Previous studies suggest that wintertime North Pacific Oscillation (NPO) is able to force an El Niño event during the following winter via the seasonal footprinting mechanism (SFM). In this study, we present evidence that springtime Arctic Oscillation (AO) has a significant modulation effect on the connection between the NPO and El Niño. When the spring AO is positive, a positive wintertime NPO can result in significant El Niño-like warming anomalies via the SFM. However, when the spring AO is negative, the connection of the NPO and the El Niño is not robust at all. Thus, the phase of spring AO should be taken into account when using NPO as a predictor for El Niño. Further analysis reveals that the mechanism of the AO modulation may be through changing the underlying SST footprinting over the subtropical northeastern Pacific.
 The El Niño–Southern Oscillation (ENSO) is one of the strongest natural variability modes in the tropical Pacific, which can exert substantial influences on the global climate anomalies [Lau and Nath, 1996; Alexander et al., 2002; Yu and Zwiers, 2007; Chen et al., 2013b]. Thus, the understanding of the dynamics underlying the ENSO has been a long-standing issue. Previous studies suggest that stochastic forcing from the midlatitudes is essential for maintaining ENSO variability [Graf, 1986; Chang et al., 1996; Vimont et al., 2001, 2003; Anderson, 2003; Chiang and Vimont, 2004; Chang et al., 2007].
 Concerning the stochastic forcing of ENSO, Vimont et al. [2001, 2003] demonstrated that the wintertime North Pacific Oscillation (NPO) [Walker and Bliss, 1932; Rogers, 1981] could affect the tropical Pacific sea surface temperature (SST) in the following winter via the seasonal footprinting mechanism (SFM). When the southern lobe of the NPO exhibits a negative anomaly, the trade winds weaken over the central and eastern subtropical Pacific. The upward surface net heat flux is reduced by the decreased wind speed and causes SST warming. These NPO-related SST warming anomalies over the subtropical (0–20°N) Northern Pacific persist into the spring via thermodynamic air-sea interaction [Xie and Philander, 1994; Vimont et al., 2009; Alexander et al., 2010] and, in turn, force a pattern of atmospheric circulation anomalies including westerly wind stress anomalies along the equatorial western Pacific, which would subsequently act as a trigger for ENSO [Barnett et al., 1989].
Alexander et al.  tested the SFM hypothesis by imposing the NPO-related surface heat flux anomaly in a coupled general circulation model. Their results showed that about 70% of the ensemble simulations generate warming anomalies in the ENSO region in the following winter. Using reanalysis data during 1949–2009, Park et al.  showed that the occurrence rate for El Niño under the conditions of a positive NPO in previous winter is about 41%. The above studies indicate that NPO-like atmospheric forcing via the SFM does not always trigger an El Niño. Hence, one question is naturally that of which factor(s) can exert an additional effect on the relationship between the wintertime NPO and the following winter El Niño?
 Several previous studies show that the spring Arctic Oscillation (AO) can influence the following East Asian summer monsoon and the following winter SST in the equatorial central-eastern Pacific [Nakamura et al., 2006; Gong et al., 2011; Chen et al., 2013a]. Further, the spring AO-related SST anomalies over the subtropical North Pacific bear some resemblance to those associated with the preceding wintertime NPO [Vimont et al., 2003; Park et al., 2013]. Motivated by these qualitatively similar structures of SST anomalies, we hypothesize that the spring AO-related SST anomalies could strengthen or weaken the SST “footprint” over subtropical North Pacific generated by the preceding winter NPO. This would then influence the formation of westerly wind anomalies generated by air-sea interaction in the following summer and finally influence the SFM-driven SST anomalies in the equatorial central-eastern Pacific. In this study, we present evidence that it is indeed the case that the spring AO exerts a robust modulation on the connection between the wintertime NPO and El Niño in the subsequent winter.
