional or accretive trends, and potential drivers of change. Shorelines with strong linear erosional or accretive trends comprise 14% of total coastal NWHS shorelines. They occur within 52 coastal NWHS and demonstrate the largest shoreline erosive and accretive trends (mean of −3.4 m yr −1 and 3.5 m yr −1 , respectively). Among the transects with strong linear behaviour, the highest recessional and accretive trends are found within low-lying unconsolidated sediments shorelines (<1 m) in vegetated tidal coastal systems, and vegetated tidal deltas at the mouth of large river systems. These extreme shoreline trends can be linked to natural coastal morphodynamics such as the opening of inlets or gradient of alongshore sediment transport. In other cases, they can be associated with direct or indirect human interferences such as land reclamation and damming of rivers upstream of a delta. Conversely, the most stable soft coasts are associated with shorelines protected by coral reefs ecosystems. In the future, these shorelines may be subject to increased instability due to the intensification of climate change and human deterioration degrading the natural protective capacity of coral reefs. A positive correlation between recessional (strong linear) shoreline change and relative sea-level change was found in the Wadden Sea, but globally, the effects of SLR on shoreline change are not apparent.
In most cases, shoreline monitoring had not been the main priority in the management of coastal NWHS. The availability of open-access datasets creates opportunities to better understand shoreline change so to inform management actions where necessary. These analyses can be repeated and refined providing new insights, as data extend in time and improve in resolution. Continued systematic monitoring is advised, especially for sites undergoing direct or indirect human interferences.

Data availability statement
The data that support the findings of this study are openly available at http://doi.org/10.5281/zenodo. 3751980.