Update of the tsunami catalogue of New Caledonia using a decision table based on seismic data and maregraphic records

14 years ago, the December 26, 2004 Indian Ocean tsunami brought to the entire World the destruction capability of tsunamis. Since then, many research programs have been initiated to try to better 10 understand the phenomenon and its related hazards, and to improve the early warning systems for the exposed coastal populations. Pacific Islands Countries and Territories (PICTs) are especially vulnerable to tsunamis. Amongst them, New Caledonia is a French overseas territory located in the South-western Pacific and exposed to several tsunamis sources. In 2010, a catalogue of tsunamis that were visually observed or measured in New Caledonia was published. Since this first study, several events occurred between 2009 and 2019 and an update of 15 this catalogue was necessary within the framework of a tsunami hazard assessment project in New Caledonia (TSUCAL). To complete this catalogue, a decision table has been designed to select potential tsunamigenic events within the USGS earthquake database, using criteria on the distance to New Caledonia, the magnitude and the hypocenter depth. Then a cross-comparison between these earthquake events, the NOAA NGDC tsunami catalogue and local tide gauge records provided 25 events that were recorded in New Caledonia for the period 20 from September 30, 2009 to January 10, 2019. These events are added to the 12 events reported with certainty during previous studies, leading to a number of 37 tsunamis triggered by earthquakes reported or recorded in New Caledonia since 1875. Six of them have been identified only thanks to local tide gauges, supporting the fact that instrumental recording of tsunamis is paramount for tsunami hazard studies, from early warning to the validation of coastal models. In addition, unpublished tide gauge data is provided for the 1960 Chile tsunami. 25

New Caledonia is an archipelago that was originally populated by Austronesians circa 3000 years ago (Forestier, 1994). It has been discovered by Europeans on September 4, 1774 (Faivre, 1950). But according to the same author, foreigners really began to settle here only after 1840, appearing with them written reports of uncommon 40 natural phenomena, including tsunamis. Thus, the reported written history of tsunamis covers only the last 180 years, and only the biggest events have left available information especially when associated to an earthquake felt by the population; it concerns mainly earthquakes occurring at the Vanuatu subduction zone. In addition, due to its oceanic location, New Caledonia is also exposed to tsunamis coming from other parts of the Pacific Ocean, from regional sources (e.g. Solomon and Tonga-Kermadec trenches ; Fig. 1) to transoceanic sources (e.g. Chile, 45 Japan or Kuril subduction zones). Most of the time these tsunamis have been reported in coeval reports by witnesses or transmitted orally in the Kanak tradition and collected in a catalogue by Sahal et al. (2010) for the period running from 1875 to 2009 as an update of the previous catalogues from Soloviev and Go (1984) and Louat and Baldassari (1989).
The present study builds on the catalogue from Sahal et al. (2010) using a decision table and local maregraphic 50 data adding events being recorded since then and adding unrevealed information concerning previously mentioned events as for example the 1960 Chile tsunami.

Methodology
This study is based on the USGS earthquake catalogue that provides accurate information on seismic events all around the World since January 1, 1900 (U.S. Geological Survey, 2019). A decision table has been created to 55 select within this database the events that were potentially tsunamigenic and with potential to reach the New Caledonia coastlines. The events extracted are then cross-compared to the reported tsunami from the NOAA NGDC tsunami database (and NOAA PTWC bulletin archives) and to local maregraphic data.

