Reconstructing and understanding the impacts of storms and surges, southern North Sea

Coastal barriers are ubiquitous globally and provide a vital protective role to valuable landforms, habitats and communities located to landward. They are, however, vulnerable to extreme water levels and storm wave impacts. A detailed record of sub‐annual to annual; decadal; and centennial rates of shoreline retreat in frontages characterized by both high (> 3 m) and low (< 1 m) dunes is established for a barrier island on the UK east coast. For four storms (2006–2013) we match still water levels and peak significant wave heights against shoreline change at high levels of spatial densification. The results suggest that, at least in the short‐term, shoreline retreat, of typically 5–8 m, is primarily driven by individual events, separated by varying periods of barrier stasis. Over decadal timescales, significant inter‐decadal changes can be seen in both barrier onshore retreat rates and in barrier extension rates alongshore. Whilst the alongshore variability in barrier migration seen in the short‐term remains at the decadal scale, shoreline change at the centennial stage shows little alongshore variability between a region of barrier retreat (at 1.15 m a−1) and one of barrier extension. A data‐mining approach, synchronizing all the variables that drive shoreline change (still water level, timing of high spring tides and peak significant wave heights), is an essential requirement for validating models that predict future shoreline responses under changing sea level and storminess. © 2016 The Authors. Earth Surface Processes and Landforms published by John Wiley & Sons Ltd.


Introduction
Barrier islands are long, thin, low-lying sand-gravel structures, typically oriented subparallel to a mainland coast and separated from it, to varying degrees, by a coastal lagoon. Accounting for 6 to 15% of the world's total shoreline length (Otvos, 2012), they protect biodiverse and ecologically valuable backbarrier wetlands as well as adjacent mainland coasts from direct storm impacts and erosion. Moreover, barriers themselves support growing residential communities [between 1990 and 2000 the US barrier island population reached 1.4 million, a decadal increase of 14% (Zhang and Leatherman, 2011)] and economically important tourist industries (Fitzgerald et al., 2008). These contexts, therefore, raise significant societal issues as to how barrier systems will respond to near-future global environmental change. Whilst many barrier systems have maintained themselves through landward migration by barrier 'roll-over' over the course of the Holocene transgression (e.g. McBride et al., 2013), there is also evidence in the recent geological record for barrier in-place drowning and subsequent sudden shifts in shoreline position (e.g. Sanders and Kumar, 1975;Mellett et al., 2012). Differing modes of barrier dynamics raise questions about (i) the interactions between rates of sea level rise, natural and anthropogenically-modified sediment supply, and regional bathymetric and topographic settings for barrier stability and (ii) future barrier maintenance, fragmentation and loss in the face of accelerating rates of sea-level rise (Church et al., 2013). A related issue, of immediate concern to lives and livelihoods on barrier islands, is the potential impact of near-future changes, still highly uncertain [Intergovernmental Panel on Climate Change (IPCC), 2013], in the magnitude and frequency of tropical and/or extratropical storms on barrier futures. Reconciling these concerns suggests a need for a more thorough synthesis of barrier dynamics, one which integrates long-term trends with short-term events.
At the process level, sediment transport to the landward side of barrier islands by overwash or through storm-induced breaches is the major driver in roll-over dynamics (Donnelly et al., 2006) and landward overwash flux is a key parameter in many coastal evolution models (e.g. Cowell et al., 1995: Jiménez andSánchez-Arcilla, 2004). However, it is clear that overwash is not a simple process in either space or time. Sallenger's (2000) storm-impact scaling model defines four storm-impact regimes based on the relative relationship between the elevation of a morphological feature (i.e. sand dune or upper beach ridge) and that of storm-induced water levels (astronomical tide + storm surge + wave runup). The model predicts non-linear morphological change as the water levels associated with larger and larger storms shift runup and wave attack higher up the profile until, in the most extreme impact category (inundation) 'massive net onshore transport occurs with landward migration of sand bodies on the order of 1 km' (Sallenger, 2000, p. 894). More recent studies, however, do not wholly support the argument that morphologic response increases monotonically with impact regime, with variable landscape change at the low impact stages of the Sallenger model and even flow and sediment transport reversal, from the backbarrier seawards, with barrier inundation (e.g. Long et al., 2014).
