The warm Blob in the northeast Pacific—the bridge leading to the 2015/16 El Niño

We address the occurrence of the warm anomaly, known as the Blob, that developed from late 2013 to 2015 in the northeast Paci ﬁ c and its connection with the El Niño – Southern Oscillation (ENSO) variability. The warm Blob results from the enhanced second ocean – atmosphere (O – A) coupled mode of variability in the tropical and North Paci ﬁ c, representing a small part of the Victoria mode (VM) in the northeast Paci ﬁ c forced by the strengthened North Paci ﬁ c Oscillation-like atmospheric pattern since 2013. We also show that this second O – A mode re ﬂ ects the meridional variability through the tropical – extratropical teleconnection and is an important precursor to the ENSO variability. The process is con ﬁ rmed by the coupled patterns that evolved from late 2013 to 2016 and the multi-year persistence of the warm Blob. We emphasize the role of evolving basin-scale VM but not the warm Blob itself prior to the


Introduction
An extremely warm sea surface temperature (SST) event developed in the northeast Pacific in late 2013. The warm SST anomaly, which became popularly known as 'the Blob' , reached temperatures more than 2.5°C higher than the climatological mean SST in a few extratropical regions of the North Pacific (Bond et al 2015) and was prominent in the south of the Gulf of Alaska (figure 1). The warm Blob has attracted much attention. Bond et al (2015) attributed the development of the Blob to strongly positive sea level pressure (SLP) anomalies over the Gulf of Alaska that suppress the local ocean heat loss in the atmosphere. As we will show later, the mechanism described resembles the typical ocean response to the strengthened northern lobe of the North Pacific Oscillation/West Pacific Pattern (NPO/ WP) (Linkin andNigam 2008, Baxter andNigam 2015) over the northeast Pacific.
From early 2014, the North Pacific gradually developed a spatial pattern with negative SST anomalies extending from the subtropical western Pacific to the central North Pacific encircled by warm SST anomalies around the North Pacific coast that reach the central tropical Pacific and the western Bering Sea (see the evolution of three-month averaged SST, SLP, and surface wind anomalies in figure 1). This SST distribution resembles the spatial pattern of the Victoria mode (VM), which is a basin-scale pattern combining the active roles from both the western (cold) and eastern (warm) North Pacific (Bond et  Original content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
part of the basin-scale VM is commonly known as the Pacific meridional mode (PMM; Chiang andVimont 2004, Chang et al 2007) or North Pacific gyre circulation (NPGO; Di Lorenzo et al 2008, Di Lorenzo et al 2015. The major differences between the VM and PMM/NPGO are described by Ding et al (2015a) and we use the VM in this paper to emphasize the important basin-scale changes instead of the regional response. A neutral to weak El Niño condition developed at the end of 2014 (Levine and McPhaden 2016). The heat content and SST anomalies in the tropical Pacific continuously evolved into a strong El Niño after April 2015 that reached its mature phase at the end of 2015 due to the phase locking to the seasonal cycle. Hereafter, we refer to the most recent strong El Niño event, starting from the end of 2014 and lasting to early 2016, as the 2015/16 El Niño (a single event). Based on the SST data, the 2015/16 El Niño surpassed the 1997/98 El Niño, making it one of the strongest El Niño events on record (e.g. more than 3°C warmer than the climatological mean in the Niño 3.4 region).
There are many questions about the two extremely warm SST events in the extratropical and tropical  To emphasize the large-scale pattern and the evolution variability, a three-month low-pass and 4°Â 4°spatial Blackman-window filter is applied to the monthly SLP and SST anomaly fields for the CEOF. Before computing the CEOF, both anomaly fields are normalized by the domain average standard deviations. Hereafter, seasons refer to those of the NH.

Community Earth System Model large ensembles (CESM-LENS)
To further support the relation between the two dominant O-A modes, we also analyze the historical simulations from the large ensembles (LENS) of CESM version 1.1 (Kay et al 2015). The LENS are forced with historical greenhouse gases and aerosol emissions from 1920 to 2005. These ensembles only differ in their 1920 initial conditions at the round-off error level (Kay et al 2015). Since all ensemble members show very similar results, we only present the control simulation here from 1850-2005 to emphasize the robust feature.

