Intensification of decadal and multi-decadal sea level variability in the western tropical Pacific during recent decades
- 1.2k Downloads
Previous studies have linked the rapid sea level rise (SLR) in the western tropical Pacific (WTP) since the early 1990s to the Pacific decadal climate modes, notably the Pacific Decadal Oscillation in the north Pacific or Interdecadal Pacific Oscillation (IPO) considering its basin wide signature. Here, the authors investigate the changing patterns of decadal (10–20 years) and multidecadal (>20 years) sea level variability (global mean SLR removed) in the Pacific associated with the IPO, by analyzing satellite and in situ observations, together with reconstructed and reanalysis products, and performing ocean and atmosphere model experiments. Robust intensification is detected for both decadal and multidecadal sea level variability in the WTP since the early 1990s. The IPO intensity, however, did not increase and thus cannot explain the faster SLR. The observed, accelerated WTP SLR results from the combined effects of Indian Ocean and WTP warming and central-eastern tropical Pacific cooling associated with the IPO cold transition. The warm Indian Ocean acts in concert with the warm WTP and cold central-eastern tropical Pacific to drive intensified easterlies and negative Ekman pumping velocity in western-central tropical Pacific, thereby enhancing the western tropical Pacific SLR. On decadal timescales, the intensified sea level variability since the late 1980s or early 1990s results from the “out of phase” relationship of sea surface temperature anomalies between the Indian and central-eastern tropical Pacific since 1985, which produces “in phase” effects on the WTP sea level variability.
KeywordsDecadal Multidecadal Sea level Pacific decadal variability Indian Ocean warming
Decadal prediction is emerging as a new priority in climate research, due to its significant societal impacts and the need to adapt to climate change (Goddard et al. 2009; Hurrell et al. 2009; Meehl et al. 2009). As an essential component of decadal prediction, predicting decadal sea level variability, particularly at the local and regional level, has large societal demand (Milne et al. 2009; Church et al. 2011; Leuliette and Willis 2011; Nicholls 2011). Skillful predictions, however, rely critically on our understanding of how and why the sea level has varied on decadal timescales, and how the variations have evolved with time. In this paper, “decadal” refers to 10–20 year variability and “multidecadal”, >20 year variability.
The rapid WTP SLR, together with the easterly trade intensification since the early 1990s, is shown to be associated with decadal climate modes, particularly the decadal variability of El Niño-Southern Oscillation (ENSO), the Pacific Decadal Oscillation (PDO; Mantua et al. 1997; Minobe 1997; Zhang et al. 1997; Garreaud and Battisti 1999; and review papers by Alexander 2010 and Liu 2012) and the Interdecadal Pacific Oscillation (IPO; Power et al. 1999; Folland et al. 2002; Meehl and Hu 2006; Power and Colman 2006; Randall et al. 2007). Zhang and Church (2012) performed multivariate statistical analysis, and showed that the large SLR in the WTP and fall in the eastern Pacific from 1993 to 2010 are correlated with the PDO (and thus IPO) and Multivariate ENSO indices. Merrifield et al. (2012) analyzed tide gauge data for the past 60 years and found significant correlations between the 5-year running mean WTP sea level variability (global mean removed) and PDO Index, and to a lesser degree, the Southern Oscillation Index. Feng et al. (2010) argued that the fast SLR in the WTP is associated with the spinup of the subtropical cells, which reversed their multidecadal weakening trends since the early 1990s (McPhaden and Zhang 2002, 2004). The spin up and spin down of the subtropical cells are associated with ENSO decadal variability (e.g., Kleeman et al. 1999). Meyssignac et al. (2012) analyzed reconstructed sea level from 1950 to 2009 and eight CMIP3 climate model solutions, and found that the 17 year reconstructed sea level trend patterns of 1959–1975 are similar to the 17 year satellite observed trends of 1993–2009, except that the SLR rates for 1993–2009 are higher in the WTP and there are some differences in detailed structures. The east–west sea level dipole variability generally follows the low-frequency Nino3.4 index. Similar dipole patterns were found in 17-year successive sea level trends in the majority of the climate model control runs with constant, preindustrial external forcing, and 4 out of the 8 models show significant correlations with low-frequency ENSO variability. Most of the twentieth century runs with varying external (natural and anthropogenic) forcing do not show significant differences with their respective control runs, even though the impact of external forcing could be seen in the peak frequency bands of the sea level variability spectra. Based on these analyses, the authors concluded that internal variability of the climate system associated with ENSO multidecadal variability dominates the 1993–2009 observed sea level trend patterns in the tropical Pacific.
In addition to the climate modes over the Pacific Ocean, satellite observations reveal coherent trend reversals of sea level and surface winds across the Indo-Pacific basin from 1993–2000 to 2000–2006, and the variability appears to occur earlier in the tropics (Lee and McPhaden 2008). These results demonstrate the close linkage between the Indian and Pacific ocean–atmosphere variability, and indicate the potential role of the tropics in determining the Indo-Pacific basin decadal climate. In contrast to the 1993–2010 sea level trend that represents multidecadal variability, the sea level trend reversal within the 1993–2006 period has a ~14 year timescale. Feng et al. (2010) analyzed a century long tide gauge record at Fremantle, west coast of Australia, and showed increased energy of 8–16 year sea level variability since the late 1980s. They suggested that the increased 8–16 year energy results from the equatorial Pacific influence via the Indonesian Throughflow. Trenary and Han (2013) performed ocean general circulation model experiments, and showed intensified Pacific impact on Indian Ocean decadal sea level variability via the Throughflow after 1990 relative to the preceding two decades (their Fig. 13).
