Coral bleaching and mortality are commonly associated with heat stress during marine heatwaves (MHWs), yet attributing these responses solely to heat stress may overlook the complexity of contemporary MHWs. Here, we document a severe MHW affecting a central Red Sea reef in mid-2024 that coincided with repeated nighttime low-oxygen events. By integrating satellite and in situ records of temperature, dissolved oxygen, and wind, alongside repeated surveys of in situ nursery-reared coral fragments, we document the temporal coincidence of these conditions with rapid coral bleaching and mortality. As temperatures approached and exceeded 33 °C during the MHW peak, nighttime oxygen minima were extreme (down to 15.2–57.8 µmol O₂ kg⁻¹), and diel temperature variability diminished (down to 0.31–0.67 °C), indicating a loss of nighttime thermal reprieves. Coincident responses of nursery corals were atypical, with transitions from healthy or bleached states to mortality occurring within 7–8 days. Bleaching and mortality also occurred across diverse coral species in nearby habitats after the MHW peak, with <1% live coral cover documented eight months later. Sustained high temperatures, reduced vertical mixing, and increased nighttime respiratory demand likely constrained oxygen availability, resulting in low-oxygen extremes. We posit that similar compound MHW extreme events may have gone undetected in the Red Sea and elsewhere, potentially obscuring the contribution of co-occurring low-oxygen extremes to coral responses attributed to heat stress.

Study site and coral nursery structure

This study was conducted at Al Fahal Reef, a mid-shore reef in the central Red Sea near Thuwal, Saudi Arabia (22°18'11” N, 38°57'53” E). On 4 June 2024, a coral nursery structure was suspended mid-water at ~9 m depth in a semi-sheltered area at the northern tip of Al Fahal Reef (22°18′20.44″ N, 38°57′54.59″ E) as part of a separate experimental study (Fig. S1) (Re and Duarte 2025). Briefly, we collected nine source colonies representing three coral species (Acropora pharaonis, Acropora hemprichii, and Stylophora pistillata; three colonies per species) at depths of 2–10 m. Ten fragments (3–8 cm in length) were produced from each colony and randomly assigned to positions on the PVC “tree” nursery structure (total n = 90). The fragments were spaced at least 4 cm apart to minimise the potential interaction between fragment microenvironments. Although deployed for a separate experimental study (Re and Duarte 2025), the nursery structure provided a standardised framework for documenting in situ coral responses during the MHW.

At approximately weekly intervals between 4 June and 15 August, we monitored all nursery fragments for health status and cleaned the bases of all coral fragments. Fragments that remained visually unchanged from their initial deployment condition were classified as ‘unbleached’. Fragments exhibiting obvious signs of paling, partial bleaching, or complete loss of pigmentation were classified as ‘bleached’. Fragments displaying bright host-derived fluorescence, a response commonly associated with physiological stress in corals (Roth and Deheyn 2013; Bollati et al. 2020), were classified as ‘fluorescent’. Fragments that showed partial tissue disintegration (or ‘tissue sloughing’) but retained most living tissues were classified as ‘tissue disintegration’ (sensu Dhillon et al. 2026), whereas fragments exhibiting substantial tissue disintegration or algal colonisation consistent with fragment mortality were classified as ‘dead’.

In situ temperature and dissolved oxygen measurements

On 17 July 2024, a miniDOT® oxygen logger (PME®) was attached to the base of the nursery structure at ~10 m depth to continuously record temperature (°C) and dissolved oxygen (mg L⁻¹) at 5-min intervals. The logger was positioned ~30 cm from the nearest coral fragment and ~2 m above the benthos, thereby capturing ambient mid-water conditions. A two-point calibration check was performed prior to deployment (using 100% air saturation and 0% oxygen (via nitrogen gas)), which confirmed no offset was required. The miniDOT® oxygen sensor was cleaned at weekly intervals between 17 July and 15 August 2024. Raw dissolved oxygen time series were quality controlled by removing missing values and applying a rolling median–median absolute deviation (MAD) de-spiking procedure to flag and replace transient high- and low-amplitude spikes using the ‘zoo’ package in R (v.4.3.0). Resulting dissolved oxygen concentrations (in mg L⁻¹) were then converted to µmol kg⁻¹ using temperature-corrected solubility relationships assuming a constant salinity of 39 psu using the ‘respirometry’ package. Additional temperature and dissolved oxygen measurements from two earlier deployments at the same reef site (27 July–11 August 2023 and 13–28 August 2023) were collected using the same sensor configuration and sampling protocol (Fig. S2). For these deployments, the loggers were attached to the reef substrate at ~10 m depth, with the sensor oriented outward into the water column. The in situ temperature and dissolved oxygen time series therefore represent measurements from a single logger deployed at the nursery site.

Wind and satellite sea surface temperature data

Wind speed data were obtained from a single SOFAR Spotter buoy deployed at Al Fahal, and daily mean wind speeds were calculated only for days with at least 12 valid records to avoid bias from incomplete daily coverage (Fig. S3). Daily SST and sea surface temperature anomaly (SSTA) data were obtained from the NOAA Coral Reef Watch CoralTemp v3.1 product (5 km spatial resolution) for the study site for 1985–2024 (Skirving et al. 2020). SST and SSTA fields were extracted from the grid cell nearest to the study location. SSTA values represent the deviation of daily SST from the CoralTemp v3.1 climatological mean for the corresponding day of the year. We used daily SST data to calculate MHW characteristics for 2024, following Hobday et al. 2016, using the ‘heatwaveR’ package in R (v.4.3.0) (Schlegel and Smit 2018). A seasonally varying 90^th-percentile threshold, calculated from the 1985–2014 baseline period, defined events lasting at least five consecutive days above this percentile. The extracted event metrics included onset date, end date, duration, peak date, maximum intensity, and mean intensity (Table S1).

Post-event photo transects

Post-event surveys were conducted on 17 April 2025 to quantify benthic composition, including coral health status. The timing allowed the assessment of longer-term outcomes and delayed coral impacts following the MHW. Three 25 m transects were first conducted in the semi-sheltered reef area near the coral nursery deployment site (22°18′20.44″ N, 38°57′54.59″ E) to assess post-event benthic condition at the focal study site. Transects followed the natural reef contours and were spaced at least 5 m apart to ensure spatial independence.

Each transect was photographed at 1-m intervals using an Olympus TG-6 camera. Benthic composition was quantified as per cent cover using CPCe, following standard protocols (Kohler and Gill 2006). One quadrat image was excluded due to poor image quality; to maintain consistency, one quadrat was removed from each transect prior to analysis (n = 24 quadrat images per transect; Table S2). For each transect, 24  1m² photo quadrats were scored using 50 randomly generated points, yielding 1,200 points per transect. We assigned each point to a benthic category. Live corals were identified to genus level whenever possible and recorded separately as either healthy or bleached (e.g., Porites sp. or Porites sp. bleached). When genus-level identification was not possible, live corals were classified as ‘other hard coral’. When colonies were dead and algal growth obscured skeletal features, we classified points as ‘dead hard coral’. We assigned remaining points to non-coral benthic categories, including turf algae, crustose coralline algae (CCA), or sponge.