Middle- to late-Holocene storminess in Brittany (NW France): Part II – The chronology of events and climate forcing

Present-day context: storm genesis versus NAO mode

European storms commonly occur during the ‘winter period’ from late September to April. Storm impact is primarily linked with wind strength as well as swash efficiency. Westerly wind strength is, in Europe, a crucial parameter acting on monthly wave heights. Furthermore, wave heights are more dependent upon storm tracks, storm age (Bertin et al., 2012) and rising wind speed (i.e. storm intensity; Wolf and Woolf, 2006). Repeated high tides during stormy periods (corresponding to clustered depressions) will enhance the morphogenetic impact associated with these storm events. Specific meteorological configurations, and especially the jetstream pattern, can also favour strong storm impact on Western European coasts (Alexander et al., 2005). Indeed, sting jets (such as the 1987 Great Storm; Browning, 2004) accelerate the deepening and rotation of the depression cell (Alexander et al., 2005; Browning, 2004) leading to freshly formed aggressive waves (Bertin et al., 2012), also favoured by a strong gradient in SST (Sanders and Gyakum, 1980).

Many studies suggest, for the pre-industrial period, a fair correlation between Northern Hemisphere climate changes and solar activity (e.g. Bond et al., 2001; Chapman and Shackleton, 2000; Mayewski et al., 2004). Furthermore, solar activity is potentially one of the drivers of the NAO variability (Abreu et al., 2012; Alvarez-Ramirez et al., 2011; Beer and Van Geel, 2008; Ineson et al., 2011). Specific NAO modes, considered to be winter phenomena, are then associated with periodic reorganization of the atmospheric and oceanic circulations in the Northern Hemisphere. During a positive decadal NAO mode, winter conditions are usually wet and rainy over NW Europe, with series of storms characterized by prevailing W–E tracks (Dawson et al., 2003; Pirazzoli et al., 2004; Rodwell et al., 1999). During a negative decadal NAO mode, cold and dry winter conditions generally prevail above NW Europe, due to the disappearance of the Westerlies (Dickson et al., 1996; Orford et al., 2000). In this schematic Northern Europe view, NAO pattern and storminess events appear to be stronger during a negative NAO mode due to the abrupt atmospheric front occurring between tropical and arctic air masses (Ayrault, 1998; Betts et al., 2004; Trouet et al., 2012). Indeed, a positive NAO mode will induce stable westerly situation and gently undulating jet streams, while a negative NAO mode will induce marked instability raised by an important swing of the jetstreams. NAO and storminess are also impacted by sea-ice cover changes in the Northern Hemisphere (Figure 2; Courtenay et al., 2009; Jaiser et al., 2012).

OPEN IN VIEWER

Figure 2.

Synoptic scheme of the classical storminess patterns linked with NAO mode.

For the following discussion, it is important to note that we will sometimes refer to ‘hurricanes’ or to ‘tropical storms’. Depressions born in a few hours from instabilities above the Bermuda Islands (Atlantic Ocean; hottest Gulf Stream waters) represent most of the storms discussed in this paper (Figure 2). However, the specific term ‘hurricane’ will be only used for hot storms coming from Africa through the Gulf of Mexico, mostly in early autumn, while the term ‘tropical storms’ will be used for storms directly rising from Africa to Europe due to an anomalous configuration of the North Atlantic jet (NAJ) streams (negative NAO; Figure 2).

NAO: From a daily to a millennial signature?

At the Holocene timescale in NW Europe, drought and dune building would be associated with repeated centennial NAO negative modes and moderate storms (Lamb, 1995). Recurrence of negative NAO configurations may have persisted long enough (few winter months) to modify temporarily the wind pattern, such as those observed during the late 20th century (Visbeck et al., 2001) with dominant westerly winds in Brittany (Hénaff, 2008; Pirazzoli et al., 2004). Over longer timescales, colder (2310–2190 cal. BP, ad 450–650, ad 1380–1420) and warmer periods (230 bc–ad 140 and ad 640–760) can be achieved in GISP2 δ¹⁸O data (central Greenland; Alley, 2004) and north of Iceland in connection with the Irminger Current (SST reconstructions; Patterson et al., 2010). The combination of these data suggests, over NW Europe, long periods (centennial scale) with prevailing negative (colder periods) as well as positive (warmer periods) winter NAO modes. Behind instrumental data, long-term NAO records have been reconstructed with various biological proxies (dendrochronology, palynology, oxygenation of water; e.g. Cook et al., 2002; Luterbacher et al., 2002; Olsen et al., 2012; Trouet et al., 2009); they reflect decadal to centennial NAO modes.

In this paper, we will also verify the NAO mode control on storminess, as commonly considered and published for the last millennia (Cook et al., 2002; Dawson et al., 2007; Trouet et al., 2009). References to daily, decadal or centennial NAO events will be used according to the duration of the period considered in this study.

