1. Introduction
Although tropical cyclone track forecasts continue to exhibit increased skill, forecasting tropical cyclone formation and intensity change remain a challenge (Elsberry et al. 2007). Tropical cyclone formation is generally defined as the process whereby an incipient disturbance forms a closed low-level circulation and develops a warm core relative to the background environment, although definitions of the precise time of genesis vary. It has been well documented that only a small number of tropical disturbances actually develop into tropical cyclones (e.g., Kerns and Chen 2013).
In the idealized simulations by Nicholls and Montgomery (2013), two pathways to genesis were observed. One pathway is similar to that described by Hendricks et al. (2004) and Montgomery et al. (2006) in that positive low-level vorticity is concentrated in vortical hot towers (VHTs) and the net diabatic heating from numerous VHTs helps drive low-level convergence and leads to the aggregation and axisymmetrization of vorticity on the system scale. This causes gradual strengthening of the low-level circulation and contraction of the radius of maximum winds. A second pathway occurs via the strengthening of the midlevel circulation prior to the formation of a small, strong low-level vortex. This second pathway matches the general description of storm formation from the idealized modeling study by Nolan (2007). Nicholls and Montgomery (2013) found that the second pathway was favored when ice was present in the midtroposphere. Increased ice production occurred when modeled convection was initially strong. The diabatic heating/cooling resulting from the presence of a substantial ice layer increased midlevel inflow that strengthened the midlevel circulation prior to the formation of the low-level vortex.
Sippel and Zhang (2008) examined a nondeveloping disturbance in the Gulf of Mexico using an ensemble forecast system and found that some members strengthened the disturbance into a tropical cyclone while others remained nondeveloping. They found that storm formation was most sensitive to deep moisture, convective available potential energy (CAPE), and the magnitude of latent heating early in the simulations, and that the final intensity of the system was strongly correlated to the initial 700-hPa mixing ratio.
For certain cases, relatively small-scale factors can determine whether tropical storm formation occurs, which supports the notion that tropical storm formation is a threshold phenomenon (Emanuel 1989; Doyle et al. 2012). This presents a challenge with regards to forecasting tropical storm formation, as often these important small-scale features are not well observed or resolved by global models. Zhang and Sippel (2009) compared nondeveloping and developing members from their earlier study (Sippel and Zhang 2008) in greater detail and found that the development outcome was sensitive to differences in the initial conditions that were smaller than the typical uncertainty present in the initial conditions. Furthermore, small variations in the timing, location, and intensity of convection led to significant differences in the larger-scale evolution of the system, and sometimes suppressed storm formation entirely. As Zhang and Sippel (2009) point out, this raises the question of whether certain cases of tropical storm formation can be skillfully predicted, and it highlights the need for an ensemble probabilistic approach for forecasting storm formation rather than relying solely on deterministic forecasts.
An observational study of the nondeveloping TCS025 disturbance (Penny et al. 2015) observed during The Observing System Research and Predictability Experiment (THORPEX) Pacific Asian Regional Campaign (T-PARC)/Tropical Cyclone Structure-2008 (TCS-08) field experiment (Elsberry and Harr 2008) suggested the system failed to develop as a result of a combination of system-scale and environmental factors. Although there were convectively active periods, convection was generally weak. Penny et al. (2015) concluded that increased system-relative flow resulting from a vertically misaligned circulation, persistent environmental vertical wind shear, and large horizontal flow deformation made the inner-core more vulnerable to midlevel low-
An original motivation for this study was to examine key processes identified by Penny et al. (2015) related to the nondevelopment of TCS025 in greater detail using high-resolution numerical simulations. However, since TCS025 was near a threshold of development, it proved difficult to correctly simulate nondevelopment, as the majority of simulations overdeveloped the disturbance. Correctly simulating nondevelopment for systems close to the threshold of development appears to be a challenge for high-resolution numerical models as Fritz and Wang (2013) also found it difficult to accurately simulate the nondevelopment of ex-Tropical Storm Gaston.
Given these challenges of simulating nondeveloping systems, this paper, which is part two of a three-part study, analyzes two high-resolution simulations of TCS025 from a multiphysics ensemble: one in which nondevelopment occurs and one in which the system unrealistically develops. The contrasts between the two simulations are used to infer key processes responsible for the evolution of TCS025 and to identify model deficiencies related to the observed tendency toward overdevelopment. It is found that the development scenario for TCS025 is extremely sensitive to the representation of convection and diabatic heating. Simulations exhibiting stronger convection and larger diabatic heating rates tend to overdevelop TCS025, while convection in the single nondeveloping simulation was limited in strength and areal coverage. In the follow-on study (part three), Penny et al. (2016) use an ensemble data assimilation system to assimilate TCS025 observations and evaluate their impact on high-resolution numerical simulations of TCS025. The remainder of this paper is outlined as follows: section 2 describes the configuration of the numerical simulations composing the multiphysics ensemble, a detailed comparison of the simulated evolution of TCS025 is provided in section 3, and a discussion of the findings and conclusions is presented in section 4.
2. Methodology
The multiphysics ensemble of TCS025 comprised numerical simulations using version 3.2.1 of the Advanced Weather Research and Forecasting (WRF-ARW) nonhydrostatic, mesoscale model (Skamarock et al. 2008). The ensemble was constructed primarily by using different cloud microphysics schemes (Table 1). Most model simulations consisted of three nested domains at 27-, 9-, and 3-km grid spacing with 33 vertical levels; however, the WSM6 1-km simulation used a fourth domain with 1-km grid spacing, and the Lin-45 simulation used 45 vertical levels.