Data selection on magnitude criteria
The first step was to collect all the available earthquakes from the catalogue for the Pacific Ocean region. We 60 decided to select only the Mw > 6.3 events according to the global tsunami databases, like the Historical Tsunami Database for the World Ocean -HTDB/WLD (http://tsun.sscc.ru/tsunami-database/index.php) or the NOAA NGDC/WDS Global Historical Tsunami Database (www.ngdc.noaa.gov/hazard/tsu_db.shtml), which empirically show that there is no tsunami triggered by earthquakes of magnitude Mw < 6.3 (Tinti, 1991).
For information, Bolt et al. (1975) indicate that the maximum run-up for a tsunami generated by a Mw = 6.5 65 earthquake would be no more than 0.5-0.75 m and Walker (2005) shows that tsunamigenic earthquakes with moment magnitudes Mw ≥ 8.6 had all a Pacific-wide impact. On the 10th of January 2019, this collection represents 4902 Mw ≥ 6.3 earthquakes on the whole Pacific Ocean (considered box: 66.2°S, 62°N, 118.4°W, 297.8°W) since July 29, 1900.
Then it is important to select events able to trigger tsunami with sufficient energy to reach New Caledonia. Ward 70 (1980) has indicated that tsunami generation is dependent upon the following criteria: the faulting mechanisms (mainly dip and slip), the magnitude (energy release) and the epicenter depth (focal depth). Thus, we consider all the faulting mechanisms without any distinction, integrating the distance from the source to New Caledonia to the earthquake magnitude and epicenter depth in the decision table.
Notice that the global tsunami catalogue is considered as a whole, and not only in the Pacific Ocean region, 75 according to the fact that catastrophic events like the 2004 Sumatra (Indian Ocean) tsunami could be recorded by tide gauges all around the World (Titov et al., 2005;Rabinovich and Thomson, 2007).

The faulting mechanisms
Although thrust and normal faults are responsible for the majority of the strong subduction earthquakes and tsunamis, Tanioka and Satake (1996) have shown that in a specific case, i.e. when the rupture occurs on a steep 80 slope with a horizontal displacement significantly larger than the vertical displacement, strike-slip faulting is also able to trigger a tsunami. In addition, Legg and Borrero (2001) and Borrero et al. (2004) have also shown that tectonic events occurring on strike-slip faults with sinuous traces could trigger tsunamis by the effect of uplift and subsidence along compressional and extensional relays. Thus, we decided to consider all faulting mechanisms because they are all potentially able to exhibit a vertical component of ocean bottom disturbance, 85 the only parameter required to disturb the water column and generate a tsunami.

Relationship between earthquake magnitude and focal depth and tsunami generation
We use the data from NOAA NGDC/WDS Global Historical Tsunami Database to plot the earthquake focal depth as a function of magnitude for 440 tsunamigenic earthquakes around the World from January 1, 1970 to August 24, 2018 (Fig. 2). Only 47 (10.7 %) of these events have Mw < 6.3 and 9 (2.0 %) have been located at a 90 depth of more than 100 km. Considering that these events are not located within the Pacific Ocean and/or they did not produce a sufficient tsunami to be recorded in New Caledonia, we decided to look only at earthquakes of magnitude Mw > 6.3 and focal depth < 100 km in the following. Note that these values are consistent with different early warning systems criteria (Tinti, 1991;UNESCO/IOC, 2009).

95
The tsunami amplitude, its wavelength and frequency components, directly linked to the earthquake magnitude and focal mechanism and rupture dynamics, determine the extent of the impacted zone: local, regional or oceanwide impact. In fact, a tsunami triggered by a landslide or a moderate earthquake is more inclined to dispersive phenomenon than a tsunami triggered by a large earthquake because dispersion is directly linked to the wavelength and frequency content, the water depth and the propagation distance (Glimsdal et al., 2013). In fact, 100 smaller the source is, quicker is the energy decay and finally the disappearance of the tsunami waves over the time and distance from the source (Rabinovich et al., 2013). For example, Tanioka et al. (2018) show the role of this dispersion phenomenon when looking at the DART buoy record of the tsunami triggered by the Mw 6.9 2016 Salvador-Nicaragua earthquake: in that specific case, the linear long waves theory overestimates considerably the numerical modeling results in comparison to the results obtained by the help of linear 105 Boussinesq equations, taking into account dispersion effect.
We decided to make an event sorting upon a magnitude and distance criteria of M = 7.5 and D = 2500 km because: i) the local and regional sources (from the Solomon, Vanuatu and Tonga-Kermadec subduction zones) are located within a circle of ~2500 km radius (Fig. 3) and ii) the far-field potential tsunami triggered by earthquakes of moment magnitude below 7.5 could not be recorded in New Caledonia according to historical data, and far-field sources could only be located between 4400 km (for the Mariana Trench, offshore Guam) and more than 10000 km away from New Caledonia (for Chile subduction zone).

