Mesophotic coral bleaching associated with changes in thermocline depth

Coral reefs host extraordinary biodiversity yet are vulnerable to thermal stress that triggers bleaching, especially as sea surface temperatures (SSTs) rise with climate change. Mesophotic coral ecosystems (MCEs; 30–150 m) were once hypothesized to be buffered from warming and to act as refugia for shallow reefs via larval dispersal. However, evidence indicates limited ecological and genetic connectivity with shallow reefs and high endemism within MCEs, underscoring their intrinsic conservation value. Mesophotic corals often reside within or near the thermocline, where abrupt temperature gradients render them sensitive to vertical thermocline displacements. Internal waves can modulate this thermocline, at times lifting cooler water to relieve stress in subsurface habitats. Yet, assessing MCE bleaching risk is challenging because widely used indicators (for example, NOAA Coral Reef Watch) track surface heat stress and do not capture subsurface dynamical processes. This study investigates whether MCEs are susceptible to deep bleaching independent of shallow-water signals, and how basin-scale climate modes and local internal wave dynamics shape bleaching risk at depth.

Prior work highlighted: (1) MCEs occupy up to two-thirds of the depth range of zooxanthellate coral environments and were hypothesized to buffer anthropogenic warming impacts; (2) subsequent studies show limited connectivity between shallow and mesophotic corals and high endemism in MCEs, emphasizing their distinct ecological roles; (3) MCEs lie near the thermocline, making them vulnerable to vertical perturbations that cause rapid temperature changes; (4) internal waves can lift the thermocline and advect cooler water upslope, potentially alleviating thermal stress; (5) commonly used SST-based bleaching metrics offer limited predictive power for subsurface habitats because thermocline dynamics are often decoupled from surface warming; (6) empirical records of mesophotic bleaching exist but are geographically variable and mechanisms remain under-documented, indicating a need for integrated oceanographic and ecological observations at depth.