List of TCS025 multiphysics ensemble simulations conducted using the WRF-ARW model. All simulations were initialized at 1200 UTC 27 Aug (except WSM6 28/06, which was initialized at 0000 UTC 28 Aug) and integrated until 1200 UTC 30 Aug. Abbreviations in the table are as follows: WSM3, WSM5, and WSM6 represent the WRF single-moment 3-, 5-, and 6-class microphysics schemes, respectively. The WDM6 is the WRF double-moment 6-class microphysics scheme. The YSU and MYJ boundary layer schemes correspond to the Yonsei University and Mellor–Yamada–Janjić schemes, respectively. The Kain–Fritsch scheme is indicated by KF, and “—” indicates no cumulus scheme was used for the highest-resolution grid. The configuration of the Lin-45 simulation is identical to the Lin simulation, except that 45 vertical levels were used instead of 33.
Cumulus convection was parameterized using the Kain–Fritsch (KF) scheme (Fritsch and Kain 1993; Kain 2004) for the two outer domains, but was explicitly represented for the higher-resolution domains, except for simulation WDM6 KF that also employed the KF scheme on the 3-km grid. Radiative processes were calculated using the Rapid Radiative Transfer Model (RRTM; Mlawer et al. 1997) for longwave radiation and Dudhia (1989) for shortwave radiation. Although the lower boundary was primarily composed of ocean surface, surface temperature was predicted for land areas using the five-layer thermal diffusion land surface scheme (Dudhia 1996).
Whereas one simulation used the Mellor–Yamada–Janjić (MYJ) boundary layer (Mellor and Yamada 1982; Janjić 1990, 1996, 2002) and Eta surface layer (Janjić 1996, 2002) schemes (MYJ simulation), all other simulations used the Yonsei University (YSU) boundary layer (Hong et al. 2006) and the fifth-generation Pennsylvania State University–NCAR Mesoscale Model (MM5) similarity theory surface-layer (Skamarock et al. 2008) schemes.
Initial and lateral boundary conditions for the WRF-ARW simulations were constructed from the 6-h European Centre for Medium-Range Weather Forecasts (ECMWF) Year of Tropical Convection (YOTC; Waliser et al. 2012) gridded analyses. Simulations were initialized at 1200 UTC 27 August 2008 so that the initial conditions could incorporate dropwindsonde data collected during the first intensive observing period (IOP) of TCS025 early on 27 August (Penny et al. 2015). There was good agreement between the ECMWF analyses and dropwindsondes deployed into TCS025 (not shown). Therefore, the minimum sea level pressure of the ECMWF analysis, which did not deepen TCS025 beyond 1002 hPa (see Fig. 1), was used to assess the degree of development in the high-resolution simulations. In addition, the observational analysis of TCS025 by Penny et al. (2015) revealed that a tropical upper-tropospheric trough (TUTT) cell modulated vertical wind shear and the thermodynamic environment surrounding TCS025. Different model initialization times were also tested, such as 0000 UTC 28 August for the WSM6 28/00 simulation, but the impacts to the development of TCS025 were minimal and the simulated evolution of the TUTT cell seemed to agree better with subsequent ECMWF analyses with a 1200 UTC 27 August 2008 initialization time. Model fields were output every 30 min for analysis.
Minimum sea level pressure (hPa) from 1200 UTC 27 Aug (except for the WSM6 28/00 simulation that was initialized at 0000 UTC 28 Aug) to 1200 UTC 30 Aug for members of the multiphysics WRF-ARW ensemble and the ECMWF YOTC analysis. Bold lines denote simulations compared in this study.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Minimum sea level pressure (hPa) from 1200 UTC 27 Aug (except for the WSM6 28/00 simulation that was initialized at 0000 UTC 28 Aug) to 1200 UTC 30 Aug for members of the multiphysics WRF-ARW ensemble and the ECMWF YOTC analysis. Bold lines denote simulations compared in this study.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Minimum sea level pressure (hPa) from 1200 UTC 27 Aug (except for the WSM6 28/00 simulation that was initialized at 0000 UTC 28 Aug) to 1200 UTC 30 Aug for members of the multiphysics WRF-ARW ensemble and the ECMWF YOTC analysis. Bold lines denote simulations compared in this study.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
3. Results
During the first 24 h, the majority of the multiphysics ensemble simulations were in good agreement with the ECMWF analysis based on minimum sea level pressure (Fig. 1). After 1200 UTC 28 August, however, most members began to exhibit a substantial decrease in minimum sea level pressure. The Kessler simulation exhibited the most rapid development, but was excluded from the detailed analysis since the Kessler microphysics scheme only accounts for warm rain processes. Although the tendency among the ensemble members was toward overdevelopment, the rate of development varied greatly. Following a period of initial deepening, several simulations exhibited very little development. For example, the minimum sea level pressure for the WDM6 simulation remained fairly constant through 1200 UTC 30 August, and the MYJ simulation only deepened slowly following 1800 UTC 28 August. The WSM3 simulation failed to develop TCS025 altogether, and was the most similar to the ECMWF analysis. Simulations that exhibited less development, such as the WSM3 simulation (including the ECMWF analysis), tended to have tracks to the west of simulations that intensified the disturbance (Fig. 2).
