SEARCH

SEARCH BY CITATION

Keywords:

  • tropical cyclones;
  • mineral aerosols;
  • climate;
  • satellite

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

[1] Recent work has shown a statistical climatological link between African dust outbreaks and North Atlantic tropical cyclone frequency and intensity. However, a definite causal link between year-to-year changes in African dust and Atlantic tropical cyclones has yet to be proven. Here we show that variability in Atlantic dust cover is linked to changes in tropical cyclone activity through the aerosols' surface radiative forcing, which has a net cooling effect on tropical Atlantic Ocean temperatures. In this manuscript we describe a new methodology for incorporating more than 25 years of satellite observations of aerosols into a simple model that estimates the aerosol direct effect and its impact on tropical eastern Atlantic Ocean temperatures. The output from our model suggests that African dust outbreaks play a nonnegligible role in the evolution of eastern Atlantic Ocean temperatures. Using the strong relationship between temperatures in the so-called main development region and the seasonal power dissipation index (PDI), we estimate that about one third of the increase in PDI over the last 25 years can be attributed to decreases in dust loadings over the same period. Our results imply that efforts aimed at attributing causality to past variability of, or at predicting future changes in, North Atlantic tropical cyclone activity must consider the important radiative influence of African dust.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

[2] It is thought that anomalously warm equatorial North Atlantic sea surface temperatures weaken the West African monsoon, decreasing precipitation across the Sahel during the boreal summer [Giannini et al., 2003; Fontaine and Janicot, 1996]. Reductions in precipitation would lead to decreases in vegetation across the Sahel [Nicholson et al., 1998], increases in source regions for mineral aerosols [Mahowald et al., 2002], and therefore an increase in dust cover over the tropical North Atlantic during the following boreal summer [Moulin and Chiapello, 2004; Prospero and Lamb, 2003]. An increase in summertime dust activity may have a direct dynamical effect on tropical cyclogenesis through enhancement of the mid level temperature inversion, increases in vertical wind shear, and reductions in relative humidity [Dunion and Velden, 2004; Wong and Dessler, 2005]. However, here we show that these dust plumes also have an indirect effect on tropical cyclone activity through cooling of sea surface temperatures. We propose that increases in dust cover also lead to cooler tropical North Atlantic Ocean temperatures (in the region of 10–20°N and 15–65°W) and reduced tropical cyclone activity [Emanuel, 1987; Gray, 1968; Goldenberg et al., 2001; Saunders and Harris, 1997; Shapiro and Goldenberg, 1998]. We suggest that the opposite argument can also be made, where anomalously cool equatorial Atlantic sea surface temperatures lead to a warming of the more northerly tropical Atlantic during the following year, via reductions in African dust cover, which gives way to increased tropical cyclone activity.

[3] In this paper we attempt to quantify this relationship between dust and ocean temperature (and ultimately tropical cyclone activity) over the last quarter century using observations and output from a radiative transfer and general circulation model. In next section of this paper we present empirical evidence demonstrating a statistical relationship between African dust storm activity and North Atlantic tropical cyclones via the oceanic cooling associated with dust cover. The subsequent section describes a simple radiative transfer model that assimilates a 25-year record of satellite aerosol observations to estimate the surface aerosol forcing, and the oceanic response to that forcing. We then utilize the output from our satellite data driven model to discuss the contribution dust has made to eastern Atlantic Ocean temperatures, and estimate what tropical cyclone activity over the last 25 years may have looked like in the absence of dust. The paper concludes with final remarks that put our results into a larger context and suggests future efforts.

