As there are currently no fixed criteria for what qualifies as a meteotsunami, in this paper we bring together various aspects used by other researchers in the field into one standardised system. Figure 1 displays a visual representation of the commonly used criteria, which we explain in more detail in Sect. 2.1.1–2.1.2. The methodologies that have been previously used by researchers and studies have variations, with some using qualitative methods that base events on eyewitness accounts (Haslett and Bryant, 2009; Haslett et al., 2009) and others using quantitative data from sea level and atmospheric observations (Tappin et al., 2013; Sibley et al., 2016). For the purpose of this paper, we have classified meteotsunami as atmospherically induced sea level oscillations having at least one sea level and one atmospheric characteristic. This allows for the distinguishing of meteotsunami from other types of waveform and is applicable to either qualitative accounts or quantitative data.

To gain a complete understanding of these events we follow Long (2015) and Haslett and Bryant (2008), who dated their historic tsunami catalogues back to approximately 1000 CE. We noted that any events preceding 1750 CE were vaguely recorded, making validation problematic, so we opted to date our catalogue back to 1750. References to meteotsunami-like events in historical accounts tend to be based on descriptions of the state of the water at the coast with a lack of instrumental tidal data. There is a lack of or limited weather data, so tracing back the atmospheric source is not straightforward. It is only in the last few decades that meteorological data with sufficient resolution have been readily available. With tide gauge data, prior to 1993 the resolution was hourly, and it was not until 1996 that all the current tide gauge sites became fully operational. Therefore, we have used 2009 as the upper limit of the historical record. The historical reports tend to be derived from newspaper articles, parish records, harbourmaster records and eyewitness accounts. Although there is reason to be sceptical of these accounts as they afford a level of biased review and sensationalism, they do still hold value in terms of a societal viewpoint and may help to fill in any gaps (Haslett and Bryant, 2009; Haslett et al., 2009).

There are certain characteristics that can be flagged up in a historical account to verify whether it refers to a meteotsunami event or not. To illustrate this, we can highlight the historical account for the event of 23 May 1847, where we can look at a letter from Robert Blight of Penzance dated 24 May and published in the Royal Cornwall Gazette on 28 May. The full extract can be found in Extract S1 in the Supplement of this paper and in Long (2015, p. 26).

The changes in the atmosphere during the day were very remarkable. In the morning, about six o'clock, we had a breeze from the southeast; by eight, it was a perfect calm; between ten o'clock and two, the mercury sunk several degrees; about three in the afternoon a breeze sprung up suddenly from the west, and the sky, as suddenly, became overcast … It is very probable that all these changes, and even the agitation of the sea, were produced by electricity.

In this particularly detailed account (Extract S1), we can identify six of the nine criteria, including a drawback and sudden inrush of water, accompanied by a rumbling noise and the water being higher than expected at 2.4 m (8 ft) (criteria 1a and 1d), all indicating a tsunami-like event. The key to the identification of a meteotsunami is in the atmospheric portion of the account; what started out as a calm morning featured a change in wind speed and direction, veering from southeasterly in the morning to westerly in the afternoon (criterion 2d). This variable wind was accompanied by a drop in temperature (criterion 2d), and finally, there was mention of the presence of a storm in terms of overcast sky, threatening rain and lightning (criterion 2a). As such, we identify this wave as a meteotsunami by applying both our oceanographic and our atmospheric criteria to the historic account.

