Bottom marine heatwave intensity, duration, and spatial extent

Within each LME, the average intensity of BMHW events can strongly vary in space, with typical BMHW anomalies ranging from as low as 0.5 °C for deeper portions of the continental shelves in all LMEs (see Fig. 1 for the bathymetric details of each region) to as high as 5 °C for large portions of the Gulf of California at ~100 m depth (Fig. 2). In LMEs with more highly varying bathymetry along the continental shelf (e.g., Gulf of Alaska, NEUS, Scotian, and Labrador LMEs), the spatial distribution of average BMHW intensity is complex, with more intense events often linked to distinct bathymetric features (compare panels of Fig. 2 with those in Fig. 1). For example, in the NEUS LME, the intensity of BMHW events clearly differs between the relatively shallow Mid-Atlantic Bight (MAB; the coastal region between ~35˚N–41˚N) and deeper Gulf of Maine (NEUS LME basin north of 41˚N; Figs. 1g and 2g). Additionally, the southwest portion of the Scotian continental shelf is characterized by deep basins (bottom depths > 200 m) that are punctuated by shallow banks (bottom depths <100 m). As a result, average BMHW intensity varies strongly in this region, with more intense events typically found along the shallow banks and weaker events in the deeper basins (Figs. 1h and 2h). These bathymetrically diverse LMEs stand in contrast to those with relatively narrow/sharp continental shelves (e.g., California Current and Gulf of California) or relatively wide/smooth continental shelves (e.g., East Bering Sea, Gulf of Mexico, and SEUS). In these less bathymetrically varying LMEs, the distribution of average BMHW intensity is more uniform in space and tends to peak in regions close to the coast with shallower bottom depths (Figs. 1 and 2).

Fig. 1: Bathymetry along the continental shelves of North America. a–i Ocean bottom depth (m) for each of the nine Large Marine Ecosystems (LMEs) along North American coastlines. Ocean grid cells with bottom depths deeper than 400 m are shaded white. Land surfaces are shaded gray. Black contours outline each LME. Full size image

Fig. 2: Average intensity of bottom marine heatwaves. a–i Bottom water temperature anomalies (°C) averaged during all bottom marine heatwave (BMHW) months from 1993–2019 in each Large Marine Ecosystem (LME). Full size image

In an effort to quantify how BMHW intensity varies with ocean bottom depth, we present scatter plots of BMHW average intensity versus the corresponding ocean bottom depth at each grid cell for each LME (Fig. S1). Next, we convert these scatter plots into two-dimensional probability histograms for each LME by binning the grid cells into various intensity and depth intervals (Fig. 3). There is a clear negative relationship between BMHW intensity and ocean bottom depth in the East Bering Sea and Gulf of Alaska LMEs (Fig. 3a, b). The Spearman correlation coefficients between the intensity and ocean bottom depth across all grid cells within each of these LMEs are R = −0.84 and −0.9, respectively. In other regions, however, the relationship between BMHW average intensity and depth is more complicated. For example, in the Gulf of California LME and for portions of the California Current LME (Fig. 3c, d), BMHW intensity and bottom depth are positively correlated from the surface to ~100 m and then negatively correlated from ~100–200 m before asymptotically leveling off from 200–400 m. As a result, the warmest BMHW intensities occur at intermediate bottom depths of 50–100 m, with average values in this depth interval of 2.9 °C and 3 °C for the California Current and Gulf of California, respectively (e.g., Fig. 3c, d, gray dots). In the California Current LME, this non-linear BMHW-depth relationship is only found in the southern portion of the domain (Fig. S1c), while BMHW intensities tend to decrease roughly linearly with depth in the northern portion of the LME (R = −0.95), similar to those in the East Bering Sea and Gulf of Alaska LMEs.

Fig. 3: Bottom marine heatwave intensity variations with ocean depth. a–i Two-dimensional histograms of bottom marine heatwave (BMHW) average intensity (°C) versus ocean bottom depth (m) in each Large Marine Ecosystem (LME). Shading indicates the probability that a grid cell (with bottom depth <400 m) falls within a given intensity-depth interval. The Spearman correlation between BMHW average intensity and bottom depth across all grid cells within each LME is shown in the top right of each panel. Gray dots indicate the BMHW average intensity averaged across regular depth intervals of 50 m. The position of each dot along the x-axis represents the center of the depth interval used for averaging. For example, the first gray dot is positioned at 25 m, and represents the BMHW average intensity averaged across all grid cells with bottom depths of 0–50 m. Full size image

In contrast, the LMEs found along the Gulf of Mexico and the North American east coast do not show as clear of a BMHW-depth relationship (Fig. 3e–i), which may be due to the complicated bathymetric and oceanographic features in these LMEs. For example, bottom temperature variability in the MAB is likely dominated by instabilities in the shelf-break front33, while bottom temperature variability in the Gulf of Maine is associated with strong tidal mixing34 and possibly advection by the Labrador Current35 or the mixing of Gulf Stream waters at depth via the Northeast Channel36. These different flow-bathymetry interactions may explain the BMHW intensity differences discussed previously (e.g., Fig. 2g) and ultimately lead to a scattered BMHW-depth relationship in which average intensities peak at ~100 m, but otherwise do not closely scale with bottom depth (Fig. 3g).