2 Data and Methods
 The SST data used in this study are derived from the National Oceanic and Atmospheric Administration Extended Reconstructed SST version 3b [Smith et al., 2008]. The National Center for Environmental Prediction-National Center for Atmospheric Research reanalysis [Kalnay et al., 1996] data are used to examine the atmospheric circulation and derive the vertically integrated diabatic heating (see supporting information) [e.g., Yu and Zwiers, 2010]. The analyzed time period is 1957–2011.
 The AO index was adopted from the Climate Prediction Center website (http://www.cpc.ncep.noaa.gov). Generally, the NPO index is defined as the principle component of the second leading empirical orthogonal function (EOF) mode of the sea level pressure (SLP) anomalies over the North Pacific [Rogers, 1981]. In this study, we choose the NPO index defined by Vimont et al. [2001, 2003] as the SLP expansion coefficient (EC) time series of the first singular value decomposition (SVD) mode between the winter (ND(−1)JFM(0)) SLP over the North Pacific (120°E–90°W, 20–80°N) and the tropical Pacific (120°E–80°W, 25°S–25°N) SST during the following winter (OND(0)JF(+1)) (Figure 1). This index is in very good agreement with that defined by EOF method, but it can better capture the seasonal footprinting mechanism. As in Vimont et al. [2001, 2003], prior to performing SVD analysis, the wintertime (ND(−1)JFM(0)) SLP and the SST during the following winter (OND(0)JF(+1)) were detrended and standardized. In addition, the SLP and SST variability that linearly correlated with the Niño3.4 index during ND(−1)JFM(0) have been removed. The Niño3.4 index (area-averaged SST anomalies over the region of 5°S–5°N and 170°–120°W) was used to characterize ENSO variability. It should be noted that the removal of Niño3.4 index is not necessary to obtain the conclusions presented in this study, but it serves to ensure that the SFM is independent of ENSO cycle. Here, and for the remainder of this study, the time notations (−1), (0), and (+1) refer to the year before, during, and after the AO year, respectively.
3 Seasonal Footprinting Mechanism for ENSO
 Figure 1 displays the first SVD mode, which accounts for 62% of the total squared covariance. The anomalous SLP pattern (Figure 1a) is marked by a meridional dipole over the North Pacific and bears a close resemblance to the NPO structure [e.g., Rogers, 1981]. The anomalous SST pattern (Figure 1b) illustrates an El Niño-like warming in the tropical central-eastern Pacific during the following winter. In addition, the EC time series of SLP and SST is highly correlated (r = 0.66, Figure 1c), indicating that the wintertime NPO-like atmospheric variability is significantly linked to the El Niño-like SST anomalies in the tropical Pacific in the following winter, consistent with the results of previous studies [Vimont et al., 2001, 2003].
 We refer to the anomalous SLP pattern shown in Figure 1a as a positive NPO (+NPO) phase. We use 0.5 standard deviations as a criterion to select anomalous NPO years and obtain 18 +NPO years in the study period (1965, 1968, 1972, 1978, 1979, 1980, 1982, 1986, 1989, 1990, 1991, 1993, 1994, 1995, 1996, 1997, 2005, and 2009). It is noted that three of these +NPO years are under the impact of strong tropical volcanic eruptions (Agung erupted in 1963, and Pinatubo erupted in 1991; hence, 1965, 1993, and 1994 are impacted by volcanic aerosol clouds, which certainly had at least an impact on radiation at surface). We excluded these three +NPO years from our analysis, since the impacts of volcanic eruptions cannot easily be subtracted from the observations. Nevertheless, the results we obtained are weakly affected by including these 3 years.
