Abstract

The tropical Atlantic interhemispheric gradient in sea surface temperature significantly influences the rainfall climate of the tropical Atlantic sector, including droughts over West Africa and Northeast Brazil. This gradient exhibits a secular trend from the beginning of the twentieth century until the 1980s, with stronger warming in the south relative to the north. This trend behavior is on top of a multidecadal variation associated with the Atlantic multidecadal oscillation. A similar long-term forced trend is found in a multimodel ensemble of forced twentieth-century climate simulations. Through examining the distribution of the trend slopes in the multimodel twentieth-century and preindustrial models, the authors conclude that the observed trend in the gradient is unlikely to arise purely from natural variations; this study suggests that at least half the observed trend is a forced response to twentieth-century climate forcings. Further analysis using twentieth-century single-forcing runs indicates that sulfate aerosol forcing is the predominant cause of the multimodel trend. The authors conclude that anthropogenic sulfate aerosol emissions, originating predominantly from the Northern Hemisphere, may have significantly altered the tropical Atlantic rainfall climate over the twentieth century.

1. Introduction

The Tropical Atlantic interhemispheric gradient in sea surface temperature (SST) significantly influences the climate of the tropical Atlantic sector, including West Africa (Giannini et al. 2003) and Northeast Brazil (Moura and Shukla 1981); the persistent Sahel drought since the late 1960s has been particularly severe (Nicholson 1993). While the warming of tropical North Atlantic temperatures has been the focus of Atlantic climate changes studies (Mann and Emanuel 2006; Evan et al. 2009)—largely because of its link to hurricane activity—our prevailing understanding of tropical Atlantic climate variability points to the thermal contrast between the tropical North and South Atlantic as the strongest determinant of climate changes there (Xie and Carton 2004). This is due to the strong control that the interhemispheric SST gradient has on the intertropical convergence zone (ITCZ) (Hastenrath and Greischar 1993) and its subsequent effect on tropical Atlantic circulation and climate (Moura and Shukla 1981). Forced changes in tropical Atlantic climate—whether from natural or anthropogenic causes— potentially manifest themselves in a similar way; indeed, climate forcings such as anthropogenic aerosols have hemispheric asymmetric distributions and may directly or indirectly drive an interhemispheric response (Williams et al. 2001; Rotstayn and Lohmann 2002; Biasutti and Giannini 2006; Ming and Ramaswamy 2009). This motivates us to investigate long-term changes—multidecadal and longer—in the Atlantic interhemispheric gradient and examine their causes.

The main result of our study is that there is a pronounced secular trend in the tropical Atlantic interhemispheric gradient over the twentieth century, with the tropical South Atlantic warming faster than the tropical North Atlantic and with resulting implications for the Atlantic ITCZ—and that the predominant cause of this trend is due to the increase in sulfate aerosol forcing, presumably due to anthropogenic sources. The trend behavior is also distinct from known multidecadal variations of North Atlantic SST by the Atlantic multidecadal oscillation (AMO).

The influence of anthropogenic sulfate emissions in causing preferential cooling to the Northern Hemisphere and shifting the tropical ITCZ southward over the twentieth century has been previously suggested (Williams et al. 2001; Rotstayn and Lohmann 2002; Rotstayn and Penner 2001; Ming and Ramaswamy 2009). However, those studies used an idealized modeling framework—an AGCM coupled to a mixed layer ocean—that did not factor in the potential role of ocean dynamics. Another study (Mann and Emanuel 2006) has suggested that anthropogenic aerosols cooled the tropical North Atlantic over the twentieth century but solely from observational inference. The previous studies also did not compare the relative roles of other climate forcings on the tropical ITCZ nor the sensitivity across different climate models to imposed sulfate aerosol forcings. In this study, we come to our conclusions based on a comprehensive analysis of a multimodel ensemble of the twentieth-century coupled model simulations.

2. Data and methods

a. Observational data and model simulations

Three reconstructed SST datasets are analyzed in this study: the Hadley Center SST dataset (HadISST; Rayner et al. 2003), the National Oceanic and Atmospheric Administration (NOAA) Extended Reconstruction SST (ERSST v3b; Smith et al. 2008), and the Kaplan Extended SST v2 (Kaplan et al. 1998). We also analyze the meridional wind dataset from the Comprehensive Ocean–Atmosphere Dataset (COADS; Slutz et al. 1985).

We utilize the output of multimodel ensembles of the Coupled Model Intercomparison Project phase 3 (CMIP3; Meehl et al. 2007) model simulations; output from the twentieth-century climate simulations and the preindustrial simulations are analyzed. Table 1 lists the models and the length of data for each model participating in the current analysis. For the twentieth-century simulations, there are a total of 71 runs across 23 different coupled models; these simulations are forced with estimated historical anthropogenic and natural forcings—greenhouse gases, anthropogenic sulfate aerosols, ozone, volcanic aerosols, and solar insolation. Preindustrial simulations are model runs where the forcings are held constant during the simulations.

Table 1.

Model simulations and ensemble size available for this analysis.

Model simulations and ensemble size available for this analysis.
Model simulations and ensemble size available for this analysis.

We also examine output of single-forcing simulations—twentieth-century simulations as in the ones above but with only one of the climate forcings applied in any one simulation—from three coupled models, the Community Climate System Model, version 3 (CCSM3) and the Parallel Climate Model version 1 (PCM1) of the National Center of Atmosphere Research (NCAR) and the National Aeronautics and Space Administration (NASA) Goddard Institute for Space Studies Model E (GISS-E). There are four types of single forcings: greenhouse gases, sulfate aerosols, volcanic aerosols, and solar insolation. Each single-forcing experiment contains seven runs—two from CCSM3, four from PCM1, and one from GISS-E. CCSM3 and PCM1 data are downloaded from the Earth System Grid Web site (http://www.earthsystemgrid.org/).

b. Methods

We derived an index of the Atlantic interhemispheric SST gradient (hereafter the AITG index) by subtracting the average SST across the entire tropical North Atlantic basin (5°–35°N, 0°–80°W) from the tropical South Atlantic (5°–35°S, 60°W–20°E) (Fig. 1a) and applying a 21-yr running mean to emphasize multidecadal and longer time-scale variations. The same data processing is applied to both observed and model-simulated monthly resolution data. Monthly anomalies are calculated by removing the long-term climatology. A 21-yr running mean is applied to the monthly anomaly, and the annual mean of this data is computed afterward. All subsequent analyses are applied on the annual-mean data.

Fig. 1.