To identify meteotsunami from 1 January 2010 to 31 December 2022, we use data records that are available at higher frequencies, meaning meteotsunami are more distinctly observable. The information for this portion of the catalogue is sourced from the British Oceanographic Data Centre (BODC) website (https://www.bodc.ac.uk, last access: 1 March 2023) and the Intergovernmental Oceanographic Commission (IOC) website (https://ioc-sealevelmonitoring.org/, last access: 8 July 2023, Sea Level Monitoring Facility, 2023), where data are displayed from the “Class A” network of tide gauges owned and funded by the Environment Agency (EA). We also use the post-processed data of Williams et al. (2021), where the raw sea level tide gauge data have been high-pass-filtered to isolate high-frequency disturbances. This removes periods of over 120 min and separates out the tidal components. In this way any signals in the tsunami frequency band (2 to 120 min) are isolated from the sea level elevations. Any remaining signals larger than the background noise are then identified and checked against our threshold criteria to verify events as potential meteotsunami. Apart from the standard processing to remove any erroneous spikes outside of the parameters, a visual quality control was carried out, where a 7 d plot of the data was evaluated to highlight any clear artificial spikes or gaps. Also, any data points that had no accompanying air pressure changes were also excluded from any further analysis.

The times of the potential meteotsunami events are noted from the tide gauge data, and they are then linked to specific precipitating convective atmospheric systems by using the meteorological C-band radar network, which is pre-processed by the UK Met Office before download (Met Office, 2003). The convective systems highlighted by the radar are classified into four distinct types (as shown in Fig. 1). These are (1) open cells which are situated behind the cold front of cyclonic weather, usually where cold dry air passes over the warm sea, creating shallow convection; (2) quasi-linear systems which tend to be multi-cellular and linearly organised with high CAPE, heavy precipitation and strong winds (this type of weather feature is sometimes called squall lines and can occur within synoptic Spanish plume events); (3) isolated, small, short-duration (<1 h) thunderstorm cells; and (4) non-linear clusters which are large, circular, long-lived clusters of precipitation and thunderstorm cells.

The atmospheric ascent soundings are obtained from the University of Wyoming website (http://weather.uwyo.edu/upperair/sounding.html, University of Wyoming, 2022) with the UK stations at Camborne (station number: 03808) and Lerwick (station number: 03005) being used. Soundings are available for 00:00 and 12:00 UTC on each day, and if a CAPE value of greater than 0 occurs, then this shows a marginally unstable atmosphere leading to convective activity. Finally, the synoptic charts allow for verification of the storm system, including the location of the pressure centres and fronts at the time of the meteotsunami wave event.

We identify 98 events as meteotsunami occurring in UK waters between January 1750 and December 2022 (Table 1), with 48 of these occurring within the historical record (1750 to 2009). This record shows that 67 % of documented meteotsunami occur in summer (April–September), with 44 % of documented meteotsunami occurring in July and August. The single year experiencing the most documented events was 1802 CE, numbering three, and the decade experiencing the most documented events was the 1840s, with six in total. The presence of a storm and/or characteristics of convective activity (thunder and lightning) at the time of the wave event was noted for 42 of the 48 events (91 %) in the historical record. There was also a defined southwest prevalence of meteotsunami in historical documents, with Devon, Cornwall and Somerset recording a combined total of 29 events. Within the historical record we have identified four new events and reclassified four tsunami events, three storm surge events and nine events of unknown origin as meteotsunami. Seven of these occurred within winter months (Table 1).

Meteotsunami have been thought to be a rare phenomenon in the UK, and it has been thought that when they do occur, it tends to be in the summer months due to the more abundant convective activity (Haslett et al., 2009; Tappin et al., 2013; Sibley et al., 2016; Thompson, 2020). However, of the 98 identified meteotsunami events verified in this paper, 50 have been interpreted as occurring since 2010, with 33 (66 %) of those occurring during the winter months and 9 of these winter events identified as new. We find that not only are UK meteotsunami more common in occurrence than previous research indicates, but also they are a year-round phenomenon, as exhibited in Table 1 and Figs. 2 and 3.

