How did nature foil all attempts to forecast the 2006 hurricane season? Undoubtedly, this question, together with others will be addressed by many studies to come. In this article, we present new insights derived from preliminary analysis using satellite data, with the intention of stimulating further studies. Data used for this study include SST, rainfall, cloud top temperature, water vapor from the Tropical Rainfall Measuring Mission (TRMM), and aerosol index (AI) from the Ozone Mapping Instrument (OMI) on Aura.
Tropical Storm and Hurricane Tracks
During July-August-September (JAS) 2005, there were 9 distinct tropical storm/hurricane tracks over the warm pool (SST > 28C) in the western Atlantic and Caribbean (WAC), and the Gulf of Mexico (Figure 1a). Five hurricanes made landfall over the Gulf coast and the eastern seaboard of the United States. In 2006, no hurricanes were found over the WAC and the Gulf region. Four distinct tropical storm/hurricane tracks were found at the northeastern edge of the warm pool far away from the eastern coast of the United States (Figure 1b). Only 2 tropical storms were formed over the WAC and the eastern seaboard of the United States, but they never developed into hurricanes.
A closer look at the full hurricane season (July through November) indicates that in 2005, the tropical storm/hurricane tracks can be classified into two groups. For Group 1, tropical storms were generated within 10-15N over the eastern Atlantic, and intensified with a clockwise track off the northwestern edge of the warm pool, east of 65W. For Group 2, tropical storms were generated and subsequently intensified to hurricane strength over the very warm water (>29C) of the WAC and the Gulf of Mexico. Long-term climatology analysis of the past 30 hurricane seasons shows that Group 1 occurs more often in early season (June-August), while Group 2 is more frequent in late season (September-November). In 2006, even though the water over the Gulf of Mexico remained very warm (SST > 29C), the WAC was substantially cooler, and there were no Group 2 tropical storms or hurricanes, contrasting with 2005 when Group 2 tropical storms and hurricanes were prevalent.
SST, Dust, Precipitation, and Cloud Anomalies
The cooling of the Atlantic in 2006 relative to 2005 was widespread, covering most of the subtropical and equatorial North Atlantic with the strongest signal (~0.6-0.8C) over the WAC (Figure 2a). This cooling was spectacular in its magnitude and extent, particularly considering that the Atlantic Ocean was supposed to be in the midst of a long-term warming trend. Also noted are substantial warming in the eastern equatorial Pacific and the Gulf of Guinea. The former can be identified with the signals of a moderate El Nino which was emerging in JAS 2006.
Concomitant to the cooling of the North Atlantic was a substantial increase in atmospheric dust loading in 2006, covering nearly all the northern tropical and subtropical Atlantic and western Africa (Figure 2b), with the oceanic maximum over the WAC, where the negative SST anomaly was most pronounced. Major reduction in aerosol loading, most likely associated with reduced biomass, was also found over the Amazon and off the coast of western South Africa between 0o and 20oS.
With the exception of the eastern Pacific, the dust anomaly pattern corresponds well, but in an inverse relationship with the SST pattern, with the areas of negative (positive) SST anomalies having positive (negative) dust anomalies overhead. This suggests that the extensive cooling over the subtropical north Atlantic may be related to the shielding of solar radiation, or the so-called solar dimming effect, by dust. Examination of the anomaly pattern of surface wind speed from the TRMM Microwave Imager indicated that the surface wind speed was also significantly higher over the negative SST anomalies in the WAC. An increase in surface wind speed could enhance surface evaporation and upper-ocean mixing, both of which would have cooled the ocean, further contributing to the extensive negative SST anomaly.
In 2006, the JAS mean rainfall over the eastern Atlantic and West Africa along 10N was generally higher than in 2005, while that over the WAC and the Gulf of Mexico was deficient compared to 2005 (Fig. 2c). Since climatologically tropical storm- and hurricane-related rainfall contributes not more than 5% of the total seasonal rain over the entire North Atlantic [Rogers et al., 2001], the reduction in total seasonal rain over the western North Atlantic and Gulf region cannot be due to tropical storms/hurricanes. Rather, the dipole-like rainfall anomaly signaled a shift in circulation regime with a more convectively active eastern Atlantic Intertropical Convergence Zone (a belt of low air pressure near the equator) and an enhanced West African monsoon, coupled to a convectively less active western Atlantic.
