Meteorological conditions are recorded by Météo France weather station, which is set on the south-east of Mangareva Island (Fig. 4), at an altitude of 91 m (Laurent and Maamaatuaiahutapu, 2019). It measured wind speeds and directions in hourly time steps. However, the station can be influenced by local orographic effects (Laurent and Maamaatuaiahutapu, 2019). For this reason, we decided to use reanalysis data from the ERA5 model (Hersbach et al., 2020) (Fig. 3). Data were extracted in a single location inside the lagoon (23∘09′11.6′′ S–134∘57′31.4′′ W) for the period June 2019 to February 2020 to match the instrument deployment period (Fig. 4).

Wind conditions observed during Leg1 are dominated by strong and short periods of SW winds probably related to local wind wave events. Conversely, Leg2 is dominated by trade wind (E to SE winds) with several SE wind events above 10 m s-1.

The RBRduet T.D sensor (RBR Ltd) is a compact logger ideal for long-term deployments and for providing measurements of temperature and pressure at high frequency. This instrument was moored on external reef slopes or inside the lagoon. Five loggers were anchored at a depth of between 9 and 13 m on five distinct sides of Gambier Islands (north-east, north, north-west, south-west, and south-east) to measure the incident waves reaching the reef crest. Post-processing provides significant wave height, mean wave period, and peak frequency. Inside the lagoon, four RBRduet T.D were moored at around 7–8 m to measure the water level and to deduce surge signals. For each station, a data logger was set-up to measure at 1 Hz intervals.

Pressure data were corrected from a constant atmospheric pressure value set to 1.01325 bar in order to avoid the influence of changes in weather conditions. To deduce wave parameters (significant wave height, peak frequency, and mean wave period), data were filtered using the Fourier transform to acquire a pressure spectrum (within the range 3–25 s). Then, the methods referenced in Aucan et al. (2017), also used in Aucan et al. (2021) and Andréfouët et al. (2023a), were applied using the linear wave theory with a homogenous cut-off frequency (set to 0.33) to filter the high- frequency spectrum. To calculate the water level, depth measurements were subtracted from the mean sea level (long-term depth-average of the temporal series). As a result, two output files were created, one at 1 h resolution containing wave parameters and another file at 1 min frequency with temperature and water level.

High accuracy (±0.002 ∘C) temperature data were recorded with SBE56 sensors designed by SEABIRD Electronics Inc. Instruments. These sensors were placed in each lagoon station at depths of approximately 2 and 20 m in order to detect vertical temperature stratification. The start time and configuration of all instruments were identical. The measurement interval was set to record every 1 min. Raw data did not need a processing stage and were directly converted into NetCDF files.

The ADCPs used during the Leg2 are of two types: the ADCP Sentinel V from Teledyne RD Instruments Inc. (TRD-I) and the Nortek Aquadopp current profiler with pressure sensor. Both instruments were bottom mounted, fixed on the nonmagnetic aluminum frames provided by the constructors, and faced upward (see photographs in Bruyère at al., 2023a, Fig. 8).

Regarding ADCPs, the two working frequency models Sentinel V50 (500 kHz) and Sentinel V20 (1000 kHz) were used to measure velocity along the water column, and the temperature and pressure at the sensor in its transducer head. ADCPs were set to measure in burst time: the V20 models measured burst each 10 min with 40 pings per ensemble. Conversely, the V50 models were recorded every 20 min with 180 pings per burst. For both models, cell size was fixed to a resolution of 1 m. The ADCPs V20 were moored at depths of 17 and 19 m and the V50s were fixed deeper, at 31 and 35 m.

Aquadopp current profilers (2 MHz version) measured three components (east, north, up) of current velocity data in shallow areas. The instrument settings were set to measure at 600 s burst intervals (10 min), each burst made of three pings. It resolved the entire water column, with a cell size of 50 cm. The ENU coordinate system was systematically used. Instruments were anchored at depths of 3.6 and 8.6 m.

For the current profilers, the near sea-surface cells contaminated by the acoustic sidelobe reflections were removed. Valid bins were then used to retrieve the magnitude and direction of the currents calculated using the zonal (u) and meridional (v) components. Pressure data were converted into depth by subtracting the depth-averaged value across the entire time series of pressure. No barometric correction was applied to depth. Processed profiles data were eventually converted into NetCDF files, which describe data at full resolution (i.e., each bin measurement is available).

The Marotte HS drag-tilt current meters manufactured by Marine Geophysics Laboratory of James Cook University allow temperature and uv components to be measured at the level of the instrument within a range of adequate conditions. Each Marotte HS used in Gambier was moored on a bottom structure and set to sample at 1 min intervals. With Python routines, vectors data (u,v) are converted in Clockwise from North convention to be consistent with oceanographic convention, and then speed and direction are deduced. Note that the Marotte HS moored at P01 was lost during Leg1.

The quality control procedure consists of visually checking each time series converted in NetCDF files with Python and Ferret and removing all remaining out-of-water data, or flag anomalies (such as out-of-range data or spikes). With this step, the final correct data coverage is determined and reported in the Global attributes of each processed file.