2. Empirical Observations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

[4] To explore the oceanic cooling exhibited by dust we created a time series of dust cover averaged over months that may be of importance to tropical cyclone activity (May through October), for the so-called main (or hurricane) development region, defined here as 10–20°N and 15–65°W [Goldenberg and Shapiro, 1996] (Figure 1, solid black line). Our dust index is a 25-year satellite record of over-water dust frequency from the Pathfinder Atmospheres Extended data set [Evan et al., 2006a]. The gray shaded region corresponds to monthly maximums and minimums in dustiness for each year's 6-month period. Here, dust activity during the first 7 years of the record is much higher than that of the remainder of the time series. Some of the highest dust concentrations occurred in 1984, 1985 and 1987, and the lowest in 1996, 2004 and 2006. A time series of tropical cyclone days [Jarvinen et al., 1984] (the number of days where a tropical cyclonic system with 1-min-averaged maximum sustained winds greater than or equal to 34 knots) averaged over the hurricane development region for each year (Figure 1, dashed line) shows that tropical cyclone activity from 1982 through 1994 is much less pronounced than over the last 12 years of this record [Goldenberg et al., 2001]. The dotted line in Figure 1 is the cyclone days time series after the application of a 1-4-7-4-1 filter, and highlights the general increase over time of tropical cyclone activity in the basin. The interannual variability in this dust series, including a general decrease in aerosols from the beginning of the record until 1996 and the subsequent leveling-off of values, is somewhat similar to the increasing time series of tropical cyclone days. This is not surprising since there is a statistically significant correlation between this dust record and tropical cyclone activity in the Atlantic basin; Evan et al. [2006b] showed a correlation coefficient between dust and tropical cyclone days of −0.7 when ENSO was taken into account.

image

Figure 1. Time series of dust and tropical cyclone days. Valued on the left-hand axis is a time series of mean May–September dust cover averaged over the hurricane development region for the period 1982–2006. The shaded regions represent the range of monthly values for each year, and the black line shows the mean. The dashed line is a time series of seasonal tropical cyclone days over the hurricane development region, valued on the right-hand axis. The dotted line is the cyclone days time series after the application of a 1-4-7-4-1 filter.

Download figure to PowerPoint

[5] We note that there is some contamination of our dust record by the volcanic eruptions of El Chichón (for 1982 values) and Mount Pinatubo (for 1991 and 1992 values). However, the year-to-year variability in our dust record has been validated against other dust data sets [Evan et al., 2006a], and surface observations of summertime dustiness from Barbados do show similar results [Chiapello et al., 2005].

[6] Monthly lead/lag correlations between our dust time series and another of Hadley ocean temperatures [Rayner et al., 2003], both averaged over the hurricane development region, show that consistently the strongest correlations between dust and temperatures occur when sea surface temperature lags behind dustiness from 1 to 3 months (Figure 2). The plot for September dustiness is the exception, where the strongest correlation occurs with September ocean temperatures; this may reflect the shoaling of the mixed layer (Figure S2 in auxiliary material) and increased ocean temperature sensitivity to changes in surface solar insolation.Generally, the dust time series can explain about 25% of the observed variance in ocean temperatures at the strongest point in the lead/lag correlations. The plots in Figure 2 are consistent with a causal relationship between dust and ocean temperature [Schollaert and Merrill, 1998], implying that dust cover has a nonnegligible effect on ocean temperature variability during the hurricane season [Evan, 2007; Lau and Kim, 2007a, 2007b]. Since the plots in Figure 2 imply an inverse relationship between dust and ocean temperatures on a year-to-year basis, we suggest that changes in dust cover may play a role in modulating the interannual and long-term variations in ocean temperature, and therefore tropical cyclone activity [Emanuel, 2005a, 2005b; Goldenberg et al., 2001; Landsea, 2005; Pielke, 2005; Webster et al., 2005]. Since aerosols may skew satellite ocean temperature retrievals, we repeated the correlation analysis in Figure 2 with the ICOADS [Woodruff et al., 2005] data set of in situ Sea Surface Temperature, finding similar results (Figure S1 in auxiliary material).

image

Figure 2. Plot of lag/lead relationship between Hadley ocean temperature (SST) and dust cover averaged over the main development region for the months of June–September. All correlations are for individual months (e.g., June SST and June dust). Asterisks denote correlations that are significant at the 95% level on the basis of the t score of a two-tailed t test.

Download figure to PowerPoint

[7] While these correlations do suggest a causal relationship between dust and ocean temperatures that is nonnegligible, they do not explicitly demonstrate that the ocean is responding to the direct aerosol effect. The following section describes a method for estimating the effect of aerosol cover on ocean temperatures using our long-term satellite record and a simple radiative transfer model.