(a) The observed AITG indices and results of EEMD analyses. The AITG index is defined as the SST difference between the tropical North Atlantic (5°N–35°N, 0°-80°W north solid-line box) and the tropical South Atlantic (5°–10°S, 60°W–20°E, south solid-line box; south minus north). (b) AITG indices from three different observational datasets: solid line: Hadley SST, long dash line: ERSST, and dot dash line: KAPLAN extended SST v2 (units: °C). (c) IMF 4 of the EEMD analysis of the indices in (b), showing multidecadal variations. (d) IMF 5, showing a trend that reverses in the 1980s. (e) The area-averaged meridional winds of the dashed-line box (0°–6°N, 10°–50°W) in (a), indicating a progressively southward ITCZ consistent with the trend behavior in the AITG (units: m s−1).

Fig. 1.

(a) The observed AITG indices and results of EEMD analyses. The AITG index is defined as the SST difference between the tropical North Atlantic (5°N–35°N, 0°-80°W north solid-line box) and the tropical South Atlantic (5°–10°S, 60°W–20°E, south solid-line box; south minus north). (b) AITG indices from three different observational datasets: solid line: Hadley SST, long dash line: ERSST, and dot dash line: KAPLAN extended SST v2 (units: °C). (c) IMF 4 of the EEMD analysis of the indices in (b), showing multidecadal variations. (d) IMF 5, showing a trend that reverses in the 1980s. (e) The area-averaged meridional winds of the dashed-line box (0°–6°N, 10°–50°W) in (a), indicating a progressively southward ITCZ consistent with the trend behavior in the AITG (units: m s−1).

Ensemble empirical mode decomposition (EEMD; Huang and Wu 2008; Wu and Huang 2009) analysis is applied to the observed AITG index to objectively extract temporal modes of low-frequency behavior. Briefly, empirical mode decomposition (EMD) is an adaptive and temporally local time-series analysis method, with the basis of the decomposition derived from the data. The time series s(t) is decomposed into intrinsic mode functions (IMFs) cj,

 
formula

where rM is the residual after M intrinsic modes are extracted. The total number of IMFs of a time series is close to log2N, where N is the number of data points in the time series. Ensemble EMD (EEMD) analysis is an improved version of EMD whereby finite random white noise is introduced initially to the analyzed time series to “fill” time scales that may not be part of the data; this is required to prevent mixing of different time scales into the same mode in the EMD analysis. In the EEMD method, the EMD is applied many times to the time series (with added noise) and the signal extracted through an ensemble average of the modes extracted. EEMD has been successfully applied to analyze nonstationary geophysical time series (e.g., Wu et al. 2008).

Detection and attribution analysis is employed to determine whether an observed change in the AITG index comes from internal variability of the climate system or as a response to the external forcing. Detection is the process of statistically demonstrating that the observed change is distinguishable from natural climate variations. It is usually done through comparing the observed change against comparable simulations of coupled general circulation models (CGCMs). Attribution is the process of establishing cause and effect in a scientifically rigorous way and is usually done through comparing climate changes in single-forcing simulations with those of the all-forcing simulations (Barnett et al. 2005). The detection and attribution method we employ is described later in the text.

3. Results

a. Observed behavior of the Atlantic interhemispheric SST gradient

The AITG indices derived from three observational SST datasets from the late 1800s exhibit common behavior throughout most of the record, except in the earliest part (Fig. 1b). The interhemispheric gradient exhibits a long-term increasing trend (south warming more than the north) throughout the twentieth century, superimposed on a pronounced multidecadal variation. This observation is supported by the results of EEMD analysis of the AITG index, which readily extracts a trend IMF (Fig. 1d) separate from the multidecadal behavior (Fig. 1c). Results of sensitivity tests show that the decompositions of these two modes are robust under changes in the end years of the AITG index. The post-80s turning points of the two modes also remain robust under sensitivity tests (see  appendix).

The multidecadal IMF (Fig. 1c) matches well with a known variation of North Atlantic SST—AMO (Kushnir 1994; Delworth and Mann 2000; Schlesinger and Ramankutty 1994)—and several previous studies (e.g., Knight et al. 2006; Ting et al. 2009) have suggested the AMO to be a natural fluctuation of the climate system. The correlation between the current EEMD-derived multidecadal IMF from Kaplan SST AITG with the AMO index derived by Ting et al. 2009 is −0.95 (−0.89 and −0.87 for Hadley SST and ERSST AITG multidecadal IMFs, respectively. Note that Kaplan SST was used in the study of Ting et al.).1 SST regression onto this multidecadal mode shows a strong interhemispheric SST gradient with opposite phases in the Southern and Northern Hemisphere in the Hadley SST dataset (Fig. 2a). SST regression patterns using ERSST or Kaplan SST datasets are similar (not shown).

Fig. 2.

Observed (Hadley) SST regressed onto the EEMD (a) IMF 4 and (b) IMF 5. Light-gray shading represents cooling, and dark-gray represents warming. Dotted areas represent areas of significant correlation (greater than 0.17 or smaller than −0.17).

Fig. 2.

Observed (Hadley) SST regressed onto the EEMD (a) IMF 4 and (b) IMF 5. Light-gray shading represents cooling, and dark-gray represents warming. Dotted areas represent areas of significant correlation (greater than 0.17 or smaller than −0.17).

The current study focuses on the century-long upward trend of the AITG index that peaks in the 1980s (Fig. 1d). In this case, an upward trend of the AITG index is caused by the tropical South Atlantic warming faster than the tropical North Atlantic, as shown by the regression of the SST onto this mode (Fig. 2b). It also implies a southward shift of the Atlantic ITCZ, given the strong dynamical control that the interhemispheric SST gradient has on the ITCZ position (Moura and Shukla 1981; Chiang et al. 2000; Chiang et al. 2002). To confirm the change to the ITCZ, we examine the meridional winds near the long-term mean position of the Atlantic ITCZ in the COADS dataset. Figure 1e represents the area-average anomalous meridional winds between 0°–6°N and 10°–50°W (the red box in Fig. 1a). A southerly trend in the meridional wind in the mean ITCZ latitudes (Fig. 1e) confirms the southward shift of ITCZ; note also that the multidecadal variation in the meridional wind roughly matches the multidecadal variation in the AITG (Fig. 1b), confirming our expectation that the meridional winds resemble the AITG behavior at multidecadal time scales.

b. Analysis of multimodel ensemble simulations of the twentieth-century interhemispheric gradient

To understand the cause of the long-term trend in the AITG, we compare the observed AITG index behavior against those in the CMIP3 multimodel ensemble–forced twentieth-century simulations and those in the preindustrial simulations of corresponding models. We first applied an empirical orthogonal function (EOF; Barnett 1977) analysis to the model AITG indices to extract the dominant ensemble behavior. The EOF is applied to the normalized AITG indices from the 71-member multimodel ensemble of the CMIP3 twentieth-century climate experiment. Prior to analysis, each index is normalized by the standard deviation of the same index from the appropriate preindustrial simulation. The first EOF (Fig. 3a), accounting for 49% of the total variance, shows an upward trend for most of the twentieth century, reversing after the 1980s. This closely resembles the trend behavior of the observed AITG index as extracted by the EEMD analysis (Fig. 1d). Most model AITG indices project positively onto EOF1 (Fig. 3b), indicating that most twentieth-century simulations reproduce the same sign in the trend of the AITG index as the observed.

