Spatial and temporal variability in geomorphic change at tidally influenced shipwreck sites: The use of time‐lapse multibeam data for the assessment of site formation processes

Shipwrecks are an integral part of our maritime archaeological landscape and are associated with diverse societal and cultural interests, yielding significant management challenges. Coupled hydrodynamic and geomorphological processes significantly impact the effective in situ preservation of these fragile sites. In this study, we assess sediment budget change and hydrodynamic triggers at metal‐hulled shipwrecks lost between 1875 and 1918, all located in the tidally dominated Irish Sea at depths between 26 and 84 m. This is conducted using time‐lapse, multibeam echosounder surveys at multiannual, annual, and weekly time steps, supported by sediment grain‐size analysis, modeled ocean currents, and shallow seismic data. Results indicate significant changes at all time steps for sites located in sand‐dominated environments, whereas the seabed around shipwrecks settled in multimodal sediments shows virtually no change outside of measurement errors (±30 cm). Variability in geomorphic change is attributed to local environmental factors, including bed shear stress, sediment supply, and spatial barriers to scour. We demonstrate that individual wrecks in similar shelf sea regions can be in very different equilibrium states, which has critical implications for the in situ management of underwater cultural heritage.

In situ preservation of shipwrecks is encouraged by the UNESCO Convention on the Protection of the Underwater Cultural Heritage, which states that assessments of environmental characteristics should be included in site investigations to examine long-term stability of the underwater cultural heritage (UNESCO, 2002). According to the Convention, these assessments should be conducted without causing disruption to sites, thus promoting the use of nondestructive methods (UNESCO, 2002).
Shipwreck sites are often considered as systems in a state of equilibrium achieved sometime after an initial wrecking incident (Astley, 2016;Quinn, 2006;Quinn & Boland, 2010; I. A. K. Ward et al., 1999;Wheeler, 2002). This equilibrium state is dynamic, meaning that it can be perturbed by external or internal forces. This perturbation in the system can result in either a new equilibrium state or maintenance of the current one, depending on the system's capacity to absorb external forces (Astley, 2016;Quinn, 2006;Quinn & Boland, 2010).
The dynamics of equilibrium varies across wreck sites, with some subject to disruptions by storms (Fernández-Montblanc et al., 2016, 2018McNinch et al., 2006), varying tidal currents (Astley, 2016;Quinn & Smyth, 2018), or anthropogenic impacts (Brennan et al., 2013;Gibbs, 2006), whereas others are nearly static, located in more stable physical environments (Eriksson & Rönnby, 2012). What is common is that all wreck sites are characterized by a negative disequilibrium trend, as they undergo a gradual degradation due to chemical (i.e., corrosion) and biological (e.g., wood-boring organisms) formation processes (Foecke et al., 2010;Gregory, 2020;Pournou, 2017;Taormina et al., 2020). Nevertheless, sites in highly dynamic environments, leading to frequent changes in equilibrium states, are prone to accelerated disintegration (Quinn & Boland, 2010). Therefore, the state of a shipwreck site as a system, and its susceptibility to disruption, should always be assessed before implementing any in situ preservation measures (Astley, 2016).
The dynamism of underwater sites is often controlled by their hydrodynamic environment and sediment budget, defined as the rate of net supply or removal of different sediments to the wreck area (I. A. K. Ward et al., 1999). One of the key processes controlling the sediment budget and determining the integrity of underwater structures is seabed scour, which occurs as a result of magnified flow velocity around objects disrupting natural near-seabed currents (Quinn, 2006;Sumer & Fredsøe, 2002). In engineering applications, erosive scour processes are considered detrimental to the stability of underwater structures such as bridges and wind farm piles, which often require special mitigation measures. For example, in the Irish Sea, significant scour developed at the base of monopiles shortly after the construction of the Arklow Bank wind park, resulting in the need to use rock armour to mitigate further scouring (Whitehouse et al., 2011). Therefore, much attention has focused on geotechnical assessments of scour and related processes for offshore engineering purposes (e.g., Matutano et al., 2013;Melling, 2015;Sumer, 2007;Whitehouse et al., 2011).
Although scouring has been researched at shipwreck sites, studies have mostly focused on single surveys, investigating intricate depositional and erosional signatures, referred to as wreck marks (Caston, 1979;Garlan et al., 2015). To fully understand the dynamics of scour development, however, high-resolution bathymetric surveys should be conducted at least two times. This approach, referred to as "time-lapse" or "repeat surveying," has not been used very often to date (e.g., Astley, 2016;Bates et al., 2011;Brennan et al., 2016;Quinn & Boland, 2010;Stieglitz & Waterson, 2013), considering that 3 million shipwrecks are estimated worldwide (Croome, 1999;UNESCO, 2017).
As changes in sediment budget due to scour can ultimately lead to the exposure or burial of shipwrecks, it therefore, also controls oxygen availability, biological encrustation, corrosion rates, and pressure gradients exerted on hulls (Quinn, 2006;I. A. K. Ward et al., 1999), mechanisms significantly influencing site formation. The presence or absence of ongoing scour processes may also indicate whether a shipwreck site is in a stable or dynamic equilibrium.
Considering all these points, it is critical that seabed change at shipwreck sites is understood at various time scales to fully assess their preservation potential.
The aim of this study is to expand the knowledge of formation processes at underwater shipwreck sites in the context of the sediment budget and the hydrodynamic environment. We achieve this by comparing the spatial and temporal development of scour signatures and other bedforms around 10 metal-hulled shipwrecks lost between 1875 and 1918, located at moderate depths (26-84 m) in contrasting hydrodynamic and sedimentary settings in the Irish Sea. Very highresolution time-lapse bathymetric data were collected and integrated with seismics, sediment samples, and modeled near-seabed tidal currents. Bathymetric survey design was optimized to collect the highest resolution data possible, resulting in difference models and analysis of geomorphic change at a resolution previously unrealized in time-lapse assessments of underwater cultural heritage.
The number of wrecks investigated and the combination of methods provide new knowledge that allows for the development of more accurate underwater site formation models. In addition, we recognize this investigation as relevant for offshore engineering applications, as it focuses on localized morphodynamic change around submerged man-made structures.

