Introduction

Heatwaves in the ocean, dubbed marine heatwaves (MHWs), are characterized by prolonged periods of extreme warm sea surface temperatures (SSTs) and are well known for adverse ecological and socio-economic impacts1,2,3,4,5,6,7,8. Notable MHWs have recently been reported from various regions of the globe9,10,11,12. In particular, a prominent MHW event developed during the winter of 2013–2014 in the Northeast Pacific, featuring unusually high SSTs in a “Blob-like” pattern in the Gulf of Alaska13,14. During the following winter, the patch of warm water shifted from the Gulf of Alaska toward the coastal regions15,16. Off the southern California coast, the maximum SST anomalies exceeded 5 °C relative to the 1982–2011 climatology. This MHW had significant impacts on marine ecosystems17 including harmful algal blooms18 and dramatic range shifts of species at all trophic levels19,20. More recently, warm SST anomalies intensified in the Northeast Pacific21,22,23,24,25 in the summer of 2019.

A variety of physical mechanisms have been proposed to explain the occurrences of Northeast Pacific MHWs3,26,27; for example, changes in the North Pacific high pressure system13,21, the presence of El Niño conditions in the tropical Pacific15,28,29,30,31 and the North Pacific Gyre Oscillation (NPGO)14. We note here that both MHW events occurred since 2013 when the Pacific Decadal Oscillation (PDO)32,33 became positive34, whereas few MHW events were reported in the prior decade or the so-called global warming hiatus period when the PDO was in a negative phase35,36,37. This fact suggests the potential importance of the PDO for decadal variations of Northeast Pacific MHWs26,38, especially when considering an inherent coupling between the NPGO and the subsequent development of PDO conditions39. Although the importance of the PDO for the development of MHWs was previously discussed on the basis of statistical analysis26,38,39, the role of the PDO in modulating decadal variations of Northeast Pacific MHWs and the associated physical mechanisms have not yet been explicitly investigated within fully coupled Earth’s system models. These aspects are then the focus of the current study.

Results

Observed decadal variations of Northeast Pacific MHWs

To explore the effect of the PDO on decadal variations of MHWs in the Northeast Pacific, we first examine the observed MHW changes (see Methods) over this region averaged over two positive PDO periods (1982–1998 and 2013–2022) and during a negative PDO period (1998–2013). We find that MHWs show generally longer (shorter) duration, higher (lower) annual frequency and larger (smaller) intensity off the Northeast Pacific coast during positive (negative) PDO periods (Fig. 1b–j). These robust MHW variations potentially reflect the effect of the PDO, but could also be influenced by many other factors. For example, the extent of background anthropogenic global warming could be different between 1982–1998 and 1998–2013, or between 1998–2013 and 2013–2022. We also note that negative PDO conditions appeared during 2020–2022 (Fig. 1a, Supplementary Fig. 1), suggesting that the positive PDO phase during 2013–2022 may have been short-lived. Finally, given that these results are based on only two positive PDO phases and one negative PDO phase, a more robust sample size is needed to confirm the relationships seen in Fig. 1.

Fig. 1: Observed decadal variations of North Pacific MHWs.
figure 1

a The monthly PDO index from 1891–2022 for the average of COBE-SST, COBE-SSTv2, ERSSTv5 and HadISST. b–d North Pacific MHW duration averaged over (b) 1998–2013 and averaged over the combined periods of (c) 1982–1998 and 2013–2022 as well as (d) the difference between the two (c minus b). The average is applied on each grid. e–g Same as (b–d) but for annual MHW frequency. h–j Same as (b–d) but MHW intensity.