Fig. 3.

Results of EOF analyses of the CMIP3 models’ AITG indices. (a) EOF1 and (b) the corresponding principal component using all 71 model-based AITG indices. The various models that make up the principal component loadings are indicated. The first EOF exhibits an upward trend, and most models project positively on EOF1, indicating that this upward trend in the AITG occurs in most twentieth-century model simulations. Since the EOF analysis was applied on normalized AITG indices, the EOFs and projection coefficients are dimensionless.

Fig. 3.

Results of EOF analyses of the CMIP3 models’ AITG indices. (a) EOF1 and (b) the corresponding principal component using all 71 model-based AITG indices. The various models that make up the principal component loadings are indicated. The first EOF exhibits an upward trend, and most models project positively on EOF1, indicating that this upward trend in the AITG occurs in most twentieth-century model simulations. Since the EOF analysis was applied on normalized AITG indices, the EOFs and projection coefficients are dimensionless.

Motivated by the EOF result, we focus on the long-term trend in the AITG index prior to the 1980s. The slope of the AITG trend from 1900 to 1982 (83-yr length) is calculated for each of the 71 CMIP3 twentieth-century simulations; the distribution of the slope magnitudes is shown in Fig. 4a. The notable aspect of this distribution is that it is significantly shifted in the positive direction. The multimodel ensemble mean AITG trend is 0.13°C (100 yr)−1; this value is close to the magnitude of the slope of the AITG trend as calculated from the KAPLAN SST dataset [0.15°C (100 yr)−1, Fig. 4a, dashed line]. It is smaller in magnitude than that computed in the Hadley SST [0.18°C (100 yr)−1] and ERSST [0.35°C (100 yr)−1] datasets.

Fig. 4.

Statistics of 83-yr trends of the modeled AITG indices from twentieth-century climate and (b) preindustrial simulations. (a) shows the distribution of slopes of the trends computed from each of the 71 twentieth-century model simulations (gray bars) is shown. Also shown is the ensemble mean slope (long dash line), slope for each of the observed AITG indices (short solid lines), and the averaged observed slope (long solid line). (b) shows the distribution of slopes computed from the corresponding preindustrial simulations; also shown is the ensemble mean slope of the preindustrial simulations (long dash-dotted line) and for the twentieth-century simulations [long dash line; units: 0.1°C (100 yr)−1].

Fig. 4.

Statistics of 83-yr trends of the modeled AITG indices from twentieth-century climate and (b) preindustrial simulations. (a) shows the distribution of slopes of the trends computed from each of the 71 twentieth-century model simulations (gray bars) is shown. Also shown is the ensemble mean slope (long dash line), slope for each of the observed AITG indices (short solid lines), and the averaged observed slope (long solid line). (b) shows the distribution of slopes computed from the corresponding preindustrial simulations; also shown is the ensemble mean slope of the preindustrial simulations (long dash-dotted line) and for the twentieth-century simulations [long dash line; units: 0.1°C (100 yr)−1].

We also calculate the distribution of slopes of 83-yr AITG trends extracted from the corresponding preindustrial simulations where the climate forcings are held constant. In computing the preindustrial AITG indices, there were indications of residual spinup effects in some of the preindustrial simulations expressed as long-term trends in the AITG index over the entire length of the simulation. To minimize these effects, we use only the AITG indices over the last 240 yr of preindustrial model output; furthermore, we detrend the resulting 240-yr-long indices prior to analysis. (Except for two models that have only 100 yr of preindustrial simulations; in these cases, their AITG indices are not detrended—note that the same statistical tests are applied on the preindustrial model output without detrending, and the conclusions remain similar.) For each preindustrial run, overlapping 83-yr-trend slopes are calculated, starting from the first year and stepping the window forward in 1-yr increments. The estimated degrees of freedom (DOF) of the preindustrial slope distribution are calculated conservatively by simply dividing, for each model, the length of the preindustrial simulation used (240 yr, in general) by 83 (i.e., the length of the window used to compute the trend slope), and then summing this value across the various models. The resulting distribution is shown in Fig. 4b.

The twentieth-century forcings have clearly affected the distribution of the AITG trend. We compare the difference in the means of the two distributions; based on the one-tailed statistic test (Fig. 4 and Table 2), the mean of the twentieth-century simulation is larger than the mean of the preindustrial simulation with well above the 99% level of confidence. These results are not sensitive to a reasonable change in the length of the record taken to compute the trend slope (results not shown). That is to say, in the average sense, the twentieth-century forcings to the climate system are making the warming in the South Atlantic stronger than that in North Atlantic.

Table 2.

83-year trend statics for the Atlantic Interhemispheric SST gradient.

83-year trend statics for the Atlantic Interhemispheric SST gradient.
83-year trend statics for the Atlantic Interhemispheric SST gradient.

Does this result mean, however, that the trend in the observed AITG—viewed as a single realization—is definitely forced? To answer this, we use the model-derived distributions to estimate the probability that the observed trend could arise purely from natural variation. Based on the preindustrial distribution, the probability that the slope of the AITG trend equals or exceeds the observed—here, as an average of the trends computed from the three SST products—is 7%. In other words, the likelihood that the observed trend arises purely from natural variation, while finite, is small. On the other hand, the same probability computed from the twentieth-century distribution is 27%. Thus, it is more reasonable to conclude that at least a portion of the observed twentieth-century upward trend of the AITG was forced. The mean observed trend is 0.22°C (100 yr)−1, while the ensemble mean trend of the twentieth-century simulations is 0.13°C (100 yr)−1; the model ensemble mean may represent the strength of the forced signal of the trend of the AITG. This suggests that at least half the observed AITG trend is forced or that the CMIP3 models’ total aerosol forcing might be underestimated.

c. Attribution

To isolate which external forcings—greenhouse gases, sulfate aerosols, volcanic aerosols, solar radiation, and ozone—contribute to the AITG trend, we examine output from single-forcing experiments of three models—CCSM3, PCM1, and GISS-E—available to us (seven members in total for each forcing). Figure 5 shows the cross-model average of the AITG indices from these runs. Clearly, the averaged AITG index behavior of the sulfate aerosol forcing-only simulations (Fig. 5c) most resembles the first EOF of the 71 model indices (Fig. 3a); more importantly, it also has a similar turning point in the 1980s. The AITG in the other single-forcing simulations, on the other hand, does not exhibit comparable behavior. Each individual AITG index from the sulfate aerosol–forcing-only simulations is highly correlated with that first EOF, and the signs of the correlation coefficients are always positive (Table 3). In the cases of other forcings, much smaller correlation coefficients are derived, and some of them are even negative, implying that there is no consistency in the sign of simulated AITG index behavior of those simulations. Thus, the single-forcing simulations indicate that sulfate aerosol changes over the twentieth century are the most likely cause of the upward trend in the AITG.

