Contributions of Interdecadal Pacific Oscillation and

Preparing to load PDF file. please wait...

0 of 0
100%
Contributions of Interdecadal Pacific Oscillation and

Transcript Of Contributions of Interdecadal Pacific Oscillation and

1 FEBRUARY 2018

HU ET AL.

1227

Contributions of Interdecadal Pacific Oscillation and Atlantic Multidecadal Oscillation to Global Ocean Heat Content Distribution
ZEYUAN HU
Department of Atmosphere and Ocean Sciences, Peking University, Beijing, China
AIXUE HU
Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, Colorado
YONGYUN HU
Department of Atmosphere and Ocean Sciences, Peking University, Beijing, China
(Manuscript received 29 March 2017, in final form 27 October 2017)
ABSTRACT
Regional sea surface temperature (SST) mode variabilities, especially the La Niña–like Pacific Ocean temperature pattern known as the negative phase of the interdecadal Pacific oscillation (IPO) and the associated heat redistribution within the ocean, are the leading mechanisms explaining the recent global warming hiatus. Here version 1 of the Community Earth System Model (CESM) is used to examine how different phases of two leading decadal time scale SST modes, namely the IPO and the Atlantic multidecadal oscillation (AMO), contribute to heat redistribution in the global ocean in the absence of time-evolving external forcings. The results show that both the IPO and AMO contribute a similar magnitude to global mean surface temperature and ocean heat redistribution. Both modes contribute warmer surface temperature and higher upper ocean heat content in their positive phase, and the reverse in their negative phase. Regionally, patterns of ocean heat distribution in the upper few hundred meters of the tropical and subtropical Pacific Ocean depend highly on the IPO phase via the IPO-associated changes in the subtropical cell. In the Atlantic, ocean heat content is primarily associated with the state of the AMO. The interconnections between the IPO, AMO, and global ocean heat distribution are established through the atmospheric bridge and the Atlantic meridional overturning circulation. An in-phase variant of the IPO and AMO can lead to much higher surface temperatures and heat content changes than an out-of-phase variation. This result suggests that changes in the IPO and AMO are potentially capable of modulating externally forced SST and heat content trends.

1. Introduction
The rate of increase in global mean surface air temperature (GMST) slowed during the early 2000s despite the rapid increase of greenhouse gases (GHGs) (Fyfe et al. 2016; Yan et al. 2016; Lewandowsky et al. 2016). This slowdown, often called the ‘‘global warming hiatus,’’ can be clearly seen in GMST from several leading observational datasets (Fyfe et al. 2016), although earlier research identified some observational uncertainty (Karl et al. 2015). Multiple studies suggest that this slowdown is contributed mostly by internal climate variability in the context of anthropogenic global
Corresponding author: Aixue Hu, [email protected]

warming (e.g., Meehl et al. 2011, 2013b; England et al. 2014; Dai et al. 2015; Meehl et al. 2016), while changes in external forcing, such as volcanic activity (Santer et al. 2014) and aerosol forcing (Smith et al. 2016), may also play a role. The major internal climate mode variability associated with this slowdown is the interdecadal Pacific oscillation (IPO) (Zhang et al. 1997; Power et al. 1999; Meehl and Hu 2006). The anomalous SST cooling and strengthened trade winds in the eastern and central equatorial Pacific associated with the negative phase of the IPO have played a dominant role in producing the observed reduction in warming. Coupled global climate model simulations can successfully reproduce a reduced warming trend by restoring either sea surface temperatures (SSTs) (Kosaka and Xie 2013, 2016) or trade winds

DOI: 10.1175/JCLI-D-17-0204.1
Ó 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).
Unauthenticated | Downloaded 06/13/22 09:12 PM UTC