Construction of the decision table
The decision table is based on the previous explanations and lays down the rules to keep or reject a considered event from the USGS catalogue with respect to specific conditions on three different parameters: the earthquake 115 magnitude (M), the focal depth (F) and the distance between the source and New Caledonia (D). Four cases are considered to select whether an event is kept or rejected. They are summarized on figure 4.

Sea-level data
The tide gauge stations of New Caledonia are located on figure 5. All the tide gauges were installed along the east coast of the Grande Terre of New Caledonia and in the Loyalty Islands, except the historical tide gauge in 125 Nouméa (Chaleix then Numbo), the capital located on the west coast of Grande Terre. Instrumental records of hourly sea level for Nouméa has been back extended to 1957 (Aucan et al., 2017a), and in the present paper we show unpublished records of digitized high frequency sea level data.
All the tide gauges started recording high frequency sea level (with sampling rates faster than 5 minutes) only after 2010 or later. The tide gauge characteristics are summarized on table 1.

130
In addition to these tide gauges, several pressure gauges have also been installed by IRD in Poindimié, Ouvéa and Uitoe (Fig. 2) within the framework of the ReefTEMPS project (Varillon et al., 2018) or the EMIL project (Aucan et al., 2017b). Their characteristics are also summarized in table 1.
Tide gauge and pressure data were detided by removing the predicted tide. The predicted tide was calculated with an harmonics analysis of the entire dataset available for each site, all of which were longer than 6 months.

Tsunami catalogue
The decision table has been able to extract 967 events (Mw ≥ 6.3) from the USGS earthquakes catalogue between January 01, 1900 and January 10, 2019.

Events after September 29, 2009
3.1.1 Extracted events present in the NOAA NGDC catalogue and recorded on tide gauges  Kirakira, Solomon, earthquake of December 8, 2016, could also trigger tsunamis that would be drowned within the main shock tsunami signal. Also small tsunami signals could be covered by the background noise, especially the coastal or offshore infragravity waves (Stephenson andRabinovich, 2009, Aucan andArdhuin, 2013).

Events before September 29, 2009
For the period before September 29, 2009, Sahal et al. (2010 have collected 18 events, including 12 events 190 related with certainty to an identified earthquake. These 12 events are detailed on table 3. There is a strong uncertainty on the accuracy of the 6 other reported events concerning the date as well as the source. 11 out of the 12 seismic events reported by Sahal et al. (2010)   111 events to the NOAA-NGDC tsunami database, it appears that 26 earthquakes triggered a tsunami on the Pacific Ocean, either local, regional or transoceanic. And from these 26 events, only 4 have been reported in New Caledonia according to Sahal et al. (2010).

Individual events during the 2009-2019 period
During this period, some events are particularly interesting as their records demonstrate the importance of local 215 tide gauges and pressure sensors on tsunami hazard studies.

February 6, 2013 Solomon tsunami
The tsunami generated by the Santa Cruz, Solomon, Mw 8.0 earthquake of February 6, 2013 at 01:12:25 UTC was recorded at the Lifou tide gauge near 3 a.m. UTC (2 p.m. local time) and well observed by local witnesses (Fig. 6). The two pictures shown on figure 6 have been taken during the first wave maximum (a) and at the 220 following minimum (b).

September 16, 2015 Chile tsunami
On September 16, 2015 at 22:54:32 UTC a magnitude Mw 8.3 earthquake in the region off Illapel, Chile triggered a transoceanic tsunami. After hours of propagation, it reached most of the South Pacific Ocean tide gauges. In New Caledonia it has been recorded on permanent tide gauge stations and pressure gauges about 16 225 hours after the earthquake as shown on Figure 7. This event is particularly interesting in the sense it confirms clearly that tsunamis are amplified near the Ouinné tide gauge, more than at the other gauges (except Hienghène where unfortunately the beginning of the record is missing). At Poindimié two pressure gauges installed outside of the lagoon (Poindimié_Fourmi) and inside the lagoon close to the shore (Poindimié_Tieti) also recorded the tsunami. Data from the two gauges shows the wave shoaling probably also the amplification (up to 5 times) due 230 to resonance inside the lagoon. Also, during this event a pressure gauge was rapidly installed in the Chaleix Naval Base at the location of the discontinued Chaleix tide gauge to compare the recorded signal to the Numbo tide gauge signal. The recorded signal at Chaleix was nearly three times higher than at Numbo's despite the close proximity of the two sites (for location details, see Aucan et al., 2017a).

