Three-layer circulation in the world deepest hadal trench

Hadal trenches (≥6000 m) are topographically confined, V-shaped depressions with unique hydrological, geological, and biological characteristics that influence deep-ocean processes. The Challenger Deep (CD) at the southern end of the Mariana Trench is the deepest known trench and the only channel below 4000 m linking the East Mariana Basin to semi-enclosed basins such as the West Mariana and West Caroline Basins. Despite its importance for Pacific deep circulation and for ventilating abyssal and hadal waters, deep circulation in the CD remains uncertain due to observational challenges. In the abyssal Pacific, the Deep Western Boundary Current (DWBC) transports Lower Circumpolar Deep Water (LCDW) from the Southern Ocean northward. Near the CD, the DWBC splits, with a portion turning east and the main branch flowing west through the trench toward the West Mariana and West Caroline Basins. Prior estimates of LCDW transport through or near the CD have varied widely and relied heavily on geostrophic calculations from sparse CTD data and limited current measurements, yielding inconsistent transports and leaving the deep circulation structure unresolved. Moreover, while cyclonic circulation has been inferred in several hadal trenches and suggested for parts of the CD from geostrophy, direct observations have also shown strong westward LCDW intrusions, sustaining debate. Turbulent mixing near rough topography, particularly amplified near the bottom and influenced by internal tides, is known to shape abyssal stratification and overturning and may play a key role in modulating trench circulation. This study addresses the gaps by deploying extra-deep moorings across the CD to quantify LCDW transport, resolve the vertical structure of circulation, assess seasonal variability, and use numerical experiments to elucidate the role of tidal-induced mixing and topography.

Previous work identified the DWBC pathway delivering LCDW into the western North Pacific and suggested that the CD lies along this pathway. Geostrophic estimates based on CTD profiles indicated westward LCDW transports of roughly -0.3 Sv through the trench, -0.55 Sv near its entrance, and -1.29 Sv within the trench. Downstream direct measurements reported transports at the Yap–Mariana Junction ranging from ~0.36 Sv northward to ~2.1 ± 1.23 Sv, and at the sill west of the CD about -2.43 ± 2.56 Sv and -2.2 ± 1.0 Sv. These disparate values highlight large uncertainties arising from sparse sampling, reference-level choices, and temporal variability. Cyclonic circulations consistent with momentum and potential vorticity constraints have been documented in other trenches (Yap, Kermadec, Izu–Bonin, Japan, Kuril–Kamchatka, Aleutian) and inferred in the CD between ~3000–7000 dbar from geostrophic analyses, while other high-resolution and direct measurements emphasized a westward LCDW intrusion dominating deeper layers. A large body of literature demonstrates bottom-intensified turbulent mixing over rough topography, with internal tides as a major energy source, and theoretical work indicates that divergences of turbulent buoyancy flux drive vertical motions (upwelling in the benthic boundary layer and downwelling above), potentially modulating circulation in enclosed basins. However, the explicit role of tidal-induced mixing in CD circulation had not been directly tested with sustained in situ currents and targeted modeling prior to this work.

