Most studies of African dust and North Atlantic climate have been limited to the short time period since the satellite era (1980 onward), precluding the examination of their relationship on longer time scales. Here a new dust dataset with the record extending back to the 1950s is used to show a multidecadal covariability of North Atlantic SST and aerosol, Sahel rainfall, and Atlantic hurricanes. When the North Atlantic Ocean was cold from the late 1960s to the early 1990s, the Sahel received less rainfall and the tropical North Atlantic experienced a high concentration of dust. The opposite was true when the North Atlantic Ocean was warm before the late 1960s and after the early 1990s. This suggests a novel mechanism for North Atlantic SST variability—a positive feedback between North Atlantic SST, African dust, and Sahel rainfall on multidecadal time scales. That is, a warm (cold) North Atlantic Ocean produces a wet (dry) condition in the Sahel and thus leads to low (high) concentration of dust in the tropical North Atlantic, which in turn warms (cools) the North Atlantic Ocean. An implication of this study is that coupled climate models need to be able to simulate this aerosol-related feedback in order to correctly simulate climate variability in the North Atlantic. Additionally, it is found that dust in the tropical North Atlantic varies inversely with the number of Atlantic hurricanes on multidecadal time scales because of the multidecadal variability of both direct and indirect influences of dust on vertical wind shear in the hurricane main development region.
Sea surface temperature (SST) variability of the North Atlantic Ocean on time scales longer than interannual mainly shows a multidecadal variation and a secular trend (e.g., Trenberth and Shea 2006). There is a debate on what causes these variations—that is, the relative roles of anthropogenic global warming, aerosols, and natural variability in controlling North Atlantic SST variability (e.g., Enfield et al. 2001; Knight et al. 2005; Trenberth and Shea 2006; Mann and Emanuel 2006; Zhang et al. 2007; Ting et al. 2009; Ottera et al. 2010; Wang and Dong 2010; Chang et al. 2011). In this paper, we focus on the multidecadal variability of the North Atlantic SST—the Atlantic multidecadal oscillation (AMO) (e.g., Delworth and Mann 2000; Enfield et al. 2001). Here the AMO index is defined as the detrended North Atlantic SST anomalies over the region of 0°–60°N and from the east coast of the Americas to 0° longitude (Fig. 1a). The extended reconstructed SST data (Smith et al. 2008) from 1950 to 2008 show that the AMO was in the cold phase from the late 1960s to the early 1990s and in the warm phases before the late 1960s and after the early 1990s.
The AMO has footprints on many parts of the global oceans (Fig. 1b; Ottera et al. 2010, and others). In addition to the basin-wide warming of the North Atlantic Ocean, the warm phase of the AMO is associated with warming in the western North and South Pacific Oceans and the tropical eastern Indian Oceans and cooling over the global Southern Oceans and in the subtropical northeastern Pacific Ocean. Opposite-signed SST anomalies occur for the cold phase of the AMO. The Atlantic SST regression shows a bipolar seesaw pattern or an interhemispheric gradient (Stocker 1998), supporting the hypothesis that a driving mechanism of the AMO may involve fluctuations of the Atlantic meridional overturning circulation (AMOC) (e.g., Delworth and Mann 2000; Knight et al. 2005; Delworth et al. 2007). As the AMOC is enhanced, a warming and a cooling will occur in the North and South Atlantic Oceans, respectively, and vice versa for a reduction of the AMOC.
AMO variability is associated with changes of climate and extreme events such as drought, flood, and hurricane activity (e.g., Enfield et al. 2001; Goldenberg et al. 2001; McCabe et al. 2004; Bell and Chelliah 2006; Wang and Lee 2009). Previous studies have shown that it is mainly the tropical portion of the AMO that is relevant for climate since the climate response to the North Atlantic SST anomalies is primarily forced at the low latitudes (e.g., Lu and Delworth 2005; Sutton and Hodson 2007). Accordingly, the influence of the AMO on climate and hurricane activity may operate via atmospheric circulation changes induced by the SST anomalies in the tropical North Atlantic and Caribbean Sea (Vimont and Kossin 2007; Wang et al. 2008). Recently, Dunstone et al. (2011) used the third climate configuration of the Met Office Hadley Centre Coupled Model version 3 (HadCM3) coupled climate model to show that the North Atlantic subpolar gyre is a potentially important region for driving the tropical atmosphere by changing the tropical SST. It is therefore important to explore and understand the mechanisms that drive multidecadal variability in the tropical North Atlantic SST.
