A method of downscaling that isolates the effect of temperature and moisture changes on tropical cyclone (TC) activity was presented in Part I of this study. By applying thermodynamic modifications to analyzed initial and boundary conditions from past TC seasons, initial disturbances and the strength of synoptic-scale vertical wind shear are preserved in future simulations. This experimental design allows comparison of TC genesis events in the same synoptic setting, but in current and future thermodynamic environments. Simulations of both an active (September 2005) and inactive (September 2009) portion of past hurricane seasons are presented.
An ensemble of high-resolution simulations projects reductions in ensemble-average TC counts between 18% and 24%, consistent with previous studies. Robust decreases in TC and hurricane counts are simulated with 18- and 6-km grid lengths, for both active and inactive periods. Physical processes responsible for reduced activity are examined through comparison of monthly and spatially averaged genesis-relevant parameters, as well as case studies of development of corresponding initial disturbances in current and future thermodynamic conditions. These case studies show that reductions in TC counts are due to the presence of incipient disturbances in marginal moisture environments, where increases in the moist entropy saturation deficits in future conditions preclude genesis for some disturbances. Increased convective inhibition and reduced vertical velocity are also found in the future environment. It is concluded that a robust decrease in TC frequency can result from thermodynamic changes alone, without modification of vertical wind shear or the number of incipient disturbances.
1. Introduction and background
In Part I of this study (Mallard et al. 2013, hereafter Part I), the capability of the nonhydrostatic Weather Research and Forecasting Model (WRF) to reproduce tropical cyclone (TC) activity during a portion of past hurricane seasons is examined. Part I also presents an approach for isolating the effects of temperature and moisture changes on regional hurricane activity in which past seasons are simulated with unmodified initial and lateral boundary conditions, and again with thermodynamic changes included. Replicating a past season with projected thermodynamic changes removes sensitivity to changes in vertical wind shear and the number of incipient disturbances, allowing direct comparison of the evolution of nearly identical initial disturbances in the same synoptic setting, but with modification of the large-scale thermodynamic environment.
Temperature and moisture changes consistent with projections of the atmosphere at the end of the twenty-first century, as derived from an ensemble of general circulation models (GCMs) run for the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) with the A1B scenario, are applied to the operational analyses that were used as initial and boundary conditions for high-resolution WRF simulations (Fig. 6 in Part I). Temperature changes are included as a function of pressure alone; therefore, the horizontal temperature gradient and vertical wind shear are not altered. Sea surface temperatures (SSTs) are also uniformly warmed. Moisture changes are accounted for by keeping relative humidity (RH) constant and recomputing the vapor content at the modified temperature. The aim of this method is to isolate differences in future TC activity based on modification of the thermodynamic profiles, apart from changes in vertical wind shear or incipient disturbances. While it is acknowledged that shear changes are likely to be important (e.g., Gualdi et al. 2008; Garner et al. 2009), the method utilized here allows separation of this effect. The goals of this study are to determine the extent of TC frequency changes due to thermodynamic changes alone and to identify which physical processes are involved.
A small physics ensemble is generated by varying the microphysical and boundary layer (BL) parameterizations in order to provide a more robust signal. The simulations are run for one month with a domain covering the North Atlantic basin (Fig. 1 in Part I). Both the Morrison et al. (2009) double-moment scheme and the WRF single-moment six-class microphysics scheme (WSM6; Hong and Lim 2006) are utilized, along with the Yonsei University (Hong et al. 2006) and the Mellor–Yamada–Janjic (Janjić 1994, 2002) BL schemes, to produce a four-member ensemble (Table 3 in Part I). The 1° National Centers for Environmental Prediction (NCEP) Global Forecast System Final Analysis (GFS FNL) is used as initial and boundary conditions, and SSTs are taken from the 0.5° Real-Time Global (RTG) SST analysis (Thiébaux et al. 2003). Ensembles are run with both 18- and 6-km grid length for present-day and modified initial and boundary conditions to determine whether the resulting changes in future TC frequency are sensitive to resolution (Table 1 in Part I). The seasonal peak of the hurricane season, September, is simulated both during an active (2005) and inactive (2009) year in order to assess whether TC activity will be more or less affected by climate change when comparing downscaled simulations from time periods of high or low hurricane activity. Additional information about the experimental setup is provided in Part I.