1.2 Present research
While previous studies have linked the intensification of WTP SLR and easterly trades since the early 1990s to multidecadal variability of the PDO, has the PDO intensity increased during recent decades to ensure the enhanced easterlies? While larger impacts of the equatorial Pacific on Indian Ocean decadal sea level variability since the late 1980s or early 1990s has been suggested, has the decadal sea level variability indeed intensified in the tropical Pacific Ocean? If yes, what are the causes?
The overall goal of this paper is to investigate the decadal and multidecadal intensifications of sea level variability in the tropical Pacific during recent decades, and to understand the causes. We will focus particularly on assessing the processes associated with ocean–atmosphere changes over the Indo-Pacific basin, given that the two basins are closely linked via both atmospheric bridge and oceanic connection, and that the Atlantic Multidecadal Oscillation has little effect on the tropical Indo-Pacific sea surface temperature (SST) variability (e.g., Sutton and Hodson 2005, their Fig. 1). Specifically, we first examine the evolution of the Pacific decadal mode, the IPO, and its changing relationship with the tropical Indian Ocean (TIO) SST. The IPO is quasi-symmetric ENSO-like interdecadal variability across the entire Pacific basin (e.g., Power et al. 1999; Meehl and Hu 2006), and is defined as the leading empirical orthogonal function (EOF) of decadal-multidecadal SST anomalies (SSTA) over the entire Pacific basin. The leading principle component (PC1) is referred to as the IPO index. Then, we assess the decadal and multidecadal sea level variability patterns and their evolution associated with different phases of the IPO since the 1950s, and identify robust changes that have occurred during recent decades. Finally, we investigate the causes for the changes. The period of the 1950s-onward is chosen because both the observational based sea level and reanalysis product are available.
Here, we focus on the IPO instead of PDO and ENSO because IPO is the basin-wide Pacific decadal climate mode, whereas PDO primarily represents North Pacific climate variability, and ENSO is dominated by interannual variability (see Sect. 3.1 for further discussion). It has been suggested that the IPO has interdecadal modulation on interannual ENSO events (e.g., Power et al. 1999), and the PDO can be regarded as the North Pacific manifestation of the IPO (e.g., Folland et al. 2002; Meehl and Hu 2006). It is debatable, however, whether IPO can be confidently treated as an independent climate mode to ENSO (e.g., Trenberth et al. 2007). The indices of IPO, PDO and decadal variability of ENSO are all highly correlated (e.g., Alexander et al. 2002; Newman et al. 2003; Deser et al. 2004; Schneider and Cornuelle 2005; Vimont 2005; Zhang and Church 2012), and their correlations are further examined in Sect. 3.1. This study has important implications for decadal prediction, which is a challenging task because decadal climate variability is determined not only by natural internal variability of the climate system, but also affected by external natural and anthropogenic forcing, and both may contribute to the skill of decadal forecast (e.g., Hoerling et al. 2011; Solomon et al. 2011).
2 Datasets and models
We use a combined approach that integrates observational analyses with model experiments. Both ocean and atmosphere model experiments are performed.
2.1 Datasets and processing
To document the basin-wide ocean–atmosphere signatures, we analyze ocean–atmosphere variables from the following datasets: weekly multi-satellite merged sea surface height (SSH) anomalies from the French Archiving, Validation, and Interpolation of Satellite Oceanographic Data (AVISO) project (Ducet et al. 2000) available since October 1992, monthly HadiSST (Rayner et al. 2006), Hurrell SST (Hurrell et al. 2008), and Kaplan SST (Kaplan et al. 1998) since 1870, the cross-calibrated multiplatform (CCMP) satellite surface winds (Atlas et al. 2008) available since July 1987, satellite derived outgoing longwave radiation (OLR; Liebmann and Smith 1996) available since 1979, and Wave- and Anemometer-Based Sea Surface Wind (WASWinds; Tokinaga and Xie 2011). To extend the sea level analysis back to the 1950s, we also analyze the upper-700 m thermosteric sea level from 1945 to 2010 calculated from updated version of Ishii and Kimoto (2009) temperature data, and sea level reconstructed specifically for the Pacific Ocean from 1950 to 2009 (Hamlington et al. 2011). The justification of using the upper-ocean thermosteric sea level is that it is the dominant component of the observed sea level variability, although salinity may also play some role at regional scales (Wunsch et al. 2007; Köhl and Stammer 2008; Lombard et al. 2009; Meyssignac et al. 2012; Nidheesh et al. 2012).
In addition to the observed and reconstructed datasets, we also analyze surface winds from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis Interim (ERAI) data available since 1989 (Simmons et al. 2007), the ECMWF operational ocean analysis/reanalysis system (ORA-S3) available for 1959–2009 (Balmaseda et al. 2008), and the National Centers for Environmental Prediction (NCEP; Kalnay et al. 1996) available since 1948. This multiple datasets approach aims to identify signals that exceed cross-dataset differences, and thus increase confidence in our analysis.
To isolate decadal-multidecadal variability signals, a Lanczos lowpass filter (Duchon 1979) with half power placed at 8 year period is used. The response curve of the 8 year lowpass filter retains ~90 % of the amplitude at 10 year period and almost full amplitudes for longer periods, which essentially retains the signals with periods of 10 years and longer. The 8 year lowpassed filter has been used to investigate the Pacific decadal variability by previous studies (e.g., Deser et al. 2012). A Lanczos 8–20 year bandpass filter is used to isolate decadal signals.