As in Fig. 1, but for storm tracks based on minimum sea level pressure positions. Starting and ending positions of the ECMWF analysis are indicated by the large open and filled black circles, respectively. Small circles connected by dashed lines denote center positions from simulations of focus in this study (bold lines) at times annotated in the figure. Latitude and longitude positions from the WRF-ARW simulations were smoothed using a five-point running average.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 1, but for storm tracks based on minimum sea level pressure positions. Starting and ending positions of the ECMWF analysis are indicated by the large open and filled black circles, respectively. Small circles connected by dashed lines denote center positions from simulations of focus in this study (bold lines) at times annotated in the figure. Latitude and longitude positions from the WRF-ARW simulations were smoothed using a five-point running average.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 1, but for storm tracks based on minimum sea level pressure positions. Starting and ending positions of the ECMWF analysis are indicated by the large open and filled black circles, respectively. Small circles connected by dashed lines denote center positions from simulations of focus in this study (bold lines) at times annotated in the figure. Latitude and longitude positions from the WRF-ARW simulations were smoothed using a five-point running average.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
To assess the key differences related to the simulated development/nondevelopment of TCS025, the Lin (developing) and WSM3 (nondeveloping) simulations were chosen for an in-depth comparison of the synoptic evolution and the dynamic and thermodynamic characteristics. These two simulations differed only in their use of microphysics schemes (see Table 1). The nondeveloping simulation that employed the WSM3 microphysics scheme (Hong et al. 2004) uses single variables to account for the mixing ratios of cloud water and cloud ice and the mixing ratios of rain and snow, which precludes mixed-phase processes. The developing simulation used the Purdue–Lin scheme (Lin et al. 1983; Chen and Sun 2002), which accounts for the mixing ratios of liquid cloud water, rain, ice, snow, and graupel.
a. Synoptic evolution comparison
Based on minimum sea level pressure (Fig. 1), the Lin simulation began to intensify the TCS025 disturbance shortly after 1200 UTC 28 August. As early as 0000 UTC 28 August (T + 12 h), notable differences between the two simulations were apparent (Fig. 3). The 850-hPa geopotential heights were lower in the Lin simulation close to the low-level circulation near 19°N, 152°E (Fig. 3d), and the updrafts at 200 hPa were much stronger surrounding the low-level circulation (Fig. 3b). A developing upper-level anticyclone and outflow channels near 18°N, 157°E and 18°N, 162°E were associated with this area of enhanced convection in the Lin simulation (Fig. 3b). These features were also present in the WSM3 simulation (Fig. 3a), but were much weaker.
Wind speed (m s−1, shading), geopotential height (m, black contours), and vertical velocity (m s−1, red contours at 0.5 m s−1 intervals) at (top) 200 and (bottom) 850 hPa valid at 0000 UTC 28 Aug (T + 12 h) for (a),(c) the WSM3 simulation and (b),(d) the Lin simulation. Contours of vertical velocity are only shown at 200 hPa, and the black lines indicate the location of the vertical cross sections in Fig. 4.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Wind speed (m s−1, shading), geopotential height (m, black contours), and vertical velocity (m s−1, red contours at 0.5 m s−1 intervals) at (top) 200 and (bottom) 850 hPa valid at 0000 UTC 28 Aug (T + 12 h) for (a),(c) the WSM3 simulation and (b),(d) the Lin simulation. Contours of vertical velocity are only shown at 200 hPa, and the black lines indicate the location of the vertical cross sections in Fig. 4.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Wind speed (m s−1, shading), geopotential height (m, black contours), and vertical velocity (m s−1, red contours at 0.5 m s−1 intervals) at (top) 200 and (bottom) 850 hPa valid at 0000 UTC 28 Aug (T + 12 h) for (a),(c) the WSM3 simulation and (b),(d) the Lin simulation. Contours of vertical velocity are only shown at 200 hPa, and the black lines indicate the location of the vertical cross sections in Fig. 4.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Northeast–southwest vertical cross sections (Fig. 4) extending from the low-level circulation to the TUTT cell reveal differences in the circulation structure and thermodynamic response to convection after 12 h into the simulation. The vorticity associated with the TCS025 circulation was vertically misaligned in the WSM3 simulation (Fig. 4a), which matches the circulation structure analyzed by Penny et al. (2015) from observations. A shallow low-level cyclonic vorticity feature was present north of a midlevel vorticity feature that was strongest near 500 hPa. Relative vorticity was weak and diffuse between the low- and midlevel vorticity features. In contrast, the midlevel vorticity in the Lin simulation was much stronger, and extended from 800 to 400 hPa as a more vertically aligned and coherent structure (Fig. 4b).
Vertical cross sections along the black lines in Fig. 3 of (top) relative vorticity (10−5 s−1, shading), and (bottom) equivalent potential temperature (K, shading) and virtual temperature anomaly (K, contours) valid at 0000 UTC 28 Aug for (a),(c) the WSM3 simulation and (b),(d) the Lin simulation. The ordinate axis is in pressure (hPa). Vectors represent wind speeds in the plane of the cross section, and vertical velocity has been rescaled by a factor of 10. The average virtual temperature out to 6° radius from the circulation center position was used as the reference state to determine the virtual temperature anomaly in (c) and (d).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Vertical cross sections along the black lines in Fig. 3 of (top) relative vorticity (10−5 s−1, shading), and (bottom) equivalent potential temperature (K, shading) and virtual temperature anomaly (K, contours) valid at 0000 UTC 28 Aug for (a),(c) the WSM3 simulation and (b),(d) the Lin simulation. The ordinate axis is in pressure (hPa). Vectors represent wind speeds in the plane of the cross section, and vertical velocity has been rescaled by a factor of 10. The average virtual temperature out to 6° radius from the circulation center position was used as the reference state to determine the virtual temperature anomaly in (c) and (d).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Vertical cross sections along the black lines in Fig. 3 of (top) relative vorticity (10−5 s−1, shading), and (bottom) equivalent potential temperature (K, shading) and virtual temperature anomaly (K, contours) valid at 0000 UTC 28 Aug for (a),(c) the WSM3 simulation and (b),(d) the Lin simulation. The ordinate axis is in pressure (hPa). Vectors represent wind speeds in the plane of the cross section, and vertical velocity has been rescaled by a factor of 10. The average virtual temperature out to 6° radius from the circulation center position was used as the reference state to determine the virtual temperature anomaly in (c) and (d).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Vertical cross sections of 
By 0000 UTC 29 August (T + 36 h), the low-level circulation in the Lin simulation was much stronger compared to the WSM3 simulation (Figs. 5c,d). The strongest winds were almost exclusively along the eastern side of the circulation. Notable upper-level differences also existed (Figs. 5a,b). Although updrafts and an upper-level anticyclone are evident near the location of the low-level circulation in the WSM3 simulation (Fig. 5a), the features were much stronger in the Lin simulation at this time (Fig. 5b). As a result, the outflow channels to the east and south of the system were also much stronger in the Lin simulation.