2. Previous work
As described in a review article by Knutson et al. (2010), a consensus of past studies has projected a decrease in TC frequency of 6% to 34% by the end of the century resulting from anthropogenic climate change. However, Knutson et al. note a large degree of uncertainty in future TC changes, as findings vary from study to study and even from basin to basin. Another review work by Grossmann and Morgan (2011) links this variability to uncertainties in both the thermodynamic and dynamic regional responses to climate change.
The downscaling work of Garner et al. (2009) assessed the impact of thermodynamic changes on TC counts in the late twenty-first century by perturbing their control experiments with expected atmospheric and sea surface temperature changes alone (with fixed RH) and then contrasting these simulations with runs where future changes in velocity fields were also included. Simulations with both thermodynamic and dynamic changes resulted in a significant reduction of TC frequency, and simulations with only the thermodynamic anomalies projected a smaller decrease in future storm counts. Garner et al. highlight the sensitivity of future TC counts to strengthening vertical wind shear. In their experimental setup, future thermodynamic changes alone are found to exert a negligible influence on TC frequency. Prior studies also have concluded that hurricanes in a future environment will be more vulnerable to the detrimental effects of wind shear (Nolan and Rappin 2008; Rappin et al. 2010, hereafter R10). In idealized simulations of a single TC in a warmed environment, Nolan and Rappin find that increases in freezing level height lead to a more vertically elongated potential vorticity tower, making the vortex more vulnerable to decoupling via vertical wind shear.
Past modeling work links projected increases in static stability to reduced TC counts (e.g., Sugi et al. 2002; Yoshimura et al. 2006; Oouchi et al. 2006; Bengtsson et al. 2007; Gualdi et al. 2008). As shown in Fig. 6 of Part I, the temperature changes included in future simulations in the present study are amplified with height throughout the troposphere. Therefore, increases in static stability are also anticipated to affect TCs in the present study, consistent with the previous results of Shen et al. (2000), Hill and Lackmann (2011), and others.
Prior studies have concluded that future decreases in TC counts result from a weakening of the tropical circulation and the corresponding reduction in vertical velocity (e.g., Bengtsson et al. 2007; Lavendar and Walsh 2011). The idealized work of Betts (1998) shows a weakening of vertical motion out of the BL as a consequence of warming SST in an environment of radiative–convective equilibrium. The rate of moisture increase in the BL as the ocean warms outpaces the increase in surface evaporation. Therefore, a decrease in mass flux out of the BL is necessary. A similar conclusion is drawn by Held and Soden (2006), when examining how latent heat release (and therefore convection and precipitation) is balanced by tropospheric radiative cooling in the global mean. As the rate of precipitation increase in a warmer climate is constrained more tightly than is the increase in water vapor, vapor transport out of the BL must decrease to compensate. The implication is that thermodynamic changes, such as those examined here, could have the effect of decreased vertical motion in the tropics, which could impact the vigor of incipient tropical disturbances. However, it is questionable whether a limited area model domain integrated over a month, as used in the present study, would establish such an equilibrium.
Past GCM studies have shown reduced TC counts as a consequence of the radiative effects of increased CO2 (e.g., Yoshimura and Sugi 2005; Held and Zhao 2011; Sugi et al. 2012). In these experiments, the future environment was composed by increasing SST and CO2, both together and then independently. While the largest reduction in TC frequency is simulated when both SST and CO2 are increased, higher CO2 is shown to produce a significant reduction in storm counts with fixed SST (10% globally, as shown in Held and Zhao with a doubling of CO2). Sugi et al. (2012) find that precipitation decreases in a high-CO2 environment are required to maintain balance with reduced radiative cooling. Consequently, upward mass flux is reduced and TC genesis is suppressed. Alternatively, if SST is increased (with greenhouse gases held fixed), increased moisture and static stability result in a reduction of upward motion. Held and Zhao (2011) propose that decreased upward mass flux in the future environment leads to suppressed TC genesis because reduced upward motion makes it easier for the advection of dry air in the free troposphere to prohibit genesis. It should be noted that, in the present work, CO2 is not explicitly increased (section 4 of Part I). Water vapor is the only greenhouse gas that is increased in the future simulations, although the imposed thermodynamic changes in the future environment are derived from GCMs that do account for higher concentrations of CO2.