 Figures 2a–2d display the mean seasonal evolution of composite SST anomalies for the 15 selected +NPO years. Correspondingly, Figures 3a–3c display the evolution of composite anomalies of the vertically integrated heating and surface wind stress. These figures well capture the SFM features reported in previous studies [Vimont et al., 2003; Alexander et al., 2010]. When low pressure prevails the southern lobe of the NPO (Figure 1a), the wind anomalies are from the west, hence opposing the trade winds over the central-eastern subtropical Pacific (Figure 3a), reducing the wind speed and upward net surface heat flux (see supporting information) and, thus, causing warming in the upper ocean. The NPO-driven SST warming anomalies (exceeding 0.1°C) over the central-eastern subtropical Pacific section of 0°–20°N persist into the spring and summer (Figures 2a and 2b) through thermodynamically coupled air-sea interaction. The subtropical warm SSTs (Figure 2b) are accompanied by anomalous atmospheric heating in the subtropical North Pacific (Figure 3b), driving an anomalous atmospheric circulation pattern including the westerly wind stress anomalies in the tropical western Pacific (Figure 3b) through a Gill-like atmospheric response [Gill, 1980]. These induced westerly wind stress anomalies subsequently excite downwelling and eastward propagating Kelvin waves, leading to SST warming anomalies in the tropical central-eastern Pacific in summer to fall (Figures 2b and 2c). The tropical SST, atmospheric heating, and atmospheric circulation anomalies sustain and develop into the following autumn and winter through air-sea interaction (Figures 2c, 2d, and 3c). Thus, an El Niño event is induced (Figure 2d).
4 AO Modulation of the Seasonal Footprinting Mechanism
 To explore the modulation of AO on the NPO-El Niño connection, we further separate the +NPO years into two groups: +NPO/+AO years when the normalized spring (March–April averaged) AO index is larger than zero and +NPO/−AO years when the AO index is less than zero. Note that the conclusions presented in this study are only weakly affected if we adopt other thresholds (such as 0.2 or 0.5 standard deviations) to define anomalous AO years, but then the ensemble becomes smaller, and this is detrimental to statistical analysis. Nine out of 15 +NPO years are selected as +NPO/+AO years, while the other 6 years are +NPO/−AO years (Table 1). Figures 2e–2h (Figures 2i–2l) display the evolution of SST anomalies during +NPO/+AO (+NPO/−AO) years. Correspondingly, Figures 3d–3f (Figures 3g–3i) display the evolution of the vertically integrated heating and surface wind stress anomalies for +NPO/+AO (+NPO/–AO) years. The evolution patterns of anomalous SST, surface wind stress, and vertically integrated heating for +NPO/+AO years are basically the same as those for all +NPO years, but generally with much stronger anomalies. In contrast, for +NPO/−AO years, the magnitudes of SST warming anomalies over the subtropical North Pacific during MAM(0) are much weaker (Figure 2i) and almost disappear during JJA(0) over the subtropical Northern Pacific (Figure 2j). Furthermore, the SST warming anomalies in the central tropical Pacific during JJA(0), SON(0), and D(0)JF(+1) are extremely weak and even exhibit negative SST anomalies over the equatorial central-eastern Pacific (Figures 2k and 2l). In addition, negative atmospheric heating anomalies can be observed over the subtropical northern and tropical central-eastern Pacific from spring to summer for +NPO/−AO years (Figures 3g and 3h). This indicates that the underlying SST footprint is too weak to force the atmosphere to trigger the westerly wind stress anomalies in the tropical western Pacific during summer that are necessary to trigger a warm ENSO event.
Table 1. List of +NPO, +NPO/+AO and +NPO/−AO, Years During 1958–2011a
El Niño Following
Non-El Niño Following
aHere 1968 refers to the 1967/1968 winter. These years are further separated into two groups, respectively, according to whether El Niño events happen in the subsequent winter. El Niño events are selected based on the definition of Trenberth .
1968, 1972, 1982, 1986, 1990, 1991, 1997, 2009
1978, 1979, 1980, 1995, 1996, 2005
 The above analyses demonstrate that when the spring AO is in its positive (negative) phase, the SFM is (not) well established and a positive wintertime NPO year can (not) result in significant El Niño-like warming anomalies. Eight out of nine +NPO/+AO years are followed by El Niño events, while none of the six +NPO/−AO years is followed by an El Niño event (Table 1). Hence, the NPO-El Niño connection is strongly modulated by the spring AO phase.