The present paper focuses on the relationships of mineral dust concentration in the tropical North Atlantic with North Atlantic SST variability and Atlantic hurricanes. Owing to the lack of long-term observations of dust over the Atlantic, most observational studies regarding mineral aerosols and North Atlantic climate have been limited to the short period since the satellite era (approximately 1980 onward). Although several modeling studies have explored the relationship between dust and climate on long-term periods (e.g., Mahowald et al. 2010), there is a high degree of disagreement in modeled dust emissions and transport between such models (Huneeus et al. 2011), underscoring the importance of observation-based studies. In this paper, we use a new dust dataset, which extends back to the 1950s, to examine dust-related North Atlantic SST variability on multidecadal time scales and propose a novel mechanism for North Atlantic SST variability. The paper is organized as follows. Section 2 briefly describes and introduces the datasets used in this study. Section 3 shows a multidecadal covariability of SST and dust in the North Atlantic, rainfall in the Sahel, and Atlantic hurricanes. Section 4 presents the linkage between dust in the tropical North Atlantic and African dust. Section 5 infers a positive feedback between North Atlantic SST, African dust, and Sahel rainfall on multidecadal time scales, and gives a discussion and summary.
Several observational and reanalysis datasets are used in this study. SST is from the National Oceanic and Atmospheric Administration (NOAA) extended reconstructed SST version 3 (Smith et al. 2008). The atmospheric reanalysis datasets include the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996), the European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40; Uppala et al. 2005), and the newly developed NOAA Earth System Research Laboratory (ESRL) Twentieth-Century Reanalysis (20CR) (Compo et al. 2011). All of these datasets are monthly and we use them for the period from January 1950 to December 2008 except for the ERA-40. For the ERA-40 dataset, we use the data from January 1958 to December 2001 since the original ERA-40 reanalysis is from September 1957 to August 2002. We also use the station wind data in the western Sahel from Mahowald et al. (2007).
The rainfall dataset is from the Global Precipitation Climatology Centre (GPCC) version 5. The monthly rainfall data are gridded from the complete GPCC station database with more than 70 000 different rain gauge stations worldwide (Rudolf et al. 1994). The gridded land precipitation is available for the period 1901–2009 in spatial resolutions of 0.5° × 0.5°, 1.0° × 1.0°, and 2.5° × 2.5°. The rainfall data are available from ftp://ftp-anon.dwd.de/pub/data/gpcc/html/fulldata_download.html.
The extended dust aerosol optical depth (DAOD) over the tropical North Atlantic from 1955 to 2008 was derived from a simple model using modern and historical data from meteorological satellites and a proxy record for atmospheric dust (Evan and Mukhopadhyay 2010). Satellite data include aerosol optical thickness retrievals from the Advanced Very High Resolution Radiometer (AVHRR) Pathfinder Extended dataset (PATMOS-x) for the period 1982–2008 and the Moderate Resolution Imaging Spectroradiometer (MODIS) on board the Aqua satellite for the period 2002–08. The proxy record for atmospheric dust is based on measurements of crustal helium-4 (4He) flux from a Porites coral at a water depth of 5 m near Pedra de Lume on Sal Island (16°45′44′′N, 22°53′23′′W), part of the Cape Verde archipelago, for the period of 1955–94 (Mukhopadhyay and Kreycik 2008). The resultant 54-yr record of DAOD has a 1.0° × 1.0° spatial resolution and a monthly temporal resolution, and covers the region 0°–30°N, 10°–65°W. Evan et al. (2011) showed that the DAOD time series is well correlated with the station dust time series observed on the island of Barbados (13°10′N, 59°30′W). Since the DAOD dataset is longer than the dust station data in Barbados, we use the DAOD data to study multidecadal variation in this paper.
Tropical cyclone data are obtained from the National Hurricane Center (NHC) Hurricane Best Track Files (HURDAT) (http://www.nhc.noaa.gov/pastall.shtml) (Jarvinen et al. 1984). Here we examine the numbers of Atlantic hurricanes, defined as the number of tropical cyclones with lifetime maximum sustained 1-min wind speeds of 64 kt or greater (Saffir–Simpson categories 1–5).