As in Fig. 3, but valid at 0000 UTC 29 Aug (T + 36 h) and the black lines indicate the locations of the cross sections in Fig. 6.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 3, but valid at 0000 UTC 29 Aug (T + 36 h) and the black lines indicate the locations of the cross sections in Fig. 6.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 3, but valid at 0000 UTC 29 Aug (T + 36 h) and the black lines indicate the locations of the cross sections in Fig. 6.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
A vertical cross section comparison valid at 0000 UTC 29 August (Fig. 6) reveals that the relative vorticity in the WSM3 simulation (Fig. 6a) originally associated with the low-level circulation of TCS025 (southernmost vorticity feature in the cross section) continued to appear weak and vertically misaligned. The strongest cyclonic vorticity was near 800 hPa, which was lower than the prior day. In contrast, relative vorticity associated with the TCS025 circulation in the Lin simulation (Fig. 6b) was notably stronger, especially in the low levels, and was well aligned in the vertical.
As in Fig. 4, but vertical cross sections along the black lines in Fig. 5 for 0000 UTC 29 Aug (T + 36 h).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 4, but vertical cross sections along the black lines in Fig. 5 for 0000 UTC 29 Aug (T + 36 h).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 4, but vertical cross sections along the black lines in Fig. 5 for 0000 UTC 29 Aug (T + 36 h).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Vertical cross sections of 
Although the absolute magnitudes of 
By 0000 UTC 30 August (T + 60 h), the minimum sea level pressure in the Lin simulation had decreased to 998 hPa (Fig. 1) and the maximum 10-m wind speed was close to 23 m s−1 (not shown). In contrast, the minimum sea level pressure and maximum 10-m wind speed in the WSM3 simulation were 1009 hPa (Fig. 1) and 13 m s−1 (not shown), respectively. The low-level wind field in the Lin simulation (Fig. 7d) had become increasingly symmetric about the center, as the strongest winds were wrapping around the north side of the system. However, winds along the western side were still weaker than along the eastern side. In contrast, the low-level circulation in the WSM3 simulation (Fig. 7c) was weaker than the day prior, and was no longer a closed circulation in the earth-relative frame.
As in Fig. 3, but valid at 0000 UTC 30 Aug (T + 60 h).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 3, but valid at 0000 UTC 30 Aug (T + 60 h).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 3, but valid at 0000 UTC 30 Aug (T + 60 h).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
In the upper levels (Figs. 7a,b), both simulations exhibited outflow channels and an upper-level ridge above and to the north of the TCS025 disturbance. The upper-level ridge was much stronger in the Lin simulation (Fig. 7b), especially to the north of the low-level circulation, whereas the upper-level ridge and anticyclonic flow in the WSM3 simulation (Fig. 7a) were even weaker than in the ECMWF analysis at this time (not shown).
b. Circulation budget analysis
The 850-hPa wind vectors (m s−1) and average 950–700-hPa relative vorticity (10−5 s−1, color shading) from (a),(c),(e) the WSM3 simulation and (b),(d),(f) the Lin simulation at 6-h intervals from 1800 UTC 27 Aug to 1200 UTC 28 Aug. The 10 m s−1 reference vectors are shown in (a) and (b). The boxed region corresponds to the area over which circulation budgets and averages in Figs. 12 and 15 were computed.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
The 850-hPa wind vectors (m s−1) and average 950–700-hPa relative vorticity (10−5 s−1, color shading) from (a),(c),(e) the WSM3 simulation and (b),(d),(f) the Lin simulation at 6-h intervals from 1800 UTC 27 Aug to 1200 UTC 28 Aug. The 10 m s−1 reference vectors are shown in (a) and (b). The boxed region corresponds to the area over which circulation budgets and averages in Figs. 12 and 15 were computed.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
The 850-hPa wind vectors (m s−1) and average 950–700-hPa relative vorticity (10−5 s−1, color shading) from (a),(c),(e) the WSM3 simulation and (b),(d),(f) the Lin simulation at 6-h intervals from 1800 UTC 27 Aug to 1200 UTC 28 Aug. The 10 m s−1 reference vectors are shown in (a) and (b). The boxed region corresponds to the area over which circulation budgets and averages in Figs. 12 and 15 were computed.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
The circulation budget was started 6 h after model initialization time to minimize contamination from model spinup. Since the low-level circulation near 20°N, 151.5°E eventually developed in the Lin simulation, the boxed region was selected to encompass this area (Fig. 8). Budget calculations were also performed with different box sizes and over different areas (not shown). Although the magnitudes of the budget terms were affected by box size and location, the relative differences between the two simulations remained consistent. The circulation budget was expressed in terms of box-average vorticity after normalizing by area, and is therefore referred to as an average vorticity budget in the following discussion.
1) Spinup: 1800 UTC 27 August–0600 UTC 28 August
The evolution of average low-level relative vorticity within the boxed regions in Fig. 8 from 1800 UTC 27 August to 0600 UTC 28 August for the WSM3 simulation (Fig. 9a) and the Lin simulation (Fig. 10a) reveal important differences. Both simulations were initialized with the strongest relative vorticity in the midlevels (~500 hPa) and exhibited a steady progression toward increased low-level vorticity. However, the increase in low-level vorticity was far greater in the Lin simulation, especially after 2100 UTC 27 August (Fig. 10a). In addition, the negative relative vorticity that developed above 250 hPa after 0000 UTC 28 August was much stronger in the Lin simulation (Figs. 9a and 10a).