3. Method for Quantifying the Aerosol Impact on Ocean Temperature

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

[8] Using observations of aerosol optical thickness from our satellite data set [Ignatov and Stowe, 2002a, 2002b], we estimate the aerosol radiative forcing over the hurricane development region from the mean difference between distributions of surface solar insolation for an atmosphere with a moderate background aerosol optical thickness of 0.1, and for one containing an aerosol layer whose optical thickness is prescribed by the satellite retrievals. Our background value of 0.1 represents average levels of sulfates and sea salt (including carbonaceous aerosols) in this region, in agreement with simulations from a global aerosol-chemistry transport model [Myhre et al., 2007]. Summertime aerosol optical thickness is generally well above the value of 0.1. With the exception of the two volcanic events mentioned in the previous section, summertime aerosol optical thickness values above the background can be attributed to mineral species. However, this should not be surprising since the hurricane development region is not downwind of any major industrial regions, but is downwind of the planet's largest dust source.

[9] We calculate the surface aerosol forcing by modeling the atmospheric transmittance accounting for extinction by stratospheric ozone, water vapor, carbon dioxide, and an aerosol layer, assuming that gas absorption and scattering can be treated independently using multiple scattering via the Meador and Weaver two stream approximation [Meador and Weaver, 1980]. We obtained monthly climatological mean and standard deviation values of stratospheric ozone and water vapor from the NCEP reanalysis [Kalnay et al., 1996]. We assumed a constant aerosol optical thickness across the visible spectrum to be the mean of the satellite 0.68 and 0.86μm channels. In the radiative transfer calculations we use a single scatter albedo (0.98 ± 0.01) and an asymmetry parameter (0.78 ± 0.01) consistent with the AVHRR aerosol optical thickness retrievals set [Ignatov and Stowe, 2002a, 2002b], and recent observational [Haywood et al., 2003] and modeling studies [Myhre et al., 2003].

[10] Cloud cover also alters the flux of shortwave radiation incident on the ocean. To account for this we scaled values of our solar insolation by climatological mean monthly values of cloudiness from the International Satellite Cloud Climatology Project [Rossow and Schiffer, 1991]. It is likely that mean monthly cloud amounts, especially those when dust activity peaks, are artificially high since optically thick dust plumes are probably classified as clouds because of their thermal contrast with the clear-sky oceanic background [Evan et al., 2006a], therefore our aerosol forcing estimates may be somewhat conservative.

[11] Since we used monthly climatological mean and standard deviation values for stratospheric ozone, water vapor, carbon dioxide, and cloudiness, and monthly mean and standard deviation values for the satellite retrievals of aerosol optical thickness, we employed a Monte Carlo technique, in order to sample distributions of these variables for creating the distributions of surface solar insolation for the dust and dust-free cases.

[12] We used climatological monthly temperature and salinity profiles from the World Ocean Atlas [Conkright et al., 2002] to calculate distributions of density profiles also using a Monte Carlo technique, and we defined the mixed layer depth to be the point at which the potential density is 0.125 kg/m3 greater than that at the surface [Levitus, 1982]. Our climatological mixed layer depth averaged over the hurricane development region has a maximum in January of 40 m, and a minimum in September of 20 m (Figure S2 in auxiliary material).

[13] We assumed all sunlight incident on the ocean surface is distributed throughout the mixed layer [Schollaert and Merrill, 1998] and calculated the local change in mixed layer temperature with respect to time in a manner similar to Schollaert and Merrill [1998, equation 1] (equation (1)).

  • equation image

[14] Here, ΔQ is the change in solar insolation between an atmosphere with aerosols and one that is clear (already accounting for the fraction of the sky that is cloudy), MLD is the depth of the mixed layer, and ρcp is the density of the ocean water times its specific heat, which we assumed to be constant at 4.1 × 106Jm−3K−1 [Schollaert and Merrill, 1998]. We estimated T/t at a 5° resolution across the hurricane development region.

[15] This methodology for estimating the aerosol surface forcing and resultant change in ocean temperature is a first-order approximation. Here we are ignoring direct and indirect impacts on the surface longwave radiation budget, oceanic heat transport, variability in the optical properties of dust across the solar spectrum, and interannual variability in the mixed layer depth and cloud cover. Although some of these assumptions may be justified by slow ocean currents in our region of interest [Schollaert and Merrill, 1998], they also are reasonable since we are looking at ocean forcing on monthly timescales, while the resultant anomalies are dissipated on timescales of 4–8 months via upward radiative and turbulent energy fluxes [Deser and Timlin, 1997].

