-
Considering the uncertainty in the historical records of TC intensity (Landsea et al., 2006; Knutson et al., 2010; Kossin et al., 2013), the basin-wide intensity is measured by the annual count of the most intense hurricanes (categories 4 and 5 on the Saffir-Simpson scale) (Webster et al., 2005; Wu, 2007; Wu and Wang, 2008). For convenience, the most intense hurricanes (categories 4 and 5) are called intense hurricanes in this study. Wu and Zhao (2012) suggested that the frequency of intense hurricanes is more sensitive to changes in large-scale parameters than other intensity indices. Since the annual counts of intense hurricanes depend on the annual tropical cyclone formation frequency, the proportion of intense hurricanes is also used in this study. The Atlantic TC track data are from the second-generation Atlantic hurricane database (HURDAT2) maintained by the National Oceanic and Atmospheric Administration (NOAA) National Hurricane Center (Landsea and Franklin, 2013).
The large-scale atmosphere/ocean environmental parameters are from the following datasets. The SST data are from the NOAA extended reconstructed SST (ERSST version 5) data with 2° latitude by 2° longitude resolution (Huang et al., 2017). The ocean MLD data are derived from the ocean reanalysis data from the ECMWF Ocean Reanalysis System 4 (ORAS4) with a resolution of 1° by 1° (Balmaseda et al., 2013). Effort has been made to establish reliable MLD calculation methods (e.g., Kara et al., 2000; Huang et al., 2018). In this study, we use the definition of the depth where the temperature is 0.5°C less than the SST (Price et al., 1986; Kelly and Qiu, 1995; Monterey and Levitus, 1997). The OFT is represented by the tropopause temperature because the tropopause can be taken as the cloud top. The tropopause is defined by where the temperature lapse rate becomes greater than 2 K km−1, and the tropopause temperature is obtained from the National Centers for Environmental Prediction/National Center for Atmospheric Research Reanalysis (NCEP/NCAR) with a resolution of 2.5° latitude by 2.5° longitude (Kalnay et al., 1996). The VWS, which is defined as the magnitude of the vector difference of the horizontal winds between 850 hPa and 200 hPa, is derived from the NCEP/NCAR reanalysis. The Niño-3.4 index is obtained from the National Center for Atmospheric Research/University Corporation for Atmospheric Research Reanalysis (NCAR/UCAR; Rasmusson and Carpenter, 1982).
The numerical experiments in the study are conducted with the TC intensity model, which is adopted from Emanuel et al. (2008). It is an axisymmetric numerical atmospheric model, coupled with a simple one-dimensional ocean model. The model is run along the observed track for each TC. In addition to the observed track, the model input includes four large-scale environmental parameters: SST, OFT, MLD, and VWS. The model is initialized with a warm-core cyclonic vortex with a maximum wind speed of 21 m s−1 because the model vortex weakens at the beginning of the simulation (Wu et al., 2018). The currently available water vapor dataset is of low confidence, so the effect of water vapor change on TC intensity is not considered, and the environmental relative humidity in the middle troposphere and boundary layer are constant (45% and 80%, respectively). The other model parameters are the same as those in Emanuel et al. (2008).
The numerical experiments are carried out for the peak season (August, September, and October). Table 1 lists all the intensity experiments in this study. E1 is designed to examine the capability of the intensity model. The monthly mean environmental parameters are used for each year. E2 is designed to examine the influence of the track changes. In E2, there are no temporary changes in the four environmental parameters, which are averaged for each month over the 60-year period (1958–2017). Four sensitivity experiments are designed to examine the contributions of the four large-scale environmental parameters. In these experiments, we hold one parameter constant averaged for each month over the 60-year period (1958–2017), while the other parameters are the same as in E1. Specifically, the MLD (SST, OFT, and VWS) is fixed in SE1 (SE2, SE3, and SE4).
Experiments Description E1 Control experiment: Monthly environmental parameters and observed TC tracks are used during 1958–2017. E2 Track effect: Same as E1, but the long-term (1958–2017) monthly mean environmental parameters are used. SE1 MLD effect: Same as E1, but the long-term (1958–2017) monthly mean MLD is used. SE2 SST effect: Same as E1, but the long-term (1958–2017) monthly mean SST is used. SE3 OFT effect: Same as E1, but the long-term (1958–2017) monthly mean OFT is used. SE4 VWS effect: Same as E1, but the long-term (1958–2017) monthly mean VWS is used. Table 1. Summary of intensity experiments.
