Subject Areas: Atmospheric Sciences, Geomorphology, Oceanology
1. Introduction: French Polynesia and the Australs Archipelago
French Polynesia is an overseas autonomous territory of France in the South Pacific, about halfway between South America and Australia  . It is composed of 118 high islands and low atolls (67 are inhabited) stretching over an expanse of more than 2000 km (55% of the surface of the United States), but its total land area is only 4167 km2 (one-third the size of Connecticut). The total population of French Polynesia is 270,000, with 80% of Polynesian ethnicity, 10% of Chinese ethnicity and 10% of Caucasian ethnicity. French Polynesia is divided into five groups of islands: The Society, Tuamotu, Gambier, Marquesas and Australs archipelagos. Tahiti, which is located within the Society Islands, is the most populous island (180,000 inhabitants). The Australs Islands  are the southernmost group of islands in French Polynesia. They consist of two separate sub-archipelagos, namely in the northwest the Tubuai islands consisting from west to east of the Maria atoll, the high islands Rimatara, Rurutu, Tubuai Island proper and Raivavae, and in the southeast the Bass Islands composed of the main island of Rapa Iti and the small Marotiri rocks. The Australs Islands population is about 6300 and has a land area of almost 150 km2. Tubuai, the main island of the Australs Archipelago, is located at 23˚23'00"S, 149˚27'00"W, 640 km south of Tahiti. Tubuai Island sustains a population of 2170 people on 45 km2 of land. Its highest summit, Mount Taita’a, culminates at 422 m. The island is ringed by a lagoon and coral reef (see Figure 1). Due to its southerly position, Tubuai has a temperate to tropical climate, with temperatures ranging from 18˚C to 29˚C. Precipitations are intense and frequent in any season, with an annual average of 1888 mm. The island of Tubuai is one of the French Polynesia islands most affected by TCs (nine events in the last seventy years)  -  . In this paper, we describe the impact of the TC that hit this island on February 5th, 2010. The sea level surge (up to 1.76 m) was the main cause of damage, as it washed off all the northern coast of the island.
2. Cyclogenesis in the South Pacific
Since 1931, 47 tropical depressions and 34 TCs have hit the islands of French Polynesia, with a number of storms ranging between a minimum of 3 and a maximum of 17 per year. The majority of TCs occur during the hot and wet season, from November to April. TCs are most likely to form in January and February, at the peak of the southern hemisphere summer. The main factor in TC formation is the South Pacific Convergence Zone (SPCZ), where the south-east trade winds, from transitory anticyclones to the south, meet with the semi-per- manent easterly flow from the eastern South Pacific anticyclone. TCs usually occur west of 160°W between 5°S and 20˚S. Nonetheless, during El Niño years, TC formation occurs further east, with some tracks extending as far as 135˚W (see Figure 2).
Earlier studies  have demonstrated the influence of El Niño Southern Oscillation (ENSO) on TC activity in the southwestern tropical Pacific basin, particularly as a function of factors associated with the Southern Oscillation (SO), such as Sea Surface Temperature (SST) and the Southern Oscillation Index (SOI)  .
3. The TC Oli Event
On January 29th, 2010, convective clusters organized into a large depression 3000 km from Tahiti, centered roughly 300 km north-west of Fiji, in the South Pacific Convergence Zone (Figure 3). On February 1st the
Figure 1. The Island of Tubuai (45 km2) is part of the Australs Archipelago of French Polynesia. It is located 640 km south of Tahiti. The Tubuai tide gauge and weather stations are located on the north shore of the island (23˚20'30"S, 149˚28'30"W)  .
Figure 2. TC tracks in the South Pacific Ocean from 1944 to 2011  . The island of Tubuai is indicated by the small triangle in the figure.