[16] We should note there is a positive correlation between cloudiness and dust across the tropical Atlantic. However, this is likely an artifact in the cloud detection (as previously discussed). While there is a long-term downward trend in Atlantic dustiness [Evan et al., 2007] and global aerosol optical thickness [Mishchenko et al., 2007], there is no trend in cloudiness over the tropical Atlantic [Heidinger et al., 2007]. Furthermore, we are ignoring aerosol indirect effects on cloud condensation nuclei number concentration [Forster et al., 2007, among others], which would also act to cool the Atlantic.

4. Model Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

[17] From Figure 3a, mean monthly aerosol forcing values, the average monthly value of ΔQ strongly reflects the seasonality of solar insolation (maximum in June, minimum in December), cloudiness (maximum in October, minimum in April), and dust activity (maximums in March and June, minimum in November). The average monthly aerosol forcing is strongest in June, and weakest in December, with the range of values being about 7 W/m2. Our climatological mean of −8 W/m2 compares well to recent model-based estimates [Yoshioka et al., 2007]. Furthermore, the variability in the dust forcing is of a larger magnitude than other surface aerosol forcings [Kvalevåg and Myhre, 2007].

image

Figure 3. Climatology the surface forcing and subsequent oceanic cooling exhibited by dust. (a) The climatological monthly aerosol surface forcing (solid line) and the annual mean value (dotted line). (b) The time series of seasonal June through October aerosol surface forcing (black line), the range of monthly values (gray shaded region), and the mean (dotted line). (c) The climatological monthly surface temperature forcing by aerosols from this study (black line) and from model output for the month following that shown on the x axis (gray line). (d) The time series of seasonal June through October surface temperature forcing by aerosols (black line), the range of monthly values (gray shaded region), and the mean (dotted line). All values are averaged over the hurricane development region.

Download figure to PowerPoint

[18] A time series of mean June–September aerosol forcing over the hurricane development region shows the strongest mean forcing of −15 W/m2 occurring at the beginning of the record in 1982 (Figure 3b), this is followed by a weakening of the magnitude of aerosol forcing until 1991, and then another weakening until the 1994 value of −7 W/m2. As previously mentioned, stratospheric aerosols associated with the eruption of El Chichón and Mount Pinatubo bias our estimated mean summertime aerosol forcing for 1982, 1991, and 1992 (and possibly to a lesser degree in 1983). However, a reduction in surface solar insolation associated with these events is well known [Minnis et al., 1993; Santer et al., 2006], and our estimates of surface cooling still appear to be reasonable (although in this case the cooling is not just due to mineral aerosols). These volcanic eruptions add to the perception of a division between aerosol activity pre- and post-1995, although a summertime reduction of dust over the Atlantic since the mid 1980s is well known [e.g., Chiapello et al., 2005]. The mean forcing values for the period of 1995–2006 shows values ranging from −7 to −9 W/m2, and the years 2005 and 2006, years of hyperactive tropical cyclone activity, exhibited the weakest mean dust forcing over our entire record. The gray shaded region in Figure 3b represents the range of values for the months of June–September for each year. These ranges are generally between 5 to 10 W/m2 and demonstrate that throughout the entire record, although the highest mean seasonal forcings occurred in the beginning of the record, some monthly values in the 1990s and 2000s were as high as those from the 80s (gray shaded region, Figure 3b).

[19] The monthly climatology of aerosol temperature forcing (equation (1) and Figure 3c) shows that ocean temperatures are most sensitive to aerosols during the boreal summer, when the mixed layer is most shallow, and is the strongest in September (0.55 K/month), during the height of the hurricane season. There is a secondary peak in aerosol forcing in March, which reflects the increased levels of dust and biomass burning that occur in this region during the boreal winter and spring [Husar et al., 1997].