| Study area
The study area encompasses the waters off the east and northeast coast of the island of Ireland from Rathlin Island in the north to Dublin Bay in the south (Figure 1 Duke. The 10 shipwreck sites were selected to represent a range of physical environments, characterized by different tidal conditions and varied geological substrates. These shipwrecks are just a few of more than 18,000 other wrecking incidents that are recorded off the island of Ireland (Brady et al., 2012;Forsythe et al., 2000). The Irish Sea is a shelf sea dominated by tidal currents that are typically rectilinear along coasts and in straits (Ozer et al., 2015).
Semidiurnal lunar (M2) and solar (S2) tides propagating from the Atlantic Ocean through the North Channel to the north and St. George's Channel to the south largely control the tidal current magnitudes (Neill et al., 2014;Ozer et al., 2015). The areas chosen for the study (Figure 1) are characterized by depth-averaged peak spring tide magnitudes ranging from 0.5 m/s around Belfast Lough (Atkins, 1997), 1 m/s off Dublin Bay (Howarth, 2001), to 3 m/s in Rathlin Sound. The latter is a candidate area for tidal turbine development (Lewis et al., 2015;Pérez-Ortiz et al., 2017), whereas offshore Dublin Bay is proposed for offshore wind development.
Flows in the Irish Sea are also influenced by surface waves, inertial currents, residual currents, and storm surges. Waves are generally characterized by a short period with a limited access of swell waves to the basin, as it is partly enclosed. In general, tidal current amplitudes are an order of magnitude greater than long-term averaged residual currents (Bowden, 1980). Inertial currents, episodically generated in the thermally stratified waters of the western Irish Sea, are observed to reach 0.2 m/s in the surface layer (Sherwin, 1987).
The circulation in this part of the basin is also affected by a densitydriven cyclonic gyre that forms during spring and summer, with modeled baroclinic currents reaching 0.14 m/s (Horsburgh & Hill, 2003). The largest nontidal, depth-averaged current magnitude has been reported to reach 0.65 m/s in the North Channel (east from Belfast Lough; Figure 1), which was attributed to a storm event (Knight & Howarth, 1999). Externally generated storm surges propagating from south and north appear to interact with tides causing twice-daily intermittent oscillations (Howarth, 2001). Nevertheless, the prevailing flows in the whole basin remain tidally generated.