For comparison, we categorize intense hurricane activity over the period 1958–2017 as the active and inactive years. The interannual variations are first obtained by removing the 5-year running average. Active (inactive) years are defined by when the amplitude of the interannual variations of the frequency of intense hurricanes is larger (smaller) than 0.5 (–0.5) standard deviation. The 18 active years are 1961, 1964, 1971, 1974, 1978, 1979, 1985, 1988, 1989, 1992, 1995, 1996, 1999, 2004, 2005, 2008, 2010, and 2011, and the 21 inactive years are 1960, 1962, 1963, 1965, 1968, 1973, 1976, 1983, 1986, 1987, 1990, 1993, 1994, 1997, 2001, 2002, 2006, 2009, 2012, 2013, and 2015. There are 46 (7) intense hurricanes in the selected active (inactive) years.
Note that the inactive years include El Niño years such as 1965, 1983, 1987, 1997, 2002, 2009, and 2015, while the active years include La Niña years such as 1988, 2008, and 2010. However, there are exceptions. The inactive year of 1973 corresponds to a La Niña year, and the active year of 1992 corresponds to a typical El Niño year. In fact, the correlation between the frequency of intense hurricanes and the Niño-3.4 index is –0.51. Murakami et al. (2018) found that the correlation coefficient between the Niño-3.4 index and the observed major hurricane frequency for the period 1979–2017 is −0.45. Using a suite of high-resolution model experiments, they found that the high number of 2017 major hurricanes was not primarily caused by La Niña conditions in the Pacific Ocean.
-
A TC track consists of a starting point (formation location) and subsequent movement. While movement is mainly controlled by large-scale steering and β-drift (Holland, 1983; Wu and Wang, 2004), TC tracks are also influenced by their own formation location, which can lead to changes in the duration of TC intensification and environmental parameters experienced by TCs. In this study, the duration of TC intensification is defined as the period between the formation and the time when a TC reaches its lifetime maximum intensity. Figure 7 shows the formation locations of intense hurricanes during active and inactive years. The formation locations for intense hurricanes can be roughly divided by the longitude of 60°W. The east region covers the tropical NA, while the west region mainly includes the Gulf of Mexico and Caribbean Sea. A total of 53 intense hurricanes formed in the east region, accounting for 65% of the total intense hurricanes in the NA basin. Figure 7 also shows the difference of TC formation frequency (contours) between the active and inactive years, which is counted in each 2.5° × 2.5° box. A positive difference indicates enhanced TC formation during active years. We can see that the east and west regions correspond to two regions with enhanced TC formation, respectively.
Figure 7. The difference of TC formation frequency (times a factor of 10) between the active and inactive years. Red (blue) dots indicate the formation location of intense hurricanes in the active (inactive) years. The unit for the TC formation frequency is the number per season (August–October) over a 2.5° × 2.5° box.
Despite enhanced TC formation, especially in the east region, it should be noted that the active years are mainly signified by the relatively high proportion of intense hurricanes compared to all TCs. In the active (inactive) years, 23.6% (4.6%) of TCs intensified into intense hurricanes. This is also indicated by the correlation between the annual frequency and the proportion of intense hurricanes (Fig. 1), which is 0.92. This suggests that the variations of the proportion of intense hurricanes can explain 85% of the variance of the frequency of intense hurricanes. In the east region, 34.0% (4.8%) of TCs intensified into intense hurricanes in the active (inactive) years. In the west region, 15.0% (4.3%) of TCs intensified into intense hurricanes in the active (inactive) years. The correlations between the annual frequency and the proportion of intense hurricanes are 0.84 and 0.81 for the west and east regions, respectively. We can see that TCs have a much greater chance of becoming intense hurricanes during active years. This strongly suggests that the enhanced activity of intense hurricanes results from factors other than the annual formation frequency in the NA.
As discussed in the last section, the importance of track changes relative to the collective contribution from environmental parameters can be evaluated by comparing the results in E1 and E2. In the east region, the correlation of the frequency of intense hurricanes between E1 and E2 is 0.81, indicating that track changes can account for 66% of the variance of the interannual variability of intense hurricanes. In the west region, the corresponding correlation is 0.68, and track changes can explain 46% of the variance of the interannual variability of intense hurricanes. Consistent with the discussion in section 4, TC track changes play an important role in regulating the interannual variations of intense hurricane activity in the west and east regions.
Track changes include changes in translation speed, duration of intensification, and environmental parameters. These environmental parameters in E2 caused by TC tracks are calculated for all TCs and averaged for the active and inactive years (Table 2), respectively. By comparing these parameters in E2, we can further understand how track changes affect intense hurricane activity. In the west region, track changes do not lead to significant changes in environmental parameters, although the increases in MLD (2.3 m) and SST (0.21°C) and the decreases in OFT (−0.44°C) and VWS (−0.08 m s−1) are favorable for TC intensification. On the other hand, during active years, TCs move faster, and the mean duration is shorter. The decreasing duration of intensification is unfavorable for TC intensification. The decreasing duration of intensification is mainly due to the westward shift of the mean formation longitude, which significantly decreases by 3.79° during active years.