There is, therefore, a critical need to determine the scales of variability in coastal change on barriers, the magnitude of change at these scales, and the processes responsible for the observed variability (Stockdon et al., 2007). These questions are being addressed through numerical modelling (e.g. Masetti et al., 2008;Moore et al., 2010;Lorenzo-Trueba and Ashton, 2014) and other approaches (e.g. Bayesian Networks: Plant and Stockdon, 2012). They have recently been explored in micro-tidal settings impacted by episodic hurricane and extratropical storm landfalls (e.g. North Carolina: Lazarus et al., 2012;Mississippi delta plain: Sherwood et al., 2014). Here, we consider these issues through a data-driven approach for a macro-tidal barrier island setting in the southern North Sea, a region where the interaction of storm surge incidence, twenty-first century sea level rise and changing coastal vulnerability is of considerable concern (Weisse et al., 2014). Since 1992, the UK Environment Agency has been collecting annual vertical aerial photography of the east coast of England and monitoring cross-shore profiles at six monthly intervals. We analyse this information alongside available tidal, surge water level and wave datasets. At longer timescales, we supplement these datasets with archival map evidence and records of historic storm surges; at the event scale, we utilize recent developments in ground survey techniques which allow rapid, accurate measurements of storm impacts. Specifically, we address these issues by the determination of: 1. shoreline change rate (calculated at high densification (10 m interval) over a 5 km barrier frontage) based on three different shoreline pairs (2006 and 2007; 2007 and 2008; 2013 and 2014), six-monthly change at a series of 1 km-spaced shore profiles (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014) and high resolution field surveys post-December 2013 storm, assessing short-term barrier morphological response to storm surge impacts; 2. shoreline change rates representing change at the approximately decadal scale based on three shorelines (1992,2000,2008) spanning, eight, eight, and 16 years, respectively; 3. shoreline change rates based on the 1891 and 2013 shorelines representing summary change at the centennial scale.

Location
The barrier island of Scolt Head Island, North Norfolk coast, UK is a National Nature Reserve that has not been subjected to interventionist management and where a long tradition of geomorphological and ecological monitoring extends back to the 1920s (Steers, 1960). The 45 km long, north-facing coastline between Old Hunstanton and Kelling Hard ( Figure 1A) experiences a macro-tidal, semi-diurnal tidal regime with a mean spring tidal range of 6.4 m at Hunstanton, declining to 4.7 m at Cromer, 15 km east of the study area. Four inshore wave stations between Scolt Head Island and Horsey recorded annual mean significant wave heights of 0.49 to 0.73 m, September 2006-September 2009(Environment Agency, 2014. Thus in the Hayes (1979) classification scheme, Scolt Head Island is a 'tide-dominated' barrier island. The combination of high tidal range and low background wave activity with low offshore slopes explains the presence of extensive intertidal environments, developing after 7 ka BP over older basal peats on glaciated surfaces (Andrews et al., 2000). These comprise subtidal mudflats, intertidal sandflats, sand and gravel barriers with ebb tide deltas, tidal channels and barrier-protected salt marshes and mudflats. Extensive reclaimed salt marsh, behind earthen embankments, characterizes much of the landward margin of the intertidal zone ( Figure 1B). The coastal topography is dominated by the Weybourne-Cley-Blakeney Point spit complex to the east and the barrier island of Scolt Head Island to the west ( Figure 1C). The history of the barrier is one of westward extension, with pauses in this extension being represented by welldeveloped dune complexes reaching 15-20 m ODN (ODN, Ordnance Datum Newlyn where 0.0 m approximates to mean sea level) and the presence of irregularly-spaced, recurved and landward-trending 'laterals' (Steers, 1960). The sequence of laterals encloses back-barrier salt marshes of progressively decreasing age from east to west (French, 1993). As well as areas of high (> 3 m) dunes, the seaward margin of the barrier is characterized by stretches of low (< 1 m) dunes; unvegetated and partially vegetated, older, inactive washover fans; and, at its western extremity, mobile spits and bars of sands and gravel.