The average duration of BMHW events also exhibit strong spatial variations across the different LMEs (Fig. 4). In several LMEs, longer duration BMHWs are associated with deeper portions of the continental shelf, such as in western portion of the East Bering Sea, the Gulf of Maine in the NEUS, the southeastern portion of the Scotian Shelf, and the northeastern portion of the Labrador LME (comparing Figs. 1a, g, i with Figs. 4a, g, i). Outside of these regions, however, the BMHW duration is highly variable and not clearly linked to specific bathymetric features. The California Current LME shows a clearer pattern of average duration, with BMHWs in the southern portion of the domain lasting longer than BMHWs in the northern portion of the domain (Fig. 4c). Similarly, the BMHW events in the Gulf of Mexico and SEUS LMEs show generally uniform average durations of ~1.5 months (Fig. 4e, f). The majority of the LMEs exhibit noisy patterns of BMHW longevity that are not strongly related to bottom depth (Fig. S2).

Fig. 4: Average duration of bottom marine heatwaves. a–i The average duration (months) of all bottom marine heatwave (BMHW) events from 1993–2019 in each Large Marine Ecosystem (LME). Full size image

In order to diagnose the prevalence of BMHWs on the broader LME-scale, we assess the spatial extent of these events with time (Fig. 5). In the East Bering Sea, there are two major periods of widespread and prolonged BMHW events, including several from 2002–2006 and again from 2015–2018 (Fig. 5a). At its peak, the BMHW event beginning in 2016 encompasses 60% of the total area of the LME with a maximum average intensity of 2.5 °C. These prolonged and widespread BMHW conditions correspond with a well-known warming event which produced myriad marine ecosystem impacts37.

Fig. 5: Spatial extent of bottom marine heatwaves. a–i The fraction of each Large Marine Ecosystem’s (LME) area experiencing bottom marine heatwave (BMHW) conditions for each month from 1993–2019. Shading denotes average BMHW intensity (°C) in a given month, as measured by bottom water temperature anomalies averaged across all grid cells experiencing BMHW conditions. Horizontal gray lines mark areal extents of 0.5 and 1. Note only grid cells with bottom depths <400 m were used for areal percentage and intensity calculations. Full size image

The Gulf of Alaska, California Current, and Gulf of California LMEs similarly feature two pronounced periods of prolonged and spatially widespread BMHW events, including 1997-1998 and 2014–2017 (Fig. 5b, d). The BMHW in 1997/1998 is the most widespread and intense of the two time periods, with peak areal extents of 0.72, 0.96, and 0.87 and peak average intensities of 1.6 °C, 3.5 °C, and 5 °C for the Gulf of Alaska, California Current, and Gulf of California LMEs, respectively. These intense BMHW conditions correspond with the 1997/1998 El Niño event, suggesting that BMHW events along the North American west coast, like their surface counterparts, may be linked by large-scale climate forcing related to the El Niño-Southern Oscillation (ENSO). We will return to this point in the Discussion section.

The BMHW events in the Gulf of California, California Current, and Gulf of Alaska LMEs from 2014–2016 feature two peaks, one in 2014 to early 2015 and another in late 2015 to 2016. The first of these peaks correspond with the evolution of a series of major Northeast Pacific MHWs12,14. During this time period, large-scale atmospheric circulation anomalies produced strong surface winds along the North American west coast, contributing to the development of MHW conditions in the California Current System12,38,39. Surface forcing associated with these wind anomalies likely contributed to the intensity and areal extent of the 2014-2015 BMHW conditions seen in the California Current LME, which peaked at 67% LME coverage and an average intensity of 2.6 °C. The intensity, spatial extent, and persistence of widespread BMHW conditions throughout the Gulf of California, California Current, and Gulf of Alaska from late 2015 into 2016 consistent with a series of downwelling coastally trapped waves that propagated up the North American coastline during the development of the 2015/2016 El Niño40. In all, BMHW conditions from 2015-2016 peaked at 0.81, 0.64, and 0.61 areal coverage and a peak average intensity of 1.4 °C, 3.0 °C, and 3.2 °C. for the Gulf of Alaska, California Current, and Gulf of California LMEs, respectively.