Figure 3. The track of TC Oli in French Polynesia. On January 29th, 2010, convective clusters organized into depression 3000 km from Tahiti, centered roughly 300 km north-west of Suva, Fiji, in the South Pacific Convergence Zone. On February 1st the depression intensified north of the Cook Islands into moderate tropical depression and was called Oli. Oli decreased in intensity on February 7th in the south Pacific seas, after hitting the Island of Tubuai (Australs Archipelago) on February 5th.
depression intensified north of the Cook Islands into a moderate tropical depression and was named Oli. It traveled 5000 km in the southwestern Pacific Ocean and through French Polynesia from February 1st to 6th, 2010 (Figure 3). The center of Oli was located south-west of the Leeward Islands, 160 km from Maupiti on February 3rd in the afternoon and 300 km from Tahiti on the night of February 3rd. TC Oli has been the most destructive TC in French Polynesia since TCs Orama (1982), Veena (1983), Osea (1997) and Kim (2000)  . It is the first severe tropical cyclone (Category 4 on the Saffir-Simpson scale) within the central basin of the South Pacific Ocean.
Oli intensified again on February 4th. The eye of the hurricane made landfall on the island of Tubuai on February 5th at about 3h00 am, local time (13h00 UTC). Preceded by a heavy swell, waves up to 8 m high battered the eastern side of Tubuai provoking a storm surge of about 2 m over-coast. The surge flooded the northern part of the island which is the most urbanized and populated area. Beaches retreated from 2.06 to 9.75 m, and the vertical lowering ranged between 27 and 40 cm  . According to the location, the volume of sand moved is estimated to be between 0.5 and 3 m3/m. The over wash deposits extended inland as far as 150 m  . The strong east wind weakened around 2h00 am (12h00 UTC), then resumed again in the opposite direction around 7h00 am (17h00 UTC). The Tubuai weather station recorded a minimum pressure of 957 hPa with a maximum mean wind of 100 km/h and gusts exceeding 180 km/h (Figure 4). There were no human casualties, but 200 houses
Figure 4. Meteorological and Zenith Total Delay (ZTD) data recorded during the landfall of the TC Oli on the Island of Tubuai (dark gray: 12h00-17h00 UTC). The meteorological station of Météo-France recorded temperature, wind intensity, wind direction, pressure and rainfall data with hourly sampling. The tropospheric mean temperature Tm (fourth line) is estimated from the surface temperature Ts and the physical characteristics of the TC. All the sensors operated normally during the cyclone landfall except the temperature sensor which ceased functioning on February 5th. The surface temperature record was reconstructed thereafter, see paragraph 4.2.
were destroyed and 172 inhabitants were made homeless  . Moreover, TC Oli was responsible for important economic losses in Tubuai (USD 11 million estimated costs), especially for farmers, whose fields were flooded by seawater. Overall in French Polynesia, the storm caused a total of USD 13 million in damages  .
4. Measurements Taken during the Course of the Cyclone
4.1. Tide Gauge Data
Although waves reached up to 8 m during the TC, the tide gauge continued recording throughout. Sea level measurements were obtained as 60 s averages both from the radar sensor which is located about one meter above the average sea level and pressure sensor which is located about one meter below the average sea level. For more information on the tide gauge network in French Polynesia, we refer the reader to  .
4.2. Meteorological Data
The meteorological station of Météo-France is located 2 km from the tide gauge along the coastline (see Figure 1). Data recorded every hour are wind speed and azimuth, rainfall, surface pressure and surface temperature (respective accuracies of 0.5 hPa and 0.5 K). All the sensors operated normally during the cyclone landfall except the temperature sensor which ceased functioning on February 5th. A least-squares interpolation is used to extend the temperature series later than February 5th. The least-squares estimates were obtained with diurnal, semi-diurnal and forced annual harmonic components from the Météo-France station data over a window of 2 months (January to February) centered on February 5th, with an rms deviation of 0.93 K.
4.3. GPS Data
The GPS observations from the GPS tide gauge station consisted of data streams of 30 s undifferentiated dual- frequency carrier-phase and pseudorange measurements from six to eight satellites. The GAMIT software  was used to process the data, along with precise orbits solutions from the Scripps Orbit and Permanent Array Center (SOPAC) to estimate zenith total delay (ZTD) with an hourly sampling data rate at the tide gauge site  . The processing window was 24 hours wide and stepped forward in increments of 12 hours. The ZTD data rms deviation was found to be less than 7 mm.
