Observations of meteotsunamis at different locations
We identified several meteotsunamis with different magnitudes from the tide gauge records (Fig. 1). Here, the focus is on describing the characteristics (heights, period) of the largest meteotsunami events (crest-to-trough wave height >0.40 m) at each location and their relation to the local meteorological conditions between January 2000 and January 2013 (Table 1). The results from six stations along south-west coastline are shown in Fig. 3 (see also Table 1) and presented below. The Bureau of Meteorology weather records were used to identify the atmospheric conditions present at the time of each meteotsunami, and detailed analyses using meteorological stations data during selected events are presented in Sect. 4.2. Time–frequency spectral plots of the sea-level records obtained during the meteotsunamis are also presented to examine the wave periods in the record.
Carnarvon: 23 October 2006
Thunderstorms with widespread wind gusts of 90–100 km h−1 were reported in WA’s central coastal region in the early morning of 23 October 2006. These thunderstorms produced a meteotsunami, whose oscillations were recorded on tide gauges at Carnarvon and Geraldton (Fig. 1). A maximum wave height of 0.60 m was recorded at Carnarvon (Fig. 3a), and the resulting high-frequency oscillations continued for ~8 h. The existing seiche oscillations were excited at ~0300, but the highest wave (crest-to-trough height) occurred at ~0800, which corresponded to low tide. Carnarvon experiences mixed tidal conditions and a spring tidal range of ~3 m (Burling et al. 2003). Meteotsunami-like events were not observed at tide stations farther north, where spring tidal ranges are between 3 and 10 m. The highest wave height recorded at Carnarvon for the 26 December 2004 tsunami was 1.14 m, whereas for the 28 March 2005 tsunami, it was only 0.3 m (Pattiaratchi and Wijeratne 2009).
The energy spectra for Carnarvon for 20–28 October 2006 showed the meteotsunami of 23 October enhanced the energy at 30- and 80-min periods (Fig. 3g). Both the 30- and 80-min periods could be related to bay oscillation. The shelf is wide (>80 km) at Carnarvon; thus, we expected the fundamental shelf oscillation to be longer than 4 h; however, oscillation periods with this range were absent from the energy spectra.
Geraldton: 25 February 2005
At ~0400 on 25 February 2005, a meteotsunami with a maximum wave height of 0.42 m was recorded at Geraldton (Fig. 3b). Tide gauges at Carnarvon and Jurien Bay also recorded meteotsunamis with small wave heights (<0.4 m) the same day; however, no extreme meteorological events in these regions were reported. The energy spectra for Geraldton for 22 February–2 March 2005 showed enhanced spectral energy at the 35-, 80-min, and 4-h periods, which was associated with the 25 February meteotsunami (Fig. 3h). The 80-min period might have been related to the first mode (n = 2) of shelf resonance periods, and the 4-h period was related to the fundamental (n = 1) shelf resonance period with a shelf width of 80 km (Eq. 1). This 4-h period was also excited during seismic tsunamis at this location (Pattiaratchi and Wijeratne 2009). The meteotsunami caused the energy bands at all frequencies to increase at the same time, with the strongest energy occurring during the 4-h period. A filtered time series and time–spectral plot of the seismic and meteorological tsunamis that occurred at Geraldton between February and April 2005 (Fig. 4) also showed that the seismic and meteorological tsunamis had enhanced existing seiche frequency-band oscillations. In contrast, the larger-amplitude shelf oscillations at the 4-h period were present almost continuously.
Fremantle: 17 October 2002
A strong cold front hits the south-west of WA at ~1300 on 17 October 2002. Several tornadoes that occurred at different sites in the study region were associated with this front. Four tornadoes were reported in the Perth metropolitan area (Courtney and Middelmann 2005). Meteorologically induced sea-level oscillations were recorded at several tide stations along the west and south coasts. At ~1330 on 17 October, the Fremantle tide gauge recorded a meteotsunami (Fig. 3c). The initial wave height (crest to trough) recorded at Fremantle was ~0.7 m, which was double the tsunami wave height of 0.33 m recorded at Fremantle as a result of the seismic tsunami of 26 December 2004 (Pattiaratchi and Wijeratne 2009).