2.2 Ocean model and experiment
To isolate the effects of winds in driving the Pacific sea level pattern, a Linear continuously stratified Ocean Model (LOM; McCreary 1981) driven solely by surface wind stress is used. The LOM has been applied to the Indian, Atlantic and Indo-Pacific basins (e.g., Han 2005; Han et al. 2008, 2010; Trenary and Han 2013). In this paper, it is set up for Indo-Pacific basin (30–290°E, 55°S–55°N) with 0.33° × 0.33° horizontal grids. The solution is well converged using the first 15 baroclinic modes. The LOM was first spun up for 30 years using ECMWF ORA-S3 monthly wind stress of 1959. Restarting from year 30, the model was integrated forward using monthly ORA-S3 wind stress forcing from 1959 to 2009.
2.3 Atmospheric models and experiments
To assess the effect of TIO warming since the 1950s, we perform two idealized experiments using the National Aeronautics and Space Administration Seasonal-to-Interannual Prediction Project (NSIPP) Atmospheric General Circulation Model (AGCM) with horizontal grids of 3.75° × 3° and 34 vertical levels (see Schubert et al. 2004 for details). One is forced by the Hurrell et al. (2008) global monthly SST climatological annual cycle of 1971–2000 with a 0.5 °C anomaly uniformly added to 10°S–10°N and linearly ramped to zero from 10–20°S and 10–20°N. This solution is based on the observed warming rates in most regions of the global tropical oceans since the 1950s (see Sect. 3.2.3), and is referred to as AGCM_0.5C. The other is the same as AGCM_0.5C except for adding an additional 0.5 °C warming to the TIO from the western boundary to 110°E with linear ramping to zero at 130°E. This experiment is based on the observations showing that the TIO warms faster than the tropical Pacific and Atlantic (Sect. 3.2.3). This solution is referred to as AGCM_TIO1.0C. Both experiments were run for 50 years, and the last 45 years results are analyzed, and their average represents the 45-member ensemble mean. The difference between the two experiments, (AGCM_TIO1.0C—AGCM_0.5C), estimates the 0.5 °C warmer TIO and is referred to as AGCM_TIO0.5C. Conversely, the—AGCM_TIO0.5C measures the effect of TIO 0.5 °C cooling.
To understand the role of tropical versus global SST in generating the Pacific basin-wide surface winds that drive the spatial patterns of sea level variability, we analyze the results from two experiments using the National Center for Atmospheric Research (NCAR) Community Atmospheric Model version 3 (CAM3) at T85 (equivalent to 1.48° × 1.48°) horizontal resolution. Two experiments are analyzed: the 5-member ensemble of Global Ocean Global Atmosphere (GOGA) experiment for 1950–2008, which is forced by global time-varying SSTs of Hurrell et al. (2008) and referred to as AGCM_GOGA, and the 5-member ensemble of Tropical Ocean Global Atmosphere (TOGA) forced by tropical (20°N–20°S) monthly SSTs from 1950 to 2008 and SST climatological seasonal cycle polewards of 30°, with linear interpolation between 20° and 30° (referred to as AGCM_TOGA). The two experiments were performed by the Climate Variability Working Group at NCAR (Deser and Phillips 2009), and the model physics and performance were reported by previous studies (e.g., Collins et al. 2006; Deser et al. 2006; Hack et al. 2006; Hurrell et al. 2006).
While the AGCM experiments are straightforward in assessing regional SST effects on atmospheric circulation, we are aware of their limitations. One key issue is that on interannual timescales, the TIO warming can be a response to remote forcing from the Pacific El Niño, which causes anomalous subsidence over the TIO, reduces convection, increases shortwave radiation and thus induces warming (Kumar and Hoerling 1998; Kumar et al. 2005; Wu et al. 2006; Deser and Phillips 2006; Copsey et al. 2006; Xie et al. 2009). The observed steady warming over the Indian Ocean since the 1950s, however, is attributed to increased greenhouse gases instead of resulting from the Pacific influence, by analyzing the results from eleven CMIP3 climate models (e.g., Du and Xie 2008). Consequently, our AGCM experiments, particularly the AGCM_TIO0.5C, will help to provide insight into the effects of TIO warming on Indo-Pacific basin climate.
In this section, we first discuss the IPO evolution since 1900, point out its distinct decadal and multidecadal timescales, and examine its changing relationship with TIO SST variability as well as their association with the 1993–2010 sea level trend (Sect. 3.1). Then we assess the Pacific sea level variability patterns associated with multidecadal variability of IPO and TIO SST since the 1950s, identify the robust changes that have occurred during recent decades, particularly the rapid SLR in the WTP from 1993 to 2010, and understand the causes (Sect. 3.2). Finally, we examine the spatial patterns and temporal evolution of 10–20 year sea level variability associated with the IPO and TIO SST decadal variability, identify robust changes and understand the causes (Sect. 3.3).
