Bottom marine heatwaves along the continental shelves of North America

Marine heatwaves (MHWs) are extreme warm ocean temperature events that impact marine ecosystems by altering species distributions, productivity, and human–wildlife interactions. Most prior work has focused on sea surface temperature (SST) extremes, aided by abundant surface observations and their ecological relevance. However, subsurface and seafloor temperature anomalies can drive distinct biological impacts on demersal species and may evolve differently from surface anomalies. It remains unclear when and where surface MHWs serve as proxies for bottom conditions. This study aims to characterize bottom marine heatwaves (BMHWs) along North American continental shelves, quantify their intensity, duration, spatial extent, and relationship to SMHWs, and assess how synchrony varies with depth and mixed layer dynamics.

Recent studies have begun to examine subsurface aspects of MHWs using Argo profiles, moored buoys, gliders, and gridded observations in regions such as the Northeast Pacific, Tasman Sea, east and north Australia, the Mediterranean, Gulf of Mexico, and Northwest Atlantic shelves. Bottom water temperature anomalies have been linked to significant ecological effects: declines in Gulf of Alaska Pacific cod, shifts in demersal fish occurrence in the California Current, invasive lionfish redistribution along the Southeast US shelf, altered cod recruitment, predation pressure on scallops in the Northeast US shelf, and lobster disease patterns. These works motivate a systematic, large-scale assessment focused specifically on bottom extremes across diverse Large Marine Ecosystems (LMEs).

Data: Monthly mean ocean bottom temperatures from the Copernicus Marine Environmental Monitoring Service GLORYS12v1 global ocean reanalysis (NEMO model, 1/12° horizontal resolution, ~8 km; 50 vertical levels) spanning 1993–2019. The reanalysis is forced by ECMWF ERA-Interim and assimilates along-track satellite altimetry, satellite SST, sea ice concentration, and in situ temperature and salinity profiles (CORA). Study domain and stratification: Analysis is organized by nine North American Large Marine Ecosystems (LMEs): East Bering Sea, Gulf of Alaska, California Current, Gulf of California, Gulf of Mexico, Southeast US (SEUS), Northeast US (NEUS), Scotian Shelf, and Labrador. Only grid cells with bottom depths <400 m are included to focus on continental shelves. BMHW statistics: For each grid cell, BMHW characteristics (average intensity, duration, spatial extent) were computed from monthly bottom water temperature anomalies. Spatial patterns were mapped and related to bathymetry. The relationship between BMHW average intensity and bottom depth was quantified using two-dimensional probability histograms and Spearman correlations across grid cells within each LME. Average BMHW duration was computed and mapped similarly. LME-scale spatial extent was quantified monthly as the fraction of shelf area in BMHW conditions, alongside the mean intensity over affected area. Comparison with SMHWs: Parallel analyses were performed for SMHWs using sea surface temperature anomalies to compare intensity, duration, and spatial extent, and to assess lags and differences in coherence between surface and bottom events. Synchrony analysis: BMHW–SMHW co-occurrence (“synchrony”) was measured as the fraction of months when both occurred at a grid cell. Synchrony was related to the ratio of mixed layer depth (MLD) to local bathymetric depth, by compositing the MLD/bathymetry ratio during co-occurring months and assessing the relationship between synchrony and this ratio. Statistical and processing choices: Monthly means were used because decorrelation timescales for bottom and surface temperature anomalies are generally >30 days in most LMEs. Analyses were repeated with and without linear detrending; results were largely robust, with stronger trend effects in the NEUS, Scotian, and Labrador LMEs. Additional diagnostics included 2D histograms of intensity vs depth, regional decorrelation timescales, and comparisons of BWTAs and SSTAs during widespread events. Validation and justification: GLORYS performance in coastal environments is supported by prior studies and additional comparisons here against bottom temperature observations at ten coastal locations around North America, showing good agreement for variability and average BMHW intensity and duration.