In this paper we focus attention largely on the dynamics of the high and low dune sections. The shallow, semi-enclosed southern North Sea is affected both by external surges, which enter the North Sea from the North Atlantic and propagate southwards as a Kelvin wave with the tide (taking nine hours to travel from Dunbar, Scotland to the outer Thames estuary) and internal surges which are generated by water level set-up from wind stress within the North Sea itself (Pugh and Woodworth, 2014). They are particularly damaging when they coincide with high spring tides and gale-force onshore winds that can generate large peak significant wave heights, resulting in coastal erosion, sea defence breaching and extensive sea flooding on low-lying coasts. Lamb and Frydendahl (1991 chronicle 26 major sea flood events along the North Sea coastline between 120 BC and AD 1978. In modern times, the benchmark event is often seen as the catastrophic storm surge of 31 January-1 February 1953, arguably, in terms of loss of life (> 2000 deaths), the most devastating natural disaster to affect northwest Europe during the last 100 years (Gerritsen, 2005).

Wave and water level datasets
Between September 2006 and September 2014, there were four major storms/storm surges on the UK east coast. These occurred on 31 October-2 November 2006;17-21 March 2007;7-8 November 2007, 2014; 5-6 December 2013 (Spencer et al., 2014). For each storm surge, the time series record of still water level at seven UK east coast tide gauges (see Figure 1A for locations), as well as the surge residual (difference between actual and predicted water level) were obtained from the British Oceanographic Data Centre (BODC) (https://www.bodc.ac.uk/data/online_delivery/ntslf/). All water level data, reported at 15 minute intervals, were converted from local Chart Datum to ODN. The nearest tide gauge to the field site is located at Cromer (52°56'3.70N 001°18'5.90E), 15 km to the east, but unfortunately there is a gap in the data for the 2007 event and little still water level data for the 2013 event, due to wave action beneath Cromer pier severely disrupting the gauge stilling well. Water level records, where data are available, are reported in Supporting Information (Figures S1A-S1D).
Co November 2007 events at the Scolt Head Island (53°00'.03N 000°41'.06E) inshore (9 m water depth) Acoustic Wave and Current (AWAC) recording station ( Figure 1B). This station had been decommissioned by the time of the December 2013 storm. It has been possible, however, to reconstruct a 2013 wave record at Scolt by cross-correlation with the Blakeney Overfalls Wave rider Buoy (53°03'.20N 001°06'.42E; 10 km offshore, 18 m water depth) record which also reports wave characteristics for the two 2007 events. Data were downloaded from the Centre for Environment, Fisheries and Aquatic Systems (CEFAS) website (http://www.cefas.defra.gov.uk/our-science/ observing-and-modelling/monitoring-programmes/wavenet.aspx). Whilst the inshore AWAC station recorded data at hourly intervals, the Blakeney Overfalls data are collected at 30 minute intervals.