Emanuel et al. (2008, hereafter E08) and R10 emphasize the importance of increased saturation deficits as an explanation for future decreases in TC counts. The earlier work of Emanuel (1989, 1995) discusses the importance of near saturated conditions (with respect to moist entropy) in the troposphere in order for TC genesis to occur. However, larger saturation deficits are expected in an environment of increased vapor with RH held constant, because the difference between any two variables will become larger if their ratio is held constant while either variable is increased. As specific humidity saturation deficits increase, so do moist entropy saturation deficits (E08; R10).
R10 study the effect of stronger surface winds and warmer SST on a preexisting disturbance in an idealized environment of radiative–convective equilibrium and imposed shear. A dimensionless incubation parameter is defined1 as
where smid and sb are the moist entropies of the midtroposphere (taken at the 600-hPa level) and the BL, respectively; is the moist entropy of the sea surface at saturation. Moist entropy is defined as in Emanuel (1994) and R10. The derivation of χ is shown in Emanuel (1995), based on assumption of BL quasi-equilibrium. As in R10, the numerator of (1) can also be expressed as the midlevel moist entropy saturation deficit, taken at 600 hPa (hereafter χmid):
Above, sb is considered to be an approximation for . In this conceptualization of the pregenesis environment, the atmosphere is assumed to be neutral to moist convection. Therefore, a parcel lifted from the boundary layer will have the same moist entropy as a saturated parcel in the midtroposphere, as further discussed in E08 (p. 363) and R10 (p. 1955). The denominator of the incubation parameter in (1) is also given its own designation, being referred to as χflux. Note that, unlike the dimensionless χ, both χmid and χflux are in units of J kg−1 K−1.
The incubation parameter χ is shown by R10 to be well correlated with the time needed for an incipient vortex to intensify into a TC. In a near-saturated column (with smaller χmid), surface fluxes of heat and moisture (corresponding with large χflux) can serve to intensify an incipient vortex, via air–sea interaction, instead of further moistening the troposphere. Therefore, as χ decreases, genesis is more likely. An increase in χmid would lengthen the time scale for genesis, lessening the likelihood of genesis, and result in decreased TC frequency.
The idealized study of R10 demonstrates how χ changes in a warmer SST environment, building on the work of Emanuel (1989, 1995). In environments where SST is higher, it is found that TC genesis is slowed and larger values of χ occur due to warming of the column and corresponding increases in χmid. R10 conclude that, in an environment of warmer SST, increased χmid (and larger incubation parameters) slowed or prevented TC genesis. The downscaling study of E08 projected a global reduction in TC counts by the end of the twenty-second century in a warmer atmosphere with increased χmid. However, when saturated specific humidity was held constant, an overall increase in global tropical cyclogenesis was simulated. This result indicates that drying (with respect to the moist entropy saturation deficit) impacts future TC genesis negatively, through the thermodynamic component of climate change.
3. Future storm counts
In the following section, we compare ensembles of monthly, basin-scale simulations run with present-day initial and boundary conditions and then run again with future temperature and moisture changes included. All physics ensembles are referred to with a designation listed in Table 1 of Part I that denotes the grid spacing and time period of the simulation and whether its initial and boundary conditions are unmodified (“current”) or GCM-derived thermodynamic changes are included (“warming”).