Evolution of average relative vorticity for the WSM3 simulation from 1800 UTC 27 Aug to 0600 UTC 28 Aug within the boxed region in Fig. 8: (a) relative vorticity (10−5 s−1), (b) total average vorticity tendency (10−5 s−1 h−1), (c) average vorticity tendency due to stretching (10−5 s−1 h−1), (d) average vorticity tendency due to horizontal eddy fluxes (10−5 s−1 h−1), (e) average vorticity tendency due to tilting (10−5 s−1 h−1), and (f) average vorticity budget terms integrated from 1800 UTC 27 Aug to 0600 UTC 28 Aug. Ordinate axes are in pressure (hPa).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Evolution of average relative vorticity for the WSM3 simulation from 1800 UTC 27 Aug to 0600 UTC 28 Aug within the boxed region in Fig. 8: (a) relative vorticity (10−5 s−1), (b) total average vorticity tendency (10−5 s−1 h−1), (c) average vorticity tendency due to stretching (10−5 s−1 h−1), (d) average vorticity tendency due to horizontal eddy fluxes (10−5 s−1 h−1), (e) average vorticity tendency due to tilting (10−5 s−1 h−1), and (f) average vorticity budget terms integrated from 1800 UTC 27 Aug to 0600 UTC 28 Aug. Ordinate axes are in pressure (hPa).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Evolution of average relative vorticity for the WSM3 simulation from 1800 UTC 27 Aug to 0600 UTC 28 Aug within the boxed region in Fig. 8: (a) relative vorticity (10−5 s−1), (b) total average vorticity tendency (10−5 s−1 h−1), (c) average vorticity tendency due to stretching (10−5 s−1 h−1), (d) average vorticity tendency due to horizontal eddy fluxes (10−5 s−1 h−1), (e) average vorticity tendency due to tilting (10−5 s−1 h−1), and (f) average vorticity budget terms integrated from 1800 UTC 27 Aug to 0600 UTC 28 Aug. Ordinate axes are in pressure (hPa).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 9, but for the Lin simulation.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 9, but for the Lin simulation.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
As in Fig. 9, but for the Lin simulation.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
In general, positive vorticity tendencies (Figs. 9b and 10b) dominated the low levels and were largest prior to 0600 UTC 28 August. A strong upper-level negative tendency was present in the Lin simulation from 2100 UTC 27 August to 0000 UTC 28 August (Fig. 10b), which is attributed to the developing upper-level anticyclone identified in the synoptic analysis during this time (see Figs. 3a,b). Negative upper-level tendencies were also present in the WSM3 simulation (Fig. 9b), but did not become large in magnitude until after 0000 UTC 28 August. This indicates that development of the upper-level anticyclone was delayed in the WSM3 simulation.
The vertical stretching tendency constituted the largest contribution to the vorticity tendency early in both simulations (Figs. 9c and 10c). The largest tendencies occurred primarily between 1800 UTC 27 August and 0000 UTC 28 August with positive tendencies below 500 hPa, and negative tendencies above this level. This pattern is associated with low-level convergence and upper-level divergence and is indicative of a convective divergence profile (Mapes and Houze 1995). Positive stretching tendencies were also evident after 0000 UTC 28 August, but were predominantly in the midlevels in the WSM3 simulation (Fig. 9c) and near 700 hPa in the Lin simulation (Fig. 10c). The positive and negative contributions from stretching were generally much larger in the Lin simulation, both in the low and upper levels, except at 0600 UTC 28 August when the positive midlevel stretching tendency was larger in the WSM3 simulation.
Tendencies due to horizontal eddy fluxes (Figs. 9d and 10d) at the periphery of the boxed region were generally positive in the upper and lower troposphere, but negative in the midtroposphere. Animations of 200-hPa vorticity (not shown) reveal that the positive upper-level tendencies from horizontal eddy fluxes, especially between 1800 UTC 27 August and 2100 UTC 27 August, were due to upper-level negative vorticity anomalies (generated from stretching processes) leaving the boxed region, while the negative tendencies in the low and midlevels were mostly due to outward fluxes of positive vorticity through the eastern boundary of the region. The contributions from horizontal eddy fluxes were larger in the Lin simulation, which suggests that the positive and negative vorticity anomalies crossing the boundaries of the boxed region were larger in magnitude in the Lin simulation. A visual inspection of Fig. 8 confirms this.
Although not as large as the stretching and horizontal eddy flux terms, contributions from the tilting of horizontal vorticity (Figs. 9e and 10e) were positive in the mid- to upper levels and predominantly negative in the low levels for both simulations. Similar to the contributions from stretching and horizontal eddy fluxes, vorticity tendencies due to tilting were more pronounced in the Lin simulation, presumably a result of stronger updrafts and increased horizontal vorticity.
Profiles of the circulation budget terms (Figs. 9f and 10f) integrated in time from 1800 UTC 27 August to 0600 UTC 28 August over the boxed region reveal that vertical stretching was the dominant process affecting vorticity throughout the depth of the troposphere in both simulations. The cumulative effect of horizontal eddy fluxes and tilting were positive in the upper levels and near the surface, but weakly negative in between. The residual was generally small in magnitude, except near the surface where it is assumed friction caused the actual tendency to be smaller than the sum of the various terms (negative residual).
2) Frictional spindown in the WSM3 simulation: 0300–1200 UTC 28 August
During an initial period of deep convection that lasted until shortly after 0000 UTC 28 August, both simulations experienced positive low-level vorticity tendencies primarily due to stretching (see Figs. 9f and 10f). Following ~0300 UTC 28 August, the convective activity in both simulations declined substantially based on azimuthally averaged model-derived cloud-top temperature (Figs. 11a,c). This time coincided with a net spindown of the low-level circulation in the WSM3 simulation.