3.1 The IPO and its changing relationship with TIO SST since 1900
The 17 year cooling/warming rates of IPO index for four representative time spans associated with IPO multidecadal phase transition periods (Fig. 2c), and the corresponding sea level changing rates (with the global mean SLR removed) since 1959 from the upper 700 m thermosteric sea level and ECMWF ORA-S3 sea level averaged in the WTP (130°E–160°E, 15°S–15°N) and central-east tropical Pacific (CETP; 160°W–110°W, 15°S–15°N) regions
IPO phase transition time span
IPO phase transition rate (°C year−1)
WTP thermo. sea level rate (mm year−1)
CETP thermo. sea level rate (mm year−1)
WTP ORA-S3 sea level rate (mm year−1)
CETP ORA-S3 sea level rate (mm year−1)
Standard deviation (STD) of 10–20 year sea level anomalies averaged in the western tropical north Pacific (WTNP; 125°E–160°E, 0°N–15°N) and central- equatorial Pacific (CEQP; 170°E–130°W, 10°S–5°N) for different decades, together with the STD of 10–20 year IPO index for the same decades
Correlation coefficients for sea level variability from the upper-700 m thermosteric sea level, IPO index, Indian Ocean SST PC1 and Nino3.4 SST
Indian Ocean SST PC1
WTNP sea level
−0.72 (99 %; 1950–2006)
−0.48 (85 %; 1950–1984)
−0.87 (99 %; 1950–2006)
0.76 (90 %; 1985–2006)
CEQP sea level
0.65 (90 %)
1.0 (99 %; 1900–2008)
0.75 (99 %; 1900–1984)
0.88 (99 %; 1900–2008)
−0.85 (95 %; 1985–2008)
0.83 (99 %; 1950–2006)
3.2 Multidecadal sea level acceleration in the WTP since 1993 and causes
3.2.1 Changes of multidecadal sea level variability patterns
3.2.2 Effects of surface winds
3.2.3 Changes in Pacific winds associated with IPO and TIO SST variability
Even though the TIO warming might be essential for intensifying the WTP SLR during recent decades, it is the tropical Pacific air–sea interaction that determines the basin-wide surface winds, which drive the basin-wide sea level trend patterns since 1993. To demonstrate this point, we analyze the results from AGCM_GOGA (global monthly SST forcing) and TOGA (tropical SST forcing) experiments. Linear trends of surface wind stress and we from 5-member ensemble mean of AGCM_GOGA (Fig. 12e) resemble those of the observed (compare Fig. 12e with a, c, d). There are, however, some quantitative differences. The most apparent difference occurs in the tropical Pacific, where the model produces strong southwesterlies and positive we in the western-central basin near 10–20°N, whereas the observations show southerlies and weak negative we. The basin-wide wind patterns shown in AGCM_GOGA are successfully simulated by AGCM_TOGA (Fig. 12f), suggesting that tropical air–sea interaction plays a deterministic role in generating the basin-wide surface winds and sea level patterns. Given that the tropical Atlantic SST variability has little effect on SST (Fig. 1 of Sutton and Hodson 2005) and surface winds in the western and central tropical Pacific basin (Fig. 6 of Kushnir et al. 2010), the observed surface winds result primarily from tropical Indo-Pacific SST forcing. Consequently, the differences between Fig. 12a, b result largely from tropical Pacific SSTAs.
As discussed earlier, in addition to the TIO warming, the 1999–2010 negative IPO phase is associated with WTP warming, whereas the 1945–1977 negative IPO phase is associated with WTP cooling (Fig. 10a, c). The WTP warming enhances east–west SST gradient and western equatorial convection, and therefore strengthens the equatorial easterlies in the central-western equatorial basin. It is unclear, however, whether this SST pattern change is part of the IPO natural variability, or it is affected by anthropogenic warming. The steady Indian Ocean warming trend since the 1950s (Figs. 2c, 11) is attributed to increased greenhouse gases (Du and Xie 2008). As shown in Fig. 2c, however, the TIO cooling from 1945 to 1977 and warming from 1978 to 1998 coincide with the variability of IPO index, with IPO−/IPO+ being associated with TIO cooling/warming. Therefore, it is possible that the IPO warm transition also plays some role in warming up the TIO from the 1950s to 1990s. This speculation is supported by the considerably faster warming rates in the central-eastern tropical Pacific but only slightly faster warming in the TIO from 1950 to 1998 (compare Fig. 11c, d–a, b).
Of particular interest is that while the IPO enters its negative phase after 1998, the TIO stays warm and does not follow the IPO index (Fig. 2c), suggesting that the warm TIO is not likely a response to IPO phase change. Rather, it is contributed from the steady TIO warming trend that is largely attributed to anthropogenic warming by previous study. This argument is consistent with the somewhat slower TIO (and WTP) warming rates from 1950 to 2010 than those for 1950–1998, but markedly reduced warming in the tropical Pacific (Fig. 11). Consequently, it is possible that anthropogenic warming plays some role in causing the changing IPO/TIO SST relationship on multidecadal timescales. Cross-spectral analysis shows that on multi-decadal timescales, coherency between the TIO and WTP SSTA is high (>0.7), and the TIO SSTA leads that of the WTP (not shown). The coherency is significantly lower with the CEQP SSTA. Meyssignac et al. (2012) attributed the 1993–2009 sea level dipole pattern in the tropical Pacific to natural variability, because similar patterns are found in climate model control simulations without anthropogenic forcing. Our results suggest that natural internal variability associated with the IPO cold transition determines the basin-scale sea level trend pattern since 1993. Warming in the WTP and TIO, however, are essential for causing the intensification of the WTP SLR, and the TIO warming may have been contributed partly from anthropogenic forcing. This regional intensification, rather than basin-wide pattern alterations, requires particular attentions when we assess the relative roles of IPO natural variability and anthropogenic forcing. The evidence provided in this paper supports, but is not sufficient to prove, the anthropogenic warming effect. Attributing the anthropogenic effect versus natural variability is beyond the scope of this study, but it is an essential next step for our future research.
