A Stable Isotope Sclerochronology-Based Forensic Method for Reconstructing Debris Drift Paths With Application to the MH370 Crash

Abstract

A flaperon belonging to Malaysian Airlines flight MH370 washed ashore on Réunion Island covered with the barnacle Lepas anatifera in July 2015, more than a year after the plane's disappearance. Here, we report the first high-precision δ¹⁸O_calcite versus temperature relationship for L. anatifera reared under laboratory conditions to unlock clues to the flaperon's drift path and origin. Using this experimental relationship and known growth rates for L. anatifera, we also demonstrate a new method for (a) converting δ¹⁸O data for one of the MH370 barnacles into a dated time series of sea surface temperatures (SSTs) experienced during the last part of the flaperon's drift and (b) identifying best fits between the observed flaperon SST time series and 50,000 SST histories generated from a particle-tracking simulation. Our new method identifies a flaperon drift path far south of a previous isotope-based reconstruction. We conclude with specific recommendations for using our method to continue the search for MH370 and other applications.

Key Points

• First experimentally derived sea surface temperature-δ¹⁸O_shell equation for the stalked barnacle, Lepas anatifera

• New numerical modeling method for reconstructing debris drift paths and origins from barnacle δ¹⁸O_shell data

• First application of these new tools to barnacle δ¹⁸O_shell data from missing flight MH370 to produce a partial drift reconstruction

Plain Language Summary

More than 8 years ago, on 8 March 2014, Malaysia Airlines Flight 370 departed from Kuala Lumpur, never to be seen again despite a 4-year extensive search using sonar imaging technology, submersible vehicles, drift models, and other high-tech methods. Pieces of plane debris were found across the Indian Ocean, with some confirmed to be of the missing plane. One of the MH370 flaperons, a part of the aircraft's wing, beached on Réunion Island with several generations of stalked barnacles attached to its surface, later identified as Lepas anatifera. At least some of the barnacles were attached and growing shortly after the crash. This study contributes the first experimentally derived equation relating oxygen isotope values of stalked barnacle L. anatifera shells to sea surface temperature during shell formation. We demonstrate how applying the new temperature-δ¹⁸O relationship to published data from small L. anatifera shells collected from the MH370 flaperon, combined with a novel particle-tracking simulation method, can be used to reconstruct the latter part of the flaperon drift path before beaching. This same method could be applied to the largest, oldest barnacles collected from the same debris to provide important information about the debris drift origin and location of the missing plane.

1 Introduction

Ocean circulation models used to reconstruct the sources of floating objects, such as plastic pollution (Lebreton et al., 2012), human remains (Mateus et al., 2015), sea turtles (Putman & Naro-Maciel, 2013), and plane crash debris (Trinanes et al., 2016), rapidly lose accuracy with increasing distance and duration of drift (Maximenko et al., 2018; Robel et al., 2011). To reconstruct longer paths, researchers are increasingly looking to detailed sea surface temperature (SST) archives encoded in the shell geochemistry of hitchhiking barnacles that colonize these floating objects as larvae (e.g., Detjen et al., 2015; Killingley, 1980; Killingley & Lutcavage, 1983; Magni et al., 2015; Pearson et al., 2020; Taylor et al., 2019). The ratio of oxygen isotopes (¹⁸O/¹⁶O) in barnacle shell calcite is temperature-dependent, with lower ratios in shell layers grown in warmer waters (Killingley & Newman, 1982). When the relationship between SST and oxygen isotope ratios in barnacle shells is known empirically, δ¹⁸O values measured from sequentially deposited shell layers can be converted into a detailed history of SSTs experienced by the barnacle during a drift, providing physical oceanographic clues to constrain where a floating object could and could not have been (e.g., Detjen et al., 2015; Killingley, 1980; Killingley & Lutcavage, 1983; Pearson et al., 2020; Taylor et al., 2019). A severe limitation of this approach is that individual SSTs do not have unique spatial solutions, especially on the scale of ocean basins over time. Simultaneously solving where and when a drifting object experienced each SST is thus the critical challenge in stitching together SSTs recorded from barnacle shells into a unique drift track leading back to a specific drift origin (Pearson et al., 2020).

One of the most recent and pressing needs for such a barnacle-based forensic reconstruction of a debris drift origin has been the search for missing Malaysian Airlines flight MH370, which disappeared with 239 passengers on 8 March 2014, after departure from Kuala Lumpur International Airport (Figure 1). One of the MH370 flaperons was recovered on Réunion Island on 29 July 2015, more than a year after the plane's disappearance, encrusted with multiple generations of the cosmopolitan stalked barnacle, Lepas anatifera, which is among the first organisms to colonize drifting objects in the open ocean and, thus, the species most likely to record a complete drift history and origin (Fraser et al., 2011; Magni et al., 2015).