Table 3.

Correlation coefficients between all-model SST index EOF1 and AITG indices from single-forcing run.

Correlation coefficients between all-model SST index EOF1 and AITG indices from single-forcing run.
Correlation coefficients between all-model SST index EOF1 and AITG indices from single-forcing run.
Fig. 5.

(a) CCSM3, PCM1, and GISS ModelE ensemble mean AITG index from twentieth-century climate simulations. Averaged AITG index from 7 single-forcing experiment runs: (b) Greenhouse gases; (c) sulfate aerosols; (d) volcanic aerosols; and (e) solar insolation. Shading represents ±1 standard deviation of the ensemble from the mean (units: °C). The AITG indices are defined in Fig. 1. The number of ensemble numbers making up the ensemble means varies: 16 runs for (a); 7 runs for (b)–(e).

Fig. 5.

(a) CCSM3, PCM1, and GISS ModelE ensemble mean AITG index from twentieth-century climate simulations. Averaged AITG index from 7 single-forcing experiment runs: (b) Greenhouse gases; (c) sulfate aerosols; (d) volcanic aerosols; and (e) solar insolation. Shading represents ±1 standard deviation of the ensemble from the mean (units: °C). The AITG indices are defined in Fig. 1. The number of ensemble numbers making up the ensemble means varies: 16 runs for (a); 7 runs for (b)–(e).

Sulfate aerosols cool the climate both by directly reflecting solar radiation and also through indirect effects on cloud reflectivity and lifetime (the first and second indirect aerosol effects, respectively) (Haywood and Boucher 2000). Anthropogenic sources of sulfates, which dominate total sulfate aerosol production over the twentieth century, arise primarily from sulfur dioxide emissions from fossil fuel burning and subsequently undergo chemical reactions to form sulfate aerosols. Because of the shorter lifetime in the atmosphere (3–7 days) (Rasch et al. 2000), atmospheric sulfate aerosol loadings are mostly concentrated in the Northern Hemisphere. When warming from the enhanced greenhouse effect is also considered, the net result is a slower rate of warming in the North as compared to the South. Indeed, both observed and modeled SST regressions onto the model-derived SST gradient index EOF1 (Fig. 3a) show a stronger warming trend in the South Atlantic than that in the North Atlantic (Figs. 6 and 7a). Figure 6 shows the regression of the Hadley SSTs onto EOF 1, which expresses stronger warming in the South Atlantic. The regressions on the other two observed SST datasets are similar (not shown).

Fig. 6.

Regression of Hadley SST onto model-derived EOF 1 of Fig. 3 (units: °C per standard deviation). Since EOF1 of Fig. 3A represents a change of about 0.25 standard deviation over 100 yr, the units of these regression coefficients are roughly equal to 0.25°C (100 yr)−1.

Fig. 6.

Regression of Hadley SST onto model-derived EOF 1 of Fig. 3 (units: °C per standard deviation). Since EOF1 of Fig. 3A represents a change of about 0.25 standard deviation over 100 yr, the units of these regression coefficients are roughly equal to 0.25°C (100 yr)−1.

Fig. 7.

Regression of the multimodel ensemble mean (top) SST and (bottom) precipitation onto EOF1 of Fig. 3 (Fig. 3A): (a),(d) all-model ensemble averages; (b),(e) AIE-model ensemble averages; and (c),(f) no-AIE model ensemble averages. Thick dark-gray contours in (d)–(f) are the multimodel ensemble mean precipitation calculated between 1900 and 1909, indicating the mean rainfall at the beginning of the twentieth century prior to being affected by the twentieth-century forcings (units: °C per standard deviation for (a)–(c); mm day−1 per standard deviation for (d)–(f). Since EOF1 of Fig. 3A represents a change of about 0.25 standard deviations over 100 yr, the units of these regression coefficients are roughly 0.25°C (100 yr)−1 for (a)–(c), and 0.25 mm day−1 (100 yr)−1.

Fig. 7.

Regression of the multimodel ensemble mean (top) SST and (bottom) precipitation onto EOF1 of Fig. 3 (Fig. 3A): (a),(d) all-model ensemble averages; (b),(e) AIE-model ensemble averages; and (c),(f) no-AIE model ensemble averages. Thick dark-gray contours in (d)–(f) are the multimodel ensemble mean precipitation calculated between 1900 and 1909, indicating the mean rainfall at the beginning of the twentieth century prior to being affected by the twentieth-century forcings (units: °C per standard deviation for (a)–(c); mm day−1 per standard deviation for (d)–(f). Since EOF1 of Fig. 3A represents a change of about 0.25 standard deviations over 100 yr, the units of these regression coefficients are roughly 0.25°C (100 yr)−1 for (a)–(c), and 0.25 mm day−1 (100 yr)−1.

The temporal behavior of historical sulfate emissions is consistent with sulfate aerosol forcing of the AITG. Over North America and Europe, sulfate emissions continuously increased until the early 1970s (Fig. 8b), after which they declined (Smith et al. 2001) because of legislation in North America and Europe in the 1960s and 1970’s that mandated reductions in atmospheric pollution.2 These result in a correspondingly large interhemispheric difference in the sulfate aerosol optical depth (which is effectively equivalent to sulfate aerosol concentration) (Fig. 9) and therefore the radiative forcing by sulfate aerosols; if we contrast the sulfate optical depth difference between the North and South Atlantic, the resulting time series also show trend behavior similar to the AITG index (Fig. 8c).3

Fig. 8.

(b) Area-averaged surface sulfate aerosol emissions from the forcing applied to the CCSM3 twentieth-century simulations. (a) Solid- and short-dashed lines indicate the values from north and south boxes, respectively). The long-dashed line is the difference between two boxes (north minus south); north box: 80°W–0°, 20°–60°N; south box: 60°W–20°E, 20°–60°S. (c) CCSM3-simulated sulfate optical depth difference between the two boxes (north minus south); eight simulations are plotted. The simulated optical depth is for the visible band.

Fig. 8.