Another mode of natural atmospheric variability: The AMO

The AMO shows longer periodicities (around 60–80 years) than the NAO (around 10 years). The AMO signs the thermal storage in the basin-wide Atlantic Ocean (Otterå et al., 2010; Schlesinger and Ramankutty, 1994) transmitted to the Northern Atlantic by the Gulf Stream (meridional heat flux and thermohaline circulation; Hoskins and Hodges (2002)). The AMO fairly records SST changes in the North Atlantic driven by insolation processes such as atmospheric circulation patterns, the Atlantic Meridional Overturning Circulation (AMOC), and sea-ice extent (Knudsen et al., 2011). The AMO, in turn, also strongly affects temperature variations over Europe (Goldenberg et al., 2001; Knight et al., 2006; Sutton and Hodson, 2007; Zhang et al., 2006).

Therefore, the AMO control will also be considered through its impact on SST, a concept very similar to the actual understanding of tropical hurricane triggering from a warm ocean and migrating above high SST (Goldenberg et al., 2001). Following what we presented above concerning NAO and AMO configurations, maximal storminess would be expected in a context of very high SSTs (positive AMO mode) combined with unsteady NAO mode (Figure 2). This hypothesis has been already proposed by Barlow et al. (1993) and also fits the Atmospheric General Circulation Model by Graff and LaCasce (2012). It supports the idea that storm tracks respond to changes in mean SST and that increasing the SST gradient at mid-latitudes causes an intensification and a poleward shift of the storm tracks, in response to a poleward shift of the jet stream (http://squall.sfsu.edu/crws/jetstream.html) and a high pressure blocking (Açores High). Moreover, storm strength increases monotonically with the magnitude of SST perturbations driven by the latent heat release (Booth et al., 2012).

For historical and pre-historical times, AMO intensity can be indirectly extracted from sea-ice cover extent and SST reconstructed north of Iceland (Berner et al., 2008). Indeed, these SSTs are linked with the intensity of the Irminger Current, a branch derived from the North Atlantic Drift, itself directly connected to the Gulf Stream (Jennings et al., 2000). For this reason, this approach cannot be applied before the total disappearance of the Laurentide Ice Sheet, due to perturbation of the SST by melt waters.

Event’s detection

Regional storm events during the 19th–20th centuries are extracted from various sources: public and administrative archives, press and vernacular archives analyzed by local historian, often available on the Internet. Historical archive data are mainly derived from the Anglo-Saxon Chronicles and from Caesar’s De Bello Gallico. A compilation of historical events from ad 300 is mainly derived from the studies of Marusek (2010) and Lamb and Frydendahl (1995; Table 1).

Furthermore, from a stratigraphical view, 47 dates from Audierne/Ile-Tudy sectors (cf. Figure 1) and 35 other pertinent dates from Brittany (Table 2) are considered. Calibration of all dates was performed at one or two sigma using the CALIB 6.0 Radiocarbon Calibration Program (Stuiver and Reimer, 1993; Stuiver et al., 1998), associated with the non-marine (IntCal09) or marine (Marine09) radiocarbon calibration (Reimer et al., 2004, 2009). For accommodating local effects, a regional marine reservoir correction Δr of about 40 ± 42 years (Sein Island; Hughen et al., 2004; Stuiver et al., 2005) has been applied.

Palaeostorm events identified and dated will be discussed in parallel with centennial NAO reconstruction, total solar irradiance (TSI) reconstruction and volcanic activity. Furthermore, SST reconstructions (Northern Iceland, REF) as well as sea-ice cover extent (chloride concentration in GISP2 ice core; Alley, 2004) will permit us to discuss possible links with different stages of AMO configuration prevailing during major palaeostorms. Given the large set of different information, we did not apply statistical analysis to discuss storm’s dating together with reconstructed NAO and AMO.

Event’s timing

Regional storminess events extracted from archive data are listed in Table 1. Combining these with the events extracted from the stratigraphical record (‘Middle- to Late-Holocene Storminess in Brittany (NW France): Part I’), it becomes possible to define a list of severe palaeostorms that occurred around Brittany (Table 2). When considering storm timing, the median (best fit) of whole congruent Brittany’s dating is used with an average error of ±30–100 years, taking into account the age of the sample in pre- or post-storm surge position (pre- or post-STS; Figure 3). This analysis has been extended to whole Brittany with the new data of Goslin (2014; Goslin et al., 2013) and previous published studies (cf. Table 2 to see all ¹⁴C dating taken into account in this study).

OPEN IN VIEWER

Figure 3.

Analysis of the median for the 9th–12th century storms. Stars represent known storms (archives).

For the historical period, the ages discussed are expressed in ad (age Anno Domini). The BA and Brittany records yield median storm ages of about ad 125 ± 50, ad 310 ± 50, ad 400 ± 50, ad 640 ± 50, ad 800 ± 50, ad 900 ± 50, ad 1010 ± 50, ad 1050 ± 50, ad 1100 ± 50, ad 1192 ± 50, ad 1335 ± 50, ad 1597 ± 50, ad 1703 ± 50, ad 1750 ± 50 and ad 1806 ± 50 (¹⁴C plateau, forced by local archives). Some dates overlap with each other when considering the uncertainty margin. However, storm ages listed above correspond to distinct events identified in the field, and their median age will be considered in this paper. Historical sources such as the Anglo-Saxon Chronicles mostly from the 8th to the 9th century, as well as compilations from ad 0 to present (Marusek, 2010) and from the 16th century (Lamb and Frydendahl, 2005) give access to information concerning some palaeostorm timing and tracks.