Each ensemble shows a robust decrease in overall future storm counts. In the case of the 18-km simulations,2 the ensemble-averaged number of named storms decreases by 28%; hurricanes decrease by 33% and major storms by 21% in the future compared with the current ensemble (Table 1). These decreases are all found to be statistically significant.3 When comparing the finer-resolution simulations (6Current05 and 6Warming05), a drop in storm frequency is again seen in all of the previously mentioned categories (Table 2). In the future, named storm, hurricane, and major hurricane frequency is reduced by 18%, 19%, and 17%, respectively. The decrease in storm frequency is smaller when comparing this set of higher-resolution ensembles relative to the decrease in storm counts shown by the 18-km simulations. In the 6-km runs for 2005, the differences in hurricane and major hurricane counts fail to pass the test for statistical significance. Percentage decreases in TC frequency are more drastic when comparing simulations during the inactive period of September 2009, as in the 6Current09 and 6Warming09 ensembles, as there are fewer TCs in the control experiment. In this case, the ensemble average of named storms decreases by 26%, or 1.25 storms, in the future; and the number of hurricanes decreases by 75%, or 1.5 storms (Table 3). Both the decreases in named storms and hurricanes do meet the previously stated statistical criterion for significance. The frequency of major storms show no change in the 6Warming09 ensemble mean compared with the 6Current09 ensemble.
Overall, a robust and statistically significant decrease in the number of named storms is predicted for the end of the century by all three sets of simulations. Two of the three sets of ensembles show a significant decrease in the number of hurricanes, with the 6Warming05 ensemble projecting a decrease that did not qualify as statistically significant. However, major TCs consistently are predicted to decrease in frequency the least, and are not projected to change in frequency for the 2009 set of simulations, despite the consistent reduction in the overall number of named storms and hurricanes. Past studies (e.g., Oouchi et al. 2006; Bengtsson et al. 2007; Knutson et al. 2008; Bender et al. 2010) have found an increase in the frequency of the most intense storms while still projecting an overall drop in storm counts.
4. Comparison of current and future TC genesis
Locations of TC genesis are concentrated in the southeastern tropical Atlantic in these simulations (Fig. 4 of Part I). Kossin et al. (2010) find the southeastern part of the basin to be a favored location for genesis during the peak months of the season. To assess what factors could be involved in the predicted decrease in future TC counts, relevant parameters are examined within a “genesis region” defined as a box covering latitudes 5°–20°N and longitudes 40°–15°W.
a. Averaged conditions
Monthly and ensemble averages within the genesis region are used to investigate the possibility of environmental changes that could explain the decrease in TC activity (Table 4). As also found by prior studies (e.g., Knutson and Tuleya 2004; Gualdi et al. 2008), convective available potential energy (CAPE) is enhanced in the future simulations by approximately 10%. An increase in CAPE is expected from the imposed changes on water vapor (Hill and Lackmann 2011). Genesis potential (as defined in Emanuel and Nolan 2004) also increases in each future ensemble, although with more variability. Therefore, the simulated decrease in future TC counts cannot be accounted for by changes in these parameters. The presence of more storms in the genesis region in the current ensemble results in small differences in the average 850–200-hPa vertical wind shear for two of the three sets of ensembles. As discussed in Part I, while momentum fields are left intact in the initial and boundary conditions, the evolving synoptic pattern in the interior, including TCs, can modify interior wind fields. As these quantities are either unchanged or do not change in a manner that could explain the simulated reduction in future TC counts, the following discussion will focus on those parameters and processes that could account for such a decrease.
In the warming simulations, systematically higher values of convective inhibition (CIN) are found, as well as decreased midtropospheric vertical velocity4 (Table 4). Enhanced values of CIN are expected in the future simulations, as the temperature changes included in future simulations feature warming that increases with height throughout the troposphere (Fig. 6 of Part I). Decreased vertical velocity is also anticipated in a future environment of increased SST and water vapor, as concluded by prior work (Betts 1998; Held and Soden 2006; Held and Zhao 2011) and discussed above. However, the TCs themselves also enhance the mean upward vertical motion in and downstream of the genesis region, resulting in stronger ascent in the simulation with more TCs (the current simulation in our experiments). It would be difficult to exclude the vertical motion contribution of TCs due to the intense updrafts and broad compensating subsidence, and we have not attempted this here. The use of ensemble and monthly averaged ω does lessen the influence of individual storms on the comparison. Additionally, while TCs do form in the genesis region, the storm tracks of strongly developed TCs with vigorous updrafts are located farther downstream (west) of the genesis region.