5. Data Analysis
5.1. Surface Winds
Figure 4 shows the variations of wind intensity and direction between February 3rd and 8th, 2010. Outside of the TC event, the dominant wind is east-southeasterly with speed of about 18 km/h (5 m∙s−1). On February 4th, as the TC is approaching Tubuai, the direction changes to north-east and the wind speed increases gradually to reach 90 km/h (25 m∙s−1). At the time of arrival of the storm eye on Tubuai, we can see a lull in wind intensity with a sudden change in direction towards north-west. However, the hourly sampling rate does not allow us to fully quantify this reduction since the drop is only 36 km/h (10 m∙s−1). The pressure gradient at sea level near the center of Oli is very large, as shown by the pressure curve in Figure 4. It is the large magnitude of the pressure gradient that accounts for the high wind velocities. Indeed, we observe a relatively constant variation with low diurnal oscillation before and after the TC and a sharp drop in pressure during its passage with a minimum value of 957 hPa on February 5th at 13h00 UTC when it was located at the station.
5.2. Precipitable Water (PW)
Precipitable water vapor is useful in predicting heavy rainfall during extreme weather conditions. However, our knowledge of the determination of PW from GPS data during a tropical cyclone remains patchy   . The determination of PW is dependent on the determination of the mean tropospheric temperature Tm  . Even for extreme weather conditions such as tropical cyclones, the mean tropospheric temperature Tm can be determined from only the surface temperature Ts  . We used this method in this paper and obtained an uncertainty in Tm of about 1.9 K, which corresponds to a 1% uncertainty in PW  . Figure 5 shows the variations of the estimated PW with time from February 1st to 8th. On the long term, the normal PW level is about 40 kg/m2 (see
Figure 5. PW recording during the landfall of the TC Oli. One can see in this time series a drying event followed by a rapid increase, then a rapid decrease shortly after the landfall. We also find sub-peaks on the 2-day PW high, suspected to be related to small-scale variability in the PW field (dark gray: from noon to 17h00 UTC).
Figure 5). Before the cyclone hit Tubuai, the PW decreased to 20 kg/m2 on February 2nd (see also Figure 5). Such decrease has been observed in similar conditions  , but not consistently  . We believe that the decrease might be caused by the TC to compensate subsidence in the mid-troposphere. From February 2nd to 5th the PW shows a gradual increase to values higher than 80 kg/m2 (at midday), which is consistent with previous studies   . The increase of PW as the TC approaches Tubuai correlates with the increase of rainfall as recorded by the rain gauge. In our case, a maximum of 80 kg/m2 PW corresponds to about 30 mm/h in precipitation rate (see Figure 4 and Figure 5).
5.3. Storm Surge
The sea level variations during the passage of the TC Oli, available from the tide gauge, are shown in Figure 6. The pressure gauge recorded a maximum rise of 150 cm. In order to analyze the signal associated with TC Oli, we separated the cyclone storm surge height from the lunisolar tide height  . The storm surge is the result of five components: the pressure effect, the direct wind effect, the effect of the Earth's rotation, the effect of waves, and the rainfall effect. The effect of the waves is negligible because of the steep external slope of the island. We assume that the rainfall effect is also negligible because of the small size of the island, however, we are not sure about the impact of the lagoon, and as we have seen, the maximum of the rain rate is coincident with the maximum of the sea level rise (Figure 4 and Figure 6). The effect of the Earth’s rotation is dependent on the orientation of the coast, and was included to model the flooded areas on the island  . The storm surge around tropical volcanic islands  is generally low compared to storm surge observed along continental coastline, where sea level rises well above 6 m  . Unlike a storm surge that affects continental coasts, the contribution of the wind in a storm surge on island is not dominant on the inverse barometer effect. However, the lagoon around an island may induce resonances, in which case the reef barrier may act as either a natural flood protection or as a flood enhancer depending on the incident wave conditions  . The bathymetric depth around these islands sharply increases from the oceanic floor (4000 m) to the edge of the lagoon (Figure 1) and the land area of the islands is relatively small (45 km2 in Tubuai). The influence of the island bathymetry on the variation of sea level is
Figure 6. The top figure shows the time series of sea level measured (gray line) and modeled (black line) and the tides (dashed line) estimated during the landfall of the TC Oli (dark gray: from noon to 17h00 UTC). The bottom figure shows the time series of the difference between the measurements and the sea level model.