For the pre-historical period, that is the Holocene long-term record from 7000 cal. BP to present, archives are not available. This stratigraphic work gives access to storm ages of about: 6600 ± 50 cal. BP, 6050 ± 50 cal. BP, 5500 ± 50 cal. BP, 4650 ± 50 cal. BP, 3400 ± 100 cal. BP, 2750 ± 50 cal. BP, 2535 ± 50 cal. BP, 2360 ± 50 cal. BP and 2060 ± 30 cal. BP, and periods close but somewhat different from those observed for the Mediterranean Sea: 6300–6100 cal. BP, 5650–5400 cal. BP, 4400–4050 cal. BP, 3650–3200 cal. BP, 2800–2600 cal. BP and 1950–1400 cal. BP (Sabatier et al., 2012).

Regional storminess record: 1850 to present

A first comparison, over the two last centuries, between storm recurrence in Brittany extracted from archives (Figure 4 and Table 1), NAO and AMO natural variability extracted from the Climatic Research Unit (CRU, University of East Anglia, UK), suggests a link between high storm frequency, positive AMO values and the onset of a NAO shift from a positive to a negative mode or inversely. Indeed, three periods of lower storminess frequency are recorded around 1850–1864, 1884–1894 and 1925–1943, and correspond to thermal minimum of AMO cycles (Figure 4), even some periods have a low validity storm record (wars period). Storm frequency is high as well with negative or positive NAO, but higher when the maximum of the AMO cycle is reached. Regional storminess seems therefore connected to a warm Northern Atlantic Ocean, associated with a thermally unstable North Atlantic air mass. This situation results from the swing of the NAJ stream that causes an abrupt contact between polar and tropical air masses (Figure 2)

OPEN IN VIEWER

Figure 4.

Comparison of the archive storminess record in Brittany (cf. Table 1), the winter NAO index (CRU, E. Anglia, UK) and the AMO (CRU, E. Anglia, UK).

As a recent example, the decadal NAO was mostly negative during the cool interval 1960–1976, further shifting to two decades 1976–1996 (solar cycles 21 and 22) dominated by positive anomalies (Rind et al., 2005; Thompson and Wallace, 1998; Visbeck et al., 2001), before shifting back to negative anomalies from 1996 to now (Figure 3). Here, the minimum in regional storminess is recorded from 1960 to 1964, during an interval indeed characterized by low AMO anomalies with mostly positive NAO anomalies. In opposition, taken also as example, Xynthia (in 2010) tropical storm track had a southern origin and was driven by enhanced swing of the NAJ streams. It occurred as a brief positive event within a period characterized by a dominant negative winter NAO mode. Joachim storm (15 December 2011) and the Great Storm (17 October 1987; Browning, 2004) occurred under similar conditions and were preceded by tropical air. Pinto et al. (2009) stressed that, for the recent period, storms mostly develop during a daily positive NAO anomaly within a monthly to yearly negative NAO mode. The process is probably similar at longer (decadal or centennial) timescales: brief heat input or magnetic perturbation that may be induced by solar eruption would temporarily destabilize the climate system by modifying the thermal budget (Damé, 2013; Mendoza, 2005).

For the last 30 years, the AMO index raised more and more positive to reach its maximal value in 2012, about 50 years after the maximum of the major solar cycle 19 (1958–1964). The NAO index turned to dominant negative values, mostly from 1995 (NOAA web site). The Arctic atmosphere is thus, today, in thermal contrast with the warm Northern Atlantic Ocean (Knudsen et al., 2011), resulting in a regional increase in storm frequency from 2004, with the onset of regional destructive storms from 2007. It seems thus possible to consider an analogous mechanism for storm triggering as well as for hurricanes. Here, it would imply rapid atmospheric instabilities (more frequent NAO anomalies), raised by the NAJ streams, over a warm ocean (accentuated convection), within a shift from or to a decadal negative NAO mode. Therefore, the driver is the thermal status of the ocean (Figure 4) and an unstable jet stream pattern driven by decadal negative NAO mode.

Historical storms: 100 bc to present

Close to 110 bc, an event is recorded both in SW Brittany and in Mont-Saint-Michel Bay (Table 2; Carter et al., 1993; Meurisse-Fort, 2009). Climate is mild and stable from 100 bc to ad 75 and sea-ice retreated; a stable, relatively cool period followed from c. 100 to c. ad 270 (McCormick et al., 2012). After 100 bc, storms are only discretely recorded. In his De Bello Gallico, Julius Cesar mentions two fierce storms damaging his fleet in Brittany in 55 bc and 54 bc (cf. Table 1), which was in a period of cooling trend but rising centennial NAO mode (Figure 4). The AMO mode is probably positive, and the NAO keeps rising (Figure 5). Some storms were then recorded in historical archives (cf. Table 1) around ad 320, ad 540–600, ad 640–670 and ad 710 (S Brittany), on the Channel in ad 429 (http://chrisagde.free.fr/merovingiens/germainauxerre.htm) and around ad 535–545, ad 567 and ad 710 (Loire’s estuary; Visset, 1982; original calibration).