The historical section of the catalogue shows an estimated return period of 5.4 years. This return period considerably decreases for the instrumental-data section, where the UK return period reduces to an estimated 0.25 years. With an average of four events per year, we can see that certain years have experienced above-average numbers and high proportions of winter events, with seven winter events out of eight in 2013, four out of seven in 2021 and five out of seven in 2022. Figure 3 displays the seasonal distribution of these events, with 34 % of meteotsunami recorded in December and January and no events being recorded in March or April. Following statistical analysis, the recorded maximum wave amplitude for each event resulted in a mean wave height of 0.33 m for winter and 0.35 m for summer. With a t-test score of 0.30 and a p value of 0.07, the tests indicate a similarity between the two sample sets, where the difference between seasonal wave heights is considered not statistically significant.

Summarising the results from the catalogue in its entirety, we suggest that there are three “hotspot” regions where meteotsunami events appear to be most frequent: (1) northwest Scotland, (2) Wales and (3) the southwest UK. Up until 2009, Penzance in the southwest UK had experienced the most meteotsunami with eight in total. Then from 2010, Kinlochbervie in northwest Scotland experienced a maximum wave height of 0.51 m during the 16 November 2016 event. This same location was exposed to 14 separate meteotsunami events in the 12 years from 2010 to 2022. Harbour-style geomorphology appears to be more susceptible to meteotsunami resonance, being associated with 71 % of the events with beach environments comprising the remaining 29 %.

In this section, we highlight two specific meteotsunami events that occurred in two consecutive winter seasons. These two events have been picked as they are new events to the catalogue, and they represent a typical winter meteotsunami hidden in the associated storm data. The winter of 2021/22 saw seven sequential named storms with five verifiable meteotsunami events, one of which was on 20 October 2021. The winter of 2022/23 saw three likely/numerically verifiable meteotsunami events, one of which was 1 November 2022. Both meteotsunami events were low profile, localised in nature and hidden within larger-scale heavily precipitating low-pressure systems.

With the identification criteria we have laid out in this paper we have verified 98 events in UK waters since 1750, of which 38 are new events containing 7 new winter events in the historical record (1750 to 2009) and 8 new winter events in the modern record (2010 to 2022).

It was found that a selection of historical events were misidentified in accounts as abnormal coastal flooding, non-tsunami, storm surge or of unknown origin. This was extended by an analysis of current data (since 2010), which allowed us to add a total of 38 new events to the catalogue, of which 15 occurred within winter months; these are highlighted in Table 1 as new (N), new winter (NWI) or verified (V) events.

The misidentified events were discovered after an attempt to highlight characteristics that match those listed in the methodology, in particular characteristics that suggested a tsunami-like phenomenon but with any associated storm-like activity or air pressure fluctuations. If the account was found to contain a lack of evidence or information to suggest a meteotsunami, it was rejected. An event occurring on 13 February 1979 was highlighted as a meteotsunami by Haslett and Bryant (2009) but was contested by Thompson et al. (2020) as being a surge caused by a winter Atlantic storm due to its seasonal placement. In their 2020 paper, Thompson et al. (2020) appear not to class Atlantic storm systems as sources of meteotsunami. They state that, from April to October, thunderstorms generate meteotsunami and, from November to March, storms generate low-pressure swells and surges. Our paper has matched descriptions in historical accounts with the criteria laid out in the methodology, and we agree with Haslett and Bryant (2009) and Haslett et al. (2009) that the 1979 winter event was a meteotsunami. This result was determined by the similarities in the pressure profile, geographical distribution and speed of anomaly to the known meteotsunami event of 26 June 2011.

In addition to the 1979 event, there were further events found that were previously labelled as meteotsunami which, according to our criteria, we found to be of alternative origin (tsunami) or to have insufficient detail or collaborative evidence to solidify a conclusion. These include the events presented in Long (2021), dated 14 October 1862 (found to be a tsunami due to an alternative source trigger), 15 August 1895 (insufficient information) and 11 May 1912 (found to be a tidal bore), and another tidal bore dated 17 May 1964 presented in Haslett and Bryant (2009) and Haslett et al. (2009).