235
Another example is given by three earthquakes that occurred during the November 19, 2017 seismic crisis located East of Mare Island: two very small tsunamis have been triggered by the Mw 6.3 and 6.6 foreshocks of the Mw 7.0 earthquake, which triggered a tsunami reaching a maximum amplitude of 0.8 m at Ouinné tide gauge. Despite their small amplitude, they are very well recorded and shown on figure 8.

240
On December 5, 2018, a magnitude Mw 7.5 earthquake occurred at 04:18:08 UTC in the South of the Vanuatu subduction zone. Widely felt by the population in New Caledonia but also in the Vanuatu, a tsunami was soon recorded, firstly on the Loyalty Is. tide gauges (Maré and Lifou) and on all the other tide gauges within 1h30 after the main shock (Fig. 9). As it reached sometimes more than 1 m high according to eye-witnesses, this tsunami was also observed by numerous people for example in Yaté, close to Ouinné, on the Southeast coast of 245 Grande Terre (Fig. 10a), and on the East coast of the Isle of Pines around the Méridien Resort and the Natural Pool touristic site where people have been evacuated (Fig. 10b).

Additional information on previously reported events
Five other events could be likely added to the herein catalogue. They come from testimonies and regional records (some of them have been already discussed in Sahal et al. (2010)).

Testimonies of tsunami events with unspecified date and not closely linked in time with any earthquake
The «1936» event reported in Northeast of the Grande Terre (north of Hienghène) with a 2 to 3 m runup could be due to a local landslide or possibly may be linked to the July 1934 North Vanuatu earthquake (Mw 7.8) which has triggered a tsunami observed in the same region (Hienghène-Touho).

255
The «May-July 1942 or 1943» large wave reported in Hienghène (2.5 m runup) and possibly the flooding «around 1940» reported on Isle of Pines (2 m runup) could be attributed to the same major event (although link between these reports is uncertain). No link could be made between these time periods and any earthquake.
These events can be the result of landslides.

260
The «December 1950 -February 1951» swelling and tidal wave reported at different localities by Sahal et al. (2010) along the east coast of the Grande Terre (Hienghène, Poindimié, Ponérihouen, Canala) and on Isle of Pines (Fig. 5) could be linked either to the December 2, 1950 South Vanuatu large earthquake (Mw 7.9) which is one the largest events located close to New Caledonia and having generated a tsunami observed in Port Vila, Vanuatu (Louat and Baldassari, 1989) or to the February 26-27, 1951 storm tide.

May 22, 1960 Chile tsunami
Although it has been already included within the catalogue from Sahal et al. (2010), the Great 1960 Chile tsunami was only reported through witness observations. Here we present a historical maregraphic record 275 recovered in the SHOM archive, that shows this transoceanic tsunami recorded in Nouméa by the Chaleix tide gauge (Fig. 11). The tsunami amplitude is about 30 cm for the two primary waves.

Limitations of the methodology
This methodology using an extraction decision table is based only on the reported or recorded tsunamis 280 generated by earthquakes and thus, do not considered tsunami triggered by landslides or, less frequently, by volcanic eruptions, representing respectively about 7% and 5% of the reported tsunamis of the NOAA tsunami database according to Harbitz et al. (2014). In the available data, there is no evidence of tsunamis related to active volcanism or landslides. Anyway, active submarine volcanoes exist in the neighborhood of New Another limitation comes from the fact that we do not consider the tsunami amplitudes in this study. In fact, it is very difficult to estimate a maximum value for each event in New Caledonia, because of obvious lack in field observations for each one and absence of tide gauge in specific places, like the Isle of Pines where important 290 tsunami run-ups have been reported for at least the December 5, 2018 event. A tsunami could have been lowly recorded by a tide gauge located in a protected area, for example in Nouméa harbor, although it has a strong impact on an exposed coast, for example on the East coast of the Isle of Pines. In addition, as detailed by Ioualalen et al. (2017), resonance phenomenon seem to play a predominant role on the tsunami behavior, especially in the Loyalty Is., depending directly on the source location and geometry. So, information concerning 295 maximum observed amplitudes mentioned in table 2 and 3 gives just an idea of what happened during the reported events.