| Geological setting
The Irish Sea was glaciated during the Last Glacial Maximum (27 ka-18 ka BP; Scourse et al., 2019;Van Landeghem & Chiverrell, 2020) when the Irish Sea Ice Stream advanced through the Irish Sea, eroding and reworking sediment, and depositing variable thicknesses of glacial diamict. This diamict was often deposited directly on the bedrock and is referred to as the upper till member (Jackson et al., 1995). The upper till member comprises sand to boulder-grade material and is often overconsolidated (Coughlan et al., 2019;Mellet et al., 2015). During the retreat phase of the Irish Sea Ice Stream during deglaciation, large amounts of meltwater were discharged along with outwash material in the form of a heterogeneous mix of sediments, predominately gravels, with mud, sand, and cobbles, referred to as the chaotic facies (Coughlan et al., 2019;Jackson et al., 1995).
The subsequent marine transgression and present-day hydrodynamics reworked much of the glacial and postglacial sediment in the south Irish Sea into a mosaic of substrates (S. L. Ward et al., 2015).
These processes created a series of dynamic bedforms, including migrating sediment waves, in an area dominated by coarse lag deposits.
The sediment waves vary in size and morphology, with the magnitude of migration highest in the central Irish Sea, with average rates of up to 35 m/year, decreasing northward (Van Landeghem, Uehara et al., 2009;Van Landeghem et al., 2012;. The northward transport of sediment eventually terminates in an area of low bed stress, referred to as the Western Irish Sea Mud Belt, where the fine-grained sediment has been accumulating since the end of the last glaciation (Belderson, 1964;Woods et al., 2019). In this area, there are thick deposits (up to 40 m) of stratified gray-brown muddy sands with silts and clays referred to as the mud facies (Jackson et al., 1995). This mud facies is typically of low shear strength and so potentially prone to scour around obstacles (Coughlan et al., 2019

| Sediment samples
Sediment samples were collected using a Shipek grab inside and outside the erosional and depositional signatures. The granulometric analysis of the sediments was performed using a MALVERN Mastersizer 3000 laser diffraction particle size analyzer for fractions with grain sizes <2.38 mm, with an exception for the samples collected at the FV St. Michan and HMS Vanguard sites, which were analyzed using a sieve stack. Sediment classification into Folk classes (Folk, 1954) and median grain size (d 50 ) calculations including >2.38 mm particle sizes (i.e., gravel) were performed using Gradistat v. 8 software (Blott & Pye, 2001). In cases where samples indicated gravelly components in the mixture or gravel as a main fraction, a nominal value of 4 mm was used to represent gravel in the median grain size calculations.

| Data analysis
The first step of the analysis procedure was to thoroughly characterize each site using the MBES and ROMS data. This entailed determining dimensions and depths of the shipwrecks and associated erosional/depositional signatures, their orientation relative to dominant tidal currents, and calculating the volume of eroded/deposited sediment. The second step was to determine sediment mobility from the collected sediment samples and ROMS data. This MAJCHER ET AL.
| 433 entailed calculating sediment mobility thresholds and the frequency of threshold exceedance for each site. The third step was to detect and quantify geomorphic change within the intervals covered by time-lapse bathymetric surveys. This included calculation of volumetric changes in sediment budget and maximum scour depths at each site. Effectively, this sequence enables us to characterize in detail the baseline oceanographic, bathymetric, and sedimentological conditions at each site (Steps 1 and 2) and, with this in hand, explore and advance secure explanations for the patterns of geomorphic change identified in Step 3. The procedure followed in each step is explained below.