1228

JOURNAL OF CLIMATE

VOLUME 31

(England et al. 2014) in the equatorial Pacific to the observations.
Moreover, this slowdown has spurred huge research interest in heat redistribution within the ocean. Observed energy imbalance at the top of the atmosphere (TOA) in the past few decades indicates a constant energy input of ;1 W m22 into Earth’s climate system (Hansen et al. 2011; Trenberth et al. 2014) and the majority of this excess heat resides in the ocean (Palmer et al. 2011). However, how this added heat is distributed within the ocean is mostly determined by internal oceanic processes. These processes can either keep most of the heat gain in the surface ocean or deposit it into the subsurface or deep ocean. Many studies have tried to identify the heat redistribution pattern associated with the hiatus (Meehl et al. 2011; Chen and Tung 2014; Lee et al. 2015; Nieves et al. 2015; Liu et al. 2016a,b). For instance, Liu et al. (2016b) showed that the anomalous warming in subsurface Indian Ocean waters associated with strengthened heat transport from the Pacific through the Indonesian Throughflow is related to the recent warming slowdown. However, large uncertainties still exist with regard to the patterns of the observed heat redistribution due to lack of consistent observations and discrepancies among different observed datasets (Chen and Tung 2016; Liu et al. 2016a).
Here, we explore how modes of internal climate variability modulate oceanic processes and how they play a role in the redistribution of heat within the ocean under preindustrial external forcing conditions. In the process, we isolate the relationship between internal variability and changes to ocean heat content (OHC) without contamination from the time-evolving external forcing changes. In fact, the interplay between internal climate variability and external forcing is an active research topic and how this interplay affects the distribution of heat in the ocean will be our future focus.
The two modes of internal variability we are testing are the IPO and the Atlantic multidecadal oscillation (AMO). These major decadal and multidecadal time scale mode variabilities in the Pacific and Atlantic are identified by both observational and modeling studies (e.g., Power et al. 1999; Meehl and Hu 2006; Deser et al. 2010; Delworth and Mann 2000; Zhang and Delworth 2006, 2007). They can both influence global-scale climate phenomena, such as rainfall in East Asian (Si and Ding 2016) or drought in the southwestern United States (Meehl and Hu 2006). To investigate the contribution of the IPO and AMO to the redistribution of ocean heat, we use the preindustrial control, a member of the Community Earth System Model (CESM1 v1; Hurrell et al. 2013) Large Ensemble project (Kay et al. 2015). In our analysis, we focus on global mean OHC as well as

OHC anomalies among ocean basins and subbasins of the Pacific and Atlantic to connect the OHC anomalies to underlying physical processes.
The physical mechanisms governing the IPO are still under intense debate due to a lack of reliable longterm observations and because of the complicated interactions among different components of the climate system. Currently, three primary mechanisms for governing the IPO have been proposed. The first theory asserts that the IPO represents a low-frequency response of the surface ocean to a stochastic atmospheric forcing (Hasselmann 1976; Frankignoul and Hasselmann 1977). A second theory posits that the IPO represents changes in SST due to advection in upper ocean circulation (Saravanan and McWilliams 1998; Meehl et al. 1998), oceanic gyre dynamics (Dewar 2001; Hogg et al. 2005; Taguchi et al. 2005; Ceballos et al. 2009), and oceanic Rossby wave adjustment (Qiu 2003; Schneider and Cornuelle 2005; Qiu et al. 2007). The third potential mechanism is that the IPO represents an air–sea coupling process, such as the unstable air–sea interactions over the North Pacific (Latif and Barnett 1994, 1996).
Similarly, the driving mechanisms of the AMO are also poorly understood and under intense debate. There are also three major mechanisms that are theorized to explain the AMO. The first theory is that the AMO is governed by oceanic processes, principally the Atlantic meridional overturning circulation (AMOC). Changes in AMOC modulate oceanic meridional heat transport, thus affecting the North Atlantic SST (Delworth et al. 1993; Delworth and Mann 2000; Latif et al. 2004; Knight et al. 2005; Semenov et al. 2010; Gulev et al. 2013; McCarthy et al. 2015; Zhang et al. 2016). A second theory is that indirect effects of anthropogenic aerosols influence long-term SST variability in the Atlantic (Mann and Emanuel 2006; Booth et al. 2012; Evan et al. 2009). The third potential mechanism is that the AMO is simply an SST response to midlatitude atmospheric stochastic forcing (Clement et al. 2015). In this paper, we do not explore the mechanisms governing the IPO or AMO, but instead focus on the relationship between the IPO/AMO states and changes in OHC.
The remainder of this paper is organized as follows. In section 2 we describe the model and the experimental design used for the results we present in this paper. Section 3 compares the fidelity of the simulated the IPO and AMO with observations. Section 4 quantifies the contributions of the IPO and AMO to global and regional OHC distribution and discusses the potential interaction between the IPO and AMO. Discussions and conclusions are given in sections 5 and 6, respectively.