Within the future genesis environment, monthly and ensemble averages show larger χmid and longer incubation parameter values (Table 4). The parameter χflux is also enhanced somewhat, but, in each ensemble these increases are outpaced by larger χmid increases and incubation values are greater in the warming runs than in the current simulations. These results are consistent with R10, where warmed SSTs resulted in similar changes to χflux, χmid, and the incubation parameter.
As discussed above, larger saturation deficits are expected in a warmer environment with increased vapor and constant RH. Specific humidity saturation deficits are increased throughout the troposphere within the future simulations discussed here [as shown in Mallard (2011), section 220.127.116.11]. As a result of increased specific humidity saturation deficits, χmid is also increased5 and larger values of the incubation parameter are present in the future environment. Increased saturation deficits are linked to a lengthening of the time until genesis and a decrease in TC counts (R10; E08).
b. Case studies
A unique aspect of our experimental design is that it allows direct comparison of highly similar initial disturbances (IDs) in the same synoptic setting, but with different thermodynamic environments. Numerous instances are available in each ensemble pair of current and future simulations where corresponding IDs enter the genesis region defined above. In some cases, the ID undergoes genesis in the current simulation but fails to do so in the corresponding future simulation. These comparisons provide insight into the physical processes that are acting to limit development in the future environment. Case studies comparing developing versus nondeveloping IDs in the current and warming runs, respectively, are referred to as “D/ND” cases. In some instances, the corresponding ID undergoes genesis in both simulations at approximately the same time. These cases, referred to as “D/D” cases, are examined to contrast with circumstances in which the warming ID fails to develop. Higher-resolution, 6-km simulations are used exclusively for all six case studies (four D/ND cases and two D/D cases). The cases emphasized here were selected as representative examples of the larger body of cases.
1) Developing/non-developing cases
The representative D/ND case (referred to as case 1) takes place over a 5-day period early in the 6Current05 and 6Warming05 simulations with identical physics options. The IDs are broad low pressure systems with some associated convection, and both encounter dry, low RH environments as they move westward (Figs. 1a,c and 2). While the RH environments of both cases look similarly dry downstream of the ID, the use of χmid to characterize the moisture of the surrounding environment shows significantly larger saturation deficits to the west of the ID in the future simulation (Fig. 3). The environment in proximity to the low pressure center also shows larger midlevel saturation deficits in the warming run.
By 1200 UTC 10 September, the current ID has continued to develop as it moves westward; however, the warming ID has not deepened and much of the convection associated with it is displaced southward from the low pressure center (Figs. 1b,d). While dry air is still downstream in both cases, large RH (low χmid) values are still present in proximity to the current ID, consistent with increased moisture convergence as the disturbance intensifies (Figs. 2b and 3b). Genesis in the 6Current05 run occurs at 0000 UTC 11 September. However, the ID in the warming simulation does not intensify and continues to show persistent dry air near the low pressure center throughout the duration of the study period (Figs. 1d, 2d, and 3d). Consistent with the differences in χmid, χ is greater for the warming case (Fig. 4). If averaged in the genesis region over the 5-day period of the case study, χ is 16% greater in the warming simulation than in the current simulation.
Vertical profiles of the moist entropy saturation deficit, spatially averaged within 300 km of the ID center, show increased dryness throughout much of the troposphere in the warming simulation, relative to the current run (Fig. 5). In both simulations, the largest deficits appear early in the period, during 8 and 9 September. The current simulation shows reduced midlevel saturation deficits in the latter half of the period, and more uniform values of between 12 and 16 J kg−1 K−1 throughout much of the troposphere. In the warming run, larger moist entropy saturation deficits of 14 to 22 J kg−1 K−1 persist in the midtroposphere, above approximately 850 hPa, throughout the period.