3. Covariability of SST, dust, rainfall, and hurricanes
Variations in Atlantic SST and its interhemispheric gradient have been recognized to contribute to drought after the 1970s in the Sahel, which is a region in Africa between the Saharan Desert and the rain forests of central Africa and the Guinean coast (e.g., Folland et al. 1986; Giannini et al. 2003). Figures 2a,b show that the Sahel rainfall anomalies covary with the AMO on multidecadal time scales. The warm (cold) phases of the AMO are associated with positive (negative) rainfall anomalies in the Sahel. Because both the SST and rainfall data extend back to the beginning of the twentieth century, we can examine the relationship over a long-term period. Figure 3 shows that the relationship between the AMO and Sahel rainfall still holds if the time series are extended for the entire twentieth century.
More interestingly, the DAOD in the tropical North Atlantic, which is proportional to mass concentration and is a proxy for atmospheric dust content, is out of phase with the AMO and the Sahel rainfall anomalies (Fig. 2c). This relationship is consistent with the analysis of Foltz and McPhaden (2008b), which used the data from 1980–2006 to show trends of Sahel rainfall and dust during the most recent upswing of the AMO. The nearly six-decade data used here show that when the North Atlantic Ocean was cold during the late 1960s to the early 1990s, the Sahel received less rainfall and the tropical North Atlantic had high concentration of dust. The opposite was true for the warm phases of the AMO after the early 1990s and before the late 1960s. We smooth each time series with a 7-yr running mean filter (the shading in Fig. 2) and calculate the correlations among these filtered indices. The maximum correlations of −0.87 and 0.85 occur when the Sahel rainfall leads the DAOD and AMO indices by 1 and 2 yr, respectively, which is consistent with the finding of Prospero and Lamb (2003) in the case of Barbados dust and Sahel rainfall. The maximum correlation between the AMO and DAOD indices is −0.67 with 0 lag. We calculate the effective degree of freedom (Quenouille 1952; Medhaug and Furevik 2011) as , where N is the number of data points for the time series of X and Y, or is the autocorrelation at lag 1, and or is the autocorrelation at lag 2. All of these correlations are statistically significant based on Student’s t test (above 95% significant levels). Although a correlation cannot tell us the cause and effect, these high correlation values suggest that dust in the tropical North Atlantic is related to Sahel rainfall and the North Atlantic SST (Prospero and Lamb 2003; Evan et al. 2011).
To emphasize AMO-related variability, we have detrended all data (i.e., removed linear trends). We also plot the data with the linear trends included—that is, data that include the components of both global warming and the AMO (Figs. 4 and 5). A comparison of Fig. 1b and Fig. 4b indicates that the global warming signal is spread throughout most of the global oceans (Barnett et al. 2005). With the long-term linear trend included, the rainfall anomalies in the Sahel show negative anomalies during the warm phase of the AMO after the early 1990s (Fig. 5b). If we compare Fig. 5b with Fig. 2b, it suggests that the impacts of the warming trend and the AMO on rainfall in the Sahel are opposite in sign. That is, the warming trend is associated with a reduction of rainfall in the Sahel region, whereas the warm phase of the AMO after the early 1990s tends to increase Sahel rainfall. This is consistent with modeling results that show that global warming leads to drought over the Sahel and a positive AMO phase enhances Sahel rainfall (e.g., Held et al. 2005; Mohino et al. 2011). The long-term drying in the Sahel may also be caused by other factors such as sulfate aerosol forcing that may lead to the cooling of the northern tropical oceans and then a decrease in Sahel rainfall (e.g., Rotstayn and Lohmann 2002; Chang et al. 2011). The DAOD anomalies in Fig. 5c are almost identical to those in Fig. 2c, indicating that the DAOD data used here do not have a significant long-term linear trend.
The spatial rainfall patterns related to the AMO and dust aerosol in the tropical North Atlantic are shown in Fig. 6. A strong positive AMO–rainfall regression stands out in the Sahel region (Fig. 6a). This indicates that on multidecadal time scales the warm (cold) North Atlantic Ocean corresponds to more (less) rainfall in the Sahel as the positive AMO phase leads to a northward shift of the intertropical convergence zone (ITCZ) in the Atlantic (e.g., Zhang and Delworth 2006; Krebs and Timmermann 2007). Since the northward shift of the Atlantic ITCZ is associated with the anomalous moisture convergence in the lower troposphere and upward motion at 500 hPa (Fig. 6c) over the Sahel, Sahel rainfall is enhanced. Figure 6a also shows the significant rainfall patterns associated with the AMO from North America to equatorial South America, consistent with previous studies (e.g., Enfield et al. 2001; McCabe et al. 2004). The dust–rainfall regression shows that high (low) concentration of dust in the tropical North Atlantic is associated with less (more) rainfall in the Sahel on multidecadal time scales (Fig. 6b). For the data with the linear trend included, the positive AMO–rainfall regression in the Sahel region almost disappears (Fig. 7a) because of the offsetting effect of global warming. The rainfall pattern from the GPCC product is consistent with the independent NCEP–NCAR reanalysis that shows the disappearance of upward motion at 500 hPa in the Sahel (Fig. 7b). Thus, our analyses here suggest that global warming decreases rainfall in the Sahel, whereas the warm phase of the AMO after the early 1990s increases rainfall in the Sahel.