Taking into account the relative contribution of all these components, we used a simple linear model  to model the storm surge S in meters with respect to the barometric pressure P as S = −2.72 (P ? 1005.5), with P expressed in hPa  . Our result (Figure 6) shows a good agreement between the modeled sea level and the data. The residuals show that up to 40 cm of the sea level variation is due to the wind effect, which mostly happened prior to the TC Oli landfall.
The storm surge due to Oli roughly coincided with the time of landfall, at around 12h00 UTC. This high sea level episode lasted 36 hours, with a maximum from 12h00 to 17h00 UTC on February 5th (see Figure 6). Taking also into account the coast orientation with respect to the TC track, we estimate that the sea level surge varied from a few centimeters on the southern shores of the island to about 2.5 meters on the northeastern coast. We propose in Figure 7 a map of the flooded areas of the island, with a comparison with the “ground truth” established by the Health Division of the municipality of Tubuai. The north coast was strongly flooded, because of its low topography and the fact that the sea level surge was higher in this area. The north-east region, which is largely used for agriculture was particularly affected from an economic point of view. The south-western and southern part of Tubuai was almost untouched by the flooding.
These floods had an important geomorphic impact. The mean horizontal retreat was important in the northern and eastern part of the island (9.75 m), medium on the north-west coast (2 m) and not apparent in the south coast  .
Figure 7. Representation of submerged areas estimated from our model (gradual gray areas) and established by the health division of the municipality of Tubuai (hatched areas), on the Island of Tubuai, during the landfall of the TC eye (February 5th, 14h00 UTC).
Our analysis has contributed to a better understanding of the impact of a TC on a small island and we are not aware of any other study of this type on small island, probably because it is always a major challenge to run equipment on an isolated island. The land based equipment combined with GPS data allow us to demonstrate a good correlation between the water vapor content and rainfall rate as also shown by   . The tide gauge data shows that for a small island the effect of the barometric pressure is far greater than the effect of the wind.
Oli is classified as one of the most intense TCs that have impacted French Polynesia in the last forty years. We have been able to describe precisely the landfall of this TC on the island of Tubuai in terms of meteorological time series. One key parameter to model TCs and atmospheric rivers  is the integrated precipitable water vapor contents of the atmosphere (PW)  . In our case, a good correlation with the maximum rainfall is observed. One striking feature of the PW time series is the low values recorded before the landfall. Due to the morphology of the island, we showed that the sea level increase is mostly due to the barometric effect. The wind effect is mostly visible before and when TC Oli hit Tubuai. Our results are coherent with ground observations.
We thank Mr. Yann Dupont, Mr. Yves-Marie Tanguy and Ms. Marie Protat, successive managers (2007-2013) of the local representation in French Polynesia of the “Service Hydrographique et Océanographique de la Marine (SHOM)”, for their help in setting up and maintaining the tide gauges network of the University of French Polynesia (UPF), Mr. Maxence Jouannet and Mr. Pascal Mainguy, successive directors (2006-2013) of the “Défense et Protection Civile de la Polynésie française” and National Disaster Risk Manager Officers (NDMO), for their help in the administrative intricacies. We thank the “Haut-Commissariat de la République en Polynésie française” and the “Gouvernement de la Polynésie française” for their support. Funding was provided by the “Contrats Etat-Polynésie française” in 2007 and 2009 (expertise by the French Agency for Research (ANR)), and by the “Fonds Pacifique” in 2009. The housing of the tide gauge in Tubuai was built by the “Groupement du Service Militaire Adapté de Polynésie française”. Additional funding was also provided by SHOM, UPF and the French Space Agency (CNES). The Geodesy Observatory of Tahiti is an observation service of the University of French Polynesia, with contributions from CNES and NASA (US National Aeronautical and Space Agency). We also thank the climatology division of Météo-France Polynésie for providing us with meteorological data.
 Dupon, J.-F., Bonvallot, J., Vigneron, E., Gay, J.C., Morhange, C., Ollier, C., Peugniez, G., Reitel, B., Yon-Cassat, F., Danard, M. and Laidet, D. (1993) Atlas de la Polynésie franç ;aise. IRD Editions, Paris.