OPEN IN VIEWER

Figure 5.

Comparison between storms (red: important, blue: major) and dune formation events in Brittany, total solar irradiance (TSI; Vieira et al., 2011), NAO (Olsen et al., 2012), SST off Northern Iceland (Berner et al., 2008) and sea-ice extent (O’Brien et al., 1996).

The regionally observed ad 125 ± 50, ad 310 ± 50, ad 400 ± 5 0 and ad 650 ± 50 storm events seem to occur during a cold spell surrounded by two positive AMO periods. This storminess interval is synchronous with a cooling in Iceland (Patterson et al., 2010) and a solar low (Stuiver et al., 1998; Vieira et al., 2011). The ad 535–545 storm results perhaps from a large volcanic event centred on ad 525 (Gao et al., 2008).

On the northern coast of Brittany, the recorded ad 800 ± 50 event or the ad 700 or 709 events (Table 2) probably include major storms. They correspond to extremely cold winters (McCormick et al., 2012). A hurricane is mentioned in ad 9 January 800 (Marusek, 2010), other storms affected South Britain in ad 17 March 803 and in ad 877 (Lamb and Frydendahl, 2005) and the Picardy in ad 838 (Verhulst, 1989). It seems thus probable that a series of storms crossed Brittany around ad 800, synchronously with positive AMO and lowering NAO centennial modes associated with extremely cold winters (Figure 5).

A major storm struck around ad 900 ± 50 at the BA, along the southern coast of Brittany and in the Bay of Brest (Loch de Landévénec) with prominent north-westerly winds. Major storms struck the western Cotentin in late ad 900 (Anglo-Saxon chronicles), as also around ad 900 in Picardy (Meurisse-Fort, 2009) followed by another major one in summer ad 921. These major storm paths may have crossed Western Europe from SW to the north similarly to the ad 1987 storm path, in a context of positive AMO mode combined with decaying centennial NAO mode. It seems thus that the post-Roman Period is rich in storm events on a still warm ocean, related to a climate cooling driven by solar activity.

A number of events with a very strong morphogenetic impact occurred in the 11th century, around ad 1014 ± 50, ad 1050 ± 50 and ad 1100 ± 50 in south of BA (Figure 1) and Brittany still with strong north-westerly winds (Figure 5). The recorded ad 1050 ± 50 storm episode may be linked to the important tidal surge and long shore drift with very high waves. It partly destroyed dunes of Treffiagat and Guidel (cf. Figure 1). This major event seems to have affected whole of Brittany (Table 2), even the Mont-Saint-Michel Bay (Billard et al., 2003) and may be considered as the millennial storm. In the literature, three huge storms are mentioned for South England: the Saint Michael Great Flood in ad 8 November 1014 (Anglo Saxon Chronicles; Lamb and Frydendahl, 2005); one each in in 1016 and 1030, which are less important; and the ‘Saint Thomas Great Wind’ in ad 10 January 1052 (Anglo-Saxon Chronicles). A late recorded event (ad 1100 ± 50) seems to fit with the clustered storm events of ad 1092–1095–1099 associated with winds from SW in southern Brittany (Marusek, 2010) and between ad 1110 and ad 1125 for South England and the English Channel (archives; Figure 3).

From the details above, it appears evident that the series of major storms, mostly winter storms (Table 1) from ad 900, 1014, 1050 and 1090–1100, had a major morphogenic impact on Brittany’s coast, up to 300 m inland when considering the present coastline. This period mostly corresponds to the Viking Solar Low (around ad 900) and the Oort Solar Minimum (around ad 1010/1040–1080) (Steinhilber et al., 2009). Northern Hemisphere temperatures reconstructed for these periods (Loehle and McCulloch, 2008; Moberg et al., 2005) suggest an abrupt cooling with severe winters. The NAO curve also shows a drop in NAO mode around ad 1010 (Trouet et al., 2009), and SSTs off Northern Iceland were still warm, suggesting a positive AMO mode (cf. Figure 6), although Dawson et al. (2007) reported a shift to dominant positive or ‘dormant’ NAO after ad 1020.

OPEN IN VIEWER

Figure 6.

Comparison between the geological storminess record (dark gray bars underline major storm episodes and light gray bars highlight error bars), written storminess archives, and three sub-recent storms (black vertical lines), in parallel with: NAO (Trouet et al., 2009) and AMO modelling (Otterå et al., 2010), North Iceland SST (Jiang et al., 2005), sea-ice extent off Northern Iceland (Massé et al., 2008), solar activity (Crowley and Lowery, 2000) and volcanic forcing (Gao et al., 2008).