Morphological datasets and analytical methods
Short-term, event-based shoreline change (2006-2014) UK Environment Agency cross-shore profile data have been collected in the field every six months ('winter' and 'summer') since 1992, using a differential global positioning system (DGPS) [horizontal and vertical precision of ±20 and ±30 mm, respectively (Lee, 2008)]. These data were used to assess shoreline change at point locations along the barrier ( Figure 1C). Initially the profile locations were at 1 km spacing, giving a total of five locations on Scolt Head Island. From summer 2011, additional locations were added to the monitoring programme. However some of the additional cross-shore transects cross washover fans where it is difficult to determine the line of the barrier margin. As a result, they were excluded from the analysis, leaving a total of eight cross-shore locations. By plotting changes in the barrier margin identified on the crossshore profiles for the most recent part of the record (since summer 2006chosen because that date marks the start of the availability of detailed storm data as discussed earlier), periods of significant landward migration, as well as stasis, in shoreline position could be identified. The ArcMap extension Digital Shoreline Analysis System [DSAS version 4.0 (Thieler et al., 2009)] methodology was then used to quantify in greater spatial detail the changing shoreline position for those years between 2006 and 2014 when significant (> 1 m) shoreline translation had been identified in the cross-shore profiles. Barrier margins were identified on georeferenced annual 'summer' vertical aerial photography and digitized within ArcMap (version 10.1). Georeferencing and digitizing errors are discussed in detail in Brooks and Spencer (2010; they suggest typical accuracy lies within 5% of total shoreline retreat). Shore-normal transects were cast at 10 m alongshore spacing and, using DSAS, the End Point Rate (i.e. rate of shoreline position change, in m a À1 ) was found for the intersection of each shoreline with each transect. A total of 455 alongshore transects was cast along the barrier, covering almost the 5 km length of shoreline. It was also possible to cross compare shoreline change evident in the cross shore profile with the end point rate (EPR) for the equivalent alongshore transect found using the DSAS methodology.
For the fourth storm (5 December 2013), cross-shore profile (3 September 2013-13 March 2014), summer 2013 and summer 2014 vertical aerial photography datasets DSAS analyses were supplemented with additional field data collected using a Real Time Kinematic (RTK) system (Leica Viva GS08 Global Navigation Satellite System; data screening to ensure three dimensional coordinate quality < 50 mm and typically < 20 mm) between 6 December 2013 and 3 March 2014.
Meso-term shoreline change (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) Late summer vertical aerial photography from 1992, 2000 and 2008, obtained from the UK Environment Agency, was used to assess shoreline change for the periods 1992-2000, 2000-2008 and 1992-2008 (1992 also represents the first year of such data collection by the Agency). These periods were chosen as analysis elsewhere on the UK East Coast, including the Weybourne cliffs at the eastern end of the study area, has shown that these first two periods reflect clearly delineated periods of enhanced and reduced storminess, respectively (Brooks and Spencer, 2013). Shoreline retreat was quantified using an identical methodology to that used for the short-term analysis described earlier.
Long-term shoreline change (1891-2013) To assess long-term shoreline change we digitized shorelines using (i) the 1891 historic Ordnance Survey map at a scale of 1:10 560 and (ii) the 2013 vertical aerial photograph obtained from the UK Environment Agency. Shorelines were approximated using the mapped mean high water mark of ordinary tides on the 1891 map. On the 2013 aerial photography, the main barrier edge could be defined clearly where the dunes were eroding, but where washover fans were present the edge was obscured so these shoreline sections (totalling a distance of 450 m) were excluded from further analysis. At the actively extending western end of the island, the change from vegetated marsh to mudflat was chosen to represent shoreline position.
Again a total of 455 transects was included in the analysis, covering a total alongshore distance of almost 5 km.

Results
Hydrodynamic conditions associated with selected storms, [2006][2007][2008][2009][2010][2011][2012][2013] Still water levels at Immingham and Cromer, the surge residual at Cromer and wave conditions associated with the four storm surges studied in detail are shown in Table I and Figure 2.
Table II sets these storms in the context of the historical record of storm impacts on the North Norfolk coast, 1883-2014. The four storms can be evaluated against the estimated return periods, based on still water levels only, from the three tide gauges at Immingham, Cromer and Lowestoft. Three of these storms were relatively modest according to this criterion, whilst it is clear that the 5-6 December 2013 event was a superstorm [although Haigh et al. (2015) believe that the 1953 storm was of comparable magnitude].