Unauthenticated | Downloaded 06/13/22 09:12 PM UTC

1 FEBRUARY 2018

HU ET AL.

1229

2. Model and experiment

The climate model used for this research is the

Community Earth System Model, version 1 (CESM1;

Hurrell et al. 2013) with nominal 18 horizontal resolution

for all components. The atmospheric component is the

Community Atmosphere Model, version 5 (CAM5); the

ocean component is the Parallel Ocean Program, ver-

sion 2 (POP2); the land component is the Community

Land Model, version 4 (CLM4), and the sea ice model is

the Community Sea Ice Code, version 4 (CICE4). The

twentieth-century climate simulated by CESM1 agrees

reasonably well with observations (Meehl et al. 2013a).

This paper leverages a fully coupled 2200-yr pre-

industrial (PI) control simulation from the CESM Large

Ensemble (CESM_LE; Kay et al. 2015). We use the PI

control because our focus is to assess the contribution

of various modes of internal variability to the re-

distribution of oceanic heat in the absence of anthro-

pogenic external forcing. We limit our analysis to the

last 1000 years of the PI control to avoid including

nonlinear trends in the deep ocean in our results. The

small trend in deep ocean temperature after year 1200 is

linear and we remove it from OHC analysis. Notably,

this trend has minimal effect on the upper ocean; using

the full 2200-yr control did not make a significant dif-

ference for the upper ocean. The observed SST data

used for this analysis is the Hadley Center reconstructed

ocean surface temperature dataset, HadISST1, from

1870 to 2014 (Rayner et al. 2003). All analyses hereafter

are based on the detrended annual mean data.

Although the CESM1 can simulate twentieth-century

observed climate reasonably well (Meehl et al. 2013a),

certain biases still exist (e.g., Bryan et al. 2007; Danabasoglu

2008; Danabasoglu et al. 2012a,b; Neale et al. 2013). Using

the 18 horizontal resolution POP2 ocean model, CESM1

does not adequately separate the Gulf Stream from Cape

Hatteras; as a result, the location of the North Atlantic

current is zonally biased (Weese and Bryan 2006). How-

ever, studies show that, in general, these biases affect the

simulations quantitatively but do not fundamentally change

the basic physical processes. Results discussed later in this

paper, however, may be affected by these biases, so a

multimodel approach may be needed in future work.

There are two terms which are used later in this study:

heat content and heat density of seawater. The heat

content of seawater is defined as

ððð

HC 5 rCpT dx dy dz,

(1)

where HC represents heat content, r is the potential density of seawater, Cp is the specific heat of seawater, T is the potential temperature of seawater, and dx, dy, and

dz represent the width of a model grid in the zonal,

meridional, and vertical directions. The heat density of

seawater is defined as

ððð

rCpT dx dy dz

HCD 5

ðð

,

(2)

dx dy

where HCD represents heat density.