(a),(c) Azimuthally averaged model-derived cloud-top brightness temperatures (°C) and (b),(d) integrated circulation budget contributions for the time period 0300–1200 UTC 28 Aug for the boxed region in Fig. 8 for (a),(b) the WSM3 simulation and (c),(d) the Lin simulation. The ordinate axis in (b),(d) is pressure (hPa).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
(a),(c) Azimuthally averaged model-derived cloud-top brightness temperatures (°C) and (b),(d) integrated circulation budget contributions for the time period 0300–1200 UTC 28 Aug for the boxed region in Fig. 8 for (a),(b) the WSM3 simulation and (c),(d) the Lin simulation. The ordinate axis in (b),(d) is pressure (hPa).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
(a),(c) Azimuthally averaged model-derived cloud-top brightness temperatures (°C) and (b),(d) integrated circulation budget contributions for the time period 0300–1200 UTC 28 Aug for the boxed region in Fig. 8 for (a),(b) the WSM3 simulation and (c),(d) the Lin simulation. The ordinate axis in (b),(d) is pressure (hPa).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Unlike the integrated circulation budget profiles in Figs. 9f and 10f, integrated profiles from 0300 to 1200 UTC 28 August (Figs. 11b,d) reveal that the vertical stretching of relative vorticity was no longer the dominant process affecting the low-level circulation tendency. In fact, the stretching tendency was slightly negative near the surface in the WSM3 simulation (Fig. 11b), but positive from 850 to 250 hPa. The negative low-level stretching tendencies indicate that organized deep convection in the WSM3 simulation was extremely weak or absent during this period, and this resulted in a spindown of the low-level circulation (Fig. 11b). The negative residual near the surface (i.e., friction) was larger in magnitude than all other factors. Note that the net circulation tendency (Fig. 11b) from the surface to 750 hPa was negative for the WSM3 simulation.
The circulation budget profiles for the Lin simulation during the convective lull (Fig. 11d) were similar in pattern to the WSM3 simulation except that friction was larger in the Lin simulation near the surface since the low-level winds were already stronger by 0300 UTC 28 August. However, despite the larger frictional dissipation, the total circulation tendency was positive in the low levels due to positive low-level stretching tendencies.
It is worth noting that both simulations experienced the largest positive stretching tendencies (Figs. 11b,d) in the midlevels during this 9-h time period from 0300 to 1200 UTC 28 August. This suggests that midlevel convergence associated with stratiform precipitation processes led to the spinup of the midlevel circulation during this time while low-level divergence likely offset positive tendencies in the low levels. Deep convection continued to reform near the low-level circulation center in the Lin simulation (Fig. 11c) following this lull in convection, but it was largely absent from the WSM3 simulation following 1200 UTC 28 August (Fig. 11a).
c. Convective intensity comparison
Since the synoptic and vorticity budget comparison of the WSM3 and Lin simulations suggest differences were primarily related to the representation of convective processes, the characteristics of convection were further examined by comparing time- and space-averaged variables within the boxed region in Fig. 8 from 1200 UTC 27 August to 1200 UTC 28 August.
Differences in the areal-averaged vertical velocity between the two experiments (Lin − WSM3) (Fig. 12a) confirm that updrafts in the Lin simulation were stronger, especially between 300 and 250 hPa. Likewise, low-level convergence and upper-level divergence (Fig. 12b) were larger, and these profiles closely match the vorticity stretching tendency patterns (Figs. 9c and 10c); enhanced low-level convergence in the Lin simulation corresponded with larger positive tendencies, while larger upper-level divergence resulted in larger negative tendencies.
Differences between the Lin and WSM3 simulations (Lin − WSM3) averaged over the boxed region in Fig. 8 from 1200 UTC 27 Aug to 1200 UTC 28 Aug for (a) vertical velocity (m s−1), (b) divergence (10−5 s−1), (c) diabatic heating rate (K h−1), (d) 
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Differences between the Lin and WSM3 simulations (Lin − WSM3) averaged over the boxed region in Fig. 8 from 1200 UTC 27 Aug to 1200 UTC 28 Aug for (a) vertical velocity (m s−1), (b) divergence (10−5 s−1), (c) diabatic heating rate (K h−1), (d)
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Differences between the Lin and WSM3 simulations (Lin − WSM3) averaged over the boxed region in Fig. 8 from 1200 UTC 27 Aug to 1200 UTC 28 Aug for (a) vertical velocity (m s−1), (b) divergence (10−5 s−1), (c) diabatic heating rate (K h−1), (d)
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Above the surface layer, the differences (Lin − WSM3) in diabatic heating rates (Fig. 12c) were almost entirely positive with a maximum difference of ~1 K h−1 near the 350-hPa level. As expected, the vertical structure and temporal evolution of the difference in the diabatic heating appeared to be closely related to differences in vertical velocity; stronger updrafts in the Lin simulation corresponded to larger diabatic heating rates, and the largest differences in the heating rates occurred immediately below the level where vertical velocity differences were largest.
Although earlier studies (e.g., Bister and Emanuel 1997) suggested the impact of evaporative downdrafts should diminish as a system progresses toward development, the developing Lin simulation experienced greater diabatic cooling near the surface than the nondeveloping WSM3 simulation (Fig. 12c). This is in agreement with more recent modeling studies (Nolan 2007; Sippel and Zhang 2008; Wang 2012) that do not find a notable decrease in the frequency or intensity of downdrafts as a system develops. The largest differences in low-level cooling rates occurred when the differences in updrafts (Fig. 12a) and upper-level diabatic heating rates were also largest.