Amplitude of tidal constituents are presented in Table 2. Tidal analysis on Leg1 was made with the “t_tide” Python package (Pawlowicz et al., 2002). The sum of the six main harmonic constituents reaches 48 cm in oceanic station (O01). For O01 station, 57 % of water elevation is related to the semi-diurnal M2 harmonic amplitude, which represents 27.7 cm. Water-level oscillations in the ocean and lagoon are similar and synchronous because of the openness of the Gambier lagoon (Fig. 5). Compared with previous works on Tuamotu atolls (Dumas et al., 2012; Andréfouët et al., 2023a), where lagoon tide signal is highly attenuated, a difference of only a few centimeters in amplitude occur in the Gambier lagoon. It was less than 1 cm between the O01 and L01 stations but more than 2.6 cm between O05 and L03) (Table 2). Those results confirm the degree of openness of the Gambier lagoon to the ocean, and explain the tide-driven circulation inside the lagoon.

Waves around Gambier were recorded on the five forereefs equipped with the RBRduet T.D sensors. Leg1 deployment occurred during the austral winter season (June to October), where distant swells generated by southern weather events are dominant. The measurements confirmed the importance of these southern swells (Fig. 6a). The station O03 located on the south-eastern side of Gambier Islands recorded several events, with wave heights > 3 m. Conversely, the significant wave heights measured at the northern station O04 averaged only ∼ 1 m, except during a few events at ∼ 2 m (i.e., July 2019). Only the two southern stations (O02 and O03) captured waves above 3 m in height (Fig. 6b). The north-west side is the calmest area owing to its protection from E-SE trade winds, with wave height under 1 m 73 % of the time. For the O03 station this class of height (0–1 m) represented only 6.6 % of the records.

Stations O04 (north) and O05 (north-east) recorded in ∼ 80 % of the time peak wave periods (Tp) under 11 s. These stations are dominated by wind waves generated by E-NE trade winds. In contrast, O01, O02, O03 measured higher Tp, which means that wave regimes are rather dominated by distant swell (Fig. 6c).

Wave measurements during the Leg2 deployment in austral summer contrast substantially with Leg1 (Fig. 7a); in particular, the wave heights measured on the south sides of the island (O02 or O03) reached a height of 3 m only once (Fig. 7b). Wave heights between 1 and 2 m were dominant (> 50 % of waves) on the southern stations (O02 and O03) and on the eastern station (O05). However, the peak period broadly matched the winter season (Fig. 7c).

Based on Fig. 8, wave events recorded during Leg1 on the south-east forereef station (O03) are also observed inside the lagoon (L02 station), but with a reduction in wave amplitude likely due to wave-energy dissipation on the subtidal reef flats in the south (bottom friction), and the wave refractions on islands and very shallow or emerged reefs. The lagoon surge amplitude (L02) reacts in the same manner as surge measured in the oceanic station (O03). This weak difference shows that the water levels in the lagoon and ocean are homogenous.

Temperature in the Gambier lagoon followed, as expected, a seasonal evolution, with lower temperatures in July (around 22 ∘C) and a maximum reaching 28 ∘C in February (Fig. 9). During the austral winter season, lagoonal temperatures are lower than in the ocean by approximately 1 ∘C. From late August to October, temperatures are similar and start to deviate from each other in November. The ocean is then cooler than the lagoon during the summer period. During Leg1, in austral winter, the temperature appeared fairly stable, and rather cool, with values restricted between 22 and 24 ∘C. Between June and August, three synchronous (all stations) temperature drops are observed on the entire water column (bottom and surface) with temperature decreasing by 1 to 2 ∘C depending on the location. The return to previous values was more or less rapid also depending on the stations, with surface waters warming up faster than deep areas. This phenomenon can be primarily related to intense wind events (Fig. 3) or wave events (Figs. 6–7).

The overall temperature dynamic during Leg2 shows an increasing trend starting in November 2019 until February 2020, and a maximum delta > 5 ∘C is observed between the maximum and minimum temperature reached per station across the period. Similar to Leg1, there are, however, several temperature drop events synchronous for all stations, in particular in December 2019. The drop is recorded by all lagoon stations, from 2 ∘C for L01, L02, and L03 to 2.5 ∘C in L04 in 6 d, and can be assigned to periods of high south-east winds.

Depth-averaged water circulation observed in the Gambier lagoon shows that currents are dominated first by the tide, and are modulated by wind orientations and intensity. The ADCP_3, for instance, recorded a depth-averaged symmetric direction-oriented NW-SE related to the natural alignment of the channel formed between Mangareva and Taravai Islands and the regular effects of the tide (Fig. 10). The same directional symmetry occurs in ADCP_1 (Fig. 10).

Rikitea lagoon is influenced by water movements on its three sides. Although in- and outflows at ADCP_4 are balanced and seem primarily dominated by the tide, inflows are more dominant in Aqua_2, owing to the effects of both the tide and the south-east wind (Fig. 3). Conversely, outflows are more persistent at the Aqua_1 station, a feature that can be interpreted as the outflow compensating for the inflows elsewhere. The Rikitea sub-lagoon circulation has been specifically studied in Bruyère et al. (2023a).

Depth-averaged current speeds and directions measured from current profilers are presented in Fig. 11. Measurements were restricted between 1 and 12 December 2019, to highlight the influence of wind intensity and direction on currents. Indeed, a strong trade wind event (SE direction and approximately 12 m s-1) was observed between 7 and 14 December (Fig. 3). This event was recorded by current profilers; Aqua2 was influenced by wind intensity and direction. In contrast, current directions observed in stations ADCP_1, ADCP_3, and ADCP_4 follow tide cycles, but a small increase in speed is remarkable.