3.3 Decadal sea level variability and amplification since late 1980s or early 1990s
3.3.1 Western tropical Pacific region
The intensification of decadal sea level since the early 1990s is particularly strong in the WTNP (western box of Fig. 14), with a STD of 10–20 year sea level variability being 2.84 cm from 1991 to 2005, comparing to 1.12 cm for 1963–1976 and 1.31 cm for 1977–1990, based on the 8 year lowpassed thermosteric sea level data (Table 2; Fig. 16e). The 10–20 year sea level variability averaged in the WTNP region exceeds 1 STD near 1988 and 2 STDs near 1993 (Fig. 16e, black curve), whereas before 1988 it is near or within 1 STD. Note that near 1980, negative sea level anomaly also exceeds 1 STD. The variability amplitude from 1980 to 1988, however, is notably larger than that from 1974 to 1980. The intensification of 10–20 year sea level variability of WTP is also detected by ORA-S3 product (Table 2), and can be visually identified in tide gauge observations at stations Malakal and Guam, which are located in the western tropical north Pacific (Fig. 5 of Merrifield 2011) where the 10–20 year sea level variability obtains large amplitudes since the late 1980s especially since the early 1990s (Figs. 14, 15, 16).
Why does the 10–20 year sea level variability in the WTP amplify, whereas the 10–20 year IPO index does not? To answer this question, we examine the Indo-Pacific SST variability particularly in the tropics, where air–sea interaction is shown to determine the basin-wide surface winds that drive the basin-wide multidecadal sea level patterns (Sect. 3.2.3). Within the 1945–1977 multidecadal IPO− phase, IPO index goes from high to low from 1968 to 1974 (Figs. 2c, 16a). The 8 year lowpassed SST difference between 1974 and 1968 shows negative IPO pattern in the Pacific, which corresponds to large-scale cooling in the TIO (Figs. 2c, 14h, 16a). This Indo-Pacific SSTA combination resembles the effect of La Niña on interannual timescales, which induces basin-wide cooling in the TIO. Our idealized AGCM experiment shows that TIO cooling induces anomalous westerlies in the WTP and along the equator (opposite pattern of Fig. 12b; Sect. 2.3), which act to reduce the WTP sea level and thus counteract the WTP SLR caused by the anomalous easterlies associated with the negative IPO (Figs. 14d, 15h).
Within the 1978–1998 multidecadal IPO+ period, IPO index changes from high to low during 1982–1989 (Figs. 2c, 16a), warm SSTA appears in the TIO and central-western tropical Pacific, and cold SSTA occurs in the eastern south Pacific (Fig. 14g). This SSTA pattern is a transition from the Indian-Pacific SSTA in phase relation (Fig. 14h) to out of phase relation (Fig. 14f). This relation change occurs near 1985, after which decadal TIO warming/cooling corresponds to Pacific cooling/warming (Figs. 2c, 16a), as discussed in Sect. 3.1. From 1989 to 1994, IPO varies from cold to warm, and the TIO changes from warm to cold (Figs. 14f, 16a). The cold TIO acts in concert with the warm Pacific, which is represented by the amplified (TIO–IPO) difference index in late 1980s (Fig. 16a, c, black curve), producing westerly wind anomalies and positive we in the tropical Pacific (Fig. 16c, blue and red curves), and therefore causes large amplitude sea level fall in the WTP (Figs. 14b, 15e, 16e).
Within the 1999–2010 multidecadal IPO− phase, IPO index varies from low near 2000 to high near 2005. The IPO variability, however, is weak relative to the previous 10–20 year variability amplitudes (Figs. 2c, 16b). Meanwhile, the TIO SST cools in most areas (e.g., the Arabian Sea, Bay of Bengal and southeast Indian Ocean), even though it warms in a region near the equator (Fig. 14e). This equatorial warming might be a response to the warm tropical Pacific via atmospheric bridge, as discussed earlier. Note that the tropical Pacific warming obtains the maximum near the central equatorial Pacific, with larger amplitudes than that from 1989 to 1994 (Fig. 14f). With this SST combination, sea level reduces rather rapidly in the WTP, with a reduction area much smaller than that from 1989 to 1994, a period when TIO exhibits basin-wide cooling (Figs. 14a, b, 15c, e). Different from the 1989 to 1994 case, sea level shows fast rise in the central-eastern equatorial basin from 2000 to 2005, whereas that for 1989–1994 is very weak (Figs. 15c and 14a versus 15e and 14b). This central equatorial Pacific decadal sea level variability will be discussed next (Sect. 3.3.2). Note that the 10–20 year data from 2000 onward are excluded in Fig. 16 (left column) by removing the filter’s end point effect. The larger variation of (TIO–IPO) difference than IPO index is seen in 8 year lowpassed data (Fig. 16b), and the amplified wind and SSHA responses can be identified in Fig. 16d, f. Given that surface wind anomalies converge to the central equatorial region where warm SSTA is large (Figs. 14e, 15a, c) and mean SST gradient is strong (mean SST >28 °C; Fig. 8), the stronger SSTA in the central-equatorial basin may have also contributed to the strong anomalous westerlies and thus the rapid WTP sea level fall.