The ‘Medieval Warm Period’ (MWP) differs in timing according to the region considered but classically ranges from ad 950 to ad 1350 in Europe. Winters remained cold until ad 1130 (Marusek, 2010). This period appears to be characterized by a persistent positive AMOC (thermohaline circulation) mode (Trouet et al., 2012) combined with a persistent ‘dormant’ or positive NAO mode from ad 1100 to ad 1450 (Dawson et al., 2007; Trouet et al., 2009), with no negative NAO events observed through biological studies. During the 13th and 14th centuries, we only record two storm events around ad 1192 ± 50 and ad 1335 ± 50. SSTs off Northern Iceland suggest dominant warm but unstable sea-surface conditions (Dawson et al., 2007; Jiang et al., 2005), with a limited extent of sea ice (Massé et al., 2008), suggesting long-lasting prevalent positive AMO conditions over at least 180 years (Figure 6). Flooding induced by storminess would, however, have occurred at Audierne at the end of the MWP (Table 2). The ad 1192 ± 50 storm event in Brittany could be attributed to an ad 1172 storm, very destructive in NW Brittany, and it could be also synchronous with the ‘old’ dune’s aggradation, the ad 1242 storm recorded at the Pointe St Mathieu or perhaps the ad 1269 ‘hurricane’ observed in the same region (archives). North Iceland SST shows a warm ocean at that time, therefore, a still positive AMO (Figure 6). However, this storm period could be linked with a low in positive NAO (Trouet et al., 2009), and occurred more or less synchronic with an important aerosols input in ad 1258 (Dawson et al., 2007; Gao et al., 2008), leading to a temporary cooling. The event recorded around ad 1335 ± 50 at Treffiagat (Figure 1) does not correspond to the hurricane occurring at low tide in summer ad 1284 (Bessemoulin, 2002). In ad 1320 and ad 1377, storms destroyed Lyme Regis (Wessex; West, 2010 (web reference)), under wet and mild winter (positive NAO). However, storms were recorded in December 1333 and January 1334 in relation with an exceptionally cold winter in Europe (negative NAO). It seems to fit the Wolf Solar Minimum (ad 1280–1350) and a negative phase of the NAO (Cook et al., 2002; partly suggested by Trouet et al. (2009)). Thermal reconstructions show an abrupt cooling at that time as well as rising sea-ice cover extent around Iceland, especially in ad 1348. This storminess also is perhaps consecutive to a huge aerosol input induced by cumulative strong volcanic activity (ad 1258, 1268, 1275 and 1285) superimposed on the Northern Hemisphere century long-term cooling (Gao et al., 2008; Otterå et al., 2010). However, despite these forcings, the centennial NAO remained positive (Trouet et al., 2009), like the AMO mode probably.

From ad 1420 to 1530, despite very cold winters, no storm is recorded, either in the literature for Brittany, or in geological studies. This anomaly is also detected in NAO reconstructions and corresponds to the Spörer Solar Minimum (Trouet et al., 2012). It is attributed to an AMOC (thermohaline circulation) reorganization after the MWP, more or less synchronic with a drastic change in atmospheric circulation (Dawson et al., 2007). SSTs are rather high and unstable, similarly to sea-ice extent (Figure 6), and much less extended than during the Maunder, Dalton and Damon Solar Minima (Massé et al., 2008). From a more global point of view, the AMO mode probably remained high, when considering the chloride curve (O’Brien et al., 1996; Figure 5), and a globally positive centennial NAO mode may have prevailed (Olsen et al., 2012; Trouet et al., 2009). AMO modelling shows a very negative event during ad 1400–1480 (Otterå et al., 2010). This lack of important storms is partly responsible for the development of maritime explorations to America, especially from 1419 to 1532 (Diffie, 1977; Weddle and Press, 1985).