The BMHW conditions in the Gulf of Mexico and SEUS are generally less coherent (e.g., more instances with fractional area <0.5) on the LME-scale (Fig. 5e, f) than those along the Pacific coast, suggesting the presence of within-LME subregions of isolated BMHW activity that may be tied to specific bathymetric features or local processes. Although the spatial extent of BMHW conditions in the NEUS, Scotian Shelf, and Labrador LMEs are generally lower than those along the Pacific coast, these LMEs do show prolonged BMHW events during 1999-2000 and 2011-2012. In particular, the BMHW in 2011-2012 peaked at 0.64, 0.5, and 0.57 areal coverage with peak average intensities of 3 °C, 3.8 °C, and 1.8 °C for the NEUS, Scotian Shelf, and Labrador LMEs, respectively. The coherent BMHW conditions in 2011-2012 may be related to large-scale atmospheric forcing associated with the broader 2012 Northwest Atlantic MHW41,42.

Comparing bottom and surface marine heatwaves

In the East Bering Sea, the Gulf of Alaska, and the northern portions of the California Current LME, SMHW intensity tends to be ~0.5 °C–1 °C warmer than BMHW intensity (Fig. 6a, c; blue shading), while in the southern California Current, Gulf of California, Gulf of Mexico, and SEUS LMEs, an average BMHW is anywhere from 0.5 °C–2.5 °C warmer than an average SMHW (Fig. 6c–f; orange/red shading). The NEUS, Scotian, and Labrador LMEs show more varied intensity difference patterns that are likely driven by complex current-bathymetry interactions. For example, there are once again clear differences between the MAB and the Gulf of Maine in the NEUS LME. In the MAB, an average BMHW is ~1 °C–2 °C warmer than an average SMHW, but SMHWs tend to be ~0.5 °C warmer than BMHWs in the interior portions of the Gulf of Maine where the seafloor is deepest (Fig. 6g).

Fig. 6: Bottom vs. surface marine heatwave intensity. a–i Difference between bottom water temperature anomalies (°C) averaged during all bottom marine heatwave (BMHW) months and sea surface temperature anomalies (°C) averaged during all surface marine heatwave (SMHW) months from 1993–2019 in each Large Marine Ecosystem (LME). Full size image

Compared to average BMHW intensity, average SMHW intensity is less variable in space within a given LME (Fig. S3). As expected, the difference between average BMHW and SMHW intensity in each LME is close to zero for the shallowest ocean bottom depths, where SMHWs and BMHWs converge (Fig. S4). In each LME, the average BMHW persists longer than the average SMHW at almost every location (Fig. 7), likely due to the fact that SMHWs are damped by the atmosphere via turbulent heat fluxes while the ocean bottom is largely insulated from such changes. This hypothesis is supported by the slower decorrelation timescale43 of BWTAs compared to SSTAs in each region (Figs. S5–S7), leading to a slower breakdown of BMHW conditions from one month to the next.

Fig. 7: Bottom vs. surface marine heatwave duration. a–i The difference in average duration (months) between bottom marine heatwaves (BMHWs) and surface marine heatwaves (SMHWs) from 1993–2019 in each Large Marine Ecosystem (LME). Full size image

Spatial coherence tends to be higher for SMHWs than for BMHWs, as indicated by the total number of months in which SMHW or BMHW conditions are widespread (i.e., exceeding 50% of an LME’s area; Fig. 8, numbers next to LME name). In every LME, there are more months with widespread SMHW conditions than with widespread BMHW conditions. Despite these differences, however, the BWTAs associated with spatially coherent BMHWs are often warmer than the concurrent overlying SSTAs (Fig. S8). Of note are the anomalous temperature differences between the ocean bottom and surface during the 1997/1998 and 2015/2016 North Pacific MHW events. In the California Current LME, BWTAs were greater than 1 °C warmer than the co-located SSTAs during the 1997/1998 event, while BWTAs in the Gulf of California LME were greater than 2.5 °C warmer than the surface anomalies during the 1997/1998 El Niño and the 2015/2016 MHW (Fig. S8c, d).

