Corresponding author: Bruce T. Anderson, Department of Earth and Environment, Boston University, 675 Commonwealth Ave., Boston, MA 02215, USA. (email@example.com)
 Sea surface temperature variations over the equatorial Pacific associated with the El Niño/Southern Oscillation (ENSO) produce changes in climates across the globe. Here we report evidence from observationally constrained ocean data for the initiation of warm ENSO events (El Niños) resulting from subsurface equatorial Pacific heat content increases related to tropical/extratropical sea level pressure (SLP) changes over the North Pacific. We hypothesize that the increase in heat content is a response to SLP-generated variations in the North Pacific trade winds, which we term trade wind charging of the equatorial Pacific. Experiments using a high-resolution numerical ocean model verify that the charging of subsurface heat content along the equatorial Pacific can be induced by SLP-generated trade wind variations. Furthermore, analysis of the numerical model results and historical observations indicates that the trade wind–induced ocean heat content increases are sufficiently large to initiate the onset of El Niño events, which mature approximately 12 months after the trade wind forcing itself.
 The El Niño/Southern Oscillation (ENSO) imposes changes to regional climates across the globe, with significant physical, biological, and socioeconomic impacts [Bradley et al., 1987; Francey et al., 1995; Stenseth et al., 2002; Schubert et al., 2004; Behrenfeld et al., 2006; Hsiang et al., 2011]. Previous research [Trenberth and Shea, 1987; Barnett et al., 1988; Vimont et al., 2003; Anderson, 2003] has shown that variations in the ENSO—here characterized by the Niño-3 index, which captures the largest ENSO-related sea surface temperature (SST) departures [Rasmusson and Carpenter, 1982; Philander, 1983]—during its season of maximum variance (e.g., November–January) [Trenberth, 1997; Jin et al., 1994]—are preceded by a mode of sea level pressure (SLP) and wind stress variability in the central tropical/extratropical North Pacific approximately 12 months earlier (Figure 1).
 As highlighted by others [Vimont et al., 2009; Di Lorenzo et al., 2010], the pattern of preceding SLPs (Figure 1a) is characteristic of the positive phase of the North Pacific Oscillation [Walker and Bliss, 1932; Rogers, 1981; Linkin and Nigam, 2008], with significant negative SLP anomalies centered over the subtropical North Pacific that extend into the tropics and the Southern Hemisphere. A similar relationship is found when considering just those anomalies preceding El Niño (positive ENSO) events (Figure 1 in supporting material—hereafter, SM). This precursor pattern of SLP variability can be captured by a sea level pressure index (SLPI) centered on 175°W–140°W and 10°N–25°N [Anderson, 2003]. The seasonal mean value of the SLPI in boreal winter, defined here as November–March, has a significant negative correlation with the November–January ENSO state 12 months later (Figure 1b) such that negative boreal winter SLPI values can be significant predictors of El Niño events at lead times of around 1 year (Figure 1b in SM).
 In this study, we use observationally constrained ocean reanalysis data along with high-resolution numerical ocean model simulations to analyze the mechanism responsible for this statistical relationship. Previous evidence suggests that variations in the North Pacific trade winds, induced by the SLP anomalies, can directly force subsurface heat content changes across the central equatorial Pacific [Anderson, 2004; Anderson and Maloney, 2006], which serve as precursors to mature El Niño (warm ENSO) events [Jin, 1997; Meinen and McPhaden, 2000; McPhaden, 2003; Newman et al., 2011]. Here we present the first process-based confirmation of this “trade wind charging” (TWC) hypothesis for initiating changes in the state of the equatorial Pacific.
2 Ocean Analysis
 The evolution of atmospheric and oceanic fields taken from the observationally constrained ORA-S3 European Centre for Medium-Range Weather Forecasts (ECMWF) ocean reanalysis [Balmaseda et al., 2008]—including wind stress, SST, and oceanic heat content—concurrent with and following negative boreal winter SLPI values indicate significant weakening of the climatological northeasterly trade winds over most of the North Pacific basin, with a strengthening of the equatorial easterlies in the eastern portion of the basin during December–March (Figure 2). Initial (December) SST anomalies are found only over the central tropical and subtropical North Pacific, colocated with the overlying wind stress anomalies (Figure 2a). By January, positive SST anomalies form in the equatorial western Pacific (Figures 2b and 2c). At the same time, subsurface temperature anomalies develop in the central equatorial Pacific (between 180°W and 150°W) and over the central tropical North Pacific along 5°N (between 165°W and 135°W), consistent with the TWC hypothesis [cf. Anderson, 2004; Anderson and Maloney, 2006].