Study area and design: Egmont Atoll (Chagos Archipelago, central Indian Ocean) was surveyed during two cruises (November 2019; March 2020). Two sites were investigated: Ile Des Rats (IDR, NW flank) and Manta Alley (MA, NE flank). November 2019 surveys covered both sites; March 2020 surveys covered MA only due to hazardous currents preventing access to IDR. Biological data collection: A Falcon Seaeye ROV equipped with LED lighting, a Seaeye camera (720p, wide-angle 91°) and a GoPro Hero 4 (2.7k, 24 fps) was used. A Valeport Modus CTD was mounted on the ROV for in-situ temperature profiling during dives. Strong currents precluded linear video transects close to the seabed; instead, still images were captured along transects at altitude <1.5 m with oblique camera angle. For each image, time, depth, latitude, and longitude were recorded. Across six depth bands (15–20; 30–40; 60–70; 80–90; 110–120; 150–160 m), 90 images per depth were collected, totaling 1080 images across the two sites. The two deepest bands were excluded from analysis due to absence of scleractinian corals. Camera angle and altitude were standardized; laser scaling was unavailable, so analyses focused on proportions (prevalence/severity) rather than area-based metrics. Image analysis and coral metrics: Still images were annotated in BIIGLE. All benthic organisms >1 cm were identified to morphospecies (OTUs) following a standardized image taxon reference framework; only zooxanthellate scleractinian corals were analyzed. Bleaching severity categories (1–4) were assigned based on color and extent: (1) unaffected, (2) light (pale or 0–20% bleached), (3) moderate (20–80% bleached), (4) severe (80–100% bleached or bleached with partial mortality). A bleaching index (BI) was computed: BI = (0·c1 + 1·c2 + 2·c3 + 3·c4)/3, where ci is the percentage of observations in category i. Bleaching prevalence (%) = (number of bleached colonies/total colonies) × 100. Multiple references (video context, within-image comparisons, physical samples from 2019 and healthy samples from 2022) were used to mitigate color-referencing limitations. Statistical analyses: Differences in BI and prevalence across site, season, and depth were tested in R using Kruskal–Wallis with post-hoc Wilcoxon rank-sum tests and Benjamini–Hochberg corrections (p < 0.05). Community composition differences were evaluated via MDS and PERMANOVA on square-root transformed abundance data using Bray–Curtis similarity with a dummy variable to address zero-inflation (PRIMER v6). Oceanographic observations: Moored arrays measured current velocity and temperature profiles. At MA, a near-bed (z = 2 m) upward-looking Nortek Signature 500 kHz ADCP provided 10-min averaged velocities (2 m bins) and hourly 1 Hz burst data including 6 mm vertical-resolution echo amplitude; a nearby taut-line mooring hosted two RBR Concerto CTDs (top/bottom) and 23 RBR solo-T thermistors at 2 m spacing from 4–50 m, logging at 1 Hz (thermistors) and 0.2 Hz (CTDs). At IDR, a taut-line mooring hosted top/bottom temperature–depth sensors and 25 thermistors at 2–4 m spacing, plus an upward-looking Aquadopp 400 kHz ADCP. Temperature data were interpolated to 2 m vertical intervals and cleaned with deviation filters and running means; echosounder data were corrected for noise floor and gain via logarithmic range decay fits. For visualization of long-term evolution (e.g., Fig. 4a), temperature was binned daily and low-pass filtered (6-day cut-off). Remote-sensed and reanalysis data: IOD index (Dipole Mode Index) was sourced from OOPC/NOAA. Basin-scale temperature evolution and thermocline depth were derived from Copernicus Marine Service Global Forecast (from 2019) and Reanalysis (pre-2019), extracted at 8.125°S, 73.875°E in deep water east of the Archipelago. NOAA Coral Reef Watch Degree Heating Weeks (DHW) time series were extracted for the Chagos region and central Indian Ocean for context. High-resolution numerical modeling: A nonhydrostatic MITgcm configuration simulated internal tide/wave dynamics around Egmont with 25 m horizontal resolution (core domain), 5 m vertical resolution, and 5 m bathymetry grid (multibeam to 400 m depth merged with GEBCO deeper than 400 m). Open boundaries used Orlanski-type radiation to minimize barotropic tide reflection. Forcing included semidiurnal M2 and diurnal K1 tidal constituents (per TPXO predictions and ADCP analyses), implemented following Vlasenko & Stashchuk. Initial stratification was set from deep-water CTD profiles (undisturbed by slope processes) from 2019–2020. A 16-day run (1-day spin-up + 15 days) captured a spring–neap cycle. Model outputs were validated against ADCP and thermistor chain measurements at MA and used to estimate bottom temperature anomalies and internal wave mode structure at MA vs IDR. In-dive temperature: The ROV-mounted CTD recorded temperature continuously during dives, enabling comparison of near-bed temperatures along transects with background deep-water profiles.

- Deepest recorded coral bleaching: Scleractinian coral bleaching was documented at mesophotic depths down to 90 m at Egmont Atoll (Chagos Archipelago) during November 2019, representing the deepest bleaching reported to date. - Absence of shallow-water bleaching: Despite deep bleaching, no bleaching was observed at shallow depths (15–20 m). ROV CTD data showed temperatures exceeded 29.5 °C in shallow water (<25 m) on only 2 of 12 dives; all other dives remained below 29.5 °C, consistent with no shallow thermal stress. - Basin-scale driver: The event coincided with the strongest positive Indian Ocean Dipole (IOD) on record (2019). Copernicus reanalysis indicated a marked deepening of the thermocline across the central/western Indian Ocean; the 22 °C isotherm deepened to 60–140 m for about 3 months centered on November 2019, exposing normally thermocline-resident mesophotic corals to anomalously warm surface-layer waters. - Recovery: By March 2020, as the IOD transitioned to weakly negative, the thermocline shoaled to typical depths (~40–50 m), and corals at depth showed less bleaching. - Local-scale modulation by internal waves: Moored observations and high-resolution modeling revealed substantial vertical excursions of the thermocline at tidal frequencies at MA (>20 m), with signals consistent with Mode 2 internal waves. Modeling and observations indicated that at MA these internal waves increased time-mean near-bed temperatures between 60–90 m, enhancing bleaching risk by advecting warmer water downslope. At IDR, internal waves tended to advect cooler sub-thermocline water upslope, potentially alleviating thermal stress. - Spatial heterogeneity at sub-km scales: Internal-wave impacts on near-bed temperature were sensitive to local bathymetry, stratification, and tidal forcing, producing differences in bleaching severity between MA and IDR over O(1 km) scales. - Community composition vs bleaching: PERMANOVA showed significant differences in coral community structure among depth bands within sites, but communities at the same depth across sites were more similar than communities at different depths within a site. Species spanning 30–40 to 60–70 (and to 80–90 m for Leptoseris spp.) exhibited more severe bleaching at mesophotic depths, and bleaching prevalence/severity were higher at MA than IDR, indicating composition alone did not explain bleaching patterns. - SST metric limitations: NOAA Coral Reef Watch DHW indicated no bleaching warning around the period, underscoring that SST-based metrics can fail to capture mesophotic bleaching risk driven by subsurface dynamics.