| Site characterization
Data integration and analysis were conducted using ESRI ArcMap v. Wessel, 1998), combined with a breakpoint classifier. Separate highpass filtered moving mean kernels were chosen to delineate erosional and depositional signatures. High-pass filtered DEMs were then masked using the extents of the erosional scour features delineated by the breakpoint classifier. The 3D Analyst surface volume tool was used on these masked high-pass filtered DEMs to calculate volumes of all the pixels with values below (erosion) and above (deposition) zero. Shipwreck structures were removed for the volumetric calculations using clipping and masking tools.

| Sediment mobility
Current velocities derived from the ROMS model and sediment samples were used to estimate how often sediment may be mobilized at the candidate sites, according to the approach prescribed by Whitehouse (1998) and Soulsby (1997). In this method, the potential mobility of sediment can be assessed by comparing the values for current-related bed shear stress (τ c ) with the critical shear stress (τ cr ) and calculating exceedance levels (τ cr < τ c ). The current-related bed shear stress is calculated using the following equation: where ρ is the water density, C D the drag coefficient, and Ū the depth-averaged current speed. The critical shear stress is then defined by the following: where θ cr is the critical Shields parameter, g is the acceleration due to gravity (9.81 m/s 2 ), ρ s (1700 kg/m 3 ; Tenzer and Gladkikh, 2014) and ρ (1027 kg/m 3 ) are densities of sediment and water, respectively, and d 50 is the median grain size.
The presence of an obstacle to a flow causes its contraction and increase in speed, hence, increasing shear stresses exerted on the bed downstream of the obstacle (Soulsby, 1997;Whitehouse, 1998).
Therefore, in this study, exceedance levels were calculated separately, assuming a fourfold shear stress amplification factor determined by Smyth and Quinn (2014) (1997), oscillatory flows induced by waves can affect the seabed when the following relation is satisfied: where H s is significant wave height and h is water depth at the seabed. The calculated 99th percentile of the distribution of significant wave heights (H s99 ) recorded at the M2 buoy was, therefore, used with Equation (3) to assess the wave influence on the sediment mobility at the sites. The wave data were analyzed separately for the period of the 2019 survey (between October 24 and September 5, 2019) to assess the wave regime for the weekly time-lapse bathymetric coverage.

| Bathymetric time-lapse analysis
The time-lapse analysis of the bathymetric data sets comprised multiple steps. First, DEMs of Difference (DoD) were obtained by a subtraction of DEMs corresponding to subsequent surveys, using a raster minus tool. Shipwreck structures were masked in the DoDs, as they introduce outliers in bathymetric time-lapse analyses (Astley, 2016). Additionally, the study concentrates on geomorphic change at the sites and not on structural changes of the shipwrecks themselves.
Recording the latter would require a different approach to survey design (e.g., Westley et al., 2019) and data analysis involving comparisons of dense point clouds rather than the DEMs. Measured depth uncertainties were approximated for each bathymetric DEM using vertical total propagated uncertainty values (vTPU) calculated by the CUBE algorithm during the multibeam data processing. These vTPU values, corresponding to any two compared DEMs, were combined using Equation (4) Modeled near-seabed current directional distributions vary slightly across these sites (Figure 2), with maximum spring tide magnitudes oscillating around 0.5 m/s (Table 2).   Gravelly muddy sand is recorded in the erosional signature stretching to the NNW, whereas the rest of the samples indicate muddy sand (Figure 4c). At this site, the volume of eroded material calculated from the relief modeling is slightly higher than for the deposited material (Table 2). Mud Belt, an area that typically experiences lower bed stresses than the surrounding sand-dominated areas (Belderson, 1964;. Nevertheless, a deep scour pit has developed around the shipwreck with a maximum depth of 4.34 m (Figure 4d). The shipwreck, which is much smaller in comparison to the other investigated vessels (Table 2), is elevated above the pit on a sediment mound (Figure 4d). Some material has been redeposited around the pit, which is, therefore, reflected in the calculated depositional volume (Table 2). Sediment samples were highly multimodal. The muddy sandy gravel is recorded close to the boundary of the scour pit. Attempted sampling was unsuccessful inside the pit, indicating a coarse sediment. Gravelly mud is recorded in the depositional mound proximal to the shipwreck, and slightly gravelly and gravelly muddy sands are present in the depositional signatures outside the scour pit and in the surrounding seabed (Figure 4d).