This picture is supported by the difference map of TRMM/VIRS (Visible and Infra-Red Scanner) brightness temperature, Tb (Figure 2d), having the eastern Atlantic and West Africa region dominated by negative signals (cold cloud top, increased deep convection), and the WAC and the Gulf region with positive signals (low cloud top, and reduced deep convection). In concert with the rainfall and the Tb patterns, the TRMM total water vapor also indicates a moister eastern Atlantic and a drier western Atlantic in 2006. The patterns of SST, rainfall and Tb are also consistent with anomalous large-scale rising of air over the eastern Atlantic and sinking over the western Atlantic found in re-analysis data from the U.S. National Centers for Environmental Prediction. This is indicative of an anomalous (in the sense of difference between 2006 and 2005) Walker-type circulation connecting the atmospheres of the two regions. The large-scale subsidence over the WAC and the Gulf of Mexico could have further suppressed deep convection and tropical storm/hurricane formation over those regions.
Covariability of North Atlantic SST and Dust
Figure 3 shows the time series of daily SST and OMI-AI for 2005 and 2006 over a region in the WAC, marked by the large rectangle in Figure 2a. This region saw the largest negative SST anomaly in 2006. It also coincides with the geographic location where Group 2 hurricanes are most likely to spawn climatologically. From the start of the seasonal warming in mid-March, the 2006 SST did not rise as fast compared to 2005, producing a relative initial cooling of 0.2-0.5C that lasted through mid-May. The most pronounced SST cooling began in mid-June 2006, reaching a maximum (>1C) in late June and mid-July, until the end of September after which the SST returned to the 2005 level (Figure 3a).
The SST cooling appeared to be closely related to the variation of Sahara dust over the region. Overall, the dust loading in 2006 was higher than 2005 (Figure 3b). An initial anomalous increase in dust began in late February and early March 2006, about 2 to 3 weeks prior to the initial SST cooling, and lasted until late May. Most striking was a major increase in dust loading at the beginning of June, 2 weeks prior to the major SST cooling. This major dust episode lasted for about a month until the end of June. The two episodes of SST cooling in the region appeared to be phase-locked to increased dust loading in the region, with a delay of about 2 to 3 weeks. Considering its episodic nature, particularly the abruptness of the SST drop at the start of the hurricane season, the 2006 Atlantic cooling was clearly not part of a long-term trend but possibly the result of a basin-wide coupled ocean-atmosphere interaction spurred by dust-induced solar dimming effect.
Historical Effects of Dust and El Nino
Table 1 shows the correlation coefficients (cc) of the dust and Nino-3 SST (monthly mean averaged over the region [5N - 5oS, 150 -90W]) with three sets of storm measures: frequency of occurrence (FOC), accumulated cyclone energy (ACE), and storm intensity (SI), for three storm categories, HC(hurricane only), TS (tropical storm only), and ALL (hurricane or tropical storm). Detailed definitions are provided in Table 1 caption. For dust, we use the Barbados record, which is the only multi-decadal dust record that exists and has been shown to be highly correlated with the outbreak of dust from North Africa over the North Atlantic [Prospero and Lamb, 2002]. All correlations are computed for two domains, the genesis region (GNR: 50-20W, 5-15N) and the intensification region (ITR: 70-40W, 15-30N) for the period 1980-2003.
As shown in Table 1, in the GNR both dust and El Nino have significant influences on tropical storms/hurricanes. Dust has stronger influence on tropical storms than hurricanes in all three measures, with the strongest signal in tropical storm intensity (cc =0.659). In contrast, El Nino has a stronger influence on hurricanes relative to tropical storms, again with strongest influence in intensity (cc=0.511). Taken over all categories and measures, El Nino has a stronger influence (more cases of significant correlations) in the GNR compared with dust, most likely through increase in the vertical wind shear [Goldenberg and Shapiro, 1996], which inhibits cyclogenesis over the GNR. Compared with the GNR, dust influences in the ITR remain strong showing significant negative correlations for hurricane frequency (cc= -0.477), hurricane strength (cc= -0.465), and ALL intensity (cc=-0.506). In contrast, the influence of El Nino is substantially diminished in the ITR, showing no significant correlations on any measures or categories.