(b) Area-averaged surface sulfate aerosol emissions from the forcing applied to the CCSM3 twentieth-century simulations. (a) Solid- and short-dashed lines indicate the values from north and south boxes, respectively). The long-dashed line is the difference between two boxes (north minus south); north box: 80°W–0°, 20°–60°N; south box: 60°W–20°E, 20°–60°S. (c) CCSM3-simulated sulfate optical depth difference between the two boxes (north minus south); eight simulations are plotted. The simulated optical depth is for the visible band.

Fig. 9.

Spatial pattern of the trend (1900–82) in sulfate optical depth in the visible band, as simulated by CCSM3 [unit: (100 yr)−1 (optical depth is dimensionless)].

Fig. 9.

Spatial pattern of the trend (1900–82) in sulfate optical depth in the visible band, as simulated by CCSM3 [unit: (100 yr)−1 (optical depth is dimensionless)].

If sulfate aerosols are the cause of the trend in the SST gradient before 1980s, the models that include both aerosol direct and indirect effects should exhibit stronger hemispheric asymmetry in the warming (Haywood and Boucher 2000; Forster et al. 2007; Lohmann and Feichter 2005) and a larger slope in the trend of the AITG index than the models that only include aerosol direct effects do. We divided the models into two subgroups—with and without aerosol indirect effect (AIE models and no-AIE models, respectively)—and repeated the analyses of section 3b on each subgroup. Among the AIE models, the Model for Interdisciplinary Research on Climate 3.2 [MIROC3.2 models [medium- (medres) and high- (hires) resolutions] and Met Office (UKMO) Hadley Centre Global Environmental Model version 1 (HadGEM1) include both first and second indirect effects; L’Institut Pierre-Simon Laplace (IPSL), UKMO Hadley Centre Global Climate Model version 3 (HadCM3), ECHAM5, and ECHAM and the global Hamburg Ocean Primitive Equation (ECHO-G) include only the first indirect effect and GISS Model E-R (GISS E-R) and Model E-H (GISS E-H) only the second indirect effect.

The results of this analysis are shown in Figs. 10 and 11. The mean AITG trend in the AIE models is 0.18°C (100 yr)−1, considerably closer to the observed value of 0.22°C (100 yr)−1 than that of the no-AIE models [0.09°C (100 yr)−1)]. The pattern of the EOF1 from the AIE models also captures better the reversal of the trend in the 1980s (Fig. 10a). The probability of the observed (and higher) value of the AITG trend to occur in the AIE models’ twentieth-century distribution is 39%, significantly higher than that in the no-AIE models (18%). The likelihood of the observed value trend to occur in the preindustrial distribution remains roughly the same, about 7%, in the two subsets. Correspondingly, the significance of the twentieth-century trend relative to model internal variations is more readily apparent in AIE model simulations than that in the no-AIE model simulations (Table 2). These results further support sulfate aerosol forcing as the leading cause of the twentieth century trend in the AITG.

Fig. 10.

As in Fig 3 for results of EOF analyses of the CMIP3 models’ AITG indices, but for (a) the AIE subset of models and (b) no-AIE subset.

Fig. 10.

As in Fig 3 for results of EOF analyses of the CMIP3 models’ AITG indices, but for (a) the AIE subset of models and (b) no-AIE subset.

Fig. 11.

As in Fig. 4, but for (a),(b) AIE subset of models and (c),(d) no-AIE models. Statistics of 83-yr trends of the modeled AITG indices from twentieth-century climate (upper panels) and from preindustrial (lower panels) simulations.

Fig. 11.

As in Fig. 4, but for (a),(b) AIE subset of models and (c),(d) no-AIE models. Statistics of 83-yr trends of the modeled AITG indices from twentieth-century climate (upper panels) and from preindustrial (lower panels) simulations.

Comparing the regression pattern of AIE model mean SST onto EOF1 in Fig. 3, against that of the no-AIE model mean SST (Fig. 7, upper panel), although both subsets show warming in most area, AIE models simulate a stronger disparity in warming across the equator than do the no-AIE models; the strength of simulated warming in the North Atlantic by AIE models is weaker than that by no-AIE models, but the strength of simulated warming in South Atlantic is roughly the same in these two subsets (Figs. 7b and 7c). AIE models also simulate a stronger southward shift of the ITCZ than the no-AIE models, as shown by the regression of the model precipitation onto EOF1 of Fig. 3 (Figs. 7e and 7f).

The magnitude of the aerosol indirect effects is somewhat unconstrained (Kiehl 2007; Anderson et al. 2003); however, we note that our conclusion regarding sulfate forcing of the AITG trend remains robust even if we only consider the no-AIE models (albeit with less significance). The time behavior of EOF1 from the no-AIE models (Fig. 10b) flattens out after 1980s, suggesting that the no-AIE models also respond to the reduction in sulfate aerosol forcing after the 1970s. We note that both CCSM3 and PCM1, whose single-forcing simulations are analyzed in this study, do not include the aerosol indirect effect.

4. Summary and discussion

a. Summary

The observed AITG (defined as the tropical South Atlantic SST minus tropical North Atlantic SST) is shown to have an upward trend throughout most of the twentieth century, with stronger warming in the South Atlantic relative to the North. The trend reverses in the 1980s. The trend is also shown to be distinct from a multidecadal variation in the AITG tied to the well-known Atlantic multidecadal oscillation.

Analysis of the same AITG index in the CMIP3 multimodel twentieth-century simulations show that there is a forced component to the AITG that closely resembles the observed trend, including the reversal in the 1980s. Through examining the distribution of the trend slopes in the CMIP3 twentieth-century and preindustrial models, we conclude that the observed trend in the AITG is unlikely to arise purely from natural variations; we suggest that at around half the observed trend in the AITG is a forced response to twentieth-century climate forcings.

Through analysis of the single-forcing twentieth-century simulations, we conclude that sulfate aerosol forcing is the most probable cause of the multimodel AITG trend. Models that include both aerosol direct and indirect effects simulate stronger AITG trend, while models that contain only aerosol direct effects simulate weaker, but still robust, AITG trend. Repeated analyses on the subgroup of models that include the indirect effect indicate a larger slope in the forced trend component, making the forcing even more distinct from the corresponding preindustrial simulations. However, the contribution of the forcing to the trend in the AITG is significant even in those models with only the direct effect, albeit with a smaller slope.

b. Discussion

Our results provide strong evidence that anthropogenic sulfate aerosols’ emissions over the twentieth century shifted the tropical Atlantic ITCZ southward, confirming previous suggestions from idealized AGCM–slab ocean model studies (Williams et al. 2001; Rotstayn and Lohmann 2002; Rotstayn and Penner 2001). It should be emphasized that those idealized model results by no means guaranteed our result. Our own experience with examining tropical convection responses to Northern Hemisphere cooling in an AGCM–slab ocean model suggested that the tropical ITCZ is particularly sensitive to change in that type of model configuration; also, Kang et al. (2009) showed that the sensitivity of the tropical convection changes to interhemispheric-like forcing depends significantly on cloud–radiative feedbacks, which tend to differ considerably from model to model. Also, it is well known that variations to the Atlantic meridional overturning circulation (AMOC) can affect the tropical Atlantic ITCZ (e.g., Cheng et al. 2007; Chang et al. 2008), and the AMOC in fully coupled models can respond nontrivially to twentieth-century climate forcings (e.g., Delworth and Dixon 2006).