Figure 1

Estimated flight path of flight MH370, search areas, and locations of identified debris.

From the earliest stages of the search investigation, particle tracking simulations and ocean drifter data were used to model debris drift paths after the crash (e.g., Davey et al., 2016; Griffin et al., 2017; Trinanes et al., 2016). However, despite these efforts, a multi-national effort to locate the missing plane has searched more than 120,000 km² but yielded no sign of its final resting spot (Australian Transport Safety Bureau, 2017; Sipalan, 2018).

The barnacles from the flaperon that washed ashore on Réunion were dead but still firmly attached with tissue inside and a strong odor, indicating that the debris must have washed ashore close to or on the date of the flaperon discovery (Poupin, 2015). Using a growth model for Lepas spp., Poupin (2015) estimated that the largest MH370 L. anatifera (∼36 mm capitulum length) individuals were 15–16 months old and conceivably recorded the complete SST drift history of the flaperon, including the drift origin. The French defense ministry, which secured the MH370 flaperon in a military laboratory in Toulouse, commissioned French scientists to develop a δ¹⁸O_shell-SST calibration for L. anatifera, enabling the first description of the SST drift history from the MH370 flaperon. To expedite the work, however, Blamart and Bassinot (2016) developed an approximated calibration, without any experimental controls, that involved comparing measured δ¹⁸O values of museum specimens of L. anatifera collected on known dates between 1984 and 2015 with multi-year, monthly average SSTs for each specimen's location. The resulting equation calculates SSTs with a relatively large standard error (roughly ±2°C) and several degrees warmer than expected based on preliminary studies of other stalked barnacles (Killingley & Newman, 1982). The first objective of this study was, therefore, to develop a more tightly constrained calcite-oxygen isotope-temperature relationship through geochemical analyses of L. anatifera shells grown in the laboratory under controlled experimental conditions.

Our second objective was to develop the first method capable of simultaneously solving the where and when of each barnacle SST, a major step toward reconstructing the entire MH370 flaperon pathway back to the crash origin. First, we use our new L. anatifera δ¹⁸O_shell–temperature calibration to convert Blamart and Bassinot (2016)'s published δ¹⁸O time series for one of the MH370 barnacles into a revised history of SSTs experienced by the MH370 flaperon during its drift. The French government provided Blamart and Bassinot (2016) with a subsample of small (<28 mm capitulum length) L. anatifera from the flaperon that recorded only the last few months of drift prior to stranding (Blamart & Bassinot, 2016; Poupin, 2015). Still, these isotope data represent an important segment of the actual flaperon drift and are invaluable for demonstrating how the new method could be used to reconstruct a complete drift track back to the crash origin.

Next, we use Poupin's (2015) length-at-age growth model (Equation S10 in Supporting Information S1) to assign specific dates to individual SSTs recorded in the barnacle time series. The date and SST of the earliest-formed sample are used to constrain possible colonization locations to a narrow band of water stretching across the Indian Ocean. A forward drift model is then run starting on this date with simulated drifters released across the constrained colonization zone; SST histories are recorded for each drift track until the flaperon's discovery on 29 July 2015. Finally, a dynamic time-warping (DWT) algorithm is used to identify individual drift tracks and origins among the 50,000 drifters whose recorded SST histories best reproduce the SST history obtained from the MH370 barnacle isotope profile. Our integrative approach is capable of reconstructing the entire MH370 flaperon drift trajectory back to the crash site location should the largest, oldest flaperon barnacles be made available for study. We conclude our methods development study with several important recommendations for future applications.

2 Materials and Methods

2.1 Experimental Design

Forty-six live juvenile L. anatifera barnacles were found attached to a freshly stranded piece of driftwood on Fanore Beach on the west coast of Ireland (53°7′16.36″N, 9°17′19.84″W) (Figure 2a, Figure S1 in Supporting Information S1) and transported alive to laboratory facilities in the Ryan Institute, National University of Ireland Galway. The driftwood was cut into three fragments, each with similar numbers of barnacles (N = 15, 15, and 16), and placed into three separate aquaria with a fixed temperature of 12°C. Barnacles were acclimated to these conditions for two weeks, at which time the temperatures were slowly increased to one of three treatment temperatures (14, 20, and 26°C) at a rate of 2°C every three days. Once treatment temperatures were reached, the barnacles were stained with a 250 mg/L concentration of tetracycline hydrochloride (cat no: T7660, Sigma-Aldrich, Wicklow, Ireland), which fluoresces under ultraviolet light, to mark the point of new experimental shell growth (Figure 2b). Water was changed during the 3-day staining, and fresh tetracycline hydrochloride was added every 24 hr. Salinity was constant at 35 PSU, within the precision of 21 refractometer measurements, in the 14 and 20°C aquaria and varied between 35 and 34.5 PSU in the 26°C aquarium.