3. Simulated and observed IPO and AMO
a. A comparison of observed and modeled IPO
The IPO pattern is defined as the first mode of the empirical orthogonal function (EOF) analysis on lowpass filtered (13-yr cutoff) and detrended Pacific SST (408S–658N, 1108E–758W) for both observed and simulated data (Power et al. 1999; Meehl and Hu 2006). The IPO index is defined as the normalized time series of the first principal component (PC) (Fig. 1a and 1b; blue lines). The observed and modeled IPO indices explain a similar percentage of the SST variance (34.9% and 35.9%, respectively) in the Pacific. The observed IPO shows spectral peaks at 12–25 years (Fig. 1c) in our analysis and ;50 years from another study (Deser et al. 2010). The simulated IPO spectrum has more peaks, such as at 13–20, 22–33, and ;50 years (Fig. 1f), which may be related to the fact that we are using a much longer modeled than observed time series (1000 vs 142 yr).
The regression patterns of SST anomalies and the normalized IPO index for both observations and model simulations are given in Figs. 2a and 2b (the IPO patterns). The pattern correlation between the modeled and observed IPO is 0.8 in the Pacific, but only 0.19 globally. This implies that the model captures the dominant signature of the IPO on observed SST in the Pacific, but was not able to reproduce the observed teleconnections. Potentially, these teleconnections could be reinforced or modulated by the time-evolving external forcings, such as anthropogenic greenhouse gases, which needs to be explored further. In the Indian Ocean, the observed positive IPO pattern of SST is characterized by a basinwide warm anomaly, whereas the modeled IPO shows a dipole pattern. In the Atlantic Ocean, both the modeled and observed IPO show, in general, warm SST anomalies over the equatorial and subtropical Atlantic regions for a positive IPO (correlation ranging from 0.2 to 0.6, which is similar to observed; figure not shown) with reduced agreement in regions north of 308N. Notably, the seesaw pattern simulated by the model does not appear in the observations. In the

Unauthenticated | Downloaded 06/13/22 09:12 PM UTC

1230

JOURNAL OF CLIMATE

VOLUME 31

FIG. 1. Analysis of time series of IPO and AMO in observation and in CESM. (a) Observed IPO index (blue line) and AMO index (red line). (b) IPO index (blue line), AMO index (red line), and AMOC index (orange line) in CESM. (c) Observed power spectrum of IPO. (d) Observed power spectrum of AMO. e) Lead–lag correlation between AMO and AMOC. (f) Power spectrum of IPO in CESM. (g) Power spectrum of AMO in CESM. ((h) Power spectrum of AMOC in CESM. In (c), (d), (f), (g), and (h), the red line indicates the ‘‘red noise’’ curve, and the blue and green lines indicate the 95% and 99% significance levels.
Unauthenticated | Downloaded 06/13/22 09:12 PM UTC

1 FEBRUARY 2018

HU ET AL.

1231

FIG. 2. Regression of normalized indices of the (top) IPO and (bottom) AMO on annual mean SSTs from (left) observations and (right) the CESM preindustrial control run. Stippling indicates regions below 95% significance.

Southern Ocean, both modeled and observed IPO patterns shows positive SST anomalies for positive IPO over the middle to eastern sections of the Pacific and negative SST anomalies for other sectors. Overall, the model reproduces the observed IPO spatial patterns and frequencies reasonably well, especially in the Pacific basin.
b. Comparing the observed and modeled AMO
Here, the AMO index is defined as the detrended and low-pass filtered SST anomaly averaged over the entire North Atlantic basin (08–708N, e.g., Kaplan et al. 1998; Buckley and Marshall 2016). A linear trend is removed from the data and a Lanczos low-pass filter (13-yr cutoff) is applied to the area-weighted mean SST to derive the AMO index as shown in Fig. 1a for observations and Fig. 1b for model simulations (red lines). The observed AMO has a significant spectral peak at ;60 years (Fig. 1d), agreeing with previous studies (e.g., Buckley and Marshall 2016). The most significant spectrum peaks for the modeled AMO are 25 and 40 years, with some other minor peaks ranging from decadal to multidecadal. Thus, the longest significant period for the modeled AMO is a bit shorter than for the observed. Besides the difference in data length, anthropogenic

external forcing may also contribute to spectral differences between the modeled and observed AMO (e.g., Booth et al. 2012; Si and Hu 2017).
The regression patterns of the normalized AMO and SST anomalies are shown for observations (Fig. 2c) and for the model (Fig. 2d). The pattern correlation between modeled and observed AMO patterns is 0.75 in the Atlantic basin, but only 0.19 globally. The low global pattern correlation between modeled and observed AMO arises primarily from the Pacific and Indian Oceans, suggesting that the teleconnections between the North Atlantic and other ocean basins are not well simulated in CESM1. In general, the modeled AMO captures the major features of the observed AMO in the Atlantic basin with much reduced impact on SST, as demonstrated by smaller regression coefficients. Therefore, although the relationship between the AMO and Atlantic SST may be well simulated by the model, the global influence of the AMO on SST is not well reproduced by CESM1, due in part to the lack of the timeevolving external forcings.
c. Relationship of modeled AMO and AMOC
Usually, the AMO is considered to be governed by both AMOC, an internal ocean process (Delworth and