3.3.2 Central-equatorial basin and relation to the WTP
By examining the decadal sea level variability patterns within each of the IPO multidecadal period, we see intensified amplitude in the central equatorial Pacific at the beginning of the twenty-first century (compare Fig. 14a with b–d, Fig. 15c with e, f, h). If we also examine the 1994–2000 period during which IPO index varies from positive to negative, which is decadal scale variability overlying on the multidecadal phase transition (compare Fig. 16a, b, red curves), we find that the 10–20 year timescale sea level fall in the central equatorial basin and the SLR in the WTP appear to be linked (Figs. 15b, d, 16e, g). Both are associated with wind anomalies in the central-western equatorial Pacific (Fig. 16e, g), which in turn are associated with tropical Pacific cooling and TIO warming (Fig. 16a, b). The correlation between WTP and CEQP 8 year lowpassed thermosteric sea level is −0.56 from 1950 to 2006. While strong WTP sea level fall (rise) occurs during 1989–1994 (1994–2000), large amplitude sea level variability in the CEQP primarily appears during 1994–2000 (Figs. 15d, e, 16e, g). This indicates that it is the pattern change of the tropical Pacific SSTA, which shifts to the central Pacific El Niño pattern—with maximum warming occurring in the central equatorial ocean and cooling in the eastern and western equatorial basins (Fig. 14e; the pattern for 1994–2000 is similar)—that may have played a critical role in causing the large amplitude decadal sea level variability in the central-equatorial basin. This structure is in contrast to the multidecadal sea level trend patterns (Figs. 1, 4a), which have large amplitudes in the WTP and eastern basin, with little sea level change in between.
From 1974 to 1981, when IPO decadal scale variability overlies on multidecadal phase transition from negative to positive, the TIO SST increases with the IPO index. This in-phase SSTA between the Indian and tropical Pacific produces a weaker (TIO–IPO) variation and an out of phase effect on surface wind and sea level, and therefore produces weaker wind and sea level variability in the tropical Pacific Ocean than the 1994–2000 period (Figs. 15d, g, 16). The 8 year lowpassed sea level variability, which includes both decadal and multidecadal variability, exceeds 1 STD near 1991 in both the WTP and CEQP region (Fig. 16f, h), and the 8–20 year sea level anomaly in the WTP exceeds 1 STD near 1988 (Fig. 16e). These results explain the increased sea level energy at 8–16 year periods since the late 1980s at the Fremantle tide gauge station (Feng et al. 2010), and the larger Pacific influence on the TIO decadal sea level variability since the early 1990s (Trenary and Han 2013). Note that the intensification of decadal sea level variability in the CEQP is not apparent in ORA-S3 sea level product (Table 2). This intensification, however, is shown in both thermosteric sea level and satellite observed AVISO sea level data, and the two have good agreements (Fig. 15a, d).
3.3.3 The changing role of the Indian Ocean
The WTP sea level variability is significantly correlated with the IPO index, with a correlation coefficient of −0.72 for the 8 year lowpassed sea level and IPO index during 1950–2006 (Table 3). It is, however, essentially uncorrelated with the TIO SST PC1 for the entire 1950–2006 period. This low correlation is because of the correlation coefficient changing sign before and after 1985. The correlation between WTP sea level variability and TIO SST PC1 is negative (r = −0.48) from 1950 to 1984, but becomes positive and increased to 0.76 for 1985–2006 (Table 3). These results indicate that decadal variability of TIO SST has played a larger role in determining the WTP decadal sea level change since 1985 than the preceding decades. The 10–20 year TIO SST variability may have resulted from the IPO influence before 1985, with warm/cold tropical Pacific corresponding to warm/cold TIO; however, it is not likely due to the IPO after 1985 since the variability is in opposite sign with the IPO effects obtained by previous studies. Note that the IPO is highly correlated with the 8 year lowpassed ENSO index, and WTP sea level variability has higher correlation with Nino3.4 SST (r = −0.87) than with the IPO (Table 3). The correlation between IPO index and TIO PC1 is 0.75 for 1900–1984 and −0.85 for 1985–2008, with both exceeding 99 % significance. To confirm these results, we also analyzed Hurrell SST and Kaplan SST, and obtain similar results on IPO and Indian Ocean SST patterns, their temporal variability (PC1), and the changing correlations between their PC1 s. These results suggest that tropical air–sea interaction over the Indo-Pacific basin may have played a deterministic role in causing the WTP decadal and multi-decadal sea level variability since the 1950s, with the TIO being more active after 1985, and air–sea coupling in the tropical Pacific associated with the IPO is crucial for generating the basin-wide surface winds that drive the basin-wide sea level patterns.
4 Summary and discussions
In this paper, we investigate the changing patterns of Pacific sea level variability on decadal (10–20 year) and multidecadal (>20 year) timescales since the 1950s, and understand their causes. Satellite observations, in situ datasets, reconstructed sea level and ocean–atmosphere reanalysis products (Sect. 2.1) are analyzed to detect the sea level pattern changes. Standalone ocean and atmosphere model experiments are performed to understand the causes.
Our results show that during the decadal and multidecadal phase transition periods of the IPO, there are basin-wide changes in surface winds, which are associated with basin-wide changes in sea level. The satellite-observed rapid SLR (global mean removed) in the WTP and fall in the eastern basin from 1993 to 2010, together with the corresponding surface winds (Figs. 1, 4a), are associated with IPO multidecadal phase transition from positive to negative (Fig. 2). The IPO index, however, has not intensified on either decadal or multidecadal timescales since 1993 relative to the preceding decades, while marked intensification of sea level variability is detected on both timescales in the WTP region, and on decadal scale in the central-equatorial basin (Tables 1, 2). The intensified sea level and surface wind signals are robust to cross-dataset differences (Figs. 4, 5, 6, 7, 14, 15, 16, 17; Tables 1, 2). In contrast to the multidecadal sea level trend patterns since 1993, which show a strong west-east dipole with little variability in the central basin (Figs. 1, 4), decadal sea level variability since the early 1990s exhibits a strong WTNP versus central-equatorial dipole (Fig. 15a–d). While the 10–20 year thermosteric sea level averaged in the WTP exceeds 1 STD near 1988, the 8 year lowpassed sea level signals (including both decadal and multidecadal variability) exceed 1 STD near 1991 in both WTP and CEQP (Fig. 16).