Figure 2

(a) Juvenile Lepas anatifera barnacle showing anatomical parts referenced in the manuscript. (b) Same shell under ultraviolet light showing (a, b) = capitulum length, (c, d) = scutum length, (b–d) = shell size at collection (pre-treatment shell growth), (d, e) = luminescent zone on scutum marking the end of the acclimation period, (e, f) = new shell growth during controlled temperature experiment (see Tables S1 and S2 in Supporting Information S1 for exact measurements, and Figures S2 and S3 in Supporting Information S1 for growth rates).

Throughout the 3-week growth experiment, animals were fed freshly hatched Artemia (brine shrimp) nauplii (Z.M. systems, Winchester, UK) daily and grown on a 12:12 hr day/night light cycle. Aquaria lids were used to minimize evaporation and control the oxygen isotopic composition of aquaria water (δ¹⁸O_water). A 20%–30% water change using temperature-acclimated water was carried out at the end of the first and second weeks (days 7 and 14). Water samples were collected by direct immersion of a labeled 50 ml tube into aquaria before and after the first and second water change (days 7, 8, 14, and 15), as well as on the last day of the experiment (day 21) to monitor variation in δ¹⁸O_water values. Animals were removed from aquaria each week (days 1, 7, and 14) for no more than 5 min and imaged using a fluorescent stereomicroscope (Olympus). At the end of the third week (day 21), all animals were anaesthetized by cooling and oven-dried for storage at 40°C for 24 hr.

2.2 Sample Treatment

Barnacle capitulum and individual plate dimensions were measured using Onde Rulers (version 1.13.1) (Table S1 in Supporting Information S1). Barnacles were then submerged in laboratory-grade 6% sodium hypochlorite solution (NaOCl) for 24 hr to detach organic tissue from the plates, which were later rinsed with de-ionized water and oven-dried at 40°C for 24 hr. New shell growth was identified under U.V. light as the zone beyond the luminescent staining (Figure 2 and Table S2 in Supporting Information S1). In preparation for stable isotope analysis, a small portion of new shell growth was broken off with a scalpel and crushed into a powder using a mortar and pestle. One of the 46 barnacles was accidently broken into pieces too small to identify new growth accurately and was not analyzed.

2.3 Stable Isotope Analyses

Powdered 100 μg samples from 45 barnacles were dissolved in 103% phosphoric acid (H₃PO₄) at 25°C for 24 hr, and the CO₂ was analyzed using a Gasbench device coupled to a Thermo Finnigan Delta V Advantage Isotope-Ratio Mass Spectrometer (IRMS) in the Stable Isotope Laboratory at the University of South Florida. Water samples were equilibrated with CO₂ standard gas and analyzed using the same Gasbench device and IRMS. The precision was <0.06‰ (1σ) for δ¹⁸O_calcite and <0.07‰ (1σ) for δ¹⁸O_water. Stable isotopic ratios are expressed in standard δ¹⁸O notation:

$$${\delta }^{18}\mathrm{O}({\perthousand})=\left(\left(\frac{{R}_{\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}}}{{R}_{\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}}}\right)-1\ast {10}^{3}\right)$$$ (1)

where R is the ratio of heavy to light isotopes in the sample and standard. All values of δ¹⁸O_calcite are reported in per mil units (‰) with respect to the Vienna Peedee Belemnite (VPDB) isotopic standard, and values of δ¹⁸O_water are reported in ‰ with respect to the Vienna Standard Mean Ocean Water (VSMOW) isotopic standard. By convention, a small offset correction of 0.27‰ (Hut, 1987) was added to the temperature-δ¹⁸O relationship to adjust for differences in CO₂ derived from the equilibration of VSMOW with CO₂ and reaction of VPDB with H₃PO₄, both at 25°C (See Bemis et al. (1998) for details).

2.4 δ¹⁸O_calcite Versus Temperature Relationships

Following Bemis et al. (1998), a linear regression model was chosen to generate a temperature-δ¹⁸O relationship, treating the controlled sequence of water temperatures as the independent variable to minimize the variance in the measured δ¹⁸O_calcite values and rearranging the generated equation to solve for T°C.

The PAST software package (version 3.16, Hammer et al., 2001) was used to develop a linear least squares regression equation for the relationship of measured δ¹⁸O_calcite values with corresponding water temperatures using measured mean δ¹⁸O_water values from the aquaria. Analysis of variance and a Tukey's test were used to test for differences in δ¹⁸O_calcite values of scutum plates between treatments. Outliers were defined as measurements of more than two standard deviations from the mean (>2σ). A p-value of 0.05 was used as the cut-off to determine significant effects.

To compare the relationship between temperature and calcite-water fractionation developed in our new relationship versus other widely referenced temperature-δ¹⁸O_carbonate relationships (Table S3 in Supporting Information S1), SSTs were computed for each relationship using δ¹⁸O_calcite values measured in this experiment, and a constant δ¹⁸O_water value of 0‰ (VSMOW).