Emanuel (1995) states that the suppression of subsaturated downdrafts (through closing midlevel saturation deficits) is a necessary precondition for TC genesis and that the negative impact of these downdrafts must be suppressed in order for enthalpy extracted from the sea surface to be used for intensification of the vortex. Therefore, as χmid is reduced, so is the incubation parameter and genesis becomes more likely. Larger midlevel saturation deficits in the warming environment would be expected to increase the impact of subsaturated downdrafts, making it more difficult to achieve saturation of the column. As shown in Fig. 6a, near-storm mid- and low-level downdrafts6 show an increased frequency of high saturation deficits in the warming run, with 53% of downdrafts having moist entropy saturation deficits greater than 30 J kg−1 K−1 compared with only 18% in the current simulation. The average saturation deficit in downdraft grid cells is increased by 56% in the warming run, relative to the current simulation, during the period plotted in Fig. 6a. Therefore, the future environment features mid- and lower-tropospheric downdrafts with enhanced saturation deficits.
During the latter part of the period, both of the moist entropy saturation deficit profiles shown in Fig. 5 exhibit reduced deficits in the lower levels with conditions closer to saturation. As noted in the previous section, χflux is somewhat enhanced in the future. However, despite moist conditions at lower levels, larger deficits at the midlevels remain present in the warming simulation. In addition to enhanced χflux and larger tropospheric saturation deficits, increased CIN and decreased ascent are also consequences of the imposed thermodynamic changes (Table 4). Therefore, the advection of moist entropy out of the BL is reduced in the future environment. Averaged during the period shown in Fig. 5, 950–800-hPa layer-averaged vertical advection of moist entropy within a radius of 200 km of the ID center is decreased by 74% in the warming simulation. When the same comparison is made over a larger area, the differences in vertical entropy advection are even more significant. If taken within 300 km of the ID center, the mean vertical moist entropy advection is negative and its absolute value is an order of magnitude larger in the warming run, compared with the current simulation. Reduced ascent out of the BL suppresses the ability of the future environment to moisten the overlying troposphere. However, the intensification of the vortex itself in the current environment would also act to modify the transport of moisture out of the BL, complicating the interpretation of our calculations. It is found in this case study that both an increased saturation deficit and decreased convective moistening are present in the future relative to the current environment.
To show consistency across all four D/ND case studies, the moisture environments in D/ND cases 2–4 are summarized here and compared with case 1. Further description of all case studies can be found in Mallard (2011, section 4.6.2). In Fig. 7, χmid from both the current and future runs is shown for each case at the time that genesis occurs in the current run. Similar to case 1, each column in Fig. 7 depicts larger saturation deficits in proximity to the ID in the warming runs. In cases 2 and 4, drier χmid values are shown wrapping into the western portion of the disturbance. Case 3 depicts both current and future IDs moving into a broad area of increased saturation deficits in the western portion of the genesis region. In each D/ND case study, dry air is present in both current and future runs; however, the moisture environment of the warming run is less favorable for development as saturation deficits are increased relative to the current simulation. If χmid in the genesis box is averaged over the 48 h prior to genesis in the current environment for cases 2–4, saturation deficits in the warming environment are 22% to 39% larger in each of the three case studies briefly discussed here.
In Fig. 8, vertical profiles of the moist entropy saturation deficit are composited for all four D/ND case studies and shown for the 48 h prior to genesis in the current simulation. Consistent with Fig. 5, saturation deficits are systematically larger in the warming runs. A composite histogram of moist entropy saturation deficits also shows that downdrafts are robustly more subsaturated immediately prior to genesis in the warming environment (Fig. 6b). In both case 1 and the composite histograms, the number of downdraft points is not significantly increased in the future. This is consistent with the TC genesis studies of Nolan (2007) and Wang (2012), which highlight the importance of midlevel moistening and near saturation but do not find a decrease in the frequency of downdrafts prior to genesis. In summary, similarly larger moist entropy saturation deficits and drier downdrafts are consistently present in warming simulations during all four D/ND case studies.