The event of 31 March 1761, which was labelled as a tsunami by both Long (2015) and Thompson et al. (2020), was found to be a winter meteotsunami due to tsunami-like waves being experienced not only along the southwest UK coast but also in Loch Ness in Scotland, with the mention of a calm sea before the arrival of thunderstorms.

The historical record (1750 to 2009) has been found to support previous studies such as Haslett and Bryant (2009) and Haslett et al. (2009) that have alluded to the positive correlation between thunderstorms and meteotsunami waves, with 71 % of summer events displaying reports of convective activity. Our results have highlighted a summer prevalence of events, with 48 % of them peaking in July and August, which reflects Thompson et al. (2020). This prevalence has been based principally on a reliance on eyewitness reports and the volume of persons present at the shoreline during these months.

These summer events tend to be associated with heat waves and so-called Spanish plumes as in the 27 June 2011 and the 18 June 2022 events along the southwest UK coast. This is where warm air moves northwards from the European continent and Iberia, during which mesoscale convective weather tends to occur. In the summer, CAPE is at its highest and over land due to warm 2 m air temperatures over landmasses (Holley et al., 2014). These types of weather event consist of single cells or clusters of small, short-duration (<1 h) thunderstorms and squall lines with more than one convective cell (Sibley, 2012; Tappin et al., 2013).

The element of risk during the summer occurs when the meteotsunami wave can become fully disconnected from its source disturbance. This effect can be particularly apparent if the meteotsunami interacts with the continental slope, where the wave can arrive hours after the original storm has dissipated or moved on. This delayed arrival of wave disturbances can surprise people who are subsequently back out on the water or near the water's edge, believing the storm has passed. This scenario was experienced during the 5 July 2021 event that occurred at Westward Ho! (north Devon), when just after midday BST a small yet powerful wave unexpectedly progressed 50 m up the beach, inundating many beachgoers.

Previous studies have suggested that winter wave anomalies such as meteotsunami are “less” likely than storm waves and surge, and winter data have not previously been interrogated for this reason. However, the present-day record (2010 to 2022) appears to contradict this, with a winter prevalence of 66 % of events peaking in December and January and with a tendency towards October and November in the 2021/22 winter season.

The results also show a geographical pattern to UK meteotsunami, with a large proportion of events occurring along the southwest UK and northwest Scotland coast in the winter, aligning with the dominant weather direction of west to east from the Atlantic Ocean, and along the southern UK coast in the summer, aligning with Spanish plumes bringing warm air polewards from the Equator with southerly winds up and along the English Channel. The geographical pattern also reflects the influence of local bathymetry, with harbours (e.g. Penzance, Plymouth, Stornoway and Port Talbot), bays (e.g. Kinlochbervie and Port Stoth) and river mouths (e.g. river Yealm and river Dart) containing conditions more favourable to meteotsunami initiation and amplification via resonance and seiching.

To further the concept presented in Williams et al. (2021), we selected two recent winter meteotsunami events and highlighted the meteotsunamigenic criteria. It has been indicated from the results that the combination of a mid-latitude depression with frontal and convective weather moving across the UK may be important in the generation of this hazard. Results have shown that, during these winter storms, convective elements are likely to be embedded around heavy rainfall (Fig. 4a and b) and strong winds associated with the cold front, leading to the potential for meteotsunami waves. This winter synoptic situation is a product of the combination of the cold maritime Arctic air being introduced into the rear side of the cold front passing over relatively warm water. The risk of flooding can be exacerbated due to surface water from precipitation as the front crosses a landmass (Masselink et al., 2015).

The results highlighted an average maximum wave height of 0.3 m, which may not seem “dangerous”, but this hazard is not purely about this single factor. The key that makes meteotsunami a potential hazard is the rapid onset of the wave (sometimes referred to as a “wall of water”) and the associated strong currents.