2.5 Particle Tracking Simulation

To demonstrate how oxygen isotope data from L. anatifera shells can be used to infer the origin and drift trajectory of the flaperon, we first used our experimentally constrained δ¹⁸O_shell–SST calibration to convert δ¹⁸O_shell data published previously by Blamart and Bassinot (2016) for specimen A2-G1 to SST. Using an approximate δ¹⁸O value of 0.4‰ for seawater in the Indian Ocean east of Reunion Island (Schmidt, 1998; Srivastava et al., 2007; Völpel et al., 2017), the last SST recorded in the A2-G1 isotope profile calculated by our relationship (24°C) matches 24–25°C SSTs recorded by the Global HYbrid Coordinate Ocean Model (HYCOM) around Réunion on 29 July 2015 (Figure S4 in Supporting Information S1), the date the MH370 flaperon was discovered and the barnacles likely died. Based on this evidence, we anchored the last SST of the A2-G1 time series to 29 July 2015.

Using a new capitulum-to-scutum-length regression developed from 45 L. anatifera in our collections at Fanore Beach, Ireland (Equation S9 in Supporting Information S1) and Poupin (2015)'s logistic age-at-scutum-length growth curve for Lepas spp. (Equation S10 sin Supporting Information S1), we converted each capitulum-length-at-sample-point datum in Blamart and Bassinot (2016) to days-of-growth-since-colonization. Based on the capitulum length at the last sample, we estimate an age of 154 days when shell growth ended.

We then assigned specific dates to each SST in the A2-G1 isotope profile by converting each length-at-sample-point datum in Blamart and Bassinot (2016) to days-of-growth-since-colonization using Poupin (2015)'s logistic age-at-length growth curve for Lepas spp. This approach yields an approximate age of 154 days at the last sample (already anchored to 29 July 2015) and, thus, an initial flaperon colonization date for A2-G1 154 days prior (25 February 2015). Using our conversion equation, we calculated the corresponding SST of the earliest δ¹⁸O_shell value of the A2-G1 profile and then located the spatial distribution of this SST (a narrow band) across the Indian Ocean on 25 February 2015 using the global HYbrid Coordinate Ocean Model (HYCOM, www.hycom.org). Any point within the spatial distribution of this SST band on 25 February 2015 was considered a possible location of the MH370 flaperon when it was colonized by A2-G1.

To identify the most likely point that A2-G1 colonized the flaperon (i.e., A2-G1's drift origin) within this SST band on 25 February 2015, we ran a particle tracking simulation with ICHTHYOP v3.3 software (Lett et al., 2008) coupled with the data-assimilative HYCOM Global Reanalysis 3.1 experiment sequence 53.x. The latter is a hybrid model, which constantly corrects itself to make the simulated ocean state more accurate with remotely sensed and in situ observations, using the Navy Coupled Ocean Data Assimilation system for data assimilation (https://www.hycom.org/dataserver/gofs-3pt1/reanalysis). Model output (current velocity and temperature) for the sea surface was extracted and was available at 3-hourly temporal resolution and 1/12° (latitude/longitude) spatial resolution. Fifty Thousand particles were released in the simulation in the Indian Ocean east of Réunion Island on 25 February 2015, evenly spaced along all possible drift starting points as defined by the starting SST (Figure 5b). Particles were restricted in vertical movement and dispersed at the sea surface only. Particle trajectories were computed using a Runge-Kutta fourth order numerical scheme and a turbulent dissipation rate of 1 × 10⁻⁹ (Monin & Ozmidov, 1981). The time step was set at 5-min intervals, and the position of the particles and SSTs experienced were recorded every 12 hr.

2.6 Dynamic Time-Warping

We used the dynamic time-warping (DTW) algorithm (Sakoe & Chiba, 1978), which compares trajectories that vary in time or speed, to measure and rank the fit of all 50,000 SST time series against the SST time series recorded in the observed A2-G1 isotope profile, that is, the true SST history of the MH370 flaperon drift path. The Python machine learning library Tslearn (Tavenard et al., 2020) was used to perform the DTW analysis. To reduce the set of temporal deformations to which DTW is invariant, we set an additional constraint named “sakoe_chiba” with its radius as 3 days. It allows the algorithm to look for the best-matched trajectories within a reasonable temporal span. Using DTW analysis, all drifters' SST tracks were sorted based on their similarity with A2-G1's δ¹⁸O-based SST. The top five best-matched drifters' SST tracks were chosen for a closer look. Sensitivity analyses were not performed due to the methods-demonstration aim of this study but would be necessary in a final application. Recommendations for future work are discussed below in Section 3.4.