During the 16th century, a main storm is recorded in Western and NE Brittany close to ad 1597 ± 50 (Table 2). It could correspond to the former tropical hurricane that crossed SW Europe during 13–15 August 1588 (Lamb and Frydendahl, 2005). Another important ‘hurricane’ is mentioned in archives for the lower Loire River in 1598. A winter storm is recorded regionally on 20 January 1606 in SW England and N Brittany (Pelé, 1973; West, 2010 (web reference)). Thermal reconstructions showed an abrupt cooling at that time (Loehle and McCulloch, 2008; Moberg et al., 2005), confirmed by very cold winters in weather compilation (Marusek, 2010). This stormy period corresponds to the end of the long Spörer Solar Minimum (1460–1550) and a shift to a prevailing negative decennial winter NAO mode (Cook et al., 2002). The period between ad 1660 and 1880 seems the richest in more intense storms, especially between ad 1660 and ad 1750 (Clarke and Rendell, 2009; Garnier, 2010; Lamb, 1995). In the 18th century, at Audierne, important storms are recorded around ad 1795 ± 50 and ad 1750 ± 50 representing more probably many superimposed events (plateau effect of the ¹⁴C curve). The nearby Sein Island was flooded many times according to the archives. A series of strong storms fits very well in this system. The 8-day long Great Storm from 26 November to 8 December 1703 (‘Daniel Defoe storm’) was the most severe storm ever recorded in the southern part of Great Britain and Sein Island (Lamb and Frydendahl, 2005; http://www.enezsun.com/Histoire/digues.htm). Similarly, the 4-day long Christmas 1717 storm, hurting most of Europe, north of the Bay of Biscay, is also a good candidate, probably additional to the 1703 event, followed by the 19–20 January 1735 and 16 February 1736 region-wide storms (Lamb and Frydendahl, 2005; Pelé, 1973). Optically stimulated luminescence (OSL) dating at Plouescat (Bay of Goulven; Figure 1) yielded ad 1723–1751 and ad 1752–1776 for two successive events on a sandy spit (Van Heteren, 2011). In the Scilly Islands (St. Agnes, Big Pole), a major sand drift has been dated by OSL between ad 1560–1680 and ad 1700–1780 (Banerjee et al., 2001). The thick gravel accumulation close to Beg-an-Dorchenn would therefore mark at that time the repetition of similar events and seems to have been reactivated between ad 1775 and 1828 (Hallégouët and Hénaff, 1993) fairly corresponding to the regional 17–19 November 1795 ‘Admiral Christians’ hurricane (Boult, 2003). These events occurred in a prevailing negative decennial winter NAO context (Cook et al., 2002). Within the Maunder (ad 1645–1715) and Dalton Minima (ad 1790–1830), exceptionally low solar activity has been related to: (1) changes in Atlantic storm tracks (Clarke and Rendell, 2009; Van der Schrier and Barkmeijer, 2005); (2) cold winters and summers in Northern Europe (Loehle, 2007; Moberg et al., 2005; Van der Schrier and Barkmeijer, 2005) and (3) extensive sea-ice cover (Massé et al., 2008; Ogilvie and Jonsson, 2001), leading to a slower AMOC (Trouet et al., 2012) that enhanced thermal contrasts on the ocean. It is related to the arrival of warm air, up to Greenland (Mayewksi et al., 2004), that we may connect with a dominant negative NAO mode (Fischer and Mieding, 2005), after a very high potential AMO mode (Figure 6). AMO modelling exercises (Otterå et al., 2010) show a series of prominent positive AMO events, especially from ad 1700 to ad 1800.

In the 19th century, the last historical storms recorded in the sedimentation at the BA are compiled from the literature, given the limitation of dating by the ¹⁴C plateau effect. The timing is mostly extracted from coastal erosion record. Four winter storms may be extracted from archives (Table 1): 11–12/02/1808 (the ‘century’ storm), 11/01/1866, 16/12/1886 and 4/12/1896. The coastal retreat at the BA was probably also enhanced by the ‘Great Gale’ (22–23/11/1824) that devastated the southern coast of Britain (West, 2010 (web reference)). It mostly reached the coasts of Brittany with south-westerly winds. In 1839, 1865/1896 and 1897, storm series breached the dunes of Kerity (Penmarc’h; Monfort, 1985); the Lescors marsh became filled with drifted sands, Kerity and La Joie marshes were flooded. The climate was initially cold (Dalton Solar Minimum, 1790–1830), and glaciers began to retreat from 1830 with milder conditions than for the main ‘Little Ice Age’ (LIA) phases. Most of large 19th century storms are associated with high SST (Figure 6) and a transitional NAO mode. A high AMO mode is suggested by the annual index (Figure 4) but a low one is inferred from modelling exercises (Otterå et al., 2010).

Pre-historic to historic storms: 6000 cal. BP to present

In this part, the validity of the observations previously made at historical timescales is tested, at the Holocene timescale. However, dating is not always as accurate as previously and data are rare. Centennial NAO reconstructions only exist from 5500 cal. BP to present (Olsen et al., 2012). AMO oscillations are deduced from North Iceland SST as well as sea-ice cover extent for the same periods.

More evidence for an early storm event are recorded around 6600 ± 50 at Pors Carn (Table 2). Then, one isolated south-westerly storm is recorded at 6050 ± 50 cal. BP at Pors Milin (Table 2), through a lagoon submersion. However, reliability is low given the available dating over these periods (storms not represented in Figure 6). This time interval is synchronous with a marked cooling as suggested by the glaciological chloride record of GISP2 (cf. O’Brien et al., 1996).