Findings demonstrate that mesophotic corals can bleach at depths up to 90 m, even when shallow reefs show no bleaching, revealing a decoupling between surface indicators and subsurface thermal stress. The basin-scale deepening of the thermocline during an extreme positive IOD exposed deep reefs to anomalously warm conditions, while local internal wave dynamics—particularly Mode 2 waves—can either exacerbate or mitigate near-bed warming depending on site-specific bathymetry and flow. Consequently, bleaching susceptibility and severity in MCEs are governed by multi-scale oceanographic processes: climate-mode modulation of thermocline depth combined with local internal tide/wave interactions. Traditional SST-based warning systems have limited utility for MCEs; accurate risk assessments must incorporate subsurface temperature variability and internal wave dynamics. The observed spatial heterogeneity over sub-kilometer scales highlights the need for site-resolved oceanographic understanding to predict where mesophotic bleaching will occur and to inform habitat-specific conservation strategies.

This study documents the deepest recorded coral bleaching (to 90 m) at Egmont Atoll, occurring without shallow-water bleaching, and links it to thermocline deepening during an extreme positive IOD, with local internal waves further modulating near-bed temperatures and bleaching severity. The work underscores that MCEs are not reliable refugia for shallow reefs and that SST indices alone are poor predictors of mesophotic bleaching risk. Integrating high-resolution oceanographic observations and models that resolve internal waves and thermocline variability is essential to assess and forecast bleaching in MCEs. Conservation planning should explicitly include MCEs, account for basin-to-local scale drivers of subsurface thermal stress, and invest in deeper reef monitoring. Future research should: (1) expand long-term subsurface temperature and current observations across diverse MCE settings; (2) develop operational models coupling tides, internal waves, and stratification at reef scales; (3) refine biological sensitivity metrics for mesophotic taxa; and (4) improve remote-sensing/data-assimilation products to infer subsurface risk from surface and limited in-situ data.

- Imaging and scaling: Due to ROV laser issues, image area could not be quantified; analyses relied on proportions (prevalence/severity), limiting area-based comparisons and absolute density estimates. - Subjective bleaching assessment: The bleaching index is based on color categories without a uniform Lambertian reference. Multiple cross-checks (video context, within-image contrasts, physical samples, and healthy reference samples from 2022) were used to mitigate, but some subjectivity remains. - Site and temporal coverage: March 2020 surveys were conducted only at MA due to hazardous currents, preventing temporal comparisons at IDR. The two deepest bands (110–120; 150–160 m) were excluded due to absence of scleractinian corals, limiting depth-range inference. - Generalizability of oceanographic mechanisms: Internal wave impacts are highly sensitive to local bathymetry, stratification, and forcing; results from Egmont may not directly transfer to other atolls or regions without site-specific validation. - Monitoring constraints: Practical and technological challenges limit continuous subsurface monitoring; some oceanographic datasets remain under analysis. Reliance on reanalysis/model outputs introduces uncertainties at local scales, and typical 5 km-resolution products cannot resolve reef-scale dynamics. - Community composition inference: While PERMANOVA indicated depth-related community differences, the contribution of rare species to detected differences and limited replication at some depths may affect interpretation of community–bleaching relationships.