3 Results and Discussion

3.1 Experimental Determination of δ¹⁸O_shell-Temperature Relationship

With two outliers (greater than two standard deviations (>2σ) from the mean of each group excluded), a linear approximation of the temperature-dependent fractionation between L. anatifera calcite and water was formulated using experimental constant temperature values (14, 20, and 26°C) and corresponding measured mean δ¹⁸O_water values from the aquaria (0.18 ± 0.1‰ (1 S.E.), 0.10 ± 0.1‰ (1 S.E.), and 0.36 ± 0.1‰ (1 S.E.), respectively) (Figure 3, Table S4 in Supporting Information S1):

$$$T{}^{\circ}\mathrm{C}=19.0\,(\pm \,0.13)-4.63\,(\pm 0.01)\ast \left({\delta }^{18}{\mathrm{O}}_{C}-\left({\delta }^{18}{\mathrm{O}}_{W}-0.27\right)\right)\left(N=43,{R}^{2}=0.97\right)$$$ (2)

Figure 3

Experimental δ¹⁸O_shell − δ¹⁸O_water values versus temperature. Linear regression of experimental δ¹⁸O_shell − δ¹⁸O_water values versus temperature for cultured Lepas anatifera in three treatments (solid black line) with 95% confidence intervals (dashed lines). Filled symbols represent mean δ¹⁸O_shell values of barnacles per temperature treatment included in the regression model (<2σ), whereas outliers (open symbols) were not considered.

From Equation 2, we find that δ¹⁸O_shell values decrease by 0.22‰ for every 1°C increase in temperature between 14 and 26°C (Figure 4). Shell δ¹⁸O values calculated using Equation 2 are 0.45 ± 0.08‰ (2σ) higher between 14 and 26°C than those calculated from the relationship of Epstein et al. (1953) for mollusks (Figure 4). This isotopic offset relative to mollusk carbonate is consistent with the magnitude and direction of offsets measured previously for other stalked barnacles (Pollicipes polymerus, 0.72‰; Calantica villosa, 0.62‰; and Neolepas zevinae, 0.75‰) (Killingley & Newman, 1982) and the lower offsets for stalked versus acorn barnacles (11 species, ∼1.3‰, (Killingley & Newman, 1982); Tetraclita serrata, ∼1.3‰, (Smith et al., 1988); and Semibalanus balanoides, ∼1.44‰, (Craven et al., 2008)) (Table S5 in Supporting Information S1).

Figure 4

δ¹⁸O_calcite-temperature relationship. Comparison of computed temperatures using the δ¹⁸O_calcite-temperature relationship from this study and several previously published δ¹⁸O-temperature calibrations from different carbonate-bearing archives (see Tables S3–S6 in Supporting Information S1). Water temperatures were determined using a constant δ¹⁸O_water value of 0‰ (Vienna Standard Mean Ocean Water).

Early work by Killingley and Newman (1982) found that δ¹⁸O_shell values from three species of stalked barnacles were intermediate between those of mollusks and acorn barnacles for a given SST, but no conversion relationships for any stalked barnacle species were developed until after the MH370 crash. The first of these studies (Blamart & Bassinot, 2016) calibrated δ¹⁸O values for L. anatifera shells using estimated rather than known SSTs and δ¹⁸O_water values at the time of the most recent shell growth, introducing two potential sources of error. SSTs calculated using the Blamart and Bassinot (2016) conversion relationship have an uncertainty of nearly ±2°C (see Text S1 in Supporting Information S1), a level that translates to exceedingly poor spatial resolution (many 1000s km²) for reconstructing drift positions in the tropics (Detjen et al., 2015; Pearson et al., 2020).

Subsequent temperature calibration work on this species by Mesaglio et al. (2021) similarly lacked experimental controls on the timing of shell growth but also modeled the δ¹⁸O_shell–SST relationship from a SST range of just 1°C (nearly an order of magnitude smaller than the SST range studied by Blamart and Bassinot (2016). The Mesaglio et al. (2021) relationship has a markedly different slope and y-intercept compared to other paleotemperature relationships and estimates SSTs up to 5°C off from those predicted by the Blamart and Bassinot (2016) relationship for the same barnacle δ¹⁸O values (Figure 4). The most likely explanation for the divergence of the Mesaglio et al. (2021) relationship is that slopes and y-intercepts cannot be estimated accurately from such a narrow range of data (da Silva & Seixas, 2017), especially when there are few data points and key experimental variables are not experimentally constrained and already noisy. Mesaglio et al. (2021) speculated that their equation differed from Blamart and Bassinot (2016)'s because the latter did not use L. anatifera but a closely related, spotted species Lepas indica. This explanation, however, is refuted by genetic evidence, which shows that spotted “L. indica” have the nuclear L. anatifera genotype and should be considered the same species (Schiffer & Herbig, 2016). This conclusion is also supported by Mesaglio et al. (2021)'s own data, which found no significant difference in the δ¹⁸O values of recent shell growth from specimens of the two morphotypes collected from the same site.