Observations: Four extra-deep current meter moorings (M1–M4) were deployed across the CD along 142.5°E from January 5, 2016 to February 2017, forming a meridional array from the south slope (M1, M2) to the north slope (M3, M4). Bottom depths were 5690, 8226, 8599, and 5842 m at M1–M4. Instruments included single-point acoustic current meters (Seaguard RCM DW and Nortek Aquadopp 6000 m) and CTDs (SBE 37-SM, SBE 16plus) at ~1000 m vertical spacing from ~3600 m to near the seafloor. Sampling was every 30 min for velocity and dissolved oxygen (DO, at 3591 m on M1), and every 10 min for temperature, salinity, and pressure. Instrument accuracies were ±1% for velocity and ±5% for DO; CTD accuracies were 0.002–0.005 °C for temperature and 0.0003–0.00055 S m⁻¹ for conductivity. Temperature and salinity drift corrections followed established procedures; all variables were averaged to daily means. One year of data (Jan 5, 2016–Jan 4, 2017) was analyzed. Data processing and transport calculation: Due to topographic channeling, analysis focused on zonal velocities. The transect was gridded at 10 m vertical and ~110 m meridional resolution, imposing zero velocity at the bottom and sidewalls. Annual-mean zonal velocities below 3591 m were obtained using natural neighbor interpolation. The upper interface of LCDW was defined by the θ = 1.2 °C isotherm, whose depth varied between ~3600–3900 m. LCDW volume transport was computed by vertically integrating zonal velocity from the θ1.2 isotherm to the bottom across the transect. Time series were filtered using a 24-day low-pass (Butterworth) to examine seasonal variability, and spectral analysis identified intra- and sub-seasonal bands; bandpass filtering quantified variance contributions. Correlations among LCDW transport, θ1.2 depth, and DO at 3591 m (M1) were computed on filtered series. Numerical modeling: A regional MITgcm configuration (9–16°N, 138–149°E) with 1/24° horizontal resolution and 112 vertical levels (10–50 m in the upper 100 m; 100 m from 100–10,800 m depth) covered the southern Mariana and northern Yap Trenches and adjacent basins. Topography was from GEBCO_08 (30 arc-second). Initial T/S fields combined WOA18 climatology (to 5500 m) and a deep CTD profile from R/V Sonne (to 10,851 m), horizontally homogeneous below 5500 m. Open boundary horizontal flows were from climatological HYCOM; a 5-gridpoint sponge relaxed thermodynamics to climatology; boundary transports were adjusted to zero net inflow. Horizontal viscosity used a Smagorinsky scheme. Two vertical mixing configurations were tested: (1) NoTM, baseline KPP with constant background diffusivity 1×10⁻⁵ m² s⁻¹ (no tidal mixing); (2) TMCD, modified KPP with background diffusivity replaced by tidal-induced diffusivity from a 3D internal tide model forced by eight primary constituents (M2, S2, N2, K2, K1, O1, P1, Q1) with boundary conditions from OTIS (1/12°). The internal tide model used 100 m layers below 5500 m, 90 s timestep, integrated 120 model days; diapycnal diffusivities were diagnosed from baroclinic energy budgets, showing bottom-intensified values O(10⁻³–10⁻²) m² s⁻¹ consistent with observations. Both NoTM and TMCD were spun up for 100 model years (600 s timestep), equilibrating after ~20 years; year-100 means were analyzed. An additional TMCD-MBF experiment applied monthly varying open-boundary flows, integrated 20 years from the TMCD year-100 state.

• First sustained, high-depth-resolution observations reveal a three-layer circulation in the Challenger Deep: (1) above ~6000 m, a strong westward LCDW flow dominates; (2) between ~6000–7500 m, a cyclonic circulation; (3) in the bottom layer, an anticyclonic circulation with westward flow over the south slope (e.g., ~0.5 cm s⁻¹ at 8001 m at M2) and eastward flow over the north slope (e.g., ~0.3 cm s⁻¹ at 8313 m at M3).
• Annual-mean LCDW transport through the CD is -1.866 ± 2.953 Sv (negative westward), larger than many previous geostrophic estimates and comparable to recent mooring-based estimates downstream.
• LCDW transport exhibits large variability: daily values ranged from about -10 to +6 Sv during 2016; seasonal (>90 days), intraseasonal (30–90 days), and subseasonal (4–30 days) bands contribute 64.53%, 11.51%, and 10.99% of variance, respectively. A seasonal reversal to eastward transport occurs in summer.
• The θ1.2 isotherm depth (3600–3900 m) and DO at 3591 m (M1) covary with LCDW transport, indicating ventilating intrusions: correlations are 0.5419 (θ1.2 depth vs. transport), -0.6899 (DO vs. transport), and -0.8621 (DO vs. θ1.2 depth).
• Above 6000 m, flows intensify westward toward the north slope with a weak countercurrent over the south slope, consistent with superposition of westward LCDW intrusion and PV-balance-driven counterclockwise circulation constrained by trench geometry and bottom slope variations.
• Numerical experiments: NoTM (no tidal mixing) reproduces westward flow above 6000 m and a weak cyclonic sense below, but fails to produce the observed bottom anticyclonic circulation. TMCD (with tidal-induced mixing) reproduces the observed cyclonic–anticyclonic structure and magnitudes; area-averaged profiles show upwelling associated with cyclonic vorticity above ~6700 m and downwelling with anticyclonic vorticity below, with the transition depth (~6700 m) matching the shift in circulation sense.
• Reanalysis (GLORYS12V1) captures the seasonal cycle of LCDW transport and its reversal (correlation 0.67 with observations; annual-cycle variance contribution 49%). Pressure differences between trench entrance and exit at deep levels correlate with transport reversals (e.g., r = -0.60 at ~3993 m), supporting a basin-scale pressure-gradient control on seasonality.