Full size image

Northeast Pacific MHWs during positive and negative PDO phases

To compare MHWs between two distinct PDO phases from many different samples, we calculate the frequency, duration and intensity of MHWs in the Northeast Pacific during 2013–2022 for the positive and negative PDO groups obtained from the large ensemble simulations of five climate models (see Methods, Supplementary Fig. 2). While internal variability in each ensemble member has its own phase, here we consider the ensemble members which are in either a positive or negative PDO phase during 2013–2022 (see Methods). Like in observations, the MHW statistics from either group include the signals due to background anthropogenic forcing, whereas the difference between the two groups excludes the influence of background anthropogenic global warming (Fig. 2). Compared to the negative PDO group, the duration and annual frequency of MHWs in the Northeast Pacific significantly increase in the positive PDO group. From the Gulf of Alaska to off the California coast, MHW duration increases by about 3 days to 13 days (Fig. 2c) and MHW frequency enhances by about 0.4 to 1.1 times per year (Fig. 2f). MHWs also become significantly stronger off the coast of the Northeast Pacific in the positive PDO phase, except at a few sites in the Gulf of Alaska and close to Baja California (Fig. 2i). These model results, which are based on the comparison of robustly defined positive and negative PDO phases, are consistent with observations, indicating a role for the PDO in modulating Northeast Pacific MHWs during the recent decade, even if the PDO did not remain positive for the entire 2013–2022 period. Furthermore, we repeat our model analysis over two earlier decades, 1986–2005 (representing the positive PDO period of 1982–1998) and 1998–2007 (representing the negative PDO period of 1998–2013) and find that the PDO effects on MHWs during these two periods (Supplementary Figs. 3 and 4) are generally consistent with those in the period 2013–2022 (Fig. 2).

Fig. 2: Positive and negative PDOs modulating North Pacific MHWs.
figure 2

a–c Composite duration of Northeast Pacific MHWs for the ensemble mean of the (a) positive and (b) negative PDO groups from the large ensemble simulations of five climate models during 2013–2022 as well as (c) the difference between the two (a, b). d–f Same as (a–c) but for annual MHW frequency. g–i Same as (a–c) but for MHW intensity. Dotted indicates that the change is not significantly different from zero at the 95% level of the Student’s t-test.

Full size image

We also elucidate the PDO effect on warm SST extremes in the Northeast Pacific coastal region in terms of probability density functions (PDFs). We first calculate the PDF of daily SSTs over 2013–2022 with an interval of 0.1oC for all ensemble members that exhibit either a positive or negative PDO during that period across the five climate models (Fig. 3a). We find that the SST PDF of all ensemble members has a mean of 15.4 °C. The PDF shows the SST pattern under background anthropogenic global warming, since the simulated PDO phase is random in the large ensemble simulations so that all of the internal variability in the different ensemble members cancels out. As a result, the PDF based on all the ensembles shows SST variations without preferred PDO phase (Supplementary Fig. 2a). On the other hand, the daily SSTs in the negative and positive PDO groups show mean changes of −0.8 °C and 0.9 °C, with PDFs skewed toward their lower and higher tails (the skewnesses are −0.44 °C3 and 0.37 °C3), respectively. These PDF patterns reflect the characteristics of SST changes due to anthropogenic global warming plus a preferred negative or positive PDO. The difference of the PDFs of the two PDO groups accordingly manifests the role of the PDO in modulating the daily SSTs in this region. By defining warm SST extremes here as those that exceed the 90th percentile, we find that the positive PDO phase significantly increases the likelihood of a warm extreme, or potentially, an MHW occurrence off the Northeast Pacific coast (Fig. 3a). However, the duration criterion also needs to be met. This result is consistent with observations at Monterey Bay40, indicating that different PDO phases can alter the temperature and fertilization of near-shore surface water and potentially induce dramatic but distinct changes in coastal ecosystems and fisheries in the California Current region41,42,43.

Fig. 3: The mean-state difference and the physical mechanisms.
figure 3

a Probability density functions (PDFs) of SSTs in the coastal region of the Northeast Pacific (within the enclosed area in b) during 2013–2022 for all ensemble members (gray), the positive (ensemble mean, red; ensemble spread, light red) and negative (ensemble mean, blue; ensemble spread, light blue) PDO groups from the five climate models simulations. The 90th percentile of the PDF from all ensemble members is denoted by the black line. The ensemble spread is defined as one standard deviation among the ensembles. b The differences of SST trend (shading) and surface wind stress (vector) during 2013–2022 between the positive and negative PDO groups (positive minus negative) for the ensemble mean of the five climate models. c, d Same as (b) but for the difference of Ekman pumping and net surface heat flux. Surface heat flux is downward positive.