Unauthenticated | Downloaded 06/13/22 09:12 PM UTC

1232

JOURNAL OF CLIMATE

VOLUME 31

TABLE 1. Composite average in different phases of IPO and AMO. The CL mean index is the climatological mean of the corresponding index in its corresponding phase; ‘‘Mean index’’ is the mean index value for the composite analysis, which is the mean of the corresponding index greater than 1 (less than 21). By comparing these two mean indices, it shows that the mean index value for the composite analysis is about twice as much as that of the climatological mean index for each phase of the IPO or AMO. GMST represents the global mean surface temperature anomaly relative to the climatological mean; GM OHC represents the global mean ocean heat content anomaly relative to the climatological mean; Pac OHC is the mean ocean heat content in the Pacific basin north of 348S; and Atl OHC is the mean ocean heat content in the Atlantic between roughly 808N and 348S. Also, (,100 m) represents the total OHC in the upper 100-m ocean,and (100–300 m) represents the total OHC in the ocean layer between 100- and 300-m depth. TheIPO/AMO index is unitless because these indices are standardized. The unit for GMST is 8C. The unit for ocean heat content is 1021 J.

CL mean index Mean index GMST GM OHC (,100 m) GM OHC (100–300 m) Pac OHC (,100 m) Pac OHC (100–300 m) Atl OHC (,100 m) Atl OHC (100–300 m)

IPO1
0.84 1.55 0.053 6 0.012 3.21 6 0.59 22.62 6 0.69 2.72 6 0.28 22.29 6 0.44 0.50 6 0.18 0.16 6 0.18

IPO2

20.76 21.50 20.054 6 23.04 6 21.91 6 22.76 6 1.60 6 20.30 6 0.16 6

0.012 0.67 0.85 0.33 0.57 0.21 0.21

AMO1
0.82 1.49 0.040 6 0.015 2.19 6 0.60 1.59 6 0.86 0.56 6 0.34 0.65 6 0.55 1.55 6 0.12 0.81 6 0.18

AMO2

20.78 21.63 20.049 6 24.32 6 22.42 6 21.48 6 20.63 6 21.85 6 20.92 6

0.016 0.70 0.57 0.40 0.44 0.13 0.20

IPO1 and AMO1
1.57/1.61 0.076 6 0.027 3.80 6 1.37 21.73 6 1.87 2.05 6 0.56 22.85 6 1.19 2.14 6 0.27 1.33 6 0.30

IPO2 & AMO2
21.33/21.74 20.087 6 0.032 26.26 6 1.45 22.96 6 1.50 23.00 6 0.69 20.78 6 0.85 22.29 6 0.37 20.85 6 0.50

Mann 2000; Zhang and Delworth 2006; Zhang et al. 2016), and the combined effect of external atmospheric forcing and intrinsic variability (e.g., changes in aerosol forcing and the North Atlantic Oscillation) (Booth et al. 2012; Clement et al. 2015; Buckley and Marshall 2016). Because our analysis is focused on the preindustrial control simulation, we lack anthropogenic forcing and examine only the relationship between the AMO and AMOC. The AMOC index is defined as the maximum of the meridional overturning streamfunction below 500-m depth in the Atlantic Ocean (Fig. 1b). The correlation of the AMO index and the AMOC index reaches 10.36 when the AMOC leads the AMO by two years (Fig. 1e). The major spectrum peaks of the AMOC are a bit longer, such as ;60 and ;45 years, and there are also peaks for a shorter period ranging from less than 20 years to about 40 years. Moreover, many of the AMOC frequencies are in good agreement with those of the AMO (Figs. 1h,g). A separate study used the same data applies wavelet spectrum analysis and the cross-wavelet transform and wavelet coherence analyses, corroborating the covariance of the AMOC and the AMO in both wave power and relative phase domain (Si and Hu 2017). Therefore, the AMOC contributes significantly to the AMO variability in our simulation and agrees well with other investigations (e.g., Delworth and Mann 2000; Zhang and Delworth 2006; Zhang et al. 2016).
4. Decadal modes and OHC distribution
a. Contributions of the IPO to global and regional OHC anomalies
We use composite analysis to assess the contribution of different IPO or AMO mean states to the regional