The spectral analysis results of the energy distribution from the tide gauge records from 15 to 22 October 2002 are shown in Fig. 3i. The spectral energy bands show waves with periods of 25 min, 1, and 2 h and 50 min. The spectral analysis results showed the meteotsunami enhanced the existing seiche energy, with the strongest energy occurring during the 2 h and 50-min period; this period was related to the shelf resonance (see Pattiaratchi and Wijeratne 2009; Pattiaratchi 2011).
Bunbury: 5 December 2002
At ~1700 on 5 December 2002, a thunderstorm and associated large hail occurred in the south-west of WA. The resulting meteotsunami was recorded at the Bunbury, Busselton, and Hillarys Harbour tide gauges. The meteotsunami-induced sea-level oscillations at Bunbury (Fig. 3d), which continued for more than a day, reached a maximum height (crest to trough) of 0.9 m; however, this wave occurred several hours after the first meteotsunami oscillation. In comparison, the highest crest-to-trough wave height recorded at Bunbury during the December 2004 tsunami was 1.75 m; this was also the highest recorded wave height along this part of the coast (Pattiaratchi and Wijeratne 2009).
The time–frequency plot for 3–11 December 2002, which includes the meteotsunami recorded at Bunbury, is shown in Fig. 3j. The four seiche frequency bands show waves with periods of 35, 60, 80 min, and 4 h, with the highest spectral energy occurring during the 4-h period and continuing past 11 December. In contrast to the meteotsunamis that occurred at Geraldton and Fremantle (see above), the energies in the lower period (<80 min) were enhanced at the start of the meteotsunami, but the shelf oscillations at the 4-h period did not increase until several hours later.
The Bunbury and Busselton tide gauges recorded the highest number of large meteotsunamis (wave heights larger than the mean tidal range of 0.5 m) during the study period. Three recent events from this region were chosen for detailed analysis, and the results are presented in Sect. 4.2.
Bremer Bay: 28 November 2006
On 28 November 2006, strong winds and cold fronts were reported at Rottnest Island (Fig. 1). The south-west WA tide gauges recorded a meteotsunami on the same day. The tide gauges at Bunbury and Esperance (not shown here) recorded a meteotsunami with initial wave heights of 0.7 m (at ~0230) and 0.6 m (at ~0320), respectively. Meteotsunami-induced sea-level oscillations were also recorded at the inner Bremer Bay tide gauge, with a maximum height of ~0.4 m recorded at ~0700 on 29 November (Fig. 3e).
The time series of spectral energy for the Bremer Bay meteotsunami on 29 November (Fig. 3k) showed the meteotsunami enhanced the frequency bands at the 28- and 65-min periods. These periods could be related to the fundamental and higher modes of the bay oscillation; however, the energy band of the shelf oscillation period could not be seen on the time–frequency plot.
Esperance: 8 January 2008
A severe thunderstorm with large hail and strong winds occurred in the south-west of WA in the late evening of 8 January 2008. This thunderstorm caused a meteotsunami along the south coast, which was recorded at the Esperance and Bremer Bay tide gauges. The meteotsunami occurred at Esperance at ~1830 and high-frequency oscillations continued for more than a day (Fig. 3f). The meteotsunami occurred during the rising mid-tide, and the first wave (with a height of ~0.7 m) was the largest in the wave group. For comparison, the highest wave recorded at Esperance on 26 December 2004 was 0.44 m (Pattiaratchi and Wijeratne 2009). The spectral energy distributions from the Esperance tide gauge records for 6–14 January 2008 are shown in Fig. 3l. The figure shows the meteotsunami enhanced the energy at the 80-min and 2-h periods.
One of the features of the time series of the meteotsunamis (Fig. 3) is such that at some locations, the first oscillation is the strongest whilst and at others the amplitude gradually increased and then decreased. Rabinovich and Monserrat (1996) classified these different types of oscillations as impulse (strong initial oscillation and fast decay), resonance (gradual build-up of oscillations), and complex (a mix of the two). The time series indicates that the meteotsunamis at Fremantle and Esperance were impulse whilst at the rest of the stations, it was resonance.