Our calibration study differs from these previous efforts in being the first to analyze the oxygen isotope composition of L. anatifera shells grown in the laboratory under controlled conditions with tight constraints on water temperature, salinity, and δ¹⁸O_water at the time of shell deposition. The new L. anatifera δ¹⁸O_shell–SST relationship differs in two critical ways from that of Blamart and Bassinot (2016). First, our relationship reduces uncertainties for SST calculations by an order of magnitude (±0.1°C vs. ±1.7°C at a SST of 27°C), which helps constrain the initial flaperon position in our drift simulation from a band of waters thousands of kilometers in latitudinal extent to just a few dozen (Figure 5b). Second, while the new relationship has a similar slope to the Blamart and Bassinot (2016) relationship, the y-intercept differs significantly, resulting in roughly 2–3°C cooler predicted temperature reconstructions throughout the drift (Figure 5a). In the Indian Ocean region where the MH370 crash is thought to have occurred, this difference results in a displacement of drift reconstructions poleward (south) by up to several thousand kilometers. In fact, the last SST recorded in the A2-G1 isotope profile calculated by our relationship (23.7 ± 0.1°C) is more than two degrees cooler than that predicted by Blamart and Bassinot (2016)'s relationship (26.8 ± 1.7°C) (Figure 5a) and closely matches 24–25°C SSTs recorded by the Global HYCOM around Réunion on 29 July 2015 (Figure S4 in Supporting Information S1), the date the MH370 flaperon was discovered.

Figure 5

(a) Sea surface temperatures (SSTs) from flaperon colonization (right) to beaching on Réunion Island (left) as reconstructed from MH370 barnacle A2-G1 δ¹⁸O_shell data and calculated using the new experimentally constrained relationship versus less-constrained, field-based relationships from Blamart and Bassinot (2016) and Mesaglio et al. (2021) (Table S7 in Supporting Information S1). (b) Black dots showing locations of 50,000 drifters (simulated flaperons) released east of Réunion Island along the distribution of 27°C waters on 25 February 2015, 154 days (estimated age of A2-G1) prior to flaperon discovery on 29 July 2015. (c) SST time series recorded for 50,000 simulated drifts between 25 February and 29 July 2015. Red line shows actual SSTs recorded by the MH370 barnacle, fit into an age-at-length model from Poupin (2015). (d) Completed 154-day paths of all 50,000 drifters released in the particle tracking simulation, initiated from 27°C waters on 25 February 2015. Colored paths highlight drifters whose recorded SSTs best fit the actual SSTs recorded by the A2-G1 barnacle (see Movie S1).

3.2 Reconstructing the Drift Path of the MH370 Flaperon

The reconstructed SST profile for A2-G1 (Figure 5a) reveals that flaperon colonization (drift origin) occurred in warmer waters around 27°C followed by a shift to continuously cooler waters around 23–24°C for a significant part of the latter drift. This is consistent with a drift modeling experiment by Griffin et al. (2017), which showed that the MH370 flaperon should have had a leftward (southward) trajectory into cooler waters as it drifted across the Indian Ocean.

In our drifter simulation with 50,000 particles, which starts on 25 February 2015 within this band, we isolated the top five best-fit drifts for the A2-G1 SST time series for further investigation (Figure 5c, colored lines in Figure S4 in Supporting Information S1). Each of these drifters spent their last five months drifting west of longitude 70°E, south of 20°S, and within 1,500 km of Réunion Island in the Indian Ocean. Only one drifter (blue track in Figure 5d and Figure S4 in Supporting Information S1) eventually reached waters around Réunion Island (within 220 km) by the end of the simulation.

3.3 Improvements in Modeling Debris Drift Pathways Over Previous Work

The aim of this study was to develop a method capable of reconstructing an unknown, high-resolution pathway and origin for drifting debris from the stable isotope data of hitchhiking barnacles. Previously, pathway and origin reconstructions have been done by visually comparing barnacle isotope values to isoscapes, that is, the spatial distribution of δ¹⁸O values predicted to occur in barnacle shells formed under different SST and δ¹⁸O_water values (e.g., Detjen et al., 2015; Killingley & Lutcavage, 1983; Pearson et al., 2020). However, because no unique spatial solutions can be identified with this approach, isoscape-based drift reconstructions have poor resolution (1000s km³), particularly in subtropical and tropical latitudes where the MH370 flaperon drift is thought to have occurred (Detjen et al., 2015; Pearson et al., 2020).