The highest still water level reached in the November 2006 event at Cromer was 2.74 m ODN, with a maximum surge residual of 1.65 m. Maximum recorded significant wave heights (H s ) at Scolt Head Island were 2.2 m, coinciding with high tides and surge residuals greater than 1.0 m. Waves remained at these levels through the long duration of the surge (18 hours 45 minutes) and there was then a steady fall in wave heights to around 0.5 m by 1800 on 3 November 2006. For the March 2007 storm, a different pattern is apparent. Surge residuals reached 1.38 m at Cromer but were over 1 m for just 2.5 hours. During this shortlived period, peak significant wave heights were just 1.7 m at Scolt. The maximum surge residual occurred at 14:45 on 18 March while the maximum still water level was reached at 18:00 on the same day when the surge residual had fallen to 0.94 m. Thus although still water levels were high (2.83 m), wave action was limited during the surge itself. The phase of high spring tides continued for the next two days, coinciding with a series of low pressure troughs bringing westerly to northerly winds sustained at Beaufort Force 5-8 (8-21 m s À1 ). Windspeeds did not drop until late on 20 March 2007. Onshore waves reached significant wave heights of 2.8 m, and a peak wave height (H max ) of 5.1 m, at Scolt and were maintained at these levels over two successive high tides on 20 March. The maximum still water level recorded during this period at Cromer was 3.35 m at 18:00 on 18 March, but high water levels occurred over successive high tides when the waves were at their highest. In November 2007 surge residuals > 1 m occurred for eight hours and 45 minutes, developing on a falling tide and coincident with low tide. Hence the maximum still water level attained during this storm at Cromer was 2.70 m. Peak significant wave heights during the surge were 2.7 m, with these waves also coinciding with low water levels. Following the passage of the surge, recorded wave heights also reached 2.7 m at time of the next high tide. Wave heights then fell away rapidly, being reduced to 1.0 m by 10 November 2007. For the December 2013 storm there is no tide data for Cromer for the peak of the storm. However, it is possible to assess the magnitude of this surge component in comparison to the other three events from still water levels recorded at Immingham. Here the storm surge produced the highest water level on record (1953-2012Spencer et al., 2014). The surge residual, at 1.97 m, was also the highest recorded for the four storms analysed here and, although it preceded the still water level peak by almost two hours it was maintained at > 1 m to encompass the water level peak. Although the Cromer tide gauge malfunctioned, the surge residual went above 1 m at 15:15 on 5 December and we assume remained above 1 m until 05:15 on 6 December by which time the tide gauge was recording again. Thus the surge residual was > 1 m for 14 hours at Cromer. The peak wave heights recorded at Blakeney Overfalls were 3.8 m, which most probably equates to wave heights of 2.7 m at Scolt (Figure 3). However, wave heights > 3 m were only maintained here for a four hour period ( [2008][2009][2010][2011][2012][2013] it was impossible to conduct a DSAS analysis because the shorelines were so close together once they had been digitized from the aerial photographs that they could not be distinguished from one another without the aerial imagery pixelating, thus blurring the precise location of the shoreline. Finally, we further refine the storm response during the latest period of activity using the RTK field data collected between December 2013 and February 2014. The shoreline position on the 2013 aerial photograph (15 July 2013) compared with the RTK field data produced a mean inland retreat (Net Shoreline Movement) of 8.14 ± 0.39 m along the barrier, while the DSAS analysis based upon the RTK data and the 2014 aerial photograph (31 October 2014) produced a mean seaward movement of the barrier of 0.11 ± 0.01 m, suggesting a slight shoreline recovery in this later period, most probably from the re-establishment of pioneer dunes in front of the eroded dune faces. The average EPR between shorelines in 2013 and 2014 was 4.56 ± 0.23 m a À1 .
Long-term shoreline change (1891-2013) From Figure 6B we see that the long-term (1891-2013) average rate of landward retreat along the whole barrier was 1.15 m a À1 . It is clear too that the western end of the island, from 1.08 km along the barrier from the westward end has been extending alongshore at an average rate of 2.53 m a À1 , with high maximum rates of almost 4 m a À1 at the most western extremity, seen in the first 12 cross-shore transects (120 m). Thus the crossover from predominantly alongshore extension to predominantly landward retreat lies at 1.08 km east of the current westernmost end of the island. Thus, in general terms, the island has elongated and moved onshore since 1891. In 1891 the island was 4.35 km long while in 2013 this had extended   to 6.00 km, with the island mid-point shifting 0.60 km to the west since 1891.