On multidecadal timescales, the basin-wide sea level patterns—including the intensified WTP SLR since 1993—result primarily from surface wind forcing, as shown by the wind-driven linear ocean model experiment (Figs. 6, 9). The observed basin-wide surface winds—including the intensified easterlies and negative Ekman pumping velocity since the early 1990s (Fig. 16)—result primarily from tropical SST forcing, as shown by the high correlation between wind and (TIO–IPO) index (Fig. 16) and demonstrated by AGCM_GOGA and AGCM_TOGA experiments (Fig. 12). These results suggest that tropical air–sea interaction—particularly in the tropical Indo-Pacific basin—may have played a deterministic role in causing the basin-wide winds and thus sea level patterns in the past few decades.
During the two negative IPO multidecadal phase periods of 1945–1977 and 1999–2010, the major SST difference over the tropical Indo-Pacific basin is warming over the Indian and WTP (west of 150°E) for the latter but cooling for the former period (Fig. 10). The TIO and WTP warm up faster than the rest of the tropical oceans from 1950 to 2010 (Fig. 11). This faster warming explains the SSTA pattern change shown in Fig. 10. Our idealized AGCM experiments show that a warmer TIO generates easterlies and negative Ekman pumping velocity in the tropical Pacific, with largest amplitudes occurring in the WTP basin (Fig. 12b). These wind anomalies act in concert with the anomalous easterlies associated with the IPO cold transition, intensify the zonal winds and Ekman pumping velocity in the western-central tropical Pacific and drive the accelerated WTP SLR from 1993 to 2010 through both local and remote processes (Sect. 3.2.3; Fig. 16). In addition, the warmer WTP increases the east–west SST gradient and enhances convection in the western equatorial basin (Figs. 8, 10, 13), and thereby strengthening equatorial easterlies and contributing to the rapid WTP SLR. In contrast, during previous decades, the TIO and WTP SSTAs are in phase with the central-eastern tropical Pacific SSTA associated with the IPO (Fig. 16b). This in-phase SSTA relation produces out of phase effects on the WTP SLR and results in weak sea level variability in the WTP region (Fig. 16).
On decadal timescales, the IPO is also associated with basin-wide sea level changing patterns, with rapid WTP SLR corresponding to IPO cold transition (index from high to low), which is analogous to the multidecadal variability (Figs. 14, 15, 16). This intensification of WTP sea level variability is associated with the out of phase relationship between the IPO and TIO SSTA since 1985, with positive (negative) IPO corresponding to cold (warm) TIO (Figs. 2c, 16a; Table 3). This SSTA combination produces “in phase” effects on tropical Pacific wind and sea level variability, because TIO warming (cooling) strengthens the anomalous easterlies (westerlies) and negative (positive) Ekman pumping velocity associated with the negative (positive) IPO (Figs. 14, 15, 16, 17; Sect. 3.3). Before 1985, the situation is opposite, with the TIO SSTA coinciding with the decadal variability of IPO index, which resembles ENSO influence on TIO SST at interannual timescales. This in-phase SSTA patterns produce out of phase effects on surface winds and decadal sea level variability, and thus weaken the WTP sea level amplitude. In contrast, the intensified sea level amplitude in the CEQP (Figs. 15, 16) is likely linked to the shifting pattern of Pacific SSTA to the central Pacific ENSO type (Fig. 14; Sect. 3.3.2).
Previous studies have attributed the steady Indian Ocean warming since the 1950s to anthropogenic greenhouse gases (e.g., Du and Xie 2008). It is likely that the IPO warm transition also has some contributions to the TIO warming trend from 1950 to 1998, given that the TIO SSTA coincides with the multidecadal variability of IPO index from the 1950s to 1990s. After 1998, the IPO enters negative phase, whereas the TIO stays warm and does not follow the IPO index (Fig. 2c), and the warm TIO is contributed from its “persistent warming trend” (Fig. 11) that is attributed to anthropogenic warming. These results indicate that anthropogenic warming may have contributed to the TIO warming and therefore WTP rapid SLR during recent decades. Meyssignac et al. (2012) attributed the fast WTP SLR to natural variability, since similar east–west dipole patterns in the tropical Pacific were found in the simulations of climate models with and without anthropogenic forcing. Our results show that while the Pacific decadal climate modes are associated with basin-wide sea level patterns, a result consistent with Meyssignac et al. (2012), the “acceleration” of SLR in the WTP region since the early 1990s is likely contributed partly from anthropogenic warming. This inference is supported, but is not proved, by the evidence provided in this paper. It is unclear whether the global climate models are able to simulate such a regional feature, and if yes whether or not this can be detected as a significant pattern change. This raises challenge for climate models for improving their ability to simulate regional SLR, in order to provide meaningful decadal predictions and future projections on regional scales.