Results
In order to demonstrate that the warm Blob is indeed the SST signature of the second coupled O-A mode in the northeast Pacific, we first show the two leading CEOF modes and their corresponding time series of principal components (PCs) in figure 2. The variances of the leading (CEOF1) and second (CEOF2) CEOFs are 36.2% and 10.5%, respectively. The third mode (8.6% variance) is not addressed here because it is irrelevant to our discussion. The CEOF1 of the SLP anomalies in the Pacific (figure 2(a)) shows mainly the Aleutian Low (AL), the semi-permanent low pressure winter center over the Aleutian Islands caused by planetary waves, in association with the typical Southern Oscillation dipole near the tropics (a strong western Pacific subtropical high and the other strong pressure low near the eastern tropics). They are the typical atmospheric expressions accompanying Pacific Decadal Oscillation (PDO) and El Niño--Southern Oscillation (ENSO) patterns. This canonical PDO/ENSO pattern also emerges in the CEOF1 of the SSTanomalies (figure 2(c)), with warm anomalies in the cold tongue from the central-eastern tropical Pacific and cold anomalies in the western Pacific, which extend to the central North Pacific in the mid-latitudes.
Environ. Res. Lett. 12 (2017) 054019 Furthermore, the temporal evolution of PC1 is strongly correlated with the Niño 3.4 index (correlation coefficient R ¼ 0.88; time series compared in figure 2 (e)) and is moderately correlated with the PDO index (R ¼ 0.67). Therefore, we can conclude that the spatial pattern of CEOF1 and its PC1 approximately correspond to the ENSO variability and the persistent teleconnection with the PDO through the atmospheric bridge (Zhang et al 1997, Newman et al 2016, Di Lorenzo and Mantua 2016. The CEOF2 of the SLP anomalies in the North Pacific (figure 2(b)) presents the typical meridional dipole structure of the NPO in the central-eastern Pacific, with positive anomalies north of 45°N over the Aleutian Islands and negative anomalies in the south between 0°and 40°N (over Hawaii). This NPO pattern is very similar to that reported previously (Linkin andNigam 2008, Furtado et al 2012) and is a robust winter atmospheric feature. In addition, the associated SST footprint resembles the VM described previously (Bond et  This consistency suggests that the Blob is the SST signature of the CEOF2 in the northeast Pacific. Spatially, the SST anomalies shown in early 2014 (figures 1(b) and (c)) qualitatively resemble figure 2(d) in the central to eastern Pacific. In winter 2013/14 (figure 1(b)), the SLP anomalies show a clear quasimeridional dipole with low/high SLP anomalies at south/north in the extratropical region, similar to the NPO dipole in figure 2(b), but have a much larger variation; see Baxter and Nigam (2015) for further discussion of the 2013/14 NPO pattern. The spatial pattern had evolved into the CEOF1-like pattern by the end of 2014 (the AL and the subtropical high in the SLP anomalies with PDO/ENSO pattern in the SST anomalies; see figure 1(f)). But the subtropical high in the western Pacific is not very intense with a weak zonal dipole in the tropics, suggesting a very weak El Niño like phenomena (Central Pacific warming signature, see Tseng et al 2017). In the extratropical Pacific, we can still observe a similar meridional dipole of SLP anomalies with an enhanced southern pole (i.e. enhanced AL) in winter 2014/15. As we mentioned earlier, the meridional dipole can enhance the Pacific VM pattern in spring which can favor the development of ENSO. Therefore, we can see the evident VM pattern in spring 2015 ( figure 1(g)). This VM pattern We next establish the connection between the CEOF2 and the ENSO variation. Figure 3(b) shows the lead-lag correlations between PC2 and two different ENSO indices (the Niño 3.4 index and the El Niño Modoki index (EMI), defined by Ashok et al (2007)). The similar correlation between PC2 and the SST anomalies over the Blob region is also shown for comparison. It is evident that PC2 consistently leads the Niño 3.4 and EMI (R ¼ 0.55 and 0.68, respectively) with slightly different lags. These results confirm that the appearance of the CEOF2 prior to ENSO occurrence (Furtado et al 2012) and the spatial pattern of CEOF2 can potentially be seen as precursors of ENSO. In general, PC1 and the Niño 3.4 index are almost in phase, but the CEOF2 pattern associated with PC2 can last several months (approximately 6-10) before PC1 matures. A similar lead of PC2 also applies to the termination of ENSO. A large phase change of the CEOF2 can be found clearly a few months before the termination of ENSO in many large events (e.g. 1972/73, 1982/83, 1997/98, 2015/16 El Niños and 1973/74, 1988/89, 1998, see figure 2(e).