A more sophisticated approach was developed by Sakamoto et al. (2019), who conducted a numerical modeling study in which simulated fish were released with random swimming behavior in an ocean current model from possible spawning grounds off of Japan. The simulation recorded environmental (SST and salinity) conditions experienced by each simulated fish and calculated agreement between otolith δ¹⁸O time series of actual harvested fish and those of simulated swimmers from the different spawning grounds that also reached the harvest point by the date of harvest. Although a significant step forward, Sakamoto et al. (2019) used independent data from a regional survey of egg abundance to constrain the migration origin to a few known starting points. In contrast, little data exist to constrain the origin of the MH370 flaperon's drift beyond the barnacle's own initial isotope value. The priority area of the Seventh Arc (Figure 1) has been used as a constraint in the search, but years of failure suggest other areas should be considered. Instead of constraining possible origins, our method instead drops simulated drifters at 50,000 independent starting locations all along the nearly 6,000 km SST band defined only by the MH370 barnacle's first isotope value. Both Sakamoto et al. (2019) and this study identify the highest probability pathway by finding the best fit between the simulated and actual δ¹⁸O time series.

Our new numerical modeling approach found a relatively small number of SST histories from among the 50,000 drifters that were strong matches for the observed barnacle SST record that recorded the actual drift of the MH370 flaperon. Only one of those drifters reached waters surrounding Réunion Island, where the actual MH370 flaperon was recovered, by the stranding date. Its recorded drift reveals that the MH370 flaperon likely spent its last several months west of longitude 70°E and within 1,500 km of Réunion Island. This result demonstrates that a best fit, high-resolution drift track and origin (i.e., a simultaneous solution to the where and when of each barnacle SST) can, in theory, be found with our combination of methods.

3.4 Next Steps in the Application of Barnacle Isotopes in the Search for Flight MH370

Although our work is a significant step forward in demonstrating how barnacle isotopes can be used to reconstruct drift origins, future applications to the search for flight MH370 can be improved in several key areas. The first and most important of these is obtaining isotope records from MH370 barnacles large and old enough to have colonized the flaperon closer to the crash location. Our study used the published isotope record for MH370 flaperon specimen A2-G1, which was one of the largest and oldest barnacles studied by Blamart and Bassinot (2016) but still relatively small and only several months in age. Data from Poupin (2015), confirmed by media photographs of the recovery of the barnacle-covered flaperon, indicate that much larger and older barnacles were attached when the flaperon was recovered. Only partial drift reconstructions are possible until the largest, oldest barnacles are released for study by the French government.

Another critical area for future work is improving the age model used to assign dates to each individual barnacle isotope value, because these dates determine the drifter start date and locations but also calculations of best fit between simulated and observed SSTs. This is easily the step of the analysis most sensitive to error. Here, we inferred dates for each isotope sample and corresponding SST using the Lepas age-at-size growth model published by Poupin (2015). Error is introduced here if actual growth rates of individual MH370 barnacles varied from Poupin's (2015) average, smoothed growth curve, as pointed out by others (De Deckker, 2017). Confidence in the reconstructed origin could be gauged with a sensitivity analysis, which would involve rerunning the drift simulation with earlier and later start dates to see how it affects reconstruction of the drift origin. The ideal approach, however, is to determine each isotope sample's age-of-formation directly by sectioning the shell, identifying and counting daily growth increments in cross-section (e.g., Bourget, 1987), and assigning isotope sample dates based on these precise day counts. To our knowledge, studies describing the periodicity of shell microstructural layer formation in stalked barnacles have not been done. However, future studies applying our methodology must aim for this level of precision.

A third area for improvement is increasing the resolution of the barnacle isotope record. Blamart and Bassinot (2016) sampled A2-G1 and other shells at roughly 1 mm increments, resulting in a data-rich SST time series from day 0 to day 57, when growth was fast, but lower resolution temporal data from day 58 until the end, when daily growth is expected to have slowed (Figure 5a: blue line). Increased sample resolution of isotope measurements would result in more detailed and complete SST records from the original flaperon drift and, thus, better discrimination of better- and worse-fit drift reconstructions using DTW analysis. Various methods, including high-precision micro-milling and laser ablation, exist for sampling carbonates down to micron- and sub-microgram scale resolution (Sakai, 2009; Sakamoto et al., 2019).

Fourth, Blamart and Bassinot (2016) found broadly similar isotope profiles between MH370 barnacles in terms of the range of isotope values recorded, warming and cooling trends, and the number, spacing, and magnitude of shorter-scale fluctuations. However, there remain slight differences between the profiles that could affect which simulated drift was the best fit to the observed isotope record. These differences are probably resolvable by simply increasing the resolution of the isotope sampling across the barnacle shells and a better sample-age calibration. If any further variation remains between profiles, error due to inter-specimen variation could be investigated by repeating the entire methodology for multiple barnacles and comparing the drift origins for each.