and global OHC changes. A composite positive IPO (IPO1) is defined as the ensemble mean of years with a normalized IPO index greater than 1 standard deviation, and a composite negative IPO (IPO2) is defined as the ensemble mean of years with the normalized IPO index less than 21 standard deviation. In this 1000-yr-long time series, there are 163 IPO1 years (with a mean IPO index of 1.55; roughly twice as large as the mean IPO index for all positive IPO years) and 155 IPO2 years (with a mean IPO index of 21.50; Table 1). We follow this definition of the composite IPO1 or IPO2 throughout the remainder of this paper and refer to the composite IPO1 (IPO2) as IPO1 (or IPO2).
The global mean OHC anomaly for an IPO1 and IPO2 state relative to the climatological mean OHC in various ocean layers is given in Fig. 3a. In general, OHC anomalies in the upper 100 m are opposite to those at 100–300-m depths, and the OHC anomalies for an IPO1 phase are opposite to those for an IPO2 phase. For example, OHC in the upper 100-m layer has a positive anomaly of 3.2 3 1021 J for an IPO1 state and the associated GMST anomaly is 0.0538C relative to the climatological mean (Table 1). But, the OHC in the 100– 300-m layer has a negative anomaly (22.6 3 1021 J), accounting for 82% of the positive OHC anomaly in the upper 100 m. The inverse relationship between the OHC anomaly at the surface to that at depths of 100–300 m implies that different phases of the IPO induce contrasting OHC anomalies in the upper few hundred meters of the ocean. For the IPO2 phase, OHC anomalies for the upper 100-m (23.0 3 1021 J) and 100–300-m (11.9 3 1021 J) layers are also opposite in sign, with an associated GMST anomaly of 20.0548C (Table 1). Below 300 m, the sign of the OHC anomaly is generally

Unauthenticated | Downloaded 06/13/22 09:12 PM UTC

1 FEBRUARY 2018

HU ET AL.

1233

FIG. 3. Composite mean global OHC anomalies at different layers for different phases of the IPO and AMO, namely (a) the IPO, with, red (blue) bar for positive (negative) IPO; (b) the AMO, with red (blue) bar for positive (negative) AMO; and (c) in-phase IPO and AMO, with red (blue) bar for in-phase positive (negative) IPO and AMO. Error bars denote 95% confidence interval.
the same as for the subsurface layer (100–300 m) of the ocean, with an OHC increase for an IPO2 phase, and a decrease for an IPO1 phase, while the magnitude of the anomaly is less significant.
The primary contributions to this pattern of global OHC anomaly associated with the IPO phase come from the Pacific Ocean (Fig. 4a and Table 1). For both