Full size image

The physical mechanisms of the PDO in modulating MHWs

The distinct MHW statistics during positive and negative PDOs from SST PDF analysis can be attributed to different mean-state SSTs between the two PDO phases. This dependence of MHWs characteristics on decadal-mean SSTs during different phases of climate variability is consistent with the MHW dependence on the long-term SST state under climate change44. Here, we calculate the difference of the 2013–2022 SST trend between the positive and negative PDO groups (Fig. 3b) and find contrasting SST changes in the western and central Pacific and along the coast of the Northeast Pacific. The coastal SSTs exhibit an anomalous warming trend of 0.6–2.3 °C/year during a positive PDO phase as compared to a negative PDO phase.

We propose two physical mechanisms that can potentially drive the warmer coastal SST during a positive PDO. Compared to a negative PDO, anomalous surface southeasterly winds prevail along the coast of the Northeast Pacific during a positive PDO (Fig. 3b), which produce anomalous downward Ekman pumping (Fig. 3c), reduce the upwelling of cold water, and result in warmer coastal SSTs. Note that climate variability over the tropical Pacific can also remotely affect this region via the propagation of coastal Kelvin waves45,46, whose offshore scale decreases with latitude. These coastal waves can excite westward propagating Rossby waves, and thus affect the offshore regions, but the Rossby waves decay rapidly away from the coast47, especially at higher latitudes where their phase speed is lower. Besides, more atmospheric heat enters through the ocean surface along the coast of the Northeast Pacific during a positive PDO (Fig. 3d). This increased net downward surface heat flux is primarily due to enhanced downward shortwave radiation and diminished upward sensible and latent heat fluxes (Supplementary Fig. 5). The latter, reduced turbulent sensible and latent heat fluxes, could be related to the anomalous surface winds that weaken the climatological winds, and in turn, decrease the surface wind speed. Noting here, relative to the negative phase of the PDO, changes in the Ekman pumping and net surface heat flux amount to 28% and 24% in magnitude during the positive phase of the PDO, which are sufficiently strong to alter the decadal-mean SSTs and impact the characteristics of MHWs. To summarize, both surface heat flux changes and ocean dynamics are important to the generation of warmer mean-state SST and hence longer, stronger and more frequent MHWs along the coast of the Northeast Pacific during a positive PDO.

We further quantify the PDO effect on Northeast Pacific MHWs during 2013–2022 by comparing it against the relative background anthropogenic global warming (Fig. 4a–c; Eqs. 1 and 2; also see Methods). We find that a positive PDO can prolong the MHW duration by up to 43%, increase the MHW annual frequency by up to 32%, and enlarge the MHW intensity by up to 10% off the Northeast Pacific coast (Fig. 4d–f). On the other hand, a negative PDO can shorten the MHW duration by up to 38%, decrease the MHW annual frequency by up to 29%, and reduce the MHW intensity by up to 10% in this region (Fig. 4g–i).

Fig. 4: Quantifications of the PDO effect on North Pacific MHWs.
figure 4

a–c The ensemble mean of (a) duration, (b) annual frequency and (c) intensity of Northeast Pacific MHWs during 2013–2022 for all ensemble members from the five climate models simulations, which reveals the effect of background anthropogenic global warming during the decade (\({\overline{{MHW}}}_{{agw}}\), see Methods). d–f The ensemble mean of \({r}_{{pdo}+}\) (see Methods) for (d) duration, (e) annual frequency, and (f) intensity of Northeast Pacific MHWs during 2013–2022 for the five climate models simulations. g–i Same as (d–f) but for the ensemble mean of \({r}_{{pdo}-}\) (see Methods).

Full size image

Discussion

In this study, we have explored the role of the PDO in driving MHWs in the Northeast Pacific on decadal timescales using both observations and climate model simulations. We find that, over recent decades, MHWs off the coast of the Northeast Pacific have become longer, more intense and more frequent under a positive PDO scenario relative to what is seen during a negative PDO scenario. These results are confirmed by the comparison of MHW statistics during positive and negative PDO conditions obtained from five climate models large ensembles. Analyses of these climate models reveal that these distinct MHW characteristics primarily resulted from the different mean-state SSTs between the two PDO phases. Compared to the negative phase of the PDO, warmer SSTs occur along the Northeast Pacific coast during the positive PDO, which are generated by suppressed coastal cold upwelling and enhanced net downward surface heat flux. We further quantify the contributions of the PDO and background anthropogenic global warming on Northeast Pacific MHWs over the past decade. Relative to background anthropogenic global warming, the positive PDO can prolong MHW duration by as much as 43% and act to increase MHW annual frequency by as much as 32% off the Northeast Pacific coast.