[20] The gray line in Figure 3c is the monthly mean surface temperature forcing over the hurricane development region from a general circulation model that includes dust direct radiative forcing [Yoshioka et al., 2007; Mahowald et al., 2006]. In the model, ocean temperature changes due to dust forcings are determined by subtracting the mean ocean temperature for runs with dust from those where the dust is absent. The temperature values observed in the model output also include longwave surface warming from dust, but this contribution is much smaller than that from the shortwave over the hurricane development region [Myhre et al., 2003; Yoshioka et al., 2007]. Additionally, the general circulation model includes possible feedbacks on ocean temperatures associated with top of the atmosphere aerosol forcing [e.g., Miller and Tegen, 1998], although more recent studies suggest that the difference between top of the atmosphere and surface forcing for dust is small over the Atlantic [Kaufman et al., 2002]. The magnitudes of the monthly temperature forcing from the model and from our satellite-based calculations are very similar (Figure 3c), suggesting a validation of our methods. The slightly larger magnitude of forcing we calculate during the months of September and October result from the stronger summertime shoaling of our mixed layer depths (see Figure S2 in auxiliary material).

[21] The exact effect aerosol forcing has on the measured sea surface temperature is difficult to estimate since temperatures for a given month are dependant on, among other factors, solar radiation received during previous months, and relative changes in ocean heat transport and radiative cooling. However, the excellent agreement between our satellite based estimates and those from the transport model speaks to the validity of the assumptions we have made, although other models that estimate surface forcing from dust disagree both with our study and the work of Mahowald et al. [2006] and Yoshioka et al. [2007] [e.g., Miller and Tegen, 1998]. Also, the data from our model are estimations of the temperature change due to dust for the top 20 to 40 m of the ocean, not just the ocean surface. Since there is a temperature gradient at the ocean's surface, it is likely that observations of sea surface temperature made with satellites, especially after the passage of a particular dust event, will show a stronger temperature response to the aerosol forcing.

[22] Our time series of average July, August and September aerosol ocean temperature forcing averaged over the hurricane development region (Figure 3d) is similar to the series for dustiness (Figure 1). Therefore, correlations between observations of ocean temperature and our aerosol forced temperatures are also similar to those for dust and ocean temperatures (Figure 2), and correlate well with tropical cyclone activity [Evan et al., 2006b]. The time series in Figure 3d shows a reduction in mean temperature forcing from about −0.9 to −0.4 K/month. Embedded in that decline is a spike of about 0.4 K/month in temperature forcings in 1991 that is enhanced by stratospheric aerosols. From 1995 onward the mean temperature changes due to aerosols are consistently around −0.4 to −0.5 K/month, and the monthly values ranging from −0.6 to −0.3 K/month. We also note a marginally weaker mean aerosol temperature forcing for 2006 relative to 2005, in disagreement with the findings of Lau and Kim [2007a, 2007b], and more in agreement with Evan [2007].

[23] Holland and Webster [2007] found that since the mid 1970s eastern tropical North Atlantic sea surface temperatures exhibited a warm anomaly that was not observed in the western tropical North Atlantic or Caribbean basins. They speculated that this anomaly of roughly 0.2°C was not related to global process and was therefore local in nature. When we account for dust's effect on ocean temperatures by subtracting our monthly estimations of oceanic cooling by dust from mean Hadley ocean temperatures, we see about a 30% reduction in the trend of mean July, August, and September ocean temperatures over the hurricane development region for the last 25 years (or a reduction in the trend of Hadley ocean temperatures of about 0.1°C per decade, after removing years with substantial stratospheric aerosol loadings: 1982, 1991, 1992). Since the magnitudes of the ocean temperature anomalies given by Holland and Webster [2007] and our estimations of oceanic cooling by dust are very similar, and given the increases in dust across the eastern tropical North Atlantic region that existed from the late 1970s through the 1980s [Evan et al., 2007; Prospero and Lamb, 2003], we believe future analysis similar to Holland and Webster [2007] that account for tropospheric aerosols are warranted.

5. Relevance to Tropical Cyclone Activity

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

[24] One way to quantify the effect dust has tropical cyclone activity is to utilize the strong linear relationship between September ocean temperatures in the hurricane development region and the seasonal tropical cyclone power dissipation index (PDI) for the entire tropical North Atlantic and Caribbean basins, as was done by Emanuel [2005a]. The PDI is a proxy for the power a tropical cyclone dissipates at the surface, and is directly proportional to the cyclone's wind speed cubed integrated over the lifetime of the storm [Emanuel, 2005a]. Since tropical Atlantic Ocean temperature serves as a proxy for basinwide circulation changes that affect tropical cyclone genesis location, track, duration, and intensity [Kossin and Vimont, 2007], this relationship between PDI and ocean temperature is probably indirect. To estimate a seasonal PDI for a dust-free atmosphere we first determine the slope of the least-squared linear best fit line between mean September ocean temperatures in the hurricane development region to seasonal PDI for the years 1982 through 2006. We then use this slope in conjunction with the time series of mean September oceanic cooling by dust, which we have already calculated, to determine the increase in seasonal PDI that would occur in a dust-free atmosphere.