2) Developing/developing case
A D/D case is now presented to provide a contrasting example of an environment in which TC genesis occurs in both the present-day and future simulations. The future ID is first identified as a TC on the 25th at 0000 UTC in the 6Warming05 run and the current ID at 1200 UTC the same day in the 6Current05 simulation. Figure 9 shows both IDs in the period proximate to genesis, at 0000 UTC on the 25th and 26th. Both waves are associated with convection throughout the case study period, and they track similarly along a zonally oriented band of high RH (not shown). The environment features reduced midlevel saturation deficits, with small incubation parameter values both up and downstream of the ID (Fig. 10). In contrast to the previous case, the genesis environment of both the current and warming IDs is moist and favorable for genesis in both simulations and genesis does occur in both.
The overall decrease in TC activity in the future simulations is attributable to several events in which current IDs were able to overcome somewhat unfavorable humidity environments, while the corresponding ID in the future simulation did not develop in dry environments of similar RH but larger χmid values. However, cases in which IDs are located in more favorably moist conditions produce TCs in both simulations. Therefore, incipient tropical disturbances in the future environment exhibited a reduced ability to overcome marginal humidity conditions.
Previous studies using a variety of statistical and dynamic methods consistently predict a decrease in the number of TCs in a warming climate (e.g., Bengtsson et al. 2007; Knutson et al. 2008; E08; Knutson et al. 2010; Held and Zhao 2011). However, a variety of mechanisms are proposed to account for the decrease. Here, the influence of climate change on TC frequency is examined using an experimental design that isolates the effect of temperature and moisture changes. This method allows a direct comparison of current and future disturbances that form in the same synoptic setting, but with differing thermodynamic environments.
Both September of 2005 and 2009 are simulated with an ensemble of WRF runs, first with unmodified operational analyses for initial and lateral boundary conditions, and again with GCM-derived end-of-the-century temperature and moisture changes included (see Table 1 and Fig. 6 in Part I). Future changes are derived from an ensemble of GCM simulations using the A1B emission scenario. This experimental design preserves the synoptic pattern, number and strength of initial disturbances, and vertical wind shear between current and future simulations. A robust decrease in TC counts is found in future simulations relative to current runs. Reductions in named storms of 18%–28% are simulated (Tables 1–3). These decreases are within the range of those projected by previous theoretical and modeling work, as summarized in Knutson et al. (2010). Our findings demonstrate that the thermodynamic component of climate change alone can account for a significant decrease in end-of-the-century TC frequency.
The physical processes that contribute to this decrease in future TC frequency are examined through comparison of monthly averages of genesis-relevant parameters in a predefined region where most of the simulated TCs form (Table 4). From these monthly averages, factors that do not favor genesis in the future environment are (i) a reduction in upward vertical motion, (ii) increased CIN, (iii) increase in the midlevel moist entropy saturation deficit [χmid as in (2)], and (iv) increase in incubation parameter [χ as in (1)]. The previous studies of E08 and R10 have shown that the incubation parameter is positively correlated with the time until TC genesis occurs and is proportional to χmid. Emanuel (1989) discusses the importance of the near saturation of the troposphere (small χmid) in the area surrounding the TC core as a necessary condition for intensification of the initial vortex and tropical cyclogenesis. Once saturation of the column occurs, fluxes of heat and moisture from the sea surface can be used for intensification of the incipient disturbance. Without near saturation of the environment proximate to the vortex, surface fluxes are instead used to close the moist entropy saturation deficit in the overlying troposphere rather than intensifying the incipient vortex. In the statistical downscaling work of E08, a decrease in future TC counts is also predicted and this change is linked to increases in the incubation parameter affected by larger values of χmid.