Storms events in Brittany are recorded around 5500 ± 50 cal. BP (Table 2 and Figure 4), mostly on the western coast of Brittany. The onset of dune building after 5500 cal. BP at Gwendrez and the breaching at Kermor (cf. Figure 1c) close to 5150 cal. BP should probably fit several storm events within this time range. This period seems also responsible for a storm terrace reaching 6–7 m NGF around 5400 ± 50 cal. BP in the Canche estuary (Meurisse-Fort, 2009; cf. Figure 1a), as well as storm impact records obtained in the Seine estuary (Sorrel et al., 2009; cf. Figure 1a) and Mont-Saint-Michel Bay (Tessier et al., 2010; cf. Figure 1a). From the chloride curve of GISP2, sea ice appears limited between 5850 cal. BP and 4950 cal. BP (O’Brien et al., 1996; Figure 5), in parallel with high SST off Northern Iceland (Berner et al., 2008; Figure 5). This period would correspond to high positive AMO, while centennial NAO values seem rather low (Olsen et al., 2012; Figure 5). Another storm is then recognized in North and Western Brittany at 4650 ± 50 cal. BP.

Consecutively, this new record for Brittany shows a major storm episode centred around 3400 ± 100 cal. BP (Table 2 and Figure 5). It would fit with the 3450 ± 50 cal. BP storm, initially published by Guilcher and Hallégouët (1991) and by Carter et al. (1993) and also preserved in the Bay of Vilaine, the Mont-Saint-Michel Bay (Billeaud et al., 2009; Tessier et al., 2010) and the Seine estuary (Sorrel et al., 2009). The impact of increased storminess around 3450 cal. BP has been more widely recognized at the European scale, but few data are commonly preserved onshore, due to the progressive sea-level rise and resulting erosion. This period would correspond to a cooling with an important Arctic glacier extent (Denton and Karlén, 1973; Nesje et al., 2008), following the Minoan Thermal Optimum (Figure 5). It could be interpreted as millennial scale storm(s) occurring during a centennial NAO shift to negative values; SST remain however high and could be related to a former peak of TSI (Steinhilber et al., 2009; Vieira et al., 2011). The AMO is therefore presumed positive from SST and sea-ice extent data, while the centennial NAO seems to shift towards negative values (Figure 5).

Other events are then recorded at 2750 and 2535 ± 50 cal. BP at Audierne and along the Western coast of Brittany. The 2750 ± 50 cal. BP corresponds to a deep solar low (Vieira et al., 2011), known as the Homeric Low (Figure 5), with a large archaeological and palaeoecological European impact (Van Geel et al., 1996). Over this interval, storms occur during a cooling spell (high sea-ice extent; Figure 5) when the centennial NAO is already mostly negative but not strong (Ólafsdóttir et al., 2013); SST north of Iceland are however still high (Berner et al., 2008). The 2535 ± 50 cal. BP storm event seems also connected with a deep solar low, a rise in SST and NAO negative values with extensive sea-ice. Finally, the 2420 ± 50 cal. BP and 2360 ± 30 cal. BP storm seem to proceed from a situation already described above: High SST, reduced sea-ice (potential positive AMO) and shift of the NAO to more positive values.

Paralleling various climate forcing with storm occurrence

The last 7000 yr BP are characterized by significant changes in climate. Greenland sea-salt data (cf. GISP2; O’Brien et al., 1996) are consistent with storminess, Northern Hemisphere glacial expansions (Denton and Karlén, 1973) and solar forcing (Bond et al., 2001). Storminess in Brittany, in a first approach, appears linked to cooling events. However, comparison between storms, NAO and AMO modes, as well as SST reconstructions, shows a much more complex system. No direct link between storms and negative NAO events intensity has been demonstrated (Burningham and French, 2012). From this storminess analysis for the 20th century (cf. Figure 4), we suggest that major storms, mostly winter events, may correspond very often to marked NAO transitions in a general positive AMO context (warm ocean surface, cf. Figure 5): brief positive NAO incursions within a negative mode period, or inversely. This is true for very recent events.

The storm formation area in the Atlantic mostly corresponds to the Gulf Stream path, commonly associated with high SST. Jet streams also control instabilities in high altitude through the development of cyclonic cells. Moreover, the NAO mode strongly depends on the latitudinal position of the NAJ prior to the NAO onset (Luo and Cha, 2012). The northward (southward) shift of the NAJ from its mean position is a precursor to the NAO+ (NAO−) event. It appears thus that the present-day storms would occur, usually in winter, when cold air masses (mobile polar anticyclones; Leroux, 2010) interrupt suddenly the jet stream (CRWS Jet Stream Map), thanks to the southern position of the Azores High, leading to an abrupt energy reloading at the contact of tropical air with polar one (maps: California Regional Weather Server, http://squall.sfsu.edu/crws/jetstream.html). This is clearly the case for Xavier (5 December 2013, Christian, 2–31 October 2013). High wind speed, warm SST and sea-water spray by waves (cf. chlorides; O’Brien et al., 1996) can significantly increase surface heat exchanges (Zhang and Delworth, 2006), accelerating storm formation. This configuration would imply, for cyclone genesis, a daily positive NAO mode within a longer (decadal to yearly) negative NAO mode, responsible for the atmospheric instability. In this situation, cyclones migrate to the North and may be accentuated at the vicinity of polar anticyclone (Oruba et al., 2012). In addition, in this case, high SST in the western Atlantic would represent the main trigger for strong storm episodes, such as hurricanes (i.e. tropical storms; Goldenberg et al., 2001).