We further list all medium to strong ENSO events (total 18 El Niños and La Niñas) after 1965 in table 1. 83% of the ENSO events (15 out of 18) are associated with large SSTanomalies in the warm Blobs more than nine months in advance (bold: larger than one standard deviation). These strong evidences indicate that the SST anomalies of the Blob (with shorter lag time than the two ENSO indices) may be a bridge between the first two dominant coupled modes and, specifically, an SST expression of the CEOF2 pattern of NPO/VM leading to the CEOF1 pattern of PDO/ ENSO. The evolution can be further seen in the correlation between the Pacific SST anomalies and the January Niño 4 index at different lags during the period 1958-2010 (figure S2 in the supporting information; the CEOF1 SST pattern is similar to figure S2(f) and the CEOF2 SST pattern is similar to figure S2(c)).
These results are consistent when the HadSST and the HadSLP2 are used to represent the coupled O-A mode (figure S3). The spatial patterns of the first two coupled models and their variances are quite similar. Although the maximum correlations are slightly lower using the HadSST/HadSLP2 data, the leading role of CEOF2 (associated with the SSTanomalies' variability in the Blob region) in the ENSO variability is still significant at the expected timing. The correlation is always better if the NCEP data is used instead of HadSLP2 regardless of SST observations. The linkage between the CEOF1 and CEOF2 can be better represented in the O-A coupled climate model. Figure 4 shows the two dominant CEOF patterns and the lead-lag correlation based on the control simulation of CESM-LENS during 1850-2005. The described connection between the CEOF2 and the ENSO variation is quite robust with correlation above 0.55 regardless of ENSO types. In fact, all CESM-LENS simulations show such a strong relationship between the PC2 of the CEOF2 and the ENSO with minor differences in the variance and pattern, supporting the close lead-lag relation between the first two dominant modes (see figure S4 for the selected CESM-LENS simulations and also the discussion in Deser et al 2012).
Several studies have indicated that the CEOF2 responds to the atmospheric forcing of the NPO.
Environ. Res. Lett. 12 (2017) 054019 Ding et al (2015a) suggested that the VM might act as an effective pathway for NPO-like atmospheric variability to drive ENSO variability via the seasonal footprinting mechanism (SFM) (Vimont et al 2003).
The warm Blob that developed after late 2013 (figure 1) is actually the northeast part of the VM and was actively involved in the dynamical process that evolved into the 2015/16 El Niño. The evolving   To verify that the warm Blob that developed after late 2013 is not a unique event but a direct and strong NPO footprint commonly seen in the northeastern Pacific before the El Niño, we further illustrate lead-lag correlations between the Pacific SLP and SST anomalies, and the January NPO index (SLP difference between 20- 30°N, 190-210°E and 55-65°N, 190-210°E, corresponding to the pattern in figure 2 (b)) from 1958 to 2016 in figures 3(c)-(h). We confirm that the patterns of the SLP and SST anomalies at lag 0 and 2 months (bottom two panels) closely match the CEOF2 patterns shown in figure 2 and the evolution in 2014 (figures 1(a)-(c)). The NPO is strengthened initially, followed by the prominent pattern of SST anomalies through the development of the CEOF2. Similar correlations of the SLP anomalies persist until March (figure 3(g)). Thereafter, the SLP anomalies diminish gradually over time. Figure S5 further shows the lead-lag correlation between the North Pacific SST anomalies (blue box in figure 2(d)) and the NPO index. The largest correlation exists in February and March, consistent with the described pattern. The southern lobe of the NPO (figure S5(c)) actually plays a larger role than the northern lobe.