Discussion
Shoreline change in a tidally-dominated barrier has occurred in clearly defined phases of activity and quiescence. The position of the dune crest has progressively been reset landwards since 2006 in three phases, with each shoreline translation averaging between 5 and 8 m. Such rates of migration under storm impacts on this coastline are by no means exceptional. In the major storm surge of 31 January to 1 February 1953 (Table II), Grove (1953) estimated 9-18 m of barrier retreat. For the 1978 surge event (Table II), Steers et al. (1979) reported even higher rates of landward migration, estimating an average of 20 m of retreat between Smuggler's Gap and Norton Hills (although this may have been be an over-estimation; see Spencer et al., 2014). Of the four events identified, three appear to have crossed water level and wave setup thresholds for significant barrier movement. However, one storm (31 October-3 November 2006) generated no shoreline change. Thus the November 2006 event generated a large surge residual on a falling tide, while the March 2007 event was characterized by a small surge residual. Both events reached similar still water levels at Cromer. However, the March storm was followed by two days of high waves at Scolt Head Island, with the peak of wave activity being experienced on two successive high tides. The six  Furthermore, the March storm was accompanied by a major morphological change on the island in the form of duneline breaching and the formation of an extensive washover fan towards the western end of the island (Michael Rooney, personal communication, 2007;). This resembles the creation of a 'breakthrough' in the centre of the island in 1953, when weather records from the Dowsing Light Vessel (53°35′N, 0°55′E) show that windspeeds in excess of 20 m s À1 (Beaufort Force 9, strong gale) were maintained for a 24 hour period over 31 January and 1 February 1953, generating significant wave heights of 7.8 m in 30 m water depths off the North Norfolk coast (Wolf and Flather, 2005). Thus echoing Sallenger (2000), a spectrum of no changedune face erosionmajor morphological change is possible, determined by both still water level and wave height at the shore, the magnitude and duration of these two components and the timing of their interaction (and see also Carter and Stone, 1989;Brooks et al., 2012).
Over the meso-term, Figure 6A suggests strongly differentiated differences in retreat rates between large recession in the 1990s and much slower landward migration in the 2000s. This is the local expression of a regional landform response signal on the UK east coast, detected not only in the retreat rates of the soft rock Weybourne-Sheringham cliffs at the eastern margin of the study frontage, but also along clifflines of similar morphology on east-facing southern North Sea coasts at Covehithe, Suffolk and Walton-on-the-Naze, Essex (Brooks and Spencer, 2013). At Scolt Head Island, these differences cannot be explained by differences in storm surge event frequency; Table II shows that there were three such events in the period of high retreat compared to four events in the low retreat period. It is interesting to note, however, that the period of accelerated shorenormal change was accompanied by reduced alongshore change and vice versa; the western terminus of the island extended westwards by 35 m between 1992 and 2000 compared with 80 m between 2000 and 2008. It is unfortunate that the Blakeney Overfalls Wave rider Buoy only became operational in 2006 and that we therefore lack the detailed record of inter-decadal incident wave climate relatively close inshore that might inform this change in barrier dynamics.
The long-term, barrier-wide mean rate of shoreline retreat at Scolt Head Island over the period 1891-2013 was 1.15 m a À1 , equating to a landward translation of the barrier of 140 m. Our findings here demonstrate that between 5 and 8 m of barrier migration has typically accompanied recent high magnitude storms; this would suggest that this degree of shoreline change could be accomplished by between 17 to 28 storm impacts. Archival evidence for the occurrence of past storms, reported in Table II (for detailed sources see Supporting Information Table  S1), shows 20 reported storm events since 1883. However, the preceding analysis of hydrodynamic forcing and morphological responses cautions against a simple correlation.