(a) East region Parameters Active years Inactive years Difference MLD (°C) 37.37 36.31 1.06 SST (°C) 27.57 27.25 0.32 OFT (°C) –73.06 –72.88 –0.18 VWS (m s−1) 9.20 9.12 0.08 Translation speed (m s−1) 6.27 6.00 0.28 Mean duration (d) 4.20 2.67 1.53 (b) West region Parameters Active years Inactive years Difference MLD (°C) 42.03 39.73 2.3 SST (°C) 28.3 27.97 0.21 OFT (°C) –72.30 –71.86 –0.44 VWS (m s−1) 9.87 9.95 –0.08 Translation speed (m s−1) 8.09 6.27 1.82 Mean duration (d) 2.17 2.55 –0.38 Table 2. Differences of the environmental parameters caused by track change (used in E2) between the active and inactive years in the east and west region. The differences at the 95% confidence level are in bold.
In the east region, almost all changes of environmental parameters are favorable for TC intensification, but the changes cannot pass the significant test at the 95% confidence level. As indicated in Table 2, the only significant difference occurs for the mean duration time, which increases by 1.53 days, or 36.7 hours. Note that the elongated duration is not due to the slow-down of TC translation. In fact, the mean translation speed is faster during active years. We examined the mean distance between the formation location and the location of maximum intensity for all TCs in the east region. The TCs travel 2048 (1310) km to reach their maximum intensity during active (inactive) years. The difference of 738 km is statistically significant. This suggests that TCs during active years travel longer distances and have longer chances for intensification.
What makes TCs during active years travel longer distances in the east region? Figure 8 shows the frequency of TC occurrence for TCs formed in the east region and the prevailing track indicated by the frequency maximum. During active years, the TCs take a westward prevailing track, which allows them to have longer chances for intensification. As shown in Fig. 3, TCs generally intensify south of 30°N. On the other hand, the TCs during inactive years take a recurving prevailing track, leading to a shorter duration of intensification.
Figure 8. The frequency of TC occurrence (contours) for TCs in the east region and the locations (dots) where TCs first reach category 4. The arrowed curves schematically show the prevailing track for (a) active and (b) inactive years. The unit for the TC occurrence frequency is the number per season (August–October) over a 2.5° × 2.5° box.
We can further demonstrate that the different prevailing tracks mainly result from the difference of formation locations between the active and inactive years. The prevailing track formation location shifts southeastward by 1.54° in longitude and 0.55° in latitude for active years, compared to inactive years. Although it is not statistically significant at the 95% confidence level, the southeastward shift in prevailing track formation location can allow TCs to travel at lower latitudes and have more time to develop. To demonstrate this, we use the TC track model developed by Wu and Wang (2004). In the track model, a TC is taken as a point vortex, and its track is a function of the translation speed and formation location. The input of the model includes the formation locations and translation speeds of TCs. We can use the track model to examine the contributions of changes in the formation locations and translation speeds, respectively. The large-scale steering flow is defined as the mean flow between 850 hPa and 300 hPa. Following Wu and Wang (2004) and Wu et al. (2005), the climatological beta drift is used in this study. In the control run, the tracks are simulated with the monthly mean steering flow and the observed formation locations for each TC. Compared with the observations (Fig. 9a), Fig. 9b shows the difference of the frequency of TC occurrence between the active and inactive years in the control run. The simulated difference pattern is very similar to the observations. Two sensitivity experiments are conducted with the track model. In the two experiments, we use the same steering flow averaged for both the active and inactive years, while the formation locations are from the active and inactive years, respectively. Figure 9c shows the difference of the frequency of TC occurrence between the active and inactive years in the sensitivity experiments. This suggests that the track differences between the active and inactive years in the east region are mainly due to the formation location differences.
Figure 9. The observed (a) and simulated (b) difference of occurrence frequency between the active and inactive years for TCs that formed in the east region, and (c) the effect of TC formation location changes in the east region. The unit for the TC occurrence frequency is the number per season (August–October) over a 2.5° × 2.5° box.
Experiments | Description |
E1 | Control experiment: Monthly environmental parameters and observed TC tracks are used during 1958–2017. |
E2 | Track effect: Same as E1, but the long-term (1958–2017) monthly mean environmental parameters are used. |
SE1 | MLD effect: Same as E1, but the long-term (1958–2017) monthly mean MLD is used. |
SE2 | SST effect: Same as E1, but the long-term (1958–2017) monthly mean SST is used. |
SE3 | OFT effect: Same as E1, but the long-term (1958–2017) monthly mean OFT is used. |
SE4 | VWS effect: Same as E1, but the long-term (1958–2017) monthly mean VWS is used. |