• BMHW intensity varies strongly with bottom depth and region. Typical average BMHW anomalies range from ~0.5 °C in deeper shelf areas to as high as 5 °C in large portions of the Gulf of California at ~100 m depth.
• Depth relationships: Strong negative correlation between BMHW intensity and depth in the East Bering Sea (Spearman R = -0.84) and Gulf of Alaska (R = -0.9). Nonlinear patterns in the Gulf of California and southern California Current with intensities increasing from surface to ~100 m, then decreasing with depth; warmest average intensities at 50–100 m (~2.9–3.0 °C). Along the Gulf of Mexico and US east coast LMEs, intensity-depth relationships are weaker and more scattered.
• BMHW duration exhibits strong spatial variability; some regions show longer BMHWs over deeper parts of the shelf (e.g., western East Bering Sea, Gulf of Maine, southeastern Scotian, northeastern Labrador). In the Gulf of Mexico and SEUS, average durations are relatively uniform at ~1.5 months; the southern California Current shows longer durations than the north.
• LME-scale spatial extent: Major, prolonged BMHWs occurred in many LMEs. Notable events include: • Gulf of Alaska, California Current, Gulf of California during 1997–1998 El Niño: peak areal extents 0.72, 0.96, 0.87; peak average intensities 1.6 °C, 3.5 °C, 5.0 °C, respectively. • 2015–2016: peak areal extents 0.81 (Gulf of Alaska), 0.64 (California Current), 0.61 (Gulf of California); peak intensities 1.4 °C, 3.0 °C, 3.2 °C, respectively. • NEUS, Scotian, Labrador during 2011–2012: peak areal extents 0.64, 0.50, 0.57; peak average intensities 3.0 °C, 3.8 °C, 1.8 °C, respectively.
• BMHW vs SMHW intensity and duration: BMHWs generally persist longer than SMHWs at almost all locations. In many regions (southern California Current, Gulf of California, Gulf of Mexico, SEUS, parts of NEUS), average BMHW intensity exceeds SMHW by ~0.5–2.5 °C, whereas in the East Bering Sea, Gulf of Alaska, and northern California Current, SMHWs tend to be ~0.5–1.0 °C warmer than BMHWs on average.
• Spatial coherence: SMHWs are more frequently widespread (>50% LME area) than BMHWs, but during widespread events, bottom anomalies can exceed surface anomalies. For example, BWTAs were >1 °C warmer than co-located SSTAs in the California Current during 1997/1998 and >2.5 °C warmer in the Gulf of California during 1997/1998 and 2015/2016.
• Timing differences: In the Gulf of Alaska and California Current during 1997/1998 (and similarly in 2014/2015), widespread BMHWs lag widespread SMHWs by several months; in the Gulf of Alaska, BMHW conditions persisted up to 7 months after SMHWs subsided.
• Synchrony and depth/MLD: BMHW–SMHW co-occurrence is higher over shallow shelves and scales with the ratio of MLD to bathymetric depth; synchrony increases as the MLD more frequently reaches the seafloor.
• Regional contrasts tied to bathymetry and circulation (e.g., differences between Mid-Atlantic Bight and Gulf of Maine; spatial heterogeneity in Scotian Shelf linked to basins and banks).

The study addresses whether bottom marine heatwaves share characteristics with surface events and to what extent surface conditions serve as proxies for bottom conditions. Findings demonstrate that BMHWs can be more intense and longer-lasting than SMHWs, and that their occurrence and properties are closely tied to bathymetry and subsurface dynamics. Synchrony between BMHWs and SMHWs is high where the mixed layer often reaches the bottom, but decreases with depth, indicating that surface observations alone may not capture seafloor extremes in deeper shelf regions. The temporal lag between widespread SMHWs and BMHWs along the US west coast suggests different governing mechanisms: SMHWs respond rapidly to atmospheric teleconnections (e.g., PNA associated with ENSO), while BMHWs reflect slower subsurface adjustments, coastally trapped waves, and thermocline displacements. Along the Northwest Atlantic, propagation and advection by currents (e.g., Labrador Current) can organize BMHWs with distinct timing and extent relative to the surface. These insights highlight the need for LME-specific understanding of flow–topography interactions driving BMHWs and have implications for ecosystem management, given that demersal species can experience extreme, prolonged warming even without strong surface signatures.

This work provides the first broad, high-resolution assessment of bottom marine heatwaves along North American continental shelves, quantifying their intensity, duration, spatial extent, and relationship to surface events across nine LMEs. BMHWs often persist longer and can be more intense than SMHWs, with synchrony strongest in shallow regions where the mixed layer reaches the seafloor. Major BMHW episodes align with large-scale climate variability (e.g., ENSO) and subsurface processes (e.g., coastally trapped waves, current-driven advection). Future research should focus on diagnosing region-specific physical drivers and flow–topography interactions within each LME, clarifying the mechanistic links to climate modes, and improving predictive capability. The results underscore the importance of maintaining and expanding subsurface observing systems and reanalysis tools to monitor and forecast BMHWs, particularly where surface proxies are unreliable.

• Reliance on a single ocean reanalysis (GLORYS12v1) may introduce biases, especially in regions with sparse historical observations; while validations are encouraging, the product may not perfectly represent subsurface temperature evolution everywhere.
• The analysis period (1993–2019) is relatively short, potentially limiting robustness of some statistical relationships as more data accrue.
• Use of monthly means, while justified by decorrelation timescales >30 days, may smooth higher-frequency variability.
• Detrending choices have limited impact overall, but regions with strong warming trends (NEUS, Scotian, Labrador) show greater sensitivity, with raw data yielding larger extents and intensities/durations in recent years.
• Detailed MHW threshold definitions and event detection specifics are not fully described in the provided text; supplementary materials may contain additional methodological details.