The study resolves the previously debated deep circulation in the Challenger Deep by combining sustained mooring observations with targeted numerical experiments. The three-layer structure reconciles geostrophic hints of cyclonic circulation with direct evidence of strong westward LCDW intrusion. Above 6000 m, the flow pattern reflects the dominance of the westward LCDW branch interacting with PV constraints and meridional variations in bottom slope, producing intensified westward currents toward the north slope and a weak counterflow over the south slope. The seasonal reversal of LCDW transport demonstrates bidirectional deep-basin connectivity and is consistent with reanalysis-derived pressure-gradient reversals between the trench entrance and exit. Below 6000 m, classic closed-contour momentum balance alone would predict cyclonic circulation, but the observed bottom anticyclonic layer requires inclusion of bottom-intensified turbulent mixing. The TMCD experiment shows that tidal-induced mixing drives coherent upwelling and downwelling patterns which, through potential vorticity conservation, force cyclonic circulation above ~6700 m and anticyclonic circulation below, thereby explaining the observed reversal of circulation sense at depth. These findings highlight the critical role of internal-tide-driven mixing, in conjunction with topography and LCDW intrusion, in setting the deep hydrodynamic environment of the world’s deepest trench and in modulating ventilation pathways for abyssal and hadal waters.

This work provides the first year-long, high-depth-range observational evidence for a three-layer circulation in the Challenger Deep: westward LCDW flow above ~6000 m, cyclonic circulation between ~6000–7500 m, and an unexpected anticyclonic circulation in the bottom layer. The annual-mean LCDW transport (-1.866 ± 2.953 Sv) is seasonally variable and reverses in summer, evidencing bidirectional deep connectivity between adjacent basins. Numerical experiments demonstrate that tidal-induced, bottom-intensified mixing is essential to reproduce the cyclonic–anticyclonic structure, with vertical motions linked to potential vorticity constraints setting the transition near ~6700 m. These results refine understanding of deep Pacific overturning pathways and the ventilation of hadal environments and underscore the importance of including realistic turbulent mixing in trench dynamics. Future work should extend spatial coverage with additional mooring lines and profiling platforms, increase temporal coverage to resolve interannual variability, obtain direct turbulence and microstructure measurements to quantify mixing energetics, and employ higher-resolution, nonhydrostatic models to resolve internal tides and boundary-layer processes that influence trench circulation.

• Observational coverage was limited to four moorings along a single transect for ~1 year, potentially missing lateral structure and interannual variability.
• The interpolation assumed zero velocities at the bottom and sidewalls, which may bias the reconstructed fields near boundaries.
• Turbulent mixing was not directly measured; its role was inferred from models constrained by an internal tide energetics scheme and prior estimates of diffusivity.
• Instrument biases and drift (CTD conductivity/temperature) required corrections; residual uncertainties remain relative to the small water-mass gradients at hadal depths.
• Numerical experiments used climatological boundary conditions (with an additional monthly-forced case) and parameterized mixing; model resolution and physics may not capture all processes (e.g., nonhydrostatic internal waves, small-scale instabilities).