Our suggestion that at least half the observed AITG trend is forced comes from comparing the value of multimodel ensemble mean trend slope [0.13°C (100 yr)−1] to the observed value [0.22°C (100 yr)−1]. However, the influence of aerosols—in particular aerosol indirect effects—is somewhat unconstrained (Kiehl 2007; Anderson et al. 2003), making a precise quantification of the forced contribution difficult. If we only used the AIE models, the forced trend is 0.18°C (100 yr)−1, considerably closer to the observed value; it suggests that, with the inclusion of appropriate physics, the twentieth-century forcing could conceivably account for most of the observed AITG trend. We note that the AIE models tend to have higher climate sensitivity and that, based on the results of Kiehl 2007, in the average sense, AIE models have smaller total anthropogenic forcing and stronger anthropogenic aerosol cooling.

It is intriguing that the forced AITG behavior—as represented by EOF1—closely resembles the trend IMF of the observed AITG, including the reversal of the trend in the 1980s, and that this reversal coincides well with the decrease in Northern Hemisphere sulfate aerosol emissions. On the other hand, it is clear that the model internal variations are also well capable of producing 83-yr-long trends (Fig. 4b). Progress on this problem awaits significant reduction of model uncertainty in its representation of aerosol effects. We note that other types of anthropogenic aerosols that may contribute to interhemispheric asymmetric warming in different ways are in general not as well constrained in the CMIP3 models, such as black carbon aerosols, which can cool the surface while warm the atmosphere (Wang 2004; Ramanathan and Carmichael 2008); these effects are likely not depicted in the current study.

The mechanisms by which sulfate aerosols affect the tropical Atlantic ITCZ remain to be determined. In particular, anthropogenic emissions occur primarily in the northern midlatitudes, yet they are able to affect the climate of the deep tropics. Sulfate aerosols advected into the northern tropics may directly cool North Atlantic SST. Alternatively, previous modeling studies have shown that cooling sourced from the mid-to-high Northern Hemisphere climate is able to progress into the northern tropics through atmospheric teleconnection mechanisms (Chiang and Bitz 2005; Broccoli et al. 2006, Kang et al. 2009), bringing about change to the interhemispheric SST gradient and ITCZ shift.

Another possibility is that anthropogenic sulfate aerosols may indirectly influence the tropical Atlantic climate through affecting the strength of the AMOC. We have examined the temporal behavior of northward heat transport by the Atlantic Ocean at the equator (as a measure of the AMOC) in the CMIP3 models; while we do see a trend to the poleward heat transport, it does not show the reversal around the 1980s (not shown). It is beyond the scope of this study to investigate details of the potential teleconnnection mechanisms; regardless, ultimately the secular trend of the AITG over the twentieth century can be sourced to anthropogenic sulfate emissions.

Variations in the AMO have been implicated in changes to the tropical Atlantic hurricane, ITCZ, and Sahel rainfall (Mann and Emanuel 2006; Zhang and Delworth 2006; Knight et al. 2006). The magnitude of the AITG trend (Fig. 1d) is comparable to the amplitude of the AMO (Fig. 1c), suggesting that the tropical Atlantic climate impacts due to the AITG trend (and, by implication, sulfate emissions) are comparable to impacts of the AMO. There has been some speculation—especially in the hurricanes and climate change literature (Mann and Emanuel 2006)—as to whether the recent (post-1980) upturn in North Atlantic SST has been due to natural or anthropogenic causes. Insofar as our results on the Atlantic interhemispheric SST gradient are applicable to this problem, they suggest that the reduction to the anthropogenic sulfate emissions post-1970s may have contributed to this upturn.

In future climate scenarios, anthropogenic greenhouse gas emissions and sulfate emissions follow very different trajectories: while greenhouse gas emissions are projected to increase monotonically, sulfate emissions over North America and Europe have decreased since the late 1970s and are projected to decrease even further in the twenty-first century (Zhao et al. 2008; Stern 2006). Moreover, increased industrial activity in developing countries—mostly located in the tropics and Southern Hemisphere (Zhao et al. 2008; Stern 2006)—may lead to increased sulfate emissions there. Combined, these lead to climate forcings continuing, into the future, the reversal in the trend of the tropical Atlantic Interhemispheric SST gradient that started in the early 1980s.

Acknowledgments

We acknowledge the modeling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) and the WCRP’s Working Group on Coupled Modelling (WGCM) for their roles in making available the WCRP CMIP3 multimodel dataset. Support of this dataset is provided by the Office of Science, U.S. Department of Energy. We thank Yochanan Kushnir, Ping Chang, Alessandra Giannini, Michela Biasutti, and Mike Wallace for useful comments and suggestions on a previous version of this manuscript. We thank Zhaohua Wu for providing the code of EEMD analysis, and Ken Lo for help accessing the GISS ModelE single forcing runs, and Mingfang Ting for the use of her AMO index. This research is supported by the Office of Science (BER), U.S. Department of Energy (Award No. DE-FG02-08ER64588 to J. Chiang).

APPENDIX

EEMD Analysis

We describe in more detail the EEMD analysis on the AITG indices.

a. Intrinsic mode functions

There are seven EEMD IMFs and a residual mode derived from EEMD analysis. The number of IMFs is log2N, where N is the number of data points. All IMFs from all three observed AITG indices are listed in Fig. A1. IMFs 1, 2, and 3 are higher-frequency signals, 4 is the multidecadal variation, and 5 is the secular trend; 6 and 7 and the residual are low-frequency signals. Magnitudes of IMF 4 and IMF 5 are comparable and are much larger than the other IMFs. Note that the residual mode is a linear trend, behaves distinctively different from the mode 5, which goes upward in most part of the twentieth century but turn around at early 80s. Panel I shows the sum of the IMFs 5, 6, 7, and residual. The combination of the higher modes is very similar to the IMF 5 alone (because IMF 5’s magnitude is much larger than the others). All the higher modes with lower frequency than the multidecadal signal possibly are responses to external forcings, and the sum of these modes may more truly account for forced responses. Nevertheless, when all the external forced responses are considered, the upward trend behavior and the turn-around in the 1980s remain intact.