An interesting topic is the relationship between African dust and Atlantic hurricane activity. Evan et al. (2006) have demonstrated a strongly negative relation between interannual variations in Atlantic tropical cyclone days and atmospheric dust measured by satellite during the years of 1982–2005. Here we focus on the multidecadal variations by performing the 7-yr running mean of all indices (the indices are detrended before the 7-yr running mean). Figure 8a shows that dust in the tropical North Atlantic is highly related to the number of Atlantic hurricanes. When dust concentration in the tropical North Atlantic is high (low), the number of Atlantic hurricanes is smaller (larger). The mechanism may be due to an increase (decrease) of tropospheric vertical wind shear (VWS) in the hurricane main development region induced by high (low) concentration of dust in the tropical North Atlantic (Fig. 8b). The spatial pattern of vertical wind shear related to dust aerosol in the tropical North Atlantic is shown in Fig. 8c. The positive wind shear–dust regression appears in the tropical Atlantic south of 20°N and the Sahel region, indicating that high (low) concentration of dust is associated with large (small) vertical wind shear, which is unfavorable (favorable) for hurricane development. Dust variability changes the meridional air temperature gradient via dust–radiation processes (e.g., the dust-related shortwave and longwave radiative heating), which, in turn, alters the strength of the zonal winds or the easterly jet (see next section) through the thermal wind balance and thus the vertical wind shear. Indeed, Chen et al. (2010) used a numerical model to show that dust–radiation processes do change vertical profiles of winds and temperatures.
The vertical wind shear variability may also be influenced by indirect effect of dust on atmospheric circulation. Dust concentration can affect other atmospheric and oceanic processes and then vertical wind shear and hurricanes, such as Sahel rainfall (Landsea and Gray 1992), the Saharan air layer (Dunion and Velden 2004), and North Atlantic SST (Goldenberg et al. 2001; Evan et al. 2008). The vertical wind shear during the summer and fall in the hurricane main development region is also related to the Atlantic warm pool (Wang et al. 2006) and the Atlantic meridional mode (Vimont and Kossin 2007). Thus, it is possible that the coupled response of winds and SST to dust radiative forcing may act to reinforce the wind shear anomalies via mechanisms described by Evan et al. (2011). All of these processes are not independent and can contribute to the changes of vertical wind shear. Further studies in this topic are needed.
4. African dust and dust in the tropical North Atlantic
Previous studies have suggested that Atlantic dust storm frequency and intensity changes are related to or are likely to be partially forced by variability in Sahelian precipitation. For example, Prospero and Lamb (2003) used the data from 1965 to 1998 to show that dust concentration observed on the island of Barbados in the western tropical North Atlantic was highly and negatively correlated with prior-year summer rainfall in the Sahel region. Our analyses of Fig. 2 show that the Sahel rainfall anomalies are inversely related to dust in the tropical North Atlantic on multidecadal time scales. The question is: What is the mechanism for the negative relationship between Sahel rainfall and dust in the tropical North Atlantic on multidecadal time scales?
One possibility is that anomalously high precipitation causes a “greening” of Africa (Nicholson et al. 1998) and shrinking of dust source regions, increase in soil moisture, and decrease in surface shear stress acting on soils where vegetation is present (Marticorena et al. 1997). It is also possible that surface winds play an important role in dust production in the Sahel region. To examine the role of surface winds, we identify the high and low dust years by the top and bottom quartiles of the DAOD time series, respectively. We then calculate the composite wind difference between the high and low dust years. The composite wind map from the NCEP–NCAR reanalysis shows an increased surface wind speed across the Sahel, centered near 15°N, during periods of high minus low DAOD in the tropical North Atlantic (Fig. 9a). An increase in surface wind speed over dust source regions such as the Bodélé Depression—the lowest point in Chad and the planet’s largest single source of dust (Washington et al. 2006)—would lead to increased atmospheric dust loading since deflation is proportional to surface shear stress (and surface shear stress is proportional to the cube of the wind speed) (Marticorena et al. 1997). Given the cubic sensitivity of dust mobilization to wind speed, a small change of winds can cause large change in dust output from the Bodélé (Washington et al. 2009) and other source regions. To confirm the results, we repeat the calculations by using the ERA-40 reanalysis and the 20CR reanalysis (Figs. 9b,c). The band of highest wind speed is shifted about 5° north for the ERA-40 and 20CR, but overall the patterns are similar: an increased surface wind speed across the Sahel and thus more dust production in the Bodélé Depression.