 A following question is the mechanism of how the NPO-El Niño connection is modulated by the spring AO phase. From the above analysis, we hypothesize that the main reason may be the strength of the SST footprint over the subtropical northeastern Pacific, which is modulated by the spring AO, being much stronger for +NPO/+AO years than that for +NPO/−AO years. To confirm this, we further display the regressions of spring SST and 850 hPa horizontal wind anomalies on the normalized spring AO index in Figure 4. Positive SST anomalies are seen in the northern Pacific midlatitudes and in the equatorial central Pacific with a northeastward extension to the subtropical Pacific, accompanied by negative anomalies in the subtropical western north Pacific and the equatorial eastern Pacific. This anomalous SST pattern bears some resemblance to the Western North Pacific pattern [Wang et al., 2012]. Of particular, noted here are the spring AO-related positive SST anomalies over the footprint regions (i.e., subtropical northeastern Pacific). The circulation anomalies depict a meridional dipole mode over the North Pacific, with an anticyclonic circulation anomaly in the North Pacific midlatitudes and a cyclonic circulation anomaly in the subtropical northern Pacific. A detailed description of the formation of spring AO-associated SST and circulation anomalies is given in Chen et al. [2013a].
 For positive spring AO years, the SST anomalies over the subtropical northeastern Pacific are positive in spring. The SST anomalies in association with positive NPO are also positive over the subtropical northeastern Pacific. Thus, the AO-related SST anomalies will superpose on the NPO-related SST anomalies to give a strengthened above-normal SST over the subtropical northeastern Pacific (Figure 2e). Subsequently, the westerly wind stress anomalies over the tropical western Pacific are effectively triggered by the enhanced SST footprint (Figure 3e) and finally generate the El Niño-like SST anomalies via the air-sea interaction in the tropical Pacific (Figure 2h). In contrast, in the case of negative spring AO, the SST over the subtropical northern Pacific is slightly below normal. This anomaly is opposite to the NPO-driven SST footprint (Figures 2i and 2j) and could not generate warming anomalies in the tropical central-eastern Pacific (Figure 2l).
5 Summary and Discussion
Vimont et al. [2001, 2003] demonstrated that wintertime NPO-like midlatitude atmospheric variability could force the tropical Pacific SST in the following winter via the SFM. Several recent studies indicated that the atmospheric forcing from the midlatitudes does not always trigger an El Niño [Alexander et al., 2010; Park et al., 2013]. In this study, we find that the spring AO has a pronounced modulating effect on the connection between NPO and El Niño. Only when the spring AO is in its positive phase, a positive phase of wintertime NPO can lead to significant El Niño-like warming anomalies in the following winter via the SFM. In contrast, a positive phase NPO year is not followed by significant warming anomalies in the tropical central-eastern Pacific when the spring AO is in its negative phase. Thus, the phase of spring AO should be taken into account when using NPO as a predictor for El Niño.
 The influence of spring AO on the NPO-El Niño connection is found to be on the constructive and destructive superposition of SST anomalies that modulate the underlying SST footprinting related to NPO over the subtropical northeastern Pacific. Yet other mechanisms by which the NPO can impact the tropical Pacific SST may also be involved in the composite analysis of present study. For example, as suggested by previous studies [Anderson and Maloney, 2006; Anderson et al., 2013], the NPO-generated westerly wind stress anomalies can effectively trigger an El Niño only with a sufficient amount of warm water in the tropical Pacific.
 We have also examined the connection of negative NPO and La Niña in the opposite spring AO phases. It is found that the spring AO also has a significant modulation effect on the negative NPO-La Niña connection. When the spring AO is negative (positive), the connection of the negative NPO and La Niña is (not) significant (not shown, but see the supporting information). Nevertheless, given the relatively small number of composite years, the results presented in this study need to be confirmed using longer time series.
 We thank two anonymous reviewers for their constructive suggestions and comments, which lead to a significant improvement in the paper. This study is supported by the National Natural Science Foundation of China grants 41025017 and 41230527.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.