Discussion
Two other super El Niño events have been observed in the last three decades (i.e. 1982/83 and 1997/98). Why, therefore, did we not see the extremely high and positive PC2 of CEOF2 (or a large warm Blob as the ocean signature) in these two El Niño events (figure 2 (e)) if it is a useful precursor? In fact, Ding et al (2015a) showed that the positive VM existed in both periods (their table 1), indicating the persistent impact of the VM on the El Niño (also seen as the positive and large PC2 in figure 2(e) before these super El Niños). However, the magnitude of the PC2 in those years was weaker than that of the most recent event. This is mainly because the defined CEOF2 only corresponds to the triggering and amplifying forcing from the NH. The other dominant ENSO amplifier results from the   1965/66, 1972/73, 1982/83, 1986/87, 1987/88, 1991/92, 1997/98, 2002/03, 2009/10, 2015/16 La Niña 1970/71, 1973/74, 1975/76, 1988/89, 1998/99, 1999/00, 2007 We note that not all positive VM events lead to El Niños. Ding et al (2015a) found that over the period 1950-2011, ∼30% of the positive VM events did not lead to ENSO development. But more than 80% of moderate to large ENSO events have the signature of positive (negative) VM prior to El Niño (La Niña), see table 1. The surface patterns associated with these VM events did not extend far enough toward the equator compared with those associated with other VM events related to ENSO. To initiate ENSO development, the preconditioned tropical subsurface heat content and its propagation are required to provide favorable tropical ocean conditions at interannual time scales (Chen et al 2015, Levine and McPhaden 2016, Tseng et al 2017. In terms of the 2015/16 El Niño, the warm water content was high from spring 2014 to early 2016 (www.pmel. ñoaa.gov/elnino/upper-ocean-heat-content-and-enso). This provides a favorable ocean condition for El Niño development. Therefore, both processes of O-A coupling and subsurface ocean temperature evolution are required to make these VM events effective in triggering warming from the central to eastern equatorial Pacific.

Conclusions
We showed that the warm Blob that developed in late 2013 and then decayed after late 2015 in the northeast Pacific was the ocean expression of the second O-A mode (CEOF2) variability in the tropical and North Pacific (i.e. a part of the VM) and was an intermediate step leading to the 2015/16 El Niño. A neutral to weak El Niño condition was initially observed at the end of 2014, similar to the CEOF1 pattern but with weak magnitude (mainly due to the suppression of the SH subtropical impact). The heat content and SST anomalies in the tropical Pacific continuously evolved into a strong El Niño after April 2015 resulting from the re-intensification of the tropical-extratropical interaction, which reaches its mature phase at the end of 2015 with a robust CEOF1 pattern. With a weakened CEOF2 pattern in 2015/16 winter, the CEOF1 pattern (i.e. El Niño) decays dramatically in 2016.
Such CEOF2 patterns are frequently associated with the development/termination of an El Niño in the following winter/spring under certain conditions. The extremely large heat content accumulation in the tropics in 2014 (interannual time scale variation) intensifies the tropical-extratropical interaction (Di Lorenzo and Mantua 2016) so as to extend the 2015/16 El Niño longer than the previous El Niño events during the observational period (i.e. from the end of 2014 to early 2016). The 2015/16 El Niño differs from the previous large El Niño events in terms of the triggering hemisphere. Indeed, the basin-scale VM is still very clear in spring 2015 (not shown) but not the NPO dipole, thereby supporting the more direct connection of the VM than the NPO with ENSO variability. Here, the NPO only triggers the VM in late 2013, which favors the development of El Niño but does not guarantee its occurrence. However, throughout 2015, the El Niño develops fully through the positive Bjerknes feedback, leading a strong CEOF1like pattern. These similarities confirm that the ENSO and the Blob are linked rather than independent phenomena. The record high recent warm Blob is a typical precursor to the strong 2015/16 El Niño. Our analysis does not focus on the causes of NPO variability, which require further model studies.