How the Saharan air layer (SAL) affects tropical cyclone intensity in the North Atlantic Ocean is an issue in debate. A composite study of 274 cases from 37 named tropical cyclones that formed during the period 2005–2007 is conducted using AIRS relative humidity between 600–700 hPa. Typically the dry SAL air is first observed within 1000 km north of the tropical cyclone center and then intrudes southward and towards the inner region of tropical cyclones along the cyclonic flow. This study provides evidence that the SAL can affect tropical cyclone intensity in both favorably and unfavorably manners by intensifying tropical cyclones when it is first found mostly in the northwest quadrant and then weakening tropical cyclones when its dry air intrudes within 360 km of the tropical cyclone center, mostly in the southwest and southeast quadrants. It appears that the SAL is favorable for the initial development of tropical cyclones but unfavorable for their subsequent intensification.
 The existence of the AEJ leads to a significant potential vorticity (PV)-gradient sign reversal with weak cyclonic or even anticyclonic PV north of the jet and strong positive PV south of the jet. Karyampudi and Carlson  suggested that such a PV pattern favors easterly wave growth via barotropic instability. They also showed that the induced ageostrophic circulation and attendant convection could contribute to wave growth and tropical cyclogenesis by supporting convection along its leading and southern borders. Thus Karyampudi and Carlson  and Karyampudi and Pierce  conclude that the outbreaks of the SAL can aid wave growth and tropical cyclone development.
 Currently how the SAL impacts tropical cyclones is still an issue of debate in the tropical cyclone research community. According to the aforementioned studies, it is likely that the SAL acts both positively and negatively on tropical cyclones since only different aspects of the influence of the SAL are previously emphasized. For example, Karyampudi and Carlson  focused on the possible dynamic instability associated with the SAL while Dunion and Velden  examined the dry air intrusion, vertical wind shear and local static instability associated with the SAL. Our goal with this paper is to address this issue by examining tropical cyclones observed over the North Atlantic and the Caribbean Sea during the period 2005–2007 because it is of practical importance for hurricane forecasters to know whether and how the positive or negative impact of the SAL occurs.
2. Data Description
 The tropical cyclone data over the North Atlantic and the Caribbean Sea during the period 2005–2007 were obtained from the National Hurricane Center (NHC). The hurricane best track data (HURDAT) include the center latitude and longitude and 1-minute-averaged maximum sustained winds for each 6-hour interval. Vertical profiles of relative humidity retrieved from the Atmospheric Infrared Sounder (AIRS) and the Advanced Microwave Sounding Unit (AMSU) are used to describe the SAL activity. The AIRS and AMSU were launched with the NASA Aqua satellite in 2002. The standard AIRS retrieved profiles (Level 2 products) are obtained twice daily (day and night) on a 1:30 p.m. sun synchronous orbit from a 705-km altitude with a horizontal resolution of ∼50 km and 28 pressure levels. The best track data are interpolated linearly to match AIRS granules. Wu et al.  showed that the AIRS retrieved profiles are very useful for investigating the influence of the SAL on tropical cyclones.
3. Intrusion of the SAL Air
 As an example, here we first show how the SAL impacted Tropical Storm Debby, which formed from a tropical wave just off the coast of Africa on 21 August 2006. The tropical depression intensified into a tropical storm about 500 km southwest of Cape Verde early on 23 August. As identified by the low relative humidity between 600 and 700 hPa in Figure 1a, the dry air associated with the SAL was located far north of the system with relative humidity as low as 10%. The dry SAL air shown by the shading in Figure 1a extended southward west of Debby and tended to intrude into 360 km of the tropical cyclone center. In the next section we will show that the radial distance of 360 km is the average intrusion location in the southwestern quadrant for all cases and is close to the average intrusion location (381 km) for weakening cases. As the system strengthened and reached its peak intensity on 23 August (Figure 1b), the SAL dry air continued intruding into the storm with the cyclonic flow. Relative humidity as low as 30% can be found around the southern part within 360 km of the tropical cyclone center. Debby maintained its peak intensity on 24 August and weakened to a tropical depression at 0600 UTC on 25 August (Figures 1c and 1d) as the dry SAL air further intruded into the central region of Debby. Although the vertical shear also affected its intensity, Figure 1 suggests that the intrusion of the dry SAL air may have inhibited the intensification of Tropical Storm Debby after 23 August.