Positive differences in areal-average upper-level 
Differences in the average profiles of temperature (Fig. 12e) and water vapor mixing ratio (Fig. 12f) reveal that the differences in low-level 
Since the solutions of the WSM3 and Lin simulations quickly diverged with regards to the system-scale development of TCS025, it was desirable to compare the convective characteristics associated with a burst of convection that occurred during the first three hours of the simulations. Although this period was likely impacted by the model cold start, it can be used to examine the convective structure since differences during this period were primarily limited to being on the convective scale and interactions with the other physical parameterizations were likely of secondary importance this early in the simulation.
Vertical cross sections at 1400 UTC 27 August (T + 2 h) (Fig. 13) reveal that the primary convective core in the Lin simulation had already started to dissipate, judging from the development of downdrafts greater than 2 m s−1 (Fig. 13f) and a large area with diabatic cooling rates greater than 10 K h−1 (Fig. 13d). In contrast, downdrafts (Fig. 13e) and diabatic cooling (Fig. 13c) were absent from the WSM3 simulation at this time, although an area of diabatic cooling developed later in the WSM3 simulation near the freezing level from the melting of snow (not shown). The total precipitating mixing ratio in the WSM3 simulation (Fig. 13a) was greater than 8 g kg−1 near 250 hPa at this time and remained this way through 1500 UTC 27 August. In contrast, the total precipitating mixing ratio in the Lin simulation (Fig. 13b) was a maximum close to 450 hPa and had decreased in magnitude since its peak at 1330 UTC, primarily due to the fallout of graupel (not shown). Despite the large area of downdrafts (Fig. 13f) and diabatic cooling (Fig. 13d) in the Lin simulation, the area occupied by updrafts and positive diabatic heating rates was still larger than in the WSM3 simulation, and the updrafts penetrated higher (cf. Figs. 13e and 13f). However, similar to 1330 UTC, the WSM3 simulation possessed a larger area with diabatic heating rates greater than 100 K h−1 (Fig. 13c).
Vertical cross sections of reflectivity (dBZ, shading) with (a),(b) total precipitating mixing ratio (g kg−1, contours); (c),(d) diabatic heating rate (K h−1, contours); and (e),(f) vertical velocity (m s−1, contours) for the (a),(c),(e) WSM3 simulation and (b),(d),(f) Lin simulation at 1400 UTC 27 Aug. Dashed contours correspond with negative values in (c)–(f), and the black horizontal dashed line denotes the freezing level.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Vertical cross sections of reflectivity (dBZ, shading) with (a),(b) total precipitating mixing ratio (g kg−1, contours); (c),(d) diabatic heating rate (K h−1, contours); and (e),(f) vertical velocity (m s−1, contours) for the (a),(c),(e) WSM3 simulation and (b),(d),(f) Lin simulation at 1400 UTC 27 Aug. Dashed contours correspond with negative values in (c)–(f), and the black horizontal dashed line denotes the freezing level.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Vertical cross sections of reflectivity (dBZ, shading) with (a),(b) total precipitating mixing ratio (g kg−1, contours); (c),(d) diabatic heating rate (K h−1, contours); and (e),(f) vertical velocity (m s−1, contours) for the (a),(c),(e) WSM3 simulation and (b),(d),(f) Lin simulation at 1400 UTC 27 Aug. Dashed contours correspond with negative values in (c)–(f), and the black horizontal dashed line denotes the freezing level.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
To ensure that the vertical cross sections in the preceding section were representative of the convective area, average profiles were computed within a boxed region surrounding the area of deep convection (roughly an area of 1° × 1°), and the WSM3 profiles were differenced from the Lin profiles at 30-min intervals from 1200 to 1500 UTC 27 August (Fig. 14). Overall there was good agreement between the differenced profiles and the vertical cross sections.
Differences between the average Lin and WSM3 (Lin − WSM3) profiles of (a) total precipitating mixing ratio (g kg−1), (b) diabatic heating rate (K h−1), and (c) vertical velocity (m s−1) at 30-min intervals from 1200 to 1500 UTC 27 Aug within the boxed region of 19°–20°N, 151.2°–152.1°E. The thick gray lines are the average profiles from 1200 to 1500 UTC 27 Aug.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Differences between the average Lin and WSM3 (Lin − WSM3) profiles of (a) total precipitating mixing ratio (g kg−1), (b) diabatic heating rate (K h−1), and (c) vertical velocity (m s−1) at 30-min intervals from 1200 to 1500 UTC 27 Aug within the boxed region of 19°–20°N, 151.2°–152.1°E. The thick gray lines are the average profiles from 1200 to 1500 UTC 27 Aug.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Differences between the average Lin and WSM3 (Lin − WSM3) profiles of (a) total precipitating mixing ratio (g kg−1), (b) diabatic heating rate (K h−1), and (c) vertical velocity (m s−1) at 30-min intervals from 1200 to 1500 UTC 27 Aug within the boxed region of 19°–20°N, 151.2°–152.1°E. The thick gray lines are the average profiles from 1200 to 1500 UTC 27 Aug.
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
Beginning at 1330 UTC 27 August, the total precipitating mixing ratio in the WSM3 simulation began to increase relative to the Lin simulation above 500 hPa (Fig. 14a) due to the production of snow (not shown). By the end of this period, the total precipitating mixing ratio of the WSM3 simulation was about 2 g kg−1 greater than in the Lin simulation near 300 hPa. Below 600 hPa, the total precipitating mixing ratio (Fig. 14a) was larger in the Lin simulation throughout almost the entire period due to rain (not shown).
The average diabatic heating profiles (Fig. 14b) in the Lin simulation were almost entirely greater in magnitude above 900 hPa, with the largest differences near 400 hPa. In addition, low-level diabatic cooling was stronger in the Lin simulation (Fig. 14b). The differenced average vertical velocity profiles (Fig. 14c) indicate that, except below 600 hPa at 1400 UTC, the average vertical velocity became progressively larger in the Lin simulation, with the largest differences being above 400 hPa when the total precipitating mixing ratios (Fig. 14a) were considerably less in the Lin simulation.