Detailed analysis of meteotsunamis
Detailed meteorological data (such as rainfall radar and sea surface pressure data, which define the propagation of thunderstorms) have been available since 2009. These data enabled us to examine three meteotsunami events (22 March 2010, 10 June 2012, and 7 January 2013) in more detail. These meteotsunamis were also recorded at more than one tide gauge station indicating the spatial extent of a single meteorological effect. For example, the meteotsunamis on 7 January 2013 (see below) were recorded in tide gages >500 km apart.
Meteotsunami: 22 March 2010
The costliest natural disaster in Western Australian history, with a damage bill estimated at A$1.08 billion, occurred in the Perth region between 15:30 and 17:00 on Monday 22 March 2010. Most of the damage was due to hail associated with thunderstorms, with coastal stations recording wind (gust) speed of more than 30 ms−1. The thunderstorms were initiated to the north of Perth and progressed south along the coast (shore parallel). Rottnest and Garden Island meteorological stations recorded a pressure change of 4 hPa associated with the thunderstorms (Fig. 5a, b), which reduced to 2 hPa as the storms progressed to Bunbury and Cape Naturaliste (Fig. 5c, d; locations in Fig. 1). The pressure jump was associated with changes in the wind speed and direction (Fig. 5), the period of pressure jump inconsistent from station to station. The timescale of the atmospheric pressure change was 3.4 and 4 hPa over 2 h at Rottnest Island and Garden Island, respectively (Fig. 5).
A time series of radar images of the rainfall (Fig. 6) shows the southward progression of the squall bands associated with the thunderstorms. Using Orlić (1980) formula (Eq. 2), including all meteorological stations, the average propagation speed of the pressure jump was ~15 ms−1 and direction ~002o (meteorological convention). However, as to be expected, the propagation speed may have changed over the large area. The radar images indicated an alongshore propagation of the thunderstorms with a speed of 9.3 ms−1, which is equivalent to a shallow-water wave celerity in 8.8 m water depth, close to the shore (Fig. 1). In the region between Bunbury and Cape Naturaliste (Fig. 1), the pressure jump propagation speed increased to 18.9 ms−1 (based on Eq. 2) and the speed of the meteotsunami between Bunbury and Busselton (Fig. 7), based on the tide gauge data, was 17.5 ms−1. Both of these examples indicated that conditions were favourable for the occurrence of Proudman resonance. However, with the available data, it is not possible to determine whether Proudman or Greenspan resonance was responsible for the generation of the meteotsunamis. The thunderstorms created a meteotsunami with maximum wave heights of 0.48, 0.55, 0.40, and 0.45 m at Hillarys, Fremantle, Bunbury, and Busselton, respectively (Fig. 7).
A wavelet transform of the Hillarys Boat Harbour sea-level (Fig. 1) record, which had a sampling interval of 6 min (Fig. 8), shows the water-level variability that occurred in the harbour. The wavelet transform also shows the presence of diurnal and semidiurnal energy and trophic/equatorial (O’Callaghan et al. 2010) and spring/neap tidal stages. The energy between the diurnal and semidiurnal components is separated because the record was obtained in March, close to the equinox when the tropic–equatorial and spring–neap tidal stages were out of phase (O’Callaghan et al. 2010). The water-level time series (Fig. 8a) shows the transformation of the tides from a diurnal to semidiurnal system and then back to a diurnal system. The time series also shows an energy peak of around 3 h, which Pattiaratchi and Wijeratne (2009) identified as the first mode of the shelf resonance with a period of 2.7 h (Fig. 8b; see also Fig. 3c, i). During the meteotsunami, oscillations with periods of 1–2 h were strongly excited (Fig. 8d).
Meteotsunami: 10 June 2012
The meteotsunami recorded on 10 June 2012 was unusual in that it occurred during winter and was generated by the passage of a low-pressure system across south-west Australia (Fig. 9). The low-pressure system was initiated as a tropical low off WA’s north-west coast, and moved almost parallel to the coast, with the centre of the low crossing the coast to the south of Perth (Bureau of Meteorology 2012). The system caused maximum (gust) wind speeds >30 ms−1 and a rapid change in wind direction from shore parallel (northerly) to onshore (westerly) (Fig. 10). The system was extreme, with a drop in sea surface pressure of 30 hPa over 36 h and a minimum of 983 hPa (Fig. 10). Sea surface pressure values <1,000 hPa are rare in these subtropical systems.