[25] Figure 4 shows the measured seasonal PDI (dark gray shaded region), and the increase in PDI that we estimate for the dust-free scenario (light gray shaded region). The most striking feature of this plot is the large increase in PDI across the entire time series that result from warmer ocean temperatures in the dust-free scenario. On average, PDI for the dust-free case is about 2.2 times greater than the measured PDI. This value of 2.2 excludes years where volcanic aerosols probably contaminate our dust estimations (1982, 1991 and 1992).

image

Figure 4. Seasonal Atlantic and Caribbean tropical cyclone power dissipation index (PDI). The dark gray region represents the measured seasonal PDI, and the light gray region represents our estimation of PDI for a “dust-free” atmosphere.

Download figure to PowerPoint

[26] It is well known that hurricane activity in the Atlantic entered a more active phase in 1995 [Goldenberg et al., 2001]. This active phase is reflected in the difference between mean values of the measured PDI (dark gray region in Figure 4); from 1995 onward the mean seasonal PDI is about 2.5 times greater than the mean seasonal PDI for 1983 through 1994 (again, ignoring 1982, 1991, and 1992 for this comparison). PDI for our dust-free scenario post-1994 is only about 1.5 times greater than it is up to this year. This reduction in the difference between mean PDI values for the dust-free case implies that the reduction in dust that has occurred after the 1980s contributed to the recent increase in (1) ocean temperatures across the hurricane development region and (2) seasonal Atlantic and PDI. We estimate that about 35% of the increase in PDI over the last quarter century can be attributed to reductions in dust from West Africa, and the subsequent loss of the cooling effect these aerosols have on ocean temperatures in the eastern tropical North Atlantic basin.

[27] Including the years of 1982, 1991, and 1992, for which the surface forcing is due to a combination of sulfates from volcanic event as well as dust, only slightly reduces the difference in PDI for post- and pre-1995 for the “dust-free” case. Here, we estimate that almost 45% of the recent increase in PDI can be explained by the oceanic cooling associated with dust and stratospheric aerosols.

6. Concluding Remarks

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

[28] Our results demonstrate that during the boreal summer dust plays a nonnegligible role in shaping the temperature of the eastern tropical Atlantic, suggesting that the exclusion of mineral aerosols in tropical Atlantic Ocean temperature attribution studies [e.g., Santer et al., 2006] may not be justified. Our results also imply that the tropical Atlantic is unique in terms of the processes that force the surface temperatures. Holland and Webster [2007] demonstrated that ocean temperatures in the eastern tropical Atlantic have been increasing more sharply than have been those in the Caribbean and Western Atlantic, after global processes have been accounted for. It is possible that this additional warming can be attributed to the decreasing dustiness that we observe over the same time period. This would imply that recent decreases in dustiness have actually helped to force the tropical cyclone activity to into a new, more destructive regime [Holland and Webster, 2007]. Therefore new analysis is warranted to determine how our results fit into what is know about changing ocean temperature across the tropics. Recent work linking Atlantic tropical cyclone activity to the Atlantic Meridional Mode [Kossin and Vimont, 2007] has shown that a marginal increase (decrease) in tropical Atlantic surface temperatures could “push” the basin into a state that is more favorable (unfavorable) for cyclogenesis (less than 0.5 K). Therefore, we also suggest that investigations into a potential relationship between dust and the Atlantic Meridional Mode are warranted.