| Gravel-dominated sites
SS Lugano and SS Santa Maria are located on coarser substrates as compared with the other sites, which is reflective of a much stronger tidal regime, with peak modeled flows of 1 m/s (Table 2). SS Lugano is situated on a sandy gravel and gravel bed, with a small depositional zone near the bow and lacking any erosional signature. A narrow (around 10-m width) low-profile (0.1-0.2 m) depositional braid extends 120 m SSE from the shipwreck, probably comprising a finer sediment (Figure 5a). In contrast, at the SS Santa Maria site located 9.6 km SE of SS Lugano, there exist significant flow-aligned erosional signatures (Table 2) (Table 2). Sediment threshold exceedance values for the multimodal sites exhibit the broadest range, most likely due to the geographic and regional differences between these sites. However, despite their relative proximity, the exceedance levels calculated for samples from the SS Chirripo site (2.3%) were significantly lower than those for the samples from SS Tiberia (50.5%). Exceedance levels for sample data at the SS Polwell are relatively low at 17.2%. Wrecks in graveldominated settings exhibit low exceedance values, 4.9% for the SS Lugano site and limited exceedance of 19% at the SS Santa Maria site (Table 2). When the fourfold shear stress amplification factor established by Smyth and Quinn (2014) is applied to these sites, exceedance levels were exceeded at least 50% of the time for all sites. Greatest exceedances are modeled for the SS WM Barkley and HMS Vanguard sites, 87.6% and 91.2% of the time, respectively.

| Wave influence
The analysis of the M2 wave buoy (Figure 1b) data shows that the calculated 99th percentile of significant wave height distribution is 3.5 m (18-year return period). According to Equation (1), only wreck sites shallower than 35 m are directly affected by storms in the study area. The SS Chirripo, RMS Leinster, and SS Polwell sites meet this criterion, but they are all located closer to the western shore of the Irish Sea than the M2 buoy, and hence, are assumed to be partly sheltered with a very limited wave influence. During a storm event that disrupted the 2019 survey, the significant wave height did not exceed 3 m; hence, none of the sites surveyed in the period were directly influenced by waves.

| Time-lapse analysis
Results of the time-lapse analysis are divided into multiannual (9, 5, and 4 years), annual, and weekly changes. Table 3 (Table 3). Results of the time-lapse volumetric change calculations for shipwrecks on sandy beds are listed in Table 4. The analysis for the multimodal sites does not indicate any significant volumetric changes outside detection thresholds, and no time-lapse data are available for the gravel-dominated substrate sites.  (Table 4). At SS Hare, major geomorphic changes occurred between 2010 and 2015 T A B L E 3 Maximum depths and annual changes of bed erosion (relative to a preceding time step) due to scour at the sites

| Weekly change
Difference models for a 1-week period before and after a spring

| DISCUSSION
In this study, we aimed to expand our knowledge of shipwreck site formation processes, focusing on sediment budgets and hydrodynamic conditions. To accomplish this, we investigated the spatial and temporal scales of geomorphic change at metalhulled historic shipwrecks in a tidally dominated environment.