Why is the TIO decadal SST variability in phase with the IPO before 1985, but out of phase since 1985? This out of phase relationship may indicate the active role of the TIO in affecting the Indo-Pacific climate, because a warm TIO favors generating a negative IPO pattern. It is clear that the TIO has played a more active role in affecting the Pacific decadal variability since the early 1990s. One hypothesis is that the accelerated warming over the TIO and WTP, where the mean SST is high (Fig. 8), makes convection more sensitive to its SST variability due to nonlinear dependent of convection on SST as the mean SST exceeds the threshold of 27–28 °C (Gadgil et al. 1984; Graham and Barnett 1987; Waliser et al. 1993). The threshold, however, may increase with the increase of tropical SST (Johnson and Xie 2010). A key issue following this study emerges: How is the decadal SST variability generated over the TIO? Note that on interannual timescales, strong ENSO variability still dominates TIO SST both before and after 1985 (Fig. 3). Finally, our analyses are based on relatively short data record. Longer-term, reliable data need to be collected to obtain more statistically significant conclusions.
We thank Dr. Clara Deser for reading the earlier version of this manuscript and providing helpful comments, and Dr. Adam Phillips for providing the CAM3 TOGA experiment results. Appreciation also goes to Dr. Martin Hoerling for the NSIPP model experiments, and Dr. James McWilliams for stimulating discussions at the earlier stage of this work. W.H. is supported by NSF CAREER award OCE 0847605. Portions of this study were supported by the Office of Science (BER), US Department of Energy, Cooperative Agreement No. DE-FC02-97ER62402, and the National Science Foundation. NCAR is sponsored by the National Science Foundation. D.Y. is supported by China 973 project. We thank NCAR CISL for computational support.
- Alexander MA (2010) Extratropical air–sea interaction, SST variability and the pacific decadal oscillation (PDO). In: Sun D, Bryan F (eds) Climate dynamics: why does climate vary. AGU Monograph #189, Washington, DC, pp 123–148Google Scholar
- Atlas R, Ardizzone J, Hoffman RN (2008) Application of satellite surface wind data to ocean wind analysis. In: Proceedings of SPIE, vol 7087, 70870B. doi:10.1117/12.795371
- Bindoff NL, Willebrand J, Artale V, Cazenave A, Gregory J et al (2007) Observations: oceanic climate change and sea level, ch5. In: Solomon S, Qin D, Manning M et al (eds) Climate change 2007: the physical science basis, contribution of working group I to the fourth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, pp 385–432Google Scholar
- Ducet N, Le Traon PY, Reverdin G (2000) Global high resolution mapping of ocean circulation from TOPEX/Poseidon and ERS-1/2. J Geophys Res 105:19477–19498Google Scholar
- Hamlington BD, Leben R, Nerem RS, Han W, Kim KY (2011) Reconstructing sea level using cyclostationary empirical orthogonal functions. J Geophys Res 116. doi:10.1029/2011JC007529
- Hoerling M, Hurrell J, Kumar A, Terray L, Eischeid J, Pegion P, Zhang T, Quan X, Xu T (2011) On North American decadal climate for 2011–2020. J Clim 24. doi:10.1175/2011JCLI4137.1
- Hurrell JW, Meehl GA, Bader D, Delworth T, Kirtman B, Wielicki B (2009) A unified approach to climate system prediction. Bull Amer Meteor Soc 90:1819–1832Google Scholar
- Kaplan A, Cane M, Kushnir Y, Clement A, Blumenthal M, Rajagopalan B (1998) Analyses of global sea surface temperature 1856–1991. J Geophys Res 103:18567–18589Google Scholar
- Köhl A, Stammer D (2008) Variability of the meridional overturning in the North Atlantic from the 50-year GECCO state estimation. J Phys Oceanogr 38:1914–1930Google Scholar
- Kumar A, Hoerling MP (1998) On the specification of regional SSTs in AGCM simulations. J Geophy Res 103:8901–8907Google Scholar
- Kushnir Y, Seager R, Ting M, Naik N, Nakamura J (2010) Mechanisms of Tropical Atlantic SST influence on North American precipitation variability. J Clim 23. doi:10.1175/2010JCLI3172.1
- Liebmann B, Smith CA (1996) Description of a complete (interpolated) outgoing longwave radiation dataset. Bull Am Meteorol Soc 77:1275–1277Google Scholar
- Liu Z, Alexander M (2007) Atmospheric bridge, oceanic tunnel, and global climatic teleconnections. Rev Geophys 45:RG2005. doi:10.1029/2005RG000172
- Luo JJ, Sasaki W, Masumoto Y (2012) Indian Ocean warming modulates Pacific climate change. Proc Natl Acad Sci 109:18701–18706Google Scholar
- Meehl GA, Goddard L et al (2009) Decadal prediction: can it be skillful? Bull Am Meteorol Soc 90:1467–1484Google Scholar
- Meyssignac B, Salas y Melia D, Becker M, Llovel W, Cazenave A (2012) Tropical Pacific spatial trend patterns in observed sea level: internal variability and/or anthropogenic signature? Clim Past Discuss 8:349–389Google Scholar
- Randall DA, Wood RA, Bony S et al (2007) Climate models and their evaluation, ch. 8. In: Solomon S, Qin D, Manning M et al (eds) Climate change 2007: the physical science basis, contribution of working group I to the fourth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, pp 589–662Google Scholar
- Simmons A, Uppala S, Dee D, Kobayashi S (2007) ERA-Interim: new ECMWF reanalysis products from 1989 onwards. ECMWF Newslett 110:25–35Google Scholar
- Trenary L, Han W (2013) Local and remote forcing of decadal sea level and thermocline depth variability in the South Indian Ocean. J Geophys Res Oceans 118. doi:10.1029/2012JC008317
- Trenberth KE et al (2007) Surface and atmospheric climate change. In: Solomon S, Qin D, Manning M et al (eds) Climate change 2007: the physical science basis, contribution of working group I to the fourth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, pp 235–336Google Scholar