Identifying meteotsunami events in winter tends to be more difficult as the waves can be hidden and overshadowed by the wave characteristics of the trigger storms and may be missed unless looking specifically at the data. We strongly consider this overshadowing to mean that many of these winter meteotsunami are not reported, and this may have been the issue in previous research where certain winter events were identified as either storm waves or surges instead of meteotsunami. As we have seen, there is a short observational record available for meteotsunami and there is evidence for severe under-recording of such events. Even though the 2010-to-2022 record has shown significant improvements in recording completeness, the current 15 min sampling interval is still too coarse. This was highlighted when certain events in the catalogue such as those on 2 October 2021, 20 October 2021, 27 November 2021 and 19 July 2022 were uncovered in the 1 min tide gauge data that were not so easy to locate in the 15 min data. This creates an issue whereby many events with a wave period of under 15 min may be potentially missed. We recommend a reduction of the sampling interval to 1 to 5 min to yield more data and to be able to draw a complete conclusion for this hazard.

Another limitation of this study linked to the sampling frequency was the treatment of wind-driven waves, which can induce infra-gravity waves of a similar wave period to meteotsunami (2 to 5 min). We did initially consider the wind and swell peak period and wave height; however, we discovered that the detection of infra-gravity waves from low-frequency tide gauge data is uncertain and was deemed to be beyond the scope of this study. To perform such an analysis and to be confident in our results, we would require 1 min per 2 Hz of data for a spectral analysis. However, it may be prudent to explore this aspect in future work.

We noted that historical accounts are not optimum for identifying and analysing meteotsunami due to their anecdotal nature, and as such the number of events represented here may be dramatically underestimated. Data before 2008 are not readily available and records are spatially sparse, which leads to incomplete data coverage and does not allow for a robust statistical analysis.

The placement of tide gauges used to provide data also affects results. The siting of UK tide gauges tends to be biased towards populated areas with harbours and river mouths for asset protection and is ideal for the capture of the resonant component of the meteotsunami wave. However, events in less populated areas may have been missed due to this placement. We suggest potential tide gauge locations (based on the occurrence rate of previous events) could include beach or estuary locations around Devon and Cornwall such as Mevagissey or Perranporth and the north of Scotland such as Dunnet or Port Stoth.

The next few decades are likely to see sea level rise push mean and extreme water levels upwards, which will subsequently increase the level of risk by bringing the height of the storm tide closer to the flood stage (Masselink et al., 2015). At many UK locations, flood defences are at the design threshold of current storm surge levels; they are not designed or built for a sudden, prolonged water flow as seen in meteotsunami (Lazarus et al., 2021). A question that has arisen from this paper is whether the winter seasons of 2013/14 and 2021/22 are outliers or whether this clustering of storms and meteotsunami will be a commonplace scenario in the future. Currently, we can detect and forecast mid-latitude depressions 9 to 10 d in advance (Penn State, 2019); knowing this, we can incorporate warnings of potential meteotsunami activity into forecasts. However, due to the localised nature of meteotsunami, the risk level in each coastal area needs to be considered on its own merits. The risks connected with a single meteotsunami event in two different bays can be quite different. One bay may suffer from inundation and flooding, whereas another bay may be impacted by strong currents. This paper provides a valuable insight into the frequency, seasonality and spatial distribution of what was a hidden hazard in the UK. These new data will need to be incorporated and taken into consideration when coastal management strategies and defences are adjusted for the future.

Meteotsunami may well have some role to play in coastal storm impacts; however, the relative contribution of meteotsunami to storm surge in the aftermath of a storm and the full extent of the risk remain unknown, and determining them is beyond the scope of this work. It is also difficult to determine if the frequency and intensity of either low-pressure winter storms or winter meteotsunami are on the increase. We invite a closer and more robust scrutiny of this hazard with a year-round perspective, bearing in mind that no solid conclusions can be drawn without high-frequency, long-term and continuous monitoring of this hazard.