Fig. 8: Bottom vs. surface marine heatwave spatial extent. a–i The fraction of each Large Marine Ecosystem’s (LME) area experiencing surface marine heatwave (SMHW) conditions for each month from 1993–2019. Shading denotes average SMHW intensity (°C) in a given month, as measured by sea surface temperature anomalies (SSTAs) averaged across all grid cells experiencing SMHW conditions. Black contours mark fraction of LME’s area in bottom marine heatwave (BMHW) conditions (i.e., bar height in Fig. 5). Horizontal gray lines mark areal extents of 0.5 and 1. Note only grid cells with bottom depths <400 m were used for areal percentage and intensity calculations. Numbers next to LME names indicate total number of months with areal extent greater than 0.5 for SMHWs and BMHWs, respectively. Full size image

In each region, there are periods of spatially widespread SMHW events with peak areal extents of greater than 0.5 that closely correspond to concurrent widespread BMHW conditions (Fig. 8). For example, in the East Bering Sea, there are widespread SMHW conditions in 2001, 2003-2004, and 2016 that occur in conjunction with widespread BMHW conditions. The same is true for the Gulf of Alaska, California Current and Gulf of California LMEs from 1997/1998 and again from 2015/2016. There are also periods of widespread SMHW and BMHW conditions in the Gulf of Mexico and SEUS in 2016/2017 and in the NEUS, Scotian, and Labrador LMEs during the well-known 2012 warming event.

There are, however, some notable differences in the spatial extent of SMHW and BMHW events. In particular, in the Gulf of Alaska and California Current LMEs, while SMHW and BMHW conditions do occur simultaneously during the development of the 1997/1998 El Niño, there is a notable lag between the most widespread (>~0.5 areal extent) SMHW conditions and the most widespread BMHW conditions (Fig. 8b, c). Specifically, a BMHW with large spatial extent follows a widespread SMHW several months later. In the Gulf of Alaska, these BMHW conditions persist as many as 7 months after SMHW conditions have subsided. A somewhat similar lagged relationship is seen in the Gulf of Alaska during the evolution of the weak 2014/2015 El Niño. The differences in timing may be related to the different physical mechanisms relevant to the formation of widespread SMHW or BMHW conditions. For example, the coastal surface ocean in the Gulf of Alaska and California Current LMEs is strongly influenced by large-scale atmospheric teleconnections associated with the Pacific-North American (PNA) pattern44, which develops rapidly in response to tropical heating associated with ENSO45. Whereas the ocean bottom is likely more sensitive to slower adjustments in subsurface currents, the propagation of coastally trapped waves, or vertical displacements of the thermocline46. These different mechanisms will be discussed in more detail in the Discussion section. In addition to the differences seen along the North American west coast, there are key differences in the spatial extent of SMHW and BMHW events along the east coast. In the Scotian and Labrador LMEs, a widespread BMHW event begins in 1999 and persists into 2000 (Fig. 8h, i). However, the corresponding surface warming is comparatively disjointed in time. Also in the Labrador LME, there is a persistent and widespread BMHW event from 2010-2013; however, from late 2011 to early 2012 there is not a corresponding SMHW. Instead, surface warming is broken up into two separate events, one in 2011 and another in the winter of 2012/2013.

Bottom and surface marine heatwave synchrony

In almost every LME, there is a clear pattern of synchrony between BMHWs and SMHWs, with the two co-occurring more often over shallower portions of the shelf (Fig. 9). As a result, for the majority of the LMEs, the synchrony of BMHWs and SMHWs scales with depth following a relationship that can be approximated using a 2nd-order polynomial model fit (Fig. S9). Therefore, the LMEs with wider continental shelves and more area at shallower depths (e.g., East Bering Sea, Gulf of Mexico, and SEUS; Fig. 1) have a larger fraction of their area in which BMHWs and SMHWs co-occur.

Fig. 9: Co-occurrence rate of bottom and surface marine heatwave events. a–i Bottom marine heatwave (BMHW) and surface marine heatwave (SMHW) temporal synchrony as measured by the fraction of months from 1993–2019 in which BMHWs and SMHWs co-occur in each Large Marine Ecosystem (LME). A value of 1 indicates that BMHWs and SMHWs co-occur 100% of the time. Full size image

The higher synchrony of BMHW and SMHW events for shallow ocean bottom depths may be expected since it is more likely in shallow regions for the mixed layer depth (MLD) to extend to the ocean floor, at which point the physical characteristics (e.g., temperature) of bottom waters would match those near the surface. We test this hypothesis by first calculating the ratio of the time varying MLD to the ocean bottom depth at each grid cell within each LME. We then composite this MLD/bathymetry ratio during months when a BMHW and SMHW are co-occurring (Fig. S10). In each LME, the synchrony of BMHWs and SMHWs is highly correlated with MLD/bathymetry ratio at each grid cell (Fig. 10), such that BMHWs and SMHWs tend to co-occur more frequently when the MLD/bathymetry ratio is high, i.e., the closer to the ocean bottom the MLD reaches during a SMHW, the more likely a BMHW is to occur.