Our results suggest that periods of longer, stronger and more frequent Northeast Pacific MHWs can last for several years due to the persistence of a positive PDO state. Nevertheless, the characteristic MHW pattern induced by the positive PDO does not need to account for every event. For example, the “Blob” in February 2014 featured a peak warming centered around 145oW as likely related to the NPGO mode, which also makes sense in light of the inherent coupling between the NPGO and PDO39. Moreover, besides the North Pacific, the PDO could also have a remote effect on the SST in the Gulf of Maine during spring and summer through atmospheric teleconnections48 and potentially affect MHWs in the Northwest Atlantic.

The contribution of the PDO under the influence of background anthropogenic global warming on Northeast Pacific MHWs may change. As our climate continues to warm, the PDO is expected to weaken and to have a shorter period49,50,51. The underlying cause is that global warming enhances the ocean stratification in the North Pacific52, which leads to increased phase speed but decreased magnitude of extratropical oceanic Rossby waves across the North Pacific53,54. The increased Rossby wave phase speed results in a shorter transit time and therefore reduced periodicity of the PDO, while the reduced Rossby wave amplitude leads to a weaker PDO relative to the present. This alteration in PDO characteristics may further affect the predictability of the PDO51,55. It is also worth mentioning that, from an alternative point of view, the PDO is not a physical mode of variability like the El Niño-Southern Oscillation (ENSO). Rather, it is the result of the superposition of different processes, including ENSO teleconnections, re-emergence, and Rossby wave propagation in the North Pacific53. Given that current global climate forecast systems have shown relatively limited skill in the prediction of MHWs56 except for those linked to ENSO57, improving the understanding of changes in ENSO and PDO predictability under climate change will be fundamental for assessing the potential predictability of MHWs on interannual to decadal timescales.

Methods

Observations

We exploit the Daily Optimum Interpolation Sea Surface Temperature (OISST) version 2.1 (v2.1) from the National Oceanic and Atmospheric Administration58, mapped onto a 0.25° × 0.25° grid and available since September 1, 1981. We adopt the OISST data from 1982 to 2022 to investigate the observed MHWs in the Northeast Pacific. To examine the historical PDO, we leverage multiple reconstructed monthly SST datasets: COBE-SST59, COBE-SSTv260, ERSSTv561 and HadISST62, which are generally of a resolution of one or two degrees and cover more than a century. For each dataset, we calculate the SST anomalies during the period from 1891–2022 by removing both the climatological annual cycle and the global-mean SST at each grid point. We determine the PDO structure using the leading empirical orthogonal function (EOF) of the SST anomalies in the North Pacific polewards of 20°N32,63 and compute the PDO index as the corresponding principal component of the leading EOF. Despite slight differences, all four reconstructed monthly SST datasets show consistent PDO structures and indices (Supplementary Fig. 1). We hence use the average of the four PDO indices when examining the historical PDO (Fig. 1a).

Climate model simulations

Previous studies3,26,27 suggest that many physical drivers, including the PDO, operate under background anthropogenic warming such that it is difficult to clearly isolate and quantify the effect of each driver simply using short-record observational data. To this end, we leverage large ensemble simulations with climate models to eliminate the influence of background anthropogenic global warming and identify the effect of the PDO. The underlying rationale is that the predictability horizon of climate models is at most one to two decades55,64,65 so that they may not capture the same timing of internal decadal climate variability as observations64. For example, during the recent decade, 2013-2022, climate models can simulate the PDO in different phases among individual ensemble members but under the same anthropogenic forcing (Supplementary Fig. 2a). Thereby, the average of these large ensembles allows a cancellation between positive and negative PDOs and hence a representation of background anthropogenic global warming (Supplementary Fig. 2a). Meanwhile, a comparison between these members during different PDO phases will help uncover the role of the PDO in influencing various climate elements, including MHWs.