Fig. A1.

All IMFs from the EEMD analysis of the AITG indices (solid line: Hadley SST, long dash line: ERSST, and dot-dash line: KAPLAN extended SST v2; units: °C).

Fig. A1.

All IMFs from the EEMD analysis of the AITG indices (solid line: Hadley SST, long dash line: ERSST, and dot-dash line: KAPLAN extended SST v2; units: °C).

b. Sensitivity tests

EEMD analysis may exhibit sensitivity to the choice of end points (Wu and Huang 2009). To test if the EEMD modes are robust under different end points, we repeat the EEMD analysis on to the AITG indices (results from Hadley SST AITG are shown as example) but vary the choice of the end year. Fig. A2 shows IMFs 4 and 5 from this sensitivity test. The variation in the IMF results is slight; thus, our EEMD results are robust.

We also tested the sensitivity of our results to the number of ensemble calculations in the EEMD analysis, varying the ensemble size (we tried 50, 60, 90, and 120). The change to the ensemble size makes little difference in the calculated IMFs.

Fig. A2.

IMFs 4 and 5 from sensitivity tests, varying the end year of the data. Only results from analysis of the Hadley AITG index are displayed.

Fig. A2.

IMFs 4 and 5 from sensitivity tests, varying the end year of the data. Only results from analysis of the Hadley AITG index are displayed.