[29] Understanding the exact causes of the year-to-year changes in Atlantic dust cover is not completely straightforward. For example, some variability in Sahelian precipitation (and therefore dust cover) may result from feedbacks associated with the dust loadings themselves [Yoshioka et al., 2007], and from nonlinear feedbacks associated with vegetation changes across the Sahel [Foley et al., 2003]. Changes in year-to-year dustiness may be modified by land use change [Moulin and Chiapello, 2006; Tegen et al., 2004a], but to what extent is still disputed [Mahowald et al., 2004; Tegen et al., 2004b]. It is also possible that anthropogenic climate change has played a role in recent variability of Sahelian precipitation [Held et al., 2005]. Additionally, while it is known that Atlantic Ocean temperatures play a strong role in modulating the strength of the West African monsoon, the oceanic regions where temperature variability has been linked to Sahelian precipitation [Giannini et al., 2003; Fontaine and Janicot, 1996] tend to lie outside the latitudes where dust is observed in the summer months [Husar et al., 1997], but the meridional gradients of Atlantic surface temperature may be more important than regional behavior [Hoerling et al., 2006]. However, it is likely that exists a feedback between oceanic forcing by mineral aerosols and the strength of the West African monsoon.

[30] Since summertime Atlantic dust cover is well correlated with Sahelian precipitation from the previous year [Prospero and Lamb, 2003; Moulin and Chiapello, 2004], reconstructions of dust cover over the last 50 years can be made with time series of Sahelian precipitation [Prospero and Lamb, 2003]. Therefore, it is possible that the noted long-term correlations between Sahel precipitation and tropical cyclones [Landsea and Gray, 1992] could be partially attributed to the cooling effect mineral aerosols have on tropical ocean temperatures. The modeling study of Mahowald et al. [2006] showed increases in vegetation across the Sahel and resultant reductions in Atlantic dust cover associated with a doubling of CO2. Therefore, not only may there be a less dusty future for the Atlantic Ocean, but coming multidecadal periods of Sahelian drought commonly associated with reductions in Atlantic tropical cyclone frequency and intensity [Landsea and Gray, 1992], may not lead to the same upswings in dust activity, cooling of ocean temperatures, and dampening of tropical cyclone activity, as they probably did in the past. However, Held et al. [2005] showed a strong drying of the Sahel with further increases in greenhouse gases, implying possible increases in dust storm activity, and therefore future cooling of the tropical North Atlantic.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

[31] We would like to thank Chris Landsea, Greg Holland, Michael Mann, and Ron Miller for their detailed review of this manuscript. Funding for this research was provided by the NOAA/NESDIS Polar Program and the NOAA/NESDIS/ORA AVHRR Reprocessing Program. The views, opinions, and findings contained in this report are those of the authors and should not be construed as an official NOAA or U.S. Government position, policy, or decision. International Satellite Cloud Climatology Project cloud data are available from http://isccp.giss.nasa.gov/. Data from the World Ocean Atlas are available from http://www.nodc.noaa.gov/, ICOADS data are available from http://dss.ucar.edu/, and Hadley HadISST1 data are available from http://www.hadobs.org. Pathfinder Atmospheres Extended satellite products are available through http://cimss.ssec.wisc.edu/clavr/; more detailed information about the project and products can also be found there. Tropical cyclone data were obtained from the National Hurricane Center Hurxricane Best Track Files.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Empirical Observations
  5. 3. Method for Quantifying the Aerosol Impact on Ocean Temperature
  6. 4. Model Results
  7. 5. Relevance to Tropical Cyclone Activity
  8. 6. Concluding Remarks
  9. Acknowledgments
  10. References
  11. Supporting Information

Auxiliary material for this article contains three figures.

Auxiliary material files may require downloading to a local drive depending on platform, browser, configuration, and size. To open auxiliary materials in a browser, click on the label. To download, Right-click and select “Save Target As…” (PC) or CTRL-click and select “Download Link to Disk” (Mac).

See Plugins for a list of applications and supported file formats.

Additional file information is provided in the readme.txt .

FilenameFormatSizeDescription
ggge1177-sup-0001-readme.txtplain text document2Kreadme.txt
ggge1177-sup-0002-fs01.epsPS document49KFigure S1. Plot of the lead/lag relationship between ICOADS in-situ ocean temperature (SST) and dust cover averaged over the hurricane development region for the months of June–September.
ggge1177-sup-0003-fs02.epsPS document10KFigure S2. Climatological monthly values of our estimated mixed layer depths averaged over the hurricane development region.
ggge1177-sup-0004-fs03.epsPS document13KFigure S3. Climatological monthly values of cloud cover from the International Satellite Cloud Climatology project(black line, left-hand axis) and dust cover from the Pathfinder Atmospheres Extended data set (gray line, right-hand axis).

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.