The short-term record of shoreline retreat of both high and low dune lines at Scolt Head Island shows a high degree of alongshore variability (Figure 4). This is carried through into the meso-term analysis ( Figure 6A) although the locations of accelerated retreat do not necessarily coincide. However, at the centennial scale this alongshore variability is lost; the eroding sections of the barrier show a remarkably consistent rate of shoreline retreat, with a clear hinge point to the area of shoreline extension ( Figure 6A). How can this time-dependent difference be explained? There is no evidence that this pattern results from the recovery of erosional hotspots by onshore sediment transport. At this timescale, Steers (1960) argued cogently that the high angle at which the laterals intersect the main barrier ( Figure 1C) can be explained by the long-term landward translation of the barrier and, in the short-term, it is clear that there is little if any barrier recovery, by seaward extension, following storm impacts. Indeed, this pattern of change is strongly reminiscent of the same time-variant signal seen in systems that are only characterized by retreat, such as soft rock cliffs (e.g. Brooks and Spencer, 2010). An alternative explanation, following arguments that have been made for the evolution of the barrier shoreline of North Carolina  is that localized storm impact signals become blurred at longer timescales by alongshore sediment transport. The exact details of such a mechanism, whilst plausible, are not well defined here. The traditional explanation for the progressive, if episodic, westward extension of Scolt Head Island is longshore transport from the east from a 'drift parting' near Sheringham, the estimated transport rate being dependent upon the particle size distribution, varying between 190 and 60 m 3 a À1 (HR Wallingford, 2002). Seabed sediments and facies mapping suggests that the transport direction is the reverse offshore, at least below the 7 m contour. This suggests that sand and shingle is being transported to the west on the beach face, but that sand is transported to the east if it is carried offshore from the steep beach face onto the extensive subtidal Burnham Flats, perhaps under storm surge conditions (East Anglian Coastal Group, 2010). Furthermore, an alternative model argues that sediment transport on the beach face is from west to east, with by-passing of the Brancaster Staithe Harbour Channel by episodic sand-waves moving through the ebb tidal delta. These easterly-moving sand waves then weld onto the western extremities of the barrier, explaining its westward growth. Using barrier lengthening and tidal inlet narrowing in the Wadden Sea as an analogy, where long-term decreases in backbarrier areas have followed several phases of historical land claim (Fitzgerald et al., 1984), post-nineteenth century extension of the Scolt barrier may also have followed the enclosure of > 200 ha of saltmarsh behind an earthen embankment east of Burnham Deepdale in 1882 ( Figure 1B). This would have reduced the tidal flow through the Harbour Channel to allow more rapid easterly passage of sandwaves across it (Royal Haskoning and Pethick, 2003).
Finally, Horsburgh and Lowe (2013) have argued that whilst there is no significant evidence for future changes in stormrelated extreme sea levels for the UK, it is not unreasonable to assume that future changes in extreme sea level will be governed by mean sea-level rise. Woodworth et al. (2009) demonstrate that absolute mean sea level (AMSL) rose by 1.4 ± 0.2 mm a À1 around the UK over the twentieth century, suggesting that the baseline water level rose by 16 ± 0.3 cm between 1891 and 2010. However, for more recent periods, Wahl et al. (2013) found a relative sea level rise of 2.7 ± 0.4 mm a À1 (1900-2011), 3.6 ± 0.5 mm a À1 (1980-2011) and 4.4 ±1.1 mm a À1 (1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011) in the Lowestoft tide gauge. If there is a broad correlation between rates of barrier retreat and sea level rise then this would suggest that a considerable acceleration in the rate of barrier migration might be expected over the remainder of the twenty-first century.

Conclusions
The analysis developed in this paper provides valuable insights into the importance of storm impacts on barrier and shoreline dynamics and shows how the richness and detail in contemporary data can be used to examine the thresholds for barrier retreat. We have mined and synchronized a range of data sets on shoreline movement and hydrodynamics to reconstruct storm impacts on a retreating barrier. We have demonstrated 862 S. M. BROOKS ET AL.
how shoreline change in the recent past (last decade) is most likely to have been driven by individual storms that cross water level and especially wave energy thresholds. The better identification of these thresholds offers the possibility of better explanations of the characteristic pattern of periods of enhanced retreat being interspersed with periods of low, or no retreat. However, explaining meso-term and long-term patterns in barrier dynamics remains challenging, particularly in the absence of corresponding forcing data.