 During late boreal winter (February–March), the initial SST anomalies, stretching from the western equatorial Pacific into the central tropical and subtropical North Pacific, intensify (Figures 2d–2g). However, their position remains relatively stationary despite significant zonal wind stress anomalies to the south, suggesting that the SST anomalies are not changing in response to surface fluxes along the equatorward edge of the wind stress anomaly field. Instead the southernmost zonal wind stress anomalies are colocated with a strengthening of off-equatorial central Pacific temperature anomalies in the subsurface, which exceed those in the equatorial region. Vertical cross sections through these surface and subsurface temperature anomalies indicate that the largest subsurface off-equatorial warming is occurring along the thermocline near 3°N in the region of climatological warm water generated by the tropical cell circulation north of the Equator (Figures 2d–2g in SM), while surface warming is confined to regions north of 10°N.
 During April, the SST anomalies in the central portion of the basin begin to shift equatorward (Figures 2 h and 2i). Concurrently, the subsurface temperature anomalies south of the equator intensify and convolve with the equatorial and off-equatorial anomalies to the north, forming a nearly symmetric, extensive subsurface temperature anomaly spanning 180°W–120°W and 5°S–5°N representative of a charging of the warm water volume of the equatorial Pacific that serves as a precursor to mature El Niño events [Meinen and McPhaden, 2000; McPhaden, 2003; Newman et al., 2011]. During April and through June, the subsurface temperature anomaly expands eastward, reaching the eastern portion of the basin approximately a month before the SST anomalies themselves (cf. Figures 2j and 2l). In addition, the zonal wind stress anomalies shift south and become more symmetric about the equator. By June, positive SST anomalies are found across the tropical Pacific basin (Figure 2l). Over the next few months, these anomalies intensify, generating a characteristic El Niño–type pattern by the end of the calendar year (Figure 3 in SM).
3 Numerical Simulation of TWC
 To ascertain the importance of TWC in generating the surface and subsurface temperature evolution of the ocean, the wind stress and heat flux anomaly fields during the middle of boreal winter are composited for years with negative SLPI values (based on a weighted average of the values during those years; Figure 4, left, in SM). These anomaly fields are then used to force a high-resolution dynamical ocean model [Perez and Kessler, 2009] centered on the equatorial Pacific basin, using a ramp-up/ramp-down technique in which the full forcing fields are added to the climatological fields and persisted for the period December–March (Figure 4f in SM). The ocean model is then allowed to evolve over the next 3 months with only climatological forcing.
 The model response to the imposed forcing is consistent with the composite analysis detailed above. Subsurface temperature anomalies first develop along the off-equatorial central North Pacific in mid-December (Movie 1 in SM). As these intensify, additional subsurface anomalies form over the equatorial Pacific near the date line as well as in the off-equatorial central South Pacific. Subsequently, the equatorial and near-equatorial subsurface temperature anomalies strengthen and propagate eastward along the thermocline, shoaling and reaching the eastern equatorial Pacific by 1 March, when they start to warm the surface waters. After the cessation of wind stress and heat flux forcing at the end of March, the subsurface temperature anomalies associated with TWC begin to dissipate; however, their continued eastward propagation along the thermocline results in an intensification of the SST anomalies through May and June, setting the stage for the onset of an El Niño event.
 Sensitivity experiments designed to isolate the influence of (1) western equatorial Pacific wind stress anomalies, (2) central equatorial and off-equatorial Pacific wind stress anomalies, and (3) heat flux anomalies (Figure 4 in SM) highlight the important role of central Pacific wind stress forcing in initiating off-equatorial as well as equatorial subsurface temperature anomalies associated with TWC (Figure 3). Furthermore, these sensitivity experiments reveal that surface heat flux forcing principally influences the tropical SSTs in the region of weakening trade winds but in the absence of any coupling with the overlying atmosphere (which is treated more fully in Vimont et al. , Vimont et al. , Alexander et al. ) surface heat flux forcing has little influence on the subsurface temperature anomalies or on the subsequent emergence of SST anomalies in the eastern equatorial Pacific (Movie 2 in SM).