The NAO shifted to strongly negative and more instable behaviour at the onset of the LIA, after a long interval of mainly positive and rather stable NAO during the MWP (Olsen et al., 2012). Both the NAO and AMO cycles vary in time and length in lacustrine record from Iceland (NAO: 2–15 years; AMO: 5–130 years) throughout the last 3000 years (Ólafsdóttir et al., 2013). AMO cyclicity seems to be missing or shorter (35 years) during the LIA. This shift in cyclicity supports a reorganization of North Atlantic circulation during the coldest phase of the LIA (Ólafsdóttir et al., 2013; Trouet et al., 2012). When the jet stream migrates to the tropics in relation with a cooling activated by solar activity, storm tracks are derived to Southern Europe. The large increase in storminess or in cyclone intensity from the end of the Spörer Solar Minimum (Figure 6) probably results from the steepening of the temperature gradient within the Northern Atlantic area, as also revealed by sea-ice extent. It would correspond, especially for the period ad 1700–1800 (Figure 5), to very high AMO pulses (modeling by Otterå et al., 2010).

NAO variance also increase in strength between ad 600 and ad 1000 similarly to the LIA (Ólafsdóttir et al., 2013), although the AMO is quite high. It resumes during the MWP to rise again from ad 1420 (Dawson et al., 2007), in relation with changes in atmospheric circulation, under jet stream control. From all these data, it appears that some indirect solar forcing is required for storms, especially during Dark Ages times, the Homeric period and the LIA, and for thermal storage in the ocean (Otterå et al., 2010; Zhang and Delworth, 2006). Indeed, the intertropical warm pool stores energy coming from direct insolation (cf. TSI), and especially UV (Damé, 2013; Otterå et al., 2010). The AMO is however often considered to be the only internal variability mode of the climate system (Wei and Lohmann, 2012). Here, we strengthen the fact that sea-ice extent and winter NAO mode are also controlled by insolation and must be taken into account (Courtenay et al., 2009; Ineson et al., 2011; Jaiser et al., 2012). In addition, volcanic activity may represent a complementary storm trigger by cooling temporarily the climate (Gao et al., 2008; Otterå et al., 2010).

Synoptic scheme of storm formation in different contexts

On the basis of all the yielded observations and of recent weather data, we thus show that storminess seems well controlled by the solar energy input and its oceanic pacing (AMOC and SST). This is not due to chance, as already shown for extratropical storms through AGCM model (Graff and LaCasce, 2012). We managed to distinguish four situations, even if configurations are less obvious than those recorded above the tropical ocean (Figure 7). These four situations (Figure 7) may succeed in different orders but the extremes configurations 7A and 7D are never directly connected. Situation 2B would often precede situation 7A (Figure 7).

OPEN IN VIEWER

Figure 7.

Synoptic scheme of the different storminess patterns linked with AMO and NAO 10.

Limits to this method

We have to keep in mind that it is difficult to discuss storm events synchronously at the Northern Atlantic scale. Indeed, storminess seems to rise from ad 1400 when considering glaciological record (GISP2; O’Brien et al., 1996). The Quiescence Period (ad 1420–1530), succeeding in Brittany to the MWP, is reduced along the North Sea only to the interval ad 1445–1515 (Bärring and Fortuniak, 2009; Marusek, 2010) but storm reappearance seems delayed from ad 1490 in the Mediterranean region (Marusek, 2010). It could be interpreted as a latitudinal shift of the jet streams. Without other reliable proxies, literature observations rather point to the discreteness in recording the storm episodes more than a regional answer to a climate shift. During the observed relative ‘calm’ in storminess, the AMO mode was low, and especially between the time interval ad 1430–ad 1480 (cf. modelling exercises by Otterå et al. (2010)). During this period when SST and NAO modes appear unsteady, NAO is however mostly positive and sea-ice cover is still rather extended. A reorganization and acceleration of the AMOC, issued from an important warming (Otterå et al., 2010; Trouet et al., 2012; Wei and Lohmann, 2012), could be thus hypothesized to explain this ‘calm’ storm trend during the 15th century by reducing ocean/atmosphere heath exchanges in favour of ocean/ocean ones. Outside quiet storm periods, most of preserved events seem to have a recurrence time of about 100–200 years and may be considered to be centennial sized events (Figures 4 and 5). Preservation concerns only events lasting several days and/or in conformity with both rather tidal coefficients.

As a limit to this method, it is necessary to keep in mind that the geological record is incomplete due to the regional subsidence and the continuous sea-level rise, which contribute to the erasing of most of the early Holocene events (Goslin, 2014; Goslin et al., 2013). As a consequence, onshore, we do not identify a 1500-year cyclicity that would drive storminess (Sorrel et al., 2012; Viau et al., 2002). Nevertheless, the period usually poorly recorded in drilling with ‘low erosional stands’, induced by mid to low tide storms, seems well to approach this cyclicity (Goslin et al., 2013; Van Vliet-Lanoë et al., 2014, this issue).