Previous studies have suggested that reanalysis product such as the NCEP–NCAR reanalysis might have large uncertainty in many variables before the satellite era (e.g., Trenberth and Guillemot 1998; Mahowald et al. 2002, 2007). Here we repeat the calculations of Fig. 9 using only the 1980–2008 period, when satellite data are available. The wind distribution pattern is similar to Fig. 9, except that the magnitude of wind speed is larger (not shown). In addition, we also analyze the observed station data in the Western Sahel of 13°–22°N, 20°W–15°E provided by Mahowald et al. (2007). Figure 10 shows the time series of the annual average of the wind speed cube from the station data in the Western Sahel. The cube of the wind speed also shows the AMO signal. When the AMO was in the cold phase from the 1970s to the early 1990s (as shown earlier, the Sahel received less rainfall and the tropical North Atlantic experienced high concentration of dust), the wind speed cube had larger values. The opposite was true when the AMO was in the warm phases before the 1960s and after the early 1990s. The correlation between the yearly time series of the wind speed cube (Fig. 10) and of dust in the tropical North Atlantic (Fig. 2c) is 0.51 (above the 95% significance level). If we choose high and low dust years used in Fig. 9, Fig. 10 shows that the wind speed cube during high dust years is 22.98 m3 s−3 larger than during low dust years. All of these results are consistent and suggest that an increased wind speed can produce more dust in the dust source region of the Bodélé Depression.
The dust in the Bodélé Depression can be transported southwestward at low levels, and then vertical mixing and vertical motion can allow the easterly flows to transport it into the Atlantic. Washington et al. (2009) computed the transport pathways from the Bodélé Depression using a Lagrangian advection model driven by the 3D ERA-40 reanalysis winds. They found that the path of the trajectories in the peak dust-production months could reach the Atlantic in 5 days. It has been shown previously that transport of Saharan dust westward from the African coast is most strongly correlated with winds at about 700 hPa (Kaufman et al. 2005b). We therefore use the NCEP–NCAR reanalysis to calculate the regression of wind at 700 hPa onto the DAOD time series (Fig. 11a). The regressed wind at 700 hPa shows that the maximum easterly wind anomalies are south of the maximum dust band between 10° and 20°N. The inconsistency in the position of the maximum easterly wind anomalies and the maximum dust band suggests that the dust changes in the tropical North Atlantic could be more due to enhanced dust production in the Sahel and Saharan regions, as evidenced by the decrease in precipitation (Fig. 6b) and the increase in surface wind over these source regions (Figs. 9 and 10). If more dust production occurs over the Sahel and Saharan regions, then the tropical North Atlantic will experience higher dust concentration because of transport by the mean zonal winds. The mean zonal winds in West Africa feature the African easterly jet (AEJ), which is a maximum near 600-hPa and 15°N (Fig. 11b) and is formed because of the temperature contrast between the Sahara and Gulf of Guinea (e.g., Cook 1999). Figure 11c shows that the strong easterly winds at 700 hPa associated with the AEJ are across the entire tropical North Atlantic between 10° and 20°N. This distribution of the mean winds is consistent with the maximum dust band between 10° and 20°N, indicating that if more dust is produced over the Sahel, the tropical North Atlantic will experience higher dust concentration due to transport by the mean zonal winds.
The low levels of the atmosphere north of the Gulf of Guinea coast are characterized by the summer monsoon westerly winds (Fig. 11b). Our composite wind difference between the high and low dust years shows surface easterly wind anomalies north of the Gulf of Guinea coast (Fig. 9a). This indicates that a high dust year in the tropical North Atlantic is associated with a weakening of the southwesterly monsoon flow, which in turn is connected to the observed decrease in Sahel precipitation (Giannini et al. 2003).