| Variability of geomorphic change at sites in different seabed environments
At all wreck sites where erosional and depositional signatures were identified, a strong correlation between their directionality and the prevalent current direction is noted, suggesting that current-induced bed stress is the dominant control on scour processes. A geomorphic change verified by difference modeling primarily takes place either inside the scour signatures or is associated with the tidally controlled migration of sediment waves.
Even during storm conditions, the influence of wave action on the sites is shown to be insignificant. However, to fully determine the influence of storm events, deployment of monitoring equipment at the sites would be necessary.

| Sand-dominated sites
Wreck sites located in sand-dominated settings are highly dynamic, with significant changes in seabed morphology recorded on a weekly, Sand-dominated sites exhibit the highest and most consistent levels of bed stress values, exceeding sediment thresholds, enabling mobilization ( Table 2). The stratigraphy at these sites is often complex, with quaternary sediments overlain by migrating bedforms and sand sheets controlled by contemporary hydrodynamic processes. Sediment waves in this part of the Irish Sea are typically formed in an upper mobile layer that comprises reworked glacial and postglacial sediments. Beneath this mobile layer lies a coarse gravel lag of glacial origin, either the chaotic facies or the upper till members described by Jackson et al. (1995).
These Late Pleistocene layers are more resistant to scour due to their coarse composition and/or overconsolidated nature . This is demonstrated at the HMS Vanguard site, where a significant geomorphic change is recorded (Figure 6h), but further vertical erosion of the scour pit is limited by the presence of the underlying glacial deposits (Figure 3). Given the similarities in regional geology, it is, therefore, possible that similar limiting layers occur at the RMS Leinster, SS Hare, and SS WM Barkley sites. Seismic data collected at these sites are not sufficient to confirm this. Barriers limiting the vertical geomorphic change at shipwrecks have previously been postulated, for example, at the HMS Scylla site (investigated by Astley, 2016) or Queen Anne's Revenge (McNinch et al., 2006). Here, we demonstrate that such limits can be detected by shallow seismic profiling, enabling accurate predictions of future scour progression at underwater sites. This technique, therefore, provides important information for site management, allowing for more accurate underwater site formation models.
In another possible scenario for the sand-dominated sites, a limit to further vertical erosion may appear when the scour pits develop to the level at which a shipwreck will be lowered below the ambient seabed and cease to perturb the flow of tidal currents (McNinch et al., 2006;Trembanis et al., 2007). In this way, a low-lying shipwreck structure halts ongoing scour processes, so it becomes gradually buried under sediments supplied with the migrating bedforms or during quiescent periods (McNinch et al., 2006). However, the 2019 survey data indicate that all the investigated shipwrecks with deep scour pits still present significant obstructions to the tidal currents (Table 2) and do not appear to be sinking within the depressions. Hence, this scenario has a low probability, unless the shipwrecks experience dramatic structural changes that would lower their height above the seabed (Astley, 2016).
Notably, the maximum scour depths do not change significantly at any of the wreck sites investigated. The largest change in the total scour depth occurred at the SS WM Barkley site between the years 2010 and 2015, when one of the pits was infilled with sediment; however, the same pit was eroded again in the subsequent years (Table 3). which locally magnify current-related shear stresses exerted on the seabed by introducing lee-wake vortices and vortex shedding (Quinn, 2006;Smyth & Quinn, 2014). The same mechanism was cited by Astley (2016)  flow to a far lesser extent than SS Tiberia, whose full length is oriented nearly perpendicular to the peak flow (Quinn & Smyth, 2018).
Local geological control on scour signature propagation is also observed at the multimodal SS Polwell site. Although extensive scour is recorded to the north of the wreck, scouring to the south of the structure is limited (Figure 4c), despite a strong bidirectional current flow. A distinctive ridge feature located to the south acts as a horizontal (or spatial) impediment to scour propagation. Given its morphology and the regional geology, this ridge is likely to be a