the IPO1 and IPO2 phases, the Pacific OHC anomalies in both the upper 100-m and 100–300-m layers account for over 85% of the global OHC anomalies. To explore this relationship, we divide the Pacific Ocean into different latitudinal bands in order to connect the underlying physical processes to the changes in OHC. These latitude bands are 1) the equatorial Pacific (158S–158N), 2) the subtropical and midlatitudinal South and North Pacific (348–158S and 158–458N), and 3) the subpolar North Pacific (458–658N). As shown in Figs. 4b–e, the upper 100-m OHC changes in the Pacific are dominated by the equatorial Pacific, accounting for 86% (76%) of the Pacific IPO1 (IPO2) OHC anomaly in the upper 100-m layer with the remainder coming from the subpolar North Pacific. In the subsurface layer (100–300 m), the contribution from the equatorial Pacific to the entire Pacific OHC anomaly is much smaller (a bit less than 40%) for both IPO1 and IPO2. On the other hand, the contribution from both the north and south subtropical and midlatitudinal regions to the upper 100-m layer Pacific OHC anomaly is minor, due in part to the opposite sign of the OHC anomaly in these two regions. In the subsurface layer (100–300 m), the OHC anomalies in both subtropical and midlatitudinal North and South Pacific are the same sign as the equatorial Pacific, and account for about 80% of the total Pacific OHC anomaly in this layer, but a portion of this anomaly is offset by the subpolar North Pacific.
The vertical distribution of the OHC anomaly in the Pacific can be understood more clearly by looking at the zonal mean ocean heat density, defined as the areaweighted mean OHC [J m22; see section 2, Eq. (2)], between the IPO1 and IPO2 (Fig. 4f). In the equatorial region (roughly 278S–278N), there is a positive heat density anomaly for the upper 100 m, but a negative anomaly in regions between 278 and 458N and between 348 and 278S in the Pacific. An examination of the heat density anomaly also explains why OHC anomalies in the subtropical north and South Pacific remain small in the upper 100-m layer due in part to the effect of averaging both positive and negative anomalies in these two regions (348–158S and 158–458N). In the 100–300-m layer, the heat density anomaly is primarily negative with the exception of a few small regions. In the subpolar North Pacific, the heat density anomaly is the same sign to at least 700 m, consistent with Fig. 4b.
The pattern of OHC distribution described above in the equatorial and subtropical Pacific can be explained by changes in subtropical cells (STCs) in association with the IPO (Figs. 4g,h). STCs are shallow meridional overturning cells in the ocean on each side of the equator that extend to depths of roughly 700 m (e.g., McPhaden and Zhang 2002; Meehl and Hu 2006; Meehl

Unauthenticated | Downloaded 06/13/22 09:12 PM UTC

1234

JOURNAL OF CLIMATE

VOLUME 31

FIG. 4. Composite mean OHC anomalies for IPO1 and IPO2 in the Pacific and its subbasins: (a) entire Pacific, (b) subpolar North Pacific, (c) subtropical North Pacific, (d) equatorial Pacific, and (e) subtropical South Pacific. Red bars represent OHC anomaly for IPO1 and blue bars represent OHC anomaly for IPO2. (f)–(h) The zonally integrated heat density anomaly between IPO1 and IPO2, climatological mean Pacific meridional streamfunction representing STCs, and the Pacific meridional streamfunction anomaly between IPO1 and IPO2, respectively. Error bars in (a)–(e) denote 95% confidence interval. Stippling in (f) and (h) indicates statistically significant changes. Sv represents units of Sverdrups, a volume transport in oceanography (106 m3 s21).
Unauthenticated | Downloaded 06/13/22 09:12 PM UTC

1 FEBRUARY 2018

HU ET AL.

1235

et al. 2013b). During a positive phase of the IPO, the SST contrast between eastern and western equatorial Pacific weakens, leading to weakened equatorial easterlies (cf. Figs. 5a and 5b), with consequently weaker equatorial Ekman upwelling and weaker STCs (cf. Figs. 4g and 4h). These weaker STCs bring less colder subsurface water to the surface, resulting in a warmer equatorial upper ocean and increased OHC. Meanwhile, associated with weaker easterlies and westerlies induced by the positive IPO in the Pacific (Fig. 5b), the Ekman convergence in the subtropics weakens, along with a reduced downwelling there. As a result, less warm surface water is subducted, leading to a cooling of subsurface layers and a negative OHC anomaly in the subtropical regions. In the subpolar North Pacific, the OHC changes are linked to a weakened (strengthened) surface Ekman divergence related to the positive (negative) IPO-induced weakening (strengthening) of the westerlies (Fig. 5b), and thus a weakened (strengthened) upwelling in this region.
The IPO contributes to ocean heat redistribution not only in the Pacific, but also in other ocean basins. However, correlated changes in OHC are less significant in other basins, possibly related to the weakly simulated IPO teleconnections in CESM. The climatological heat density (Fig. 6a) of the upper 100 m resembles the SST pattern. The heat density anomaly between the IPO1 and IPO2 for this layer (Fig. 6b) resembles the regression pattern of the IPO and SST (Fig. 2b). Compared to the heat density anomaly in the Pacific, the anomaly in other basins is less than 50% of that seen in the Pacific. In the eastern Indian Ocean, the heat density anomaly has the same sign as in the western Pacific, suggesting an influence of the Pacific on the Indian Ocean, potentially by way of the Indonesian Throughflow (Liu et al. 2016b). In the Pacific sector of the Southern Ocean, the heat density anomaly is positive (or negative if this anomaly is defined as IPO2 minus IPO1). Meehl et al. (2016) suggest the negative IPO has contributed to the recent sea ice expansion in the Pacific sector of the Southern Ocean. Our analysis indicates that the negative OHC anomaly in the same sector during the IPO2 in recent hiatus years along with strengthened westerlies (Fig. 5b) may have also contributed to sea ice expansion there.
b. The AMO contributions to global and regional OHC anomaly
Analogous to the IPO, a composite positive AMO (AMO1) is defined as the ensemble mean of years with a normalized AMO index greater than 1, and a composite negative AMO (AMO2) is defined as the ensemble mean of the years with a normalized AMO