 In order to understand the impact of the SAL, 37 named tropical cyclones that formed in the North Atlantic during the period 2005–2007 are further examined. Some tropical cyclones during this period are not included because they were located west of 85°W or north of 45°N. The intrusion of the SAL air into a tropical cyclone is defined with the nearest location of the dry air with the AIRS relative humidity between 600 and 700 hPa. Examination of the relative humidity in the inner region of tropical cyclones shows that the SAL intrusion into tropical cyclone circulation is well identified with the grid points of 30% or lower relative humidity. Though the 30% relative humidity threshold is selected somewhat subjectively, we believe it is more helpful to distinguish the intrusion of the dry SAL air into tropical cyclones. It is also required that at least three non-isolated grid points meet the criterion simultaneously to exclude the low relative humidity possibly associated with convective downdrafts.
 With the presence of the dry SAL air near tropical cyclones, 274 cases are yielded for the three-year period, including 30% cases in the tropical depression stage. The dry air defined with 30% or lower relative humidity within 1000 km of the tropical cyclone center was observed in about 93% of the sample. The identified nearest locations of the intrusion of the dry SAL air are composited with respect to the tropical cyclone center. Figure 2 shows the spatial distribution of the nearest intrusion locations for the 274 cases. Although the dry SAL air can appear in any quadrant of the tropical cyclone, 64.5% (174) of them are found in the northwest and southwest quadrants, indicating a typical pathway for the SAL intrusion, which can also be identified with the count of the closest locations in each 200 km×200 km box (contours in Figure 2). As shown in Figure 2, since the SAL is usually located to the north of the tropical cyclone, the intrusion is first observed about 500 km north of the tropical cyclone center and then the dry air is steered southward with the cyclonic flow and drawn toward the tropical cyclone center.
 Examination of the 37 tropical cyclones discussed in this study also suggests such a typical intrusion pathway. While in 16.2% of the tropical cyclones intrusions only occurred in the northeast and northwest quadrants, most of the intrusions (in 78.4% of the tropical cyclones) extended to the southwest quadrant. It is found that intrusions of the dry SAL air occurred successively in northeast, northwest, southwest and southeast quadrants in 45.9% of the tropical cyclones.
4. Influence of the SAL Air
 We assume that a distinction can be found between the intensifying and weakening cases if the SAL plays both positive and negative roles in tropical cyclones. For this reason, the 24-hour intensity change in the maximum sustained wind speed is categorized into weakening, neutral (no change) and intensifying and we select the intensifying and weakening cases, which account for 40% (110) and 33% (90) of the sample, respectively, to discuss the impact of the SAL on tropical cyclone intensity. Note that doing so does not mean that the SAL dry air has little influence on the neutral cases, but the cases with intensity change should contain clearer signals of the influence of the SAL than the neutral cases that may be affected by the combination of the SAL, large-scale atmospheric environments, underlying surface condition and internal dynamics.
Figure 3 shows the composite of the nearest locations of the dry SAL intrusion and their counts within a 200 km × 200 km box for the weakening and intensifying cases, respectively, indicating significant differences between the weakening and intensifying cases. First the nearest intrusion locations in the weakening cases are generally within 360 km of the tropical cyclone center, suggesting that the intrusion of the dry SAL air into the inner region is important for weakening tropical cyclones. In contrast, the nearest intrusion locations in the intensifying cases are generally beyond 360 km of the tropical cyclone center. Second since the dry SAL air usually intrudes into tropical cyclones in a cyclonic pathway (Figures 1 and 2), Figure 3 suggests that the SAL air may first intensify tropical cyclones when it is found in the northwest quadrant and then weaken tropical cyclones when its dry air intrudes into the southwest and southeast quadrants. Considering the difference in radial distance, it is likely that the intensification of tropical cyclones is due to the induced transverse circulation that promotes the upward motion and convection in the tropical cyclones [Karyampudi and Carlson, 1988; Karyampudi et al., 1999; Karyampudi and Pierce, 2002] while the weakening may result from the dry SAL intrusion.