| Gravel-dominated sites
Although no time-lapse data of sufficient resolution are available for the gravel-dominated sites, MBES coverage and sediment samples, coupled with modeled near-seabed current information, allow for some observations. For example, at the SS Santa Maria site located off the north coast of Ireland, a significant volume of material has been eroded from around the bow of the wreck (Figure 5b and

| Equilibrium states of shipwrecks and their implications for in situ preservation
It is clear that sand-dominated sites are highly dynamic as compared with the gravel-dominated and multimodal sites (Figure 11a). This dynamism is manifested in all aspects of the sites' analyses, from the initial volumes calculated for the erosional and depositional sig-  (Table 4). Similar to the Burgzand Noord shipwrecks in the North Sea described by Manders (2009) (Bates et al., 2011;Pascoe, 2012).
Significant changes (±2-3 m) on a multiannual timescale were also noted at the more modern (World War 2) shipwreck site of SS Richard Montgomery (Astley, 2016). The wreck is also situated on a sandy seabed (fine sand), with an external supply of sediments, and the reorganization of the erosional/depositional signatures was observed around it (Astley, 2016).
Such an open, high-energy system may have a profound impact on in situ preservation of archaeological material. Tonnes of sediments are regularly eroded/deposited around the shipwrecks, acting as scrambling devices (Muckelroy, 1978) and causing rapid shifts in pressures acting on their hulls. This mechanism probably affects the SS WM Barkley site. In the time-lapse data we collected, a build-up of sediment is initially recorded to the north of the shipwreck, partly covering its portside (years 2010-2015; Figure 6a).
Subsequently, substantial erosion of the site uncovered part of its starboard side (years 2015-2019; Figure 6b). This erosion event also gave rise to the maximum bathymetric change (−4.9 m) recorded in any of the DoD. Such dramatic bed level change is likely to impact the structural stability of shipwrecks, resulting in mechanical damage and accelerated collapse (Quinn, 2006). Therefore, such open sites, in highly dynamic equilibrium with their environment, should be a priority when it comes to introducing in situ preservation measures, planning regular monitoring or authorized emergency excavation.
An exception among the sand-dominated wrecks is the HMS Vanguard site, where scouring appears to have reached its maximum depth. A non-erodible coarser sediment layer exposed in the scour pit limits the vertical scour extent and sediment wave migration occurs only outside of it. As sediment supply to the site has ended, the site appears to have reached a static equilibrium phase ( Figure 11a). This stability has possibly aided the remarkable preservation of the shipwreck, despite it being older than the other sites investigated. HMS Vanguard is nearly fully intact on the seabed in comparison to the other sand-dominated sites (Table 2), which are partly buried and scattered. Although the site appears stable, change may be triggered in the future by events like severe storms, anthropogenic activities, or parts of the wreck itself collapsing ( Figure 11a). HMS Vanguard is still in a negative disequilibrium trend like any other underwater site and will undergo a gradual degradation due to corrosion and other formation processes, as it is almost entirely exposed to the seawater (I. A. K. Ward et al., 1999;Figure 11c).
The sites dominated by multimodal sediments expressed virtually no significant geomorphic change throughout all the time-lapse periods, and may, therefore, be described as static, closed systems, which have achieved a stable equilibrium (Figure 11b). These shipwrecks are only slightly buried in the seabed, and the scour processes have settled analogous to artificial reef structures on a mixed seafloor described by Raineault et al. (2013). At the SS Tiberia, FV St.
Michan, and SS Polwell sites, scour processes caused the removal of fine-grained sediments, leaving only coarser, non-erodible substrates in the pits. SS Chirripo must have reached the equilibrium state soon after the wrecking incident, as no scour is recorded around it. As a result, these shipwrecks are generally largely intact and have a higher preservation potential, akin to HMS Vanguard.
Although no time-lapse data are available for the SS Lugano site, it is also assumed to be in a static equilibrium with the environment.