index less than 21. There are 159 sample years that meet the AMO1 definition in the 1000-yr-long control run, with a mean AMO index of 1.49, and 135 AMO2 sample years with a mean AMO index of 21.63. We adhere to these definitions for the remainder of the text and refer the composite AMO1 (AMO2) as AMO1 (AMO2).
Figure 3b shows the global mean OHC distribution at different layers for the AMO1 and AMO2 phases. In contrast to the IPO, the OHC anomalies in the upper 100-m and 100–300-m layers are of the same sign. Moreover, there is a significant asymmetry in the OHC anomaly for the AMO2 (26.7 3 1021J), when compared to the anomaly for the AMO1 (3.8 3 1021 J; Table 1). The OHC anomaly for the AMO2 is roughly twice that of the AMO1. The difference in corresponding GMST changes is much smaller (0.0408C for AMO1 vs 20.0498C for AMO2; Table 1). Changes in OHC in deeper layers are small, with the exception of the 1500-m layer for the negative AMO.
Regionally, the Atlantic Ocean contributes a significant portion of the AMO-related global OHC anomaly in the upper 300 m (Fig. 7a), but the asymmetry of the OHC changes in this layer between AMO1 and AMO2 is mainly from the Pacific and Southern Oceans (figure not shown). The Atlantic OHC anomaly in the upper 300 m accounts for about 63% (12.4 3 1021 J) of the global mean OHC anomaly in this layer for the AMO1, but 41% (22.8 3 1021 J) of the global OHC anomaly in this layer for the AMO2. In the Pacific Ocean, the OHC anomaly in the upper 300 m contributes 32% (11.2 3 1021 J) of the total OHC anomaly for the AMO1, and a similar percentage (32%, or 22.1 3 1021 J) for the AMO2. In the Southern Ocean, the OHC anomaly in the upper 300 m is small for the AMO1 but reaches 22% (21.5 3 1021 J) for the AMO2. Thus, all ocean basins contribute to the global OHC anomalies for both the AMO1 and AMO2 although the contribution from the Atlantic is larger than from other ocean basins. In deeper ocean layers, the OHC changes are less significant, except for in the Southern and Atlantic Oceans (figure not shown). These changes in OHC may not be directly related to the AMO; rather, they may be a delayed response to surface changes. The OHC changes in these layers generally have opposite signs between the AMO1 and AMO2.
Similar to our investigation of the Pacific, we divide the Atlantic into four subbasins, namely 1) the subpolar North Atlantic (458–808N), 2) the subtropical North Atlantic (458–158N), 3) the subtropical South Atlantic (158–348S), and 4) the equatorial Atlantic (158S–158N).
Depth-related changes in OHC in the Atlantic vary by latitude and are associated with different physical

Unauthenticated | Downloaded 06/13/22 09:12 PM UTC

1236

JOURNAL OF CLIMATE

VOLUME 31

FIG. 5. (a) Climatological mean surface wind, (b) mean wind anomaly between positive and negative IPO, and (c) the mean wind anomaly between positive and negative AMO.
Unauthenticated | Downloaded 06/13/22 09:12 PM UTC
IpoAmoFigOhc AnomalyOcean