 The differences in the average tropical cyclone parameters such as the maximum sustained surface wind (TCVMAX), the center latitude (TCLAT) and longitude (TCLON) and the average SAL parameters including the closest latitude (SALLAT) and longitude (SALLON) of the intruded SAL dry air and the distance of the intrusion location to tropical cyclone center (SALDIS) are further examined. Table 1 shows the characteristics for intensifying and weakening cases. Note that 26.4% (19.8%) of the intensifying (weakening) cases are in the tropical depression stage. The statistical test shows that the differences in the average TCVMAX, TCLAT, SALLAT and SALDIS are significant at the 99% level using a two-sided t test. As shown in Table 1, on average a tropical cyclone of 28.7 m/s centered at 30.7°N tends to weaken in the next 24 hours if the dry SAL air is 381 km from the tropical cyclone center. In contrast, a tropical cyclone of 23.4 m/s centered at 22.0°N tends to intensify in the next 24 hours if the dry SAL air is about 581 km from the tropical cyclone center. That is, weakening tropical cyclones have a stronger initial intensity on average and are located more northward than intensifying tropical cyclones, and the dry SAL air intrudes closer to the tropical cyclone center and is located farther northward in the weakening cases than in the intensifying cases. Considering that the intrusion is occurring mostly in the southwest quadrant, the higher latitude is favorable for weakening cases.
Table 1. Average Characteristics of the SAL Dry Air Intrusion for Intensifying and Weakening Cases, Respectively With Bold Numbers Indicating That the Differences Between the Two Cases Are Statistically Significant at the 99% Level
 The influence of the SAL on the intensity of Atlantic tropical cyclones is investigated through a composite study of intrusions of the dry SAL air with respect to the tropical cyclone center. The nearest intrusion locations in 37 named tropical cyclones that formed during the period 2005–2007 are identified using the relative humidity between 600–700 hPa retrieved from the AIRS suite on the NASA's Aqua satellite. Based on the 24-hour intensity change of these tropical cyclones, 274 selected cases are categorized into intensifying, neutral and weakening cases. The characteristics of the weakening and intensifying cases are computed and compared to address how the SAL can play positive and negative roles in tropical cyclone development.
 In the case that the SAL is initially located to the north of tropical cyclones or tropical cyclones track northward, usually the dry SAL is first observed within 1000 km of the tropical cyclone center, steered southward with the cyclonic flow, and drawn toward the tropical cyclone center. This study suggests that the SAL can impact tropical cyclone intensity in both favorably and unfavorably manners during the intrusion of the dry air toward the tropical cyclone center. The SAL air may first intensify tropical cyclones when it is found in the northwest quadrant and then weaken tropical cyclones when its dry air intrudes into the southwest quadrant within 360 km of the tropical cyclone center. Weakening (intensifying) tropical cyclones have a strong (weak) average intensity and are located relatively northward (southward) latitude, suggesting that the SAL tends to be favorable for the initial development of tropical cyclones but unfavorable for their subsequent intensification.
 Note that in this study we only investigated the humidity aspect of tropical cyclone intensity change. As mentioned in section 1, the vertical wind shear and static stability associated with the SAL activity can affect tropical cyclone intensity. These aspects should be addressed in future study. In addition, uncertainty should be considered in the AIRS/Aqua relative humidity profiles.
 The authors thank Ramesh Kakar (NASA HQ) for his support through the NASA EOS and NAMMA-06 projects. This study is also supported by the National Science Foundation of China (NSFC 40875038). The authors would like to thank two anonymous reviewers for their valuable comments.