 Instead, the positive SST anomalies that develop over the eastern equatorial Pacific (Figure 5 in SM) are initially (from approximately 1 April to 1 June) a response to the shoaling of subsurface temperature anomalies initiated by western Pacific wind stress forcing (Movie 2 in SM), which propagate to the eastern equatorial Pacific as Kelvin waves [McCreary, 1976]. As the surface temperature anomaly signature of the western Pacific forcing begins to dissipate (June 1), the eastern equatorial Pacific SST anomalies are sustained and augmented by the shoaling of off-equatorial and equatorial subsurface temperature anomalies induced by central Pacific wind stress forcing (Movie 2 in SM)—the result of negative wind stress curl anomalies at 5°N (as well as positive wind stress curl anomalies at 5°S), which tend to produce a charging of the thermocline via equatorward vertically integrated mass transport as well as vortex stretching [cf. Clarke et al., 2007]—such that by the end of the integration the SST anomalies exceed 0.5 °C over a broad region of the eastern equatorial Pacific (Figure 5l in SM).
 As noted before, similar surface temperature features—including the initial warm anomalies in the central tropical North Pacific and the subsequent emergence and intensification of warm anomalies in the eastern equatorial Pacific—are found in the observationally constrained ocean reanalysis as well (Figure 6 in SM). Indeed, a comparison of SST anomalies in the eastern equatorial Pacific at the end of the numerical model integration, as characterized by the Niño-3 index, with those found in the historical record at the same time of year indicates that the wind stress–induced SST anomalies are (a) consistent with ones that develop into mature El Niño events over the course of the subsequent 4–6 months and (b) likely to evolve into mature El Niños over the course of the subsequent 4–6 months (Figure 4), even absent further equatorial wind stress forcing over the western Pacific during the May–June period.
 Observational evidence suggests that El Niños (in contrast to La Niñas) may require an initiation of subsurface warming across the Pacific basin distinct from that supplied by the system's oscillatory dynamics [Kessler, 2002]. Here we show that such initiation can be triggered through the wind stress–induced modification of the subsurface equatorial Pacific ocean temperatures via trade wind charging of the equatorial Pacific, a process driven by tropical/extratropical SLP changes over the North Pacific. In our study, we find no evidence that these SLP changes are related to the concurrent or preceding ENSO cycle (Figure 7 in SM). We emphasize that the mechanism outlined here is fundamentally different from previous ones investigating extratropical atmospheric forcing of the tropical Pacific. These include the oscillatory modification of the equatorial Pacific warm water volume associated with recharge/discharge mechanisms, which given the weakening of the trade winds analyzed here would produce a discharge of warm water from the equatorial Pacific [Jin, 1997]; the equatorward shift of atmospheric circulations induced by the seasonal persistence of subtropical SST anomalies associated with the Seasonal Footprinting Mechanism, which does not influence the structure of the subsurface equatorial Pacific until the following boreal spring [Vimont et al., 2001; Vimont et al., 2003]; trade wind–induced variations in evaporation associated with the Meridional Mode, which to first order involves only the surface temperatures [Alexander et al., 2010]; and Rossby wave dynamics, which on short time scales (<12 months) involve the initiation and westward propagation of subsurface ocean anomalies distinct from those analyzed here (as evidenced by their geographic location—west of 170E° at latitudes between 20°N–30°N [Solomon et al., 2008]).
 The TWC mechanism elucidated here is not considered to be the sole initiator of ENSO events—other initiating mechanisms, including those listed above, can contribute to the onset of a particular event [Neelin et al., 1998]. However, these results do support the idea that precursor changes in North Pacific extratropical sea level pressures and central tropical Pacific trade winds play a role in El Niño initiation. Atmospheric variability in this region is in part related to internal variability of the extratropical Pacific storm track [Linkin and Nigam, 2008; Caballero and Anderson, 2009; Riviere, 2010] and as such is inherently unpredictable [Di Lorenzo et al., 2010]. However, it has also been linked to deterministic processes in the tropical Pacific on interannual [Ashok and Yamagata, 2009] to decadal time scales [Di Lorenzo et al., 2010], suggesting the potential for enhanced predictability of ENSO even prior to the 12–15 months studied here.
 This work was supported by the Department of Energy (DE-SC0004975 to B.T.A.). R.C.P. was supported by the NOAA/Atlantic Oceanographic and Meteorological Laboratory and the Cooperative Institute of Marine and Atmospheric Studies at the University of Miami. A.K. was supported by National Center for Atmospheric Research, which is sponsored by NSF. We also extend our thanks to Ekaterina Oleinik at Boston University's Scientific Computing and Visualization group for all her work in designing and producing the accompanying animations.