 Salvat, B., Bambridge, T., Tanret, D. and Petit, J. (2015) Environnement marin des î ;les Australes, Polynésie franç ;aise, Institut Récifs Coralliens Pacifique, CRIOBE and The Pew Charitable Trusts.
 Larrue, S. and Chiron, T. (2010) Les îl ;les de Polynésie française face à l’aléa cyclonique. Vertig O, 10, No. 3.
 Adapted from
 Chand, S.S. and Walsh, K.J.E. (2009) Tropical Cyclone Activity in the Fiji Region: Spatial Patterns and Relationship to Large-Scale Circulation. Journal of Climate, 22, 3877-3893.
 Troup, A.J. (1965) The Southern Oscillation. Quarterly Journal of the Royal Meteorological Society, 91, 490-506.
 Etienne S. (2012) Marine Inundation Hazards in French Polynesia: Geomorphologic Impacts of Tropical Cyclone Oli in February 2010. In: Terry, J. and Goff, J., Eds., Natural Hazards in the Asia-Pacific Region: Recent Advances and Emerging Concepts, The Geological Society of London Special Publication, London, Vol. 361, 21-39.
 Barriot, J.P., Serafini, J., Sichoix, L., Reymond, D. and Hyvernaud, O. (2012) The Tsunami of March 11, 2011 as Observed by the Network of Tide Gauges of French Polynesia. Journal of Marine Science and Technology (Taiwan), 20, 639-646.
 Song, D.-S. and Grejner-Brzezinska, D.A. (2009) Remote Sensing of Atmospheric Water Vapor Variation from GPS Measurements during a Severe Weather Event. Earth Planets Space, 61, 1117-1125.
 Liou, Y.A. and Huang, C.Y. (2000) GPS Observations of PW during the Passage of a Typhoon. Earth Planets and Space, 52, 709-712.
 Whiteman, D.N., Evans, K.D., Demoz, B., O’C. Starr, D., Eloranta, E.W., Tobin, D., Feltz, W., Jedlovec, G.J., Gutman, S.I., Schwemmer, G.K., Cadirola, M., Melfi, S.H. and Schmidlin, F.J. (2001) Raman Lidar Measurements of Water Vapor and Cirrus Clouds during the Passage of Hurricane Bonnie. Journal of Geophysical Research, 106, 5211-5225.
 Stern, D.P. and Nolan, D.S. (2012) On the Height of the Warm Core in Tropical Cyclones. Journal of the Atmospheric Sciences, 69, 1657-1680.
 Drews, C. and Galarneau, T.J. (2015) Directional Analysis of the Storm Surge from Hurricane Sandy 2012, with Applications to Charleston, New Orleans, and the Philippines. PLoS ONE, 10, e0122113.
 Brakenridge, G.R., Syvitski, J.P.M., Overeem, I., Higgins, S.A., Kettner, A.J., Stewart-Moore, J.A. and Westerhoff, R. (2012) Global Mapping of Storm Surges and the Assessment of Coastal Vulnerability. Natural Hazards, 1-18.
 Torres-Freyermuth, A., Mariño-Tapia, I., Coronado, C., Salles, P., Medellín, G., Pedrozo-Acuña, A., Silva, R., Candela, J. and Iglesias-Prieto, R. (2012) Wave-Induced Extreme Water Levels in the Puerto Morelos Fringing Reef Lagoon. Natural Hazards and Earth System Sciences, 12, 3765-3773.
 Dacre, H.F., Clark, P.A., Martinez-Alvarado, O., Stringer, M.A. and Lavers, D.A. (2015) How Do Atmospheric Rivers Form, in Diabatic Influence on Mesoscale Structures in Extratropical Storms (DIAMET) Special Collection. Bulletin of the American Meteorological Society, 1243-1255.
 Foster, J., Bevis, M., Chen, Y.L., Businger, S. and Zhang, Y. (2003) The Ka‘u Storm (Nov. 2000): Imaging Precipitable Water Using GPS. Journal of Geophysical Research, 108, No. D18.