In conclusion, our results suggest that the increased atmospheric loading of Saharan dust over the North Atlantic during the 2006 hurricane season might have been instrumental in initiating a climate regime shift in the coupled ocean-atmosphere system of the tropical and subtropical Atlantic. Associated with the shift were rapid basin-scale cooling in the North Atlantic, and suppressed tropical storm/hurricane activities in the western Atlantic and the Caribbean. Historical data also support this thesis, indicating that Saharan dust may have stronger influence than El Nino on hurricane statistics in the subtropical West Atlantic/Caribbean region, while El Nino influence may be stronger in the tropical eastern Atlantic.
Finally, we should point out that the possible roles of Saharan dust on tropical storms and hurricanes in cooling SST described here are climatic impacts, which are distinct from the dynamic effects of Saharan Air Layer (SAL) in suppressing cyclogenesis and individual tropical storm/hurricane development [Dunion and Veldon 2004]. How SAL modulated individual tropical storms or hurricanes in 2005 and 2006, and how the modulations are affected by large-scale circulation and Atlantic SST changes are clearly important subjects for further studies.
Acknowledgments
This work was supported jointly by the NASA Precipitation Measuring Mission, the NASA Africa Monsoon Multiscale Analysis, and the Interdisciplinary Investigation Program of the Earth Science Division, NASA Headquarters. Professor J. Prospero kindly provided the Barbados dust data for this study.
References
Dunion, J. P., and C. S. Velden, (2004), The impact of the Saharan air layer on Atlantic tropical cyclone activity. Bull. Am. Meteor. Soc., 85, 353-365.
Goldenberg, S. B., and L. J. Shapiro, (1996), Physical mechanisms for the association of El Nino and West African rainfall with Atlantic major hurricane activity. J. Climate, 9, 1169-1187.
Propsero, J. M. and P. J. Lamb, (2003), African droughts and dust transport to the Caribbean: Climate Change Implications, Science, 302, 1024-1027.
Rogers, E. B., R. F. Adler, and H.F. Pierce, (2001), Contribution of tropical cyclones to the North Atlantic rainfall climatology as observed from satellites. J. Appl. Meteorology, 40, 1785-1800.
Author Information
William K. M. Lau, Laboratory for Atmospheres, NASA Goddard Space Flight Center, Greenbelt, Md.; E-mail: William.k.lau@nasa.gov; and Kyu-Myong Kim, Goddard Earth Sciences and Technology Center, University of Maryland at Baltimore County, Baltimore.
Table 1. Correlation of Barbados dust index and Nino-3 SST with three sets of storm measures: FOC, ACE and SI. (Correlations exceeding the 5% confidence level are shown in bold.). All correlations are computed separately for three storm categories, TS (tropical storm only), HC (hurricane only), and ALL (tropical storm or hurricane).
FOC is defined as the total number of days from July through November (JASON) in which a given storm category is found within the region.
ACE is the square of the maximum sustained wind speed of storm at every 6-hour.
SI is defined as the ACE divided by FOC.
Figure 1. Tropical storm (open circles) and hurricane (dark circles) tracks superimposed on seasonal mean (JAS) fields of TMI (TRMM Microwave Instrument) SST (C) for (a) 2005 and (b) 2006.
Figure 2. Maps show difference between JAS 2006 and 2005 (2006 minus 2005) of (a) SST (C),( b) OMI-AI (units are non-dimensional), (c) TMI rainfall (millimeters per hour), and (d) VIRS cloud top temperature (Tb; C).
Figure 3. Time series showing daily (smoothed by a 5-day running mean) variation of (a) SST and (b) OMI/AI, in the WAC region (70-40W, 15-30N) in 2005 and 2006. Blue (red) color indicates negative (positive) anomaly in 2006 relative to 2005.