5. Discussion and summary
The observed results presented in this paper suggest that a feedback process between the AMO and dust in the tropical North Atlantic may operate through Sahel rainfall variability. An initially warm North Atlantic Ocean is associated with a northward shift of the Atlantic ITCZ and southwesterly surface wind anomalies, resulting in an increase of rainfall in the Sahel. The increased rainfall leads to a decrease in source regions for mineral aerosols (Mahowald 2007). Associated with the decrease of aerosols in the semiarid regions of Africa is a decrease of atmospheric wind-blown dust over the tropical North Atlantic. This, in turn, is a positive feedback onto tropical North Atlantic SST via the aerosol direct effect by changing the amount of solar radiation reaching the ocean’s surface in agreement with earlier studies (Foltz and McPhaden 2008a; Evan et al. 2009, 2011). However, further studies are needed to quantify the impact of dust forcing on North Atlantic SST and to determine the processes controlling African dust emission on multidecadal time scales. In particular, coupled ocean–atmosphere–land models instead of ocean-only models are needed to address the issue.
Our analyses suggest that surface processes over Africa may be more important than changes in the low-level winds over the tropical North Atlantic for dust cover in the Atlantic. That is, the dust changes in the tropical North Atlantic could be more due to increased dust production in the Sahel and Saharan regions and subsequent westward transport by the mean winds. Dust production depends on the surface shear stress, which is proportional to the cube of the wind speed. If the wind speed is larger in the dust source region of Africa, more dust is produced there and thus more dust is transported to the tropical North Atlantic by the mean easterly winds.
Almost all coupled climate models have difficulty in simulating the magnitude and phase of the AMO (Knight 2009; Medhaug and Furevik 2011), and one reason may be that they are missing the positive dust–AMO feedback proposed here. For example, for the few model experiments included in the Coupled Model Intercomparison Project (CMIP) phase 3 database that include dust, dust concentration is prescribed and does not vary from one year to the next. Furthermore, model dust concentrations are strongly dependent upon the global surface area over which deflation occurs, and there is considerable model spread in global desert area in the future warming scenarios (Mahowald 2007). Lastly, examination of a suite of coupled model experiments show large disagreement in even the long-term mean dust emissions and transport (Huneeus et al. 2011). We thus propose that one way to examine the effectiveness of coupled models in the future CMIP database, of which some will include interactive dust, is to identify the extent to which the connection between AMO-like variability and Atlantic dust cover, as observed here, is reproduced in twentieth-century experiments. Coupled climate models need to capture the aerosol-related feedback for a realistic climate simulation in the Atlantic.
African dust also affects cloudiness by changing condensation nuclei, the Saharan air layer, and atmospheric vertical motion. For example, Kaufman et al. (2005a) used the MODIS satellite data to show that aerosols increase the coverage of shallow clouds in the tropical Atlantic by increasing the number of condensation nuclei and reducing the cloud droplet size. The aerosol-induced cloud cover change can in turn affect the tropical North Atlantic SST by changing the heat fluxes into the ocean. Thus, the indirect effect of aerosols on cloud acts in the same sense as the direct effect in terms of its impact on tropical North Atlantic SST.
In summary, this paper uses a relatively long-term dust dataset to show a multidecadal covariability of the AMO, dust in the tropical North Atlantic, rainfall in the Sahel and Atlantic hurricanes, and to propose a novel mechanism for North Atlantic SST variability on multidecadal time scales. During the cold phase of the AMO from the late 1960s to the early 1990s, the Sahel received less rainfall and the tropical North Atlantic experienced high concentration of dust. The opposite was true when the AMOs were during the warm phases before the late 1960s and after the early 1990s. The observed results suggest that a positive feedback exists between North Atlantic SST, African dust, and Sahel rainfall on multidecadal time scales. An implication of this study is that coupled climate models need to be able to simulate this aerosol-related feedback in order to correctly simulate climate variability and change in the North Atlantic. In addition to the relationships with the AMO and Sahel rainfall, dust in the tropical North Atlantic also varies multidecadally with vertical wind shear in the hurricane main development region, helping to explain the inverse relationship between dust and the number of Atlantic hurricanes.
We thank three anonymous reviewers for their comments and suggestions that helped improve the manuscript. We also thank Dr. Natalie Mahowald for providing us the observed station data in the Western Sahel. This work was supported by grants from National Oceanic and Atmospheric Administration (NOAA) Climate Program Office, the base funding of NOAA Atlantic Oceanographic and Meteorological Laboratory (AOML), and a grant from National Science Foundation. The findings and conclusions in this report are those of the author(s) and do not necessarily represent the views of the funding agency.