Case studies are drawn from the 6-km simulations when the same ID is present in both the current and future simulations, where the ID does not develop in the warming simulation while the current ID does undergo genesis. Such circumstances tend to occur where both genesis environments are similarly dry with respect to RH and only marginally favorable for genesis (Fig. 2). However, the current ID does intensify and undergo genesis, while the future disturbance does not (Fig. 1). The future ID enters an environment of enhanced χmid (relative to the current simulation) and fails to develop (Figs. 3 and 7). Consistently, saturation deficits are larger in the warming run and the incubation parameter indicates a longer time period until genesis occurs (Table 4, Fig. 4). Low-level vertical advection of moist entropy out of the BL is also reduced in the warming run, relative to the current simulation. Low- and midlevel downdrafts are consistently more subsaturated (with respect to moist entropy) in the area surrounding the warming IDs (Fig. 6). Ultimately, the future simulation fails to produce a TC. In circumstances where both the present-day and future environments produce a TC, both simulations consistently show a favorable moisture environment for development with respect to both RH and χmid (Figs. 9 and 10). Therefore, it is concluded that circumstances of nondevelopment in future simulations, which ultimately result in decreased TC frequency, occur when both environments are only marginally favorable for TC genesis because of the presence of dry air.
Such a mechanism for the suppression of storms would be present in a future environment where temperatures are increased while RH is approximately constant, and both of these conditions are commonly found in GCM projections for the tropical environment at the end of the century (e.g., Allen and Ingram 2002). However, it is notable that current and future TC development occurred similarly in high humidity environments. Therefore, it is expected that thermodynamic changes might not affect the frequency of TC genesis as significantly in a moisture-rich region of the tropics where dryness does not typically hamper development. Owing to the presence of the Sahara Desert and frequent dry-air advection across the tropical North Atlantic, we speculate that this mechanism of future TC suppression could be especially important in the North Atlantic basin relative to others. It is unknown whether the presence of larger saturation deficits would be of primary importance in suppressing future TC genesis in an environment of increased wind shear. However, it has been shown in the present work that significant decreases in storm counts can result from thermodynamic changes alone, without modification of the vertical wind shear.
Past studies (Held and Zhao 2011; Sugi et al. 2012) projecting decreased TC counts as a result of increased CO2 have concluded that decreased vertical velocities are of primary importance. While vertical motion is decreased in the future environment in the present work, this finding would also be expected simply because there are fewer storms in the warming runs. In our simulations, increases in future CIN are an imposed thermodynamic change that would lead to decreased vertical motion (Table 4). Ultimately, a more idealized experimental design would be necessary in order to separate the effects of decreased ascent and increased saturation deficit in suppressing future genesis.
This research was supported by DOE Grant ER64448, awarded to North Carolina State University. The authors thank the Renaissance Computing Institute (RENCI) for making available their computing resources and technical support. Thanks to Profs. Walt Robinson and Fred Semazzi and Dr. Kevin Hill, as well as Dr. Isaac Held and two anonymous reviewers, who provided helpful guidance on this work. We also greatly appreciate the editorial guidance of Dr. Kevin Walsh. The WRF Model is made available by NCAR, funded by the National Science Foundation. We also thank the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for collecting and archiving the CMIP3 model output, and the WCRP's Working Group on Coupled Modeling (WGCM) for organizing the model data analysis activity. The WCRP CMIP3 multimodel dataset is supported by the Office of Science, U.S. Department of Energy.
Current affiliation: National Exposure Research Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina.
In the current work, the incubation parameter is defined as it is in R10 and E08. It should be noted that in Emanuel (1995), a similar quantity is defined, but with the numerator appearing as the midlevel moist entropy saturation deficit. The physical basis for this substitution is briefly described in the proceeding discussion.
Storm counts from each current ensemble are compared with verification in Part I and the decrease in storm counts in 6Current05 relative to 18Current05 is also discussed.
The criterion applied for statistical significance is a paired Student's t test using a p value threshold of 0.10.
500-hPa ω is used to compare vertical velocity here, as was also used in Lavendar and Walsh (2011).
Here, downdraft grid cells are defined as having a positive ω anomaly, relative to the mean ω averaged over the genesis region.