Contributions for tsunami hazard assessment and risk management
The 25 events are added to the 12 events from Sahal et al. (2010), leading to a list of 37 tsunamis reported or recorded for New Caledonia over the last 144 years. Besides the 1875 event (no exact location available), the 36 300 earthquake epicenters are shown on figure 12. As expected, it highlights 5 different tsunamigenic zones able to trigger tsunamis toward New Caledonia: locally, the Vanuatu subduction zone is responsible of 17/37 tsunamis, i.e. 45.94 %; at a regional scale, the Tonga-Kermadec subduction zone triggered 3/37 tsunamis, i.e. 8.1 % and the Solomon / Papua New Guinea subduction zone is responsible of 9/37, i.e. 24,32 %; and at an ocean scale the transoceanic tsunamis represents 8/37 events, i.e. 21.62 %.

305
Besides the December 17, 2016 PNG Mw 7.9 earthquake which hypocenter has been located ~100 km deep, the other earthquakes are located not deeper than 50 km.
Concerning tsunami amplitudes, their observed range varies to a few centimeters to several meters. Local tsunamis issued from South Vanuatu earthquakes impact mainly the Loyalty Is. and the south-east coast of Grande Terre (including the Isle of Pines) and are the most frequent and stronger. The north-eastern part of 310 Grande Terre is more impacted by regional tsunamis coming from the North (Solomon and north Vanuatu subduction zones). Transoceanic tsunamis are also important to be considered in New Caledonia, able to produce locally wave amplification of 1-2 m high.
A graphic representation of the 36 earthquakes showing magnitude (Mw) as a function of the distance between the epicenter location and the center of New Caledonia highlights 3 different groups (Fig. 13): the group of 8 315 earthquakes on the right corresponds without surprise to the transoceanic tsunamis; the group on the left corresponds to the local events from the southern Vanuatu subduction zone and the central group corresponds to the regional events (Solomon, Northern Vanuatu and Tonga). In terms of risk managing, this study brings new constraints for alert thresholds: -In a local field, with an epicenter located within a distance less than 500 km, only a magnitude Mw ≥ 6.3 320 earthquake is able to trigger a tsunami that could be reported along New Caledonia coastlines.
-At a regional scale, i.e. at a distance of more than 1000 km, only the earthquakes with magnitude Mw ≥ 6.7 would be considered as potentially hazardous for New Caledonia in terms of tsunami waves and currents.
-At a far field location, i.e. at a distance of more than 6000 km, only earthquakes with magnitude Mw ≥ 7.7 would be considered as potentially hazardous for New Caledonia in terms of tsunami waves and currents.

Conclusion
This study allows to complete the tsunami catalogue of New Caledonia with 25 new events of seismic origin for the period between September 30, 2009 and January 10, 2019: 19 already identified in the NOAA NGDC tsunami catalogue and 6 others recorded on New Caledonia gauges but not reported either in the NOAA NGDC catalogue nor within the NOAA PTWC bulletins. The New Caledonia tsunami catalogue is now reaching a number of 37 tsunamis. It also emphasizes that there is a considerable lack of tsunami information in New Caledonia for the pre-September 2009 period concerning medium-magnitude events, due to the fact that there was less or no tide gauges and DART buoys able to record small tsunamis. Note that there is no study dealing with paleotsunami in New Caledonia.
Finally, this study highlights clearly the value of tide gauges records, including old paper ones, and the necessity 335 to settle the gauges in well identified locations, i.e. not always in sheltered areas but more in places facing main tsunami pathways. It also brings to light the necessity to add more sensors in exposed areas like on the East coast of the Isle of Pines.