Fifth, adding differential wave (Stokes drift) and wind (windage) motions relative to ambient current may improve the accuracy of the drift model (Bosi et al., 2021; Maximenko et al., 2018). Stokes drift effects on the surface velocity of floating objects may result in a larger probable drift area, as shown in simulations including and excluding Stokes drift for simulated MH370 flaperon-like objects (Durgadoo et al., 2021). Buoyancy tests and the presence of barnacles on the MH370 flaperon indicate that the flaperon floated in a near-horizontal, slightly submerged position, where the effects of windage would be negligible (Durgadoo et al., 2021). However, at-sea testing of a simulated flaperon showed that in high wind conditions, the flaperon pitched end over end, producing velocities that drifted slightly left of the wind (i.e., southward trajectories) (Griffin et al., 2017). Interestingly, the cooler SST reconstructions from our new barnacle isotope-SST conversion equation relative to those predicted by Blamart and Bassinot (2016) provide the first confirmation of a more southerly drift track for the MH370 flaperon.

Future work might also apply our approach to isoscapes with high temporal resolution similar to Sakamoto et al. (2019), rather than SSTs, although the benefits are uncertain. Here, we compared SSTs calculated from the barnacle isotope data directly to ocean SST records because salinity variation across the central Indian Ocean where the MH370 flaperon drift occurred is low, with a mean of 34.5–35.5 PSU as recorded by The European Space Agency Climate Change Initiative for Sea Surface Salinity (CCI + SSS) project (available at https://www.esa.int/ESA_Multimedia/Images/2019/05/Global_sea-surface_salinity). We estimate that salinity variation from rainfall would contribute to uncertainty in our barnacle SST record of less than 0.5°C.

Last, we recommend exploratory geochemical tools beyond stable oxygen isotopes. Although, De Deckker (2017) found no reliable environmental signal in Mg/Ca ratios from L. anatifera shells, drift pathways through upwelling zones should result in characteristically depleted ¹⁴C_shell due to the ocean reservoir effect. Distinctive ocean productivity patterns may also be recorded in carbon isotopes of amino acids incorporated into barnacle shells from the animal's diet (Ellis et al., 2014).

3.5 Applications Beyond the Search for Flight MH370

Our integrated methodology can, in theory, be used to source any drifting object in the ocean, including plastic pollution (Bosi et al., 2021; Lebreton et al., 2012) and human remains (Mateus et al., 2015), as stalked barnacles tend to colonize floating objects rapidly. Application to biological questions, such as tracking marine animal migrations (e.g., whales, sea turtles), is more complicated. Although large marine animals are often covered with barnacle epibionts, they usually swim rather than drift (Putman & Naro-Maciel, 2013). In such cases, our new δ¹⁸O-SST conversion equation could still be used to generate a SST history of the animal's route and constrain origins and pathways, but at low resolution (e.g., Detjen et al., 2015; Pearson et al., 2020; Taylor et al., 2019). Specific trajectories could not be modeled and compared for a swimming animal as we have done here.

However, there are some situations where large migratory marine animals do drift. For example, sea turtles infected with pathogens often become lethargic (Frick & Pfaller, 2013), leading them to spend increased time floating at the sea surface. Although stalked barnacles occur on healthy sea turtle carapaces, particularly in species such as Hawksbill sea turtles (Eretmochelys imbricata) (Eckert & Eckert, 1987; Fuller et al., 2010), the presence of stalked barnacles on the tips of the flippers is generally associated with turtles weakened and rendered inactive at the surface by disease (Fernández et al., 2015). The geography of sea turtle disease infection is still poorly understood, hindering implementation of effective environmental policy. We suggest that our new method could be applied here too by providing a way to reconstruct the past movements of diseased sea turtles back to the initiation of infection.

Acknowledgments

We want to express our gratitude to Dr. Lukasz Wojtas and Gaurav Verma from the X-ray Facility at The University of South Florida for their assistance with using the X-ray diffractometer, to Jessica Wilson at the University of South Florida Stable Isotope Laboratory for processing the stable isotope samples, and to Dr. Mark Flint of the College of Veterinary Medicine at Ohio State University and Kathy Heym of the Florida Aquarium in Tampa for support and helpful discussions about applications of our method to sea turtle conservation. We also wish to thank our three reviewers and the editor, who provided helpful feedback. This project was funded by The Florida Aquarium Conservation Fund and The Professional Association of Diving Instructors (CGA App: 28832).

Conflict of Interest

The authors declare no conflicts of interest relevant to this study.

Open Research

Data Availability Statement

The authors confirm that the data supporting the findings of this study are available within the article and its Supporting Information S1. Isotope data are included in Supporting Information S1. The data used for numerical modeling in the study are available via Herbert (2023). Numerical modeling of MH370 flaperon drift based on barnacle geochemical data [Dataset]. University of South Florida, V4, https://doi.org/10.17632/nr4vt4h8gt.4.