The meteotsunami generated from the low-pressure system produced the highest water level ever recorded at Fremantle, which has a continuous record spanning over 110 years and is the longest sea-level record in the southern hemisphere (Pattiaratchi 2011). The meteotsunami wave height was 0.61 m; however, this was not the highest meteotsunami ever recorded at this site (Fig. 11; Table 1).
The Fremantle tide record has several components that contribute to sea-level variability: tides, storm surges, and annual and inter-annual variability due to large-scale oceanic changes (Pattiaratchi 2011). The meteotsunami occurred at high tide (Fig. 11) and coincided with the local storm surge; it was also a La Niña year, so the meteotsunami also coincided with a strong Leeuwin current. These factors all contributed to producing the highest water level recorded at Fremantle. In contrast, the corresponding meteotsunamis at Bunbury (wave height of 1.03 m) and Busselton (wave height of 1.10 m) had the highest wave heights recorded at these stations (Fig. 12; Table 1) and were equivalent to the maximum storm surges recorded at these stations during tropical cyclone Alby in 1978 (Bureau of Meteorology 2012; Haigh et al. 2014).
The meteotsunami was generated during onshore winds. Onshore winds, generated when the front crossed the coast and the wind speed rapidly increased and then decreased over 1–2 h (as recorded at the meteorological stations) (Fig. 10), most likely caused the meteotsunami. It is also possible that Proudman resonance may have contributed to the generation of the meteotsunami. In a numerical model study, Vilibić (2008) demonstrated that pressure jumps travelling across the shelf, where the depths decrease towards the shoreline, can create conditions for Proudman resonance. However, there is insufficient data to clearly identify the mechanism which was responsible for the generation of the meteotsunami.
Meteotsunami: 7 January 2013
Thunderstorms, which affected an area of over 500 km from Geraldton to Bunbury (Fig. 1), caused the meteotsunami recorded on 7 January 2013. Two pressure jumps of 4 hPa were recorded at North Island meteorological station at 0000 and 07:30 (Fig. 13); however, only the first pressure jump generated a meteotsunami at Geraldton at 00:45 (Fig. 14). The initial pressure jump, in the form of a thunderstorm system, moved south and arrived at Rottnest Island at 06:30 (Fig. 13). This pressure jump is reflected in the high rainfall squall band in the rainfall radar image of the Perth region obtained at 06:00 on 7 January (Fig. 15).
The pressure jump speed, estimated using meteorological s data, between North Island and Rottnest Island was 19 ms−1, which increased to 32 ms−1 between Rottnest Island and Bunbury (Fig. 13). If all the meteorological stations were used together with Eq. 2, the mean propagation speed of the pressure jump was ~22 ms−1 and direction ~357o. This indicated that over the propagation distance (>500 km), the speed of the pressure jumps was not constant. The pressure jumps at each station were associated with peaks in the wind speed (with wind gusts exceeding 15 ms−1) and northerly winds with periods <1 h (Fig. 13). The maximum meteotsunami wave height of 0.81 m was recorded at Geraldton (Table 1), with lower values recorded at Fremantle (0.48 m), Bunbury (0.3 m), and Jurien Bay (0.43 m). The pressure jump speed between Rottnest Island and Bunbury was 32 ms−1, which was similar to the meteotsunami speed between Fremantle and Bunbury of ~30 ms−1, which confirmed that thunderstorms caused the meteotsunami.
The wavelet transform of the Geraldton (Fig. 1) sea-level record, which had a sampling interval of 5 min, shows the dominant energy periods during the meteotsunami (Fig. 16). The time series shows an energy peak of around 4 h, which Pattiaratchi and Wijeratne (2009) identified as the first mode of the shelf resonance (Fig. 16b; see also Figs. 3b, h, 4). Oscillations with periods of 1–2 h were strongly excited during the meteotsunami (Fig. 16b). A smaller meteotsunami with similar periods was identified 24 h earlier (6 January; Fig. 16b).