No erosion is recorded in the DEM (Figure 5a) and no sediment waves are present; thus, the geomorphological context is assumed static. However, SS Santa Maria's broken bow caused the advance of significant erosional signatures (Figure 5b), despite resting on a coarse seabed dominated by gravel and boulders. The tidal currents F I G U R E 11 Conceptual site dynamics models for the investigated sand-dominated (a) and multimodal sites (b) with a corresponding site preservation estimation for the two categories (c). Rapid changes in the site dynamics due to continuous sediment supply and/or single events cause accelerated degradation and lower site preservation. In contrast, stable periods favor preservation, which is, however, always decreasing due to chemical and biological deterioration [Color figure can be viewed at wileyonlinelibrary.com] at the location are among the strongest in the Irish Sea (Pérez-Ortiz et al., 2017); hence, there is a probability of some geomorphic change and the associated system's instability.
Nevertheless, it is important to understand that the stability of underwater sites is not solely controlled by their geomorphological and hydrodynamic settings. A disruption of any equilibrium state can be triggered by external anthropologic influences like digging (Manders, 2009), trawling (Brennan et al., 2016), or dredging operations (Quinn & Boland, 2010). In this study, a direct indication that anthropogenic activity impacts the sites is evidenced by the trawl marks at FV St. Michan. Although no obvious significant reorganization of material has occurred at that site, it is possible that continuation of the trawling may damage the shipwreck and trigger further changes as new nuclei for scour may be introduced, increasing site dynamics (Figure 11b; Quinn, 2006).
In summary, we suggest that knowledge about sediment types, bathymetry, and hydrodynamic processes at underwater sites can provide invaluable information for cultural resource managers. More important, such high-quality hydrographic, oceanographic, and geological data are becoming increasingly publicly available through various studies, initiatives, and projects like INFOMAR in Ireland (Guinan et al., 2020;O'Toole et al., 2020). Such open-sourced data sets can be used for pilot studies, assessing the preservation potential of individual sites and targeting the shipwrecks that need more detailed investigation, for example, involving high-resolution time-lapse surveys.
Further research is needed in understanding vertical propagation of scour at shipwreck sites, for example, using shallow seismic profiling, which is proven to be an effective method for detecting Although we successfully investigated the geomorphic change at shipwreck sites at weekly, annual, and multiannual time steps, these time series should be extended at both ends, allowing the capture of very short-term (daily tidal oscillations and storms) and long-term (10+ years) changes at sites. Additionally, we recognize the need to evaluate the direct influence of dynamic geomorphic change on the structural integrity of shipwrecks. Although such geomorphic changes are known to have detrimental effects on frequently monitored offshore engineering structures (Melling, 2015;Whitehouse et al., 2011), the scale of potential damage to historic shipwrecks remains unclear. Filling these knowledge gaps would result in further refinements of shipwreck site formation models.

| CONCLUSIONS
This study offers new knowledge about site formation processes and long-term stability of metal-hulled shipwrecks, which have remained largely unresearched (Keith, 2016), despite their increasing archaeological value. The management of such shipwrecks is at the crossroads of diverse interests and factors, arising from cultural heritage management, environmental risks, sport diving accessibility, and others (Firth, 2018;Tomalin et al., 2000). As geomorphic changes at the sites are demonstrated to have a direct link to their in situ preservation, we recognize this investigation as highly pertinent to support site management decisions. In the end, it is also applicable for informing future offshore developments and marine spatial planning, as shipwreck sites can be inversely used as proxies of local geomorphological and hydrodynamic conditions (Caston, 1979;Geraga et al., 2020).

ACKNOWLEDGMENTS
The authors specially thank Andrew Conway and the MI Oceanographic Services Team for their assistance with the ocean current data compilation. They thank the reviewers and the associate editor for their insights, which helped them to improve the manuscript. They would also like to thank Carlos Loureiro for advice regarding the buoy data. Additionally, they express their gratitude to Alex Braun, Annika

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

DATA AVAILABILITY STATEMENT
The bathymetric and seismic data that support the findings of this study are available upon request from the corresponding author at