In this study, we consider large ensemble simulations conducted with five climate models that have daily SST outputs available: ACCESS-ESM1.566, CanESM567, CESM168, CESM269,70, and EC-Earth371, i.e., we use 172 ensemble members in total (Supplementary Table 1). The model PDO is calculated using the same approach as for the observed PDO. For each ensemble member, we calculate its PDO index as the principal component of the leading EOF based on the simulated SST anomalies over the period 1920–2022. We adopt a similar approach to that in ref. 72 by selecting the ensemble members that simulate a positive (negative) PDO during 2013–2022 into a positive (negative) PDO group (Supplementary Fig. 2b, Supplementary Table 1). We calculate the difference for the ensemble means between the positive and negative PDO groups to explore the PDO effect on MHWs (Supplementary Fig. 2). Here, it is worth noting that, although the five models follow different future anthropogenic warming scenarios after 2005 or 2014 (Supplementary Table 1), the inter-model difference in external forcing between 2013–2022 is fairly small and seldom affects the result in the current study.

Besides 2013–2022, we also selected the ensemble members that simulated a positive (negative) PDO during two earlier decades—1986–1995 and 1998–2007—into positive (negative) PDO groups, respectively (Supplementary Fig. 2c, d, Supplementary Table 2). For either decade, we calculated the difference in the ensemble means between the positive and negative PDO groups to explore the PDO effect on MHWs (Supplementary Figs. 3 and 4).

Detection of marine heatwaves

We adopt the MHW definition from ref. 73 in which a MHW represents an event in which daily SST exceeds the local seasonal threshold (i.e., the 90th percentile of daily SST for the same day during the climatology period) for at least five consecutive days within a given area. Two events with an interruption of <3 days are considered as one MHW event. For observations, we probe the MHWs during 1982–2022 using OISSTv2.1 so that the climatological period for MHW detection is 1982–2022. For climate model simulations, we focus on the PDO effect on MHWs during the recent decade. For each model, we use ensemble mean daily SST climatology during 2013–2022 for MHW detections. Though this 2013–2022 climatology or the neutral PDO state lasts only one decade, our ensemble mean approach in model simulations helps capture the background anthropogenic warming and allows for a better representation of daily SST climatology under global warming74. Similarly, we use ensemble mean daily SST climatology during 1986–1995 or 1998–2007 for MHW detections in either period. For all the reference periods in both observations and model simulations, the daily SST climatology and the 90th percentile threshold corresponding to each day are calculated using the daily SST for all years within 11 days centered on that day, and the results obtained are then smoothed for 31 days.

We explore MHW characteristics such as frequency, duration and intensity from observations and model simulations. Frequency is defined as the number of events per year; duration is defined as the time between the start date and end date of the event; and intensity is defined as the maximum amplitude of each MHW, i.e., the maximum SST anomaly relative to the seasonally varying climate mean over the duration of the event. We apply the Student’s t-test to the difference of MHW characteristics between positive and negative PDO groups to examine the statistical significance of the PDO effect on MHWs.

To quantify the PDO effect on Northeast Pacific MHWs during 2013–2022, we first calculate MHW duration, frequency and intensity averaged over all the ensembles of the five models (\({\overline{{MHW}}}_{{agw}}\)), which reveals the effect of background anthropogenic global warming during the decade. We then calculate MHW duration, frequency and intensity averaged over the ensembles of the positive (\({\overline{{MHW}}}_{{pdo}+}\)) and negative (\({\overline{{MHW}}}_{{pdo}-}\)) groups. Relative to the background anthropogenic global warming, the effects of the positive and negative PDOs on MHWs can be denoted by the ratios of \({r}_{{pdo}+}\) and \({r}_{{pdo}-}\), i.e.,

$${r}_{{pdo}+}=({\overline{{MHW}}}_{{pdo}+}-{\overline{{MHW}}}_{{agw}})/{\overline{{MHW}}}_{{agw}}$$
(1)
$${r}_{{pdo}-}=({\overline{{MHW}}}_{{pdo}-}-{\overline{{MHW}}}_{{agw}})/{\overline{{MHW}}}_{{agw}}$$
(2)