REFERENCES

REFERENCES
Anderson
,
T.
, and
Coauthors
,
2003
:
Climate forcing by aerosols—A hazy picture
.
Science
,
300
,
1103
1104
.
Barnett
,
T. P.
,
1977
:
Principal time and space scales of Pacific trade wind fields
.
J. Atmos. Sci.
,
34
,
221
236
.
Barnett
,
T. P.
, and
Coauthors
,
2005
:
Detecting and attributing external influences on the climate system: A review of recent advances
.
J. Climate
,
18
,
1291
1314
.
Biasutti
,
M.
, and
A.
Giannini
,
2006
:
Robust Sahel drying in response to late 20th century forcings
.
Geophys. Res. Lett.
,
33
,
L11706
,
doi:10.1029/2006GL026067
.
Broccoli
,
A. J.
,
K. A.
Dahl
, and
R. J.
Stouffer
,
2006
:
Response of the ITCZ to Northern Hemisphere cooling
.
Geophys. Res. Lett.
,
33
,
L01702
,
doi:10.1029/2005GL024546
.
Chang
,
P.
, and
Coauthors
,
2008
:
Oceanic link between abrupt changes in the North Atlantic Ocean and the African monsoon
.
Nat. Geosci.
,
1
,
444
448
.
Cheng
,
W.
,
J. C. H.
Chiang
, and
C. M.
Bitz
,
2007
:
Adjustment of the global climate to an abrupt slowdown of the Atlantic meridional overturning circulation
.
Ocean Circulation: Mechanisms and Impacts, Geophys. Monogr., Vol. 173, Amer. Geophys. Union, 295–314
.
Chiang
,
J. C. H.
, and
C.
Bitz
,
2005
:
Influence of high latitude ice cover on the marine intertropical convergence zone
.
Climate Dyn.
,
25
,
477
496
.
Chiang
,
J. C. H.
,
Y.
Kushnir
, and
S. E.
Zebiak
,
2000
:
Interdecadal changes in eastern Pacific ITCZ variability and its influence on the Atlantic ITCZ
.
Geophys. Res. Lett.
,
27
,
3687
3690
.
Chiang
,
J. C. H.
,
Y.
Kushnir
, and
A.
Giannini
,
2002
:
Deconstructing Atlantic ITCZ variability: Influence of the local cross-equatorial SST gradient, and remote forcing from the eastern equatorial Pacific
.
J. Geophys. Res.
,
107
,
4004
,
doi:10.1029/2000JD000307
.
Delworth
,
T. L.
, and
M.
Mann
,
2000
:
Observed and simulated multidecadal variability in the Northern Hemisphere
.
Climate Dyn.
,
16
,
661
676
.
Delworth
,
T. L.
, and
K. W.
Dixon
,
2006
:
Have anthropogenic aerosols delayed a greenhouse gas-induced weakening of the North Atlantic thermohaline circulation?
Geophys. Res. Lett.
,
33
,
L02606
,
doi:10.1029/2005GL024980
.
Evan
,
A. T.
,
D. J.
Vimont
,
A. K.
Heidinger
,
J. P.
Kossin
, and
R.
Bennartz
,
2009
:
The role of aerosols in the evolution of tropical North Atlantic Ocean temperature anomalies
.
Science
,
324
,
778
781
.
Forster
,
P.
, and
Coauthors
,
2007
:
Changes in atmospheric constituents and in radiative forcing
.
Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 129–234, 755
.
Giannini
,
A.
,
R.
Saravanan
, and
P.
Chang
,
2003
:
Oceanic forcing of Sahel rainfall on interannual to interdecadal time scales
.
Science
,
302
,
1027
1030
.
Hastenrath
,
S.
, and
L.
Greischar
,
1993
:
Circulation mechanisms related to northeast Brazil rainfall anomalies
.
J. Geophys. Res.
,
98
,
5093
5102
.
Haywood
,
J.
, and
O.
Boucher
,
2000
:
Estimates of the direct and indirect radiative forcing due to tropospheric aerosols: A review
.
Rev. Geophys.
,
38
,
513
543
.
Huang
,
N.
, and
Z.
Wu
,
2008
:
A review on Hilbert–Huang transform: Method and its applications to geophysical studies
.
Rev. Geophys.
,
46
,
RG2006
,
doi:10.1029/2007RG000228
.
Kang
,
S. M.
,
D. M. W.
Frierson
, and
I. M.
Held
,
2009
:
The tropical response to extratropical thermal forcing in an idealized GCM: The importance of radiative feedbacks and convective parameterization
.
J. Atmos. Sci.
,
66
,
2812
2827
.
Kaplan
,
A.
, and
Coauthors
,
1998
:
Analyses of global sea surface temperature 1856–1991
.
J. Geophys. Res.
,
103
,
18 567
18 589
.
Kiehl
,
J.
,
2007
:
Twentieth-century climate model response and climate sensitivity
.
Geophys. Res. Lett.
,
34
,
L22710
,
doi:10.1029/2007GL031383
.
Knight
,
J. R.
,
C. K.
Folland
, and
A. A.
Scaife
,
2006
:
Climate impacts of the Atlantic multidecadal oscillation
.
Geophys. Res. Lett.
,
33
,
L17706
,
doi:10.1029/2006GL026242
.
Kushnir
,
Y.
,
1994
:
Interdecadal variations in North Atlantic sea surface temperature and associated atmospheric conditions
.
J. Climate
,
7
,
141
157
.
Lohmann
,
U.
, and
J.
Feichter
,
2005
:
Global indirect aerosol effects: A review
.
Atmos. Chem. Phys.
,
5
,
715
737
.
Mann
,
M. E.
, and
K. A.
Emanuel
,
2006
:
Atlantic hurricane trends linked to climate change
.
Eos, Trans. Amer. Geophys. Union
,
87
,
233
.
Meehl
,
G. A.
,
C.
Covey
,
T.
Delworth
,
M.
Latif
,
B.
McAvaney
,
J. F. B.
Mitchell
,
R. J.
Stouffer
, and
K. E.
Taylor
,
2007
:
The WCRP CMIP3 multi-model dataset: A new era in climate change research
.
Bull. Amer. Meteor. Soc.
,
88
,
1383
1394
.
Ming
,
Y.
, and
V.
Ramaswamy
,
2009
:
Nonlinear climate and hydrological responses to aerosol effects
.
J. Climate
,
22
,
1329
1339
.
Moura
,
A. D.
, and
J.
Shukla
,
1981
:
On the dynamics of droughts in northeast Brazil—Observations, theory, and numerical experiments with a general circulation model
.
J. Atmos. Sci.
,
38
,
2653
2675
.
Nicholson
,
S.
,
1993
:
An overview of African rainfall fluctuations of the last decade
.
J. Climate
,
6
,
1463
1466
.
Ramanathan
,
V.
, and
G.
Carmichael
,
2008
:
Global and regional climate changes due to black carbon
.
Nat. Geosci.
,
1
,
221
227
.
Rasch
,
P.
,
M.
Barth
,
J.
Kiehl
,
S.
Schwartz
, and
C.
Benkovitz
,
2000
:
A description of the global sulfur cycle and its controlling processes in the National Center for Atmospheric Research Community Climate Model, version 3
.
J. Geophys. Res.
,
105
,
1367
1385
.
Rayner
,
N. A.
, and
Coauthors
,
2003
:
Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century
.
J. Geophys. Res.
,
108
,
4407
,
doi:10.1029/2002jd002670
.
Rotstayn
,
L. D.
, and
J. E.
Penner
,
2001
:
Indirect aerosol forcing, quasi forcing, and climate response
.
J. Climate
,
14
,
2960
2975
.
Rotstayn
,
L. D.
, and
U.
Lohmann
,
2002
:
Tropical rainfall trends and the indirect aerosol effect
.
J. Climate
,
15
,
2103
2116
.
Schlesinger
,
M. E.
, and
N.
Ramankutty
,
1994
:
An oscillation in the global climate system of period 65–70 years
.
Nature
,
367
,
723
726
.
Slutz
,
R. J.
,
S. J.
Lubker
,
J. D.
Hiscox
,
S. D.
Woodruff
,
R. L.
Jenne
,
D. H.
Joseph
,
P. M.
Steurer
, and
J. D.
Elms
,
1985
:
Comprehensive Ocean-Atmosphere Data Set; Release 1
.
NOAA Environmental Research Laboratories, Climate Research Program, 268 pp
.
Smith
,
S. J.
,
H.
Pitcher
, and
T. M. L.
Wigley
,
2001
:
Global and regional anthropogenic sulfur dioxide emissions
.
Global Planet. Change
,
29
,
99
119
.
Smith
,
T. M.
,
R. W.
Reynolds
,
T. C.
Peterson
, and
J.
Lawrimore
,
2008
:
Improvements to NOAA’s historical merged land-ocean surface temperature analysis (1880–2006)
.
J. Climate
,
21
,
2283
2296
.
Stern
,
D. I.
,
2006
:
Reversal of the trend in global anthropogenic sulfur emissions
.
Global Environ. Change
,
16
,
207
220
.
Ting
,
M. F.
,
Y.
Kushnir
,
R.
Seager
, and
C. H.
Li
,
2009
:
Forced and internal twentieth-century SST trends in the North Atlantic
.
J. Climate
,
22
,
1469
1481
.
Wang
,
C.
,
2004
:
A modeling study on the climate impacts of black carbon aerosols
.
J. Geophys. Res.
,
109
,
D03106
,
doi:10.1029/2003JD004084
.
Williams
,
K. D.
,
A.
Jones
,
D. L.
Roberts
,
C. A.
Senior
, and
M. J.
Woodage
,
2001
:
The response of the climate system to the indirect effects of anthropogenic sulfate aerosol
.
Climate Dyn.
,
17
,
845
856
.
Wu
,
Z.
, and
N. E.
Huang
,
2009
:
Ensemble Empirical Mode Decomposition: A noise-assisted data analysis method
.
Adv. Adapt. Data Anal.
,
1
,
1
41
.
Wu
,
Z.
,
E. K.
Schneider
,
B. P.
Kirtman
,
E. S.
Sarachik
,
N. E.
Huang
, and
C. J.
Tucker
,
2008
:
The modulated annual cycle: An alternative reference frame for climate anomalies
.
Climate Dyn.
,
31
,
823
841
.
Xie
,
S.-P.
, and
J. A.
Carton
,
2004
:
Tropical Atlantic variability: Patterns, mechanisms, and impacts
.
Earth Climate: The Ocean–Atmosphere Interaction, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 121–142
.
Zhang
,
R.
, and
T. L.
Delworth
,
2006
:
Impact of Atlantic multidecadal oscillations on India/Sahel rainfall and Atlantic hurricanes
.
Geophys. Res. Lett.
,
33
,
L17712
,
doi:10.1029/2006GL026267
.
Zhao
,
T. X. P.
,
I.
Laszlo
,
W.
Guo
,
A.
Heidinger
,
C.
Cao
,
A.
Jelenak
,
D.
Tarpley
, and
J.
Sullivan
,
2008
:
Study of long-term trend in aerosol optical thickness observed from operational AVHRR satellite instrument
.
J. Geophys. Res.
,
113
,
D07201
,
doi:10.1029/2007JD009061
.

Footnotes

1

A 21-yr running mean filter was applied on the AMO index derived by Ting et al. 2009 since the AITG indices in the current study are derived from 21-yr running mean filtered data.

2

The only sulfate aerosol emission forcing data available to us is that of the CCSM3. There might be differences among models regarding their reconstruction of the sulfate aerosol emission, but they in general have a qualitatively similar temporal evolution since they often base the temporal histories on similar emission data (Forster et al. 2007).

3

The sulfate aerosol optical depth is from the output of CCSM3 twentieth century simulations.