4. Discussion and conclusions
The in-depth comparison of the WSM3 (nondeveloping) and Lin (developing) simulations of TCS025 revealed that the Lin simulation possessed the strongest vertical velocity and the largest diabatic heating rates. Rapid pressure (height) falls in the lower troposphere coupled with the development of an upper-level anticyclone and outflow aloft are expected responses to diabatic heating associated with the strong convection in the Lin simulation. Evidence that strong convective overturning had redistributed 
During an initial period of convective activity present in both simulations, low-level vorticity increased primarily through the stretching of vorticity. However, this process was stronger in the Lin simulation. As the convective intensity diminished early on 28 August, the low-level vorticity tendency remained positive in the Lin simulation as low-level convergence was still contributing to spinup. This genesis pathway resembles that described by Hendricks et al. (2004) and Montgomery et al. (2006). In contrast, the low-level stretching tendency in the WSM3 simulation became negative during this convective lull, which allowed for the spindown of the low-level circulation.
A comparison of vertical cross sections and average profiles during the first three hours of the simulations revealed important differences in the convective characteristics that appear to have been primarily the result of the explicit representation of graupel in the Lin simulation. Updrafts in the Lin simulation strengthened more quickly than in the WSM3 simulation, and the strongest updrafts occurred higher in the troposphere. A considerable amount of graupel formed above the freezing level in the Lin simulation shortly after the start of the simulation, but the mixing ratios quickly decreased from graupel fallout and conversion to rain. In contrast, the largest precipitating mixing ratios (snow) in the WSM3 simulation occurred much higher in the atmosphere, and seemed to persist for a longer period of time. Although there were times when the WSM3 simulation possessed the largest area with diabatic heating rates greater than 100 K h−1, the average positive diabatic heating rates were larger in the Lin simulation. In addition, evaporative downdrafts were more prevalent in the Lin simulation and this led to the development of low-level cold pools that were largely absent from the WSM3 simulation.
These differences in the convective characteristics were consistent with analyses conducted over a much larger area and longer time period (see Fig. 12), and had important implications for the development of TCS025. Since the fall speed of graupel is greater than that of snow, a larger amount of hydrometeor mass was able to fall out of the updrafts in the Lin simulation, which is evidenced by the larger total precipitating mixing ratios (rain) below 600 hPa in the Lin simulation (Fig. 14a). Zhu and Zhang (2006) examined the impact of different microphysics parameterizations in numerical simulations of Hurricane Bonnie (1998) and also found that the presence of graupel reduced hydrometeor mass loading. The rapid fallout of graupel in the Lin simulation reduced the hydrometeor mass loading, which allowed for stronger updrafts in the mid- to upper troposphere. This helped to increase the diabatic heating rates and accelerate the convective cycle. In addition, the explicit representation of graupel in the Lin simulation likely contributed to the larger diabatic heating rates observed in the Lin simulation (Fig. 14b). As Zhu and Zhang (2006) point out, the depositional growth processes of graupel tend to occur more quickly than for snow, which increases the rate of latent heat release.
Another important difference was the development of downdrafts in the Lin simulation, which were associated with the rapid fallout of graupel and conversion to rain below the freezing level. Evaporative cooling in the lower troposphere resulted in the formation of a cold pool, which likely played an important role in helping to initiate new convection in the Lin simulation. In contrast, the only appreciable diabatic cooling that occurred during the first three hours of the WSM3 simulation resulted from the melting of snow near the freezing level.
In summary, the reduction in hydrometeor mass loading and additional latent heat release from graupel caused mid- to upper-level warming and surface pressure falls in the Lin simulation. This helped the secondary circulation to organize and strengthen, which in turn helped to accelerate these processes associated with deep convection. In addition, evaporative downdrafts in the Lin simulation led to cold pool formation that likely helped sustain and initiate new convection. In contrast, the secondary circulation in the WSM3 simulation did not strengthen enough to promote the deep convection and latent heating necessary for the storm to develop.
The fact that these two high-resolution simulations were identical except for the choice of microphysics parameterization scheme, but varied considerably in the degree by which the TCS025 system developed (see Fig. 1) suggests that the model outcome may have been overly sensitive to the representation of microphysical processes. This conclusion is supported by the study of Rogers et al. (2007) that compared the microphysical quantities from mesoscale models with observations within the tropical cyclone environment. Rogers et al. (2007) found that the correlation between the hydrometeor mixing ratio and vertical motion was much higher in the models than the observations suggest.
Given that the choice of microphysics scheme led to considerable differences in the observed convective characteristics, impacts on the development of TCS025 were examined more directly by exploring the sensitivity to the diabatic heating rate. To facilitate this, an additional set of experiments (Fig. 15a) was conducted using the Purdue–Lin microphysics scheme in which the diabatic heating was multiplied by a constant factor throughout the simulations that ranged from 0.85 to 1.1. Interestingly, experiments with a diabatic factor of 0.9 and less (10% reduction in the diabatic heating rate) failed to develop, while factors of 0.95 and greater resulted in development. In addition, the degree of development was closely related to the magnitude of the adjustment factors (Fig. 15a), which indicates that the diabatic heating rate was extremely important for modulating storm development in the TCS025 simulations.
(a) Minimum sea level pressure (hPa) for simulations that used the Purdue–Lin microphysics scheme in which the diabatic heating rate was multiplied by a factor ranging from 0.85 to 1.1. The minimum sea level pressure from the ECMWF analysis (black) is shown for reference. (b) Vertical profiles of the average diabatic heating rate (K h−1) over the boxed region in Fig. 8 for the period 1800 UTC 27 Aug–0600 UTC 28 Aug for the Lin simulation (red line), the WSM3 simulation (purple line), and for simulations that used the Purdue–Lin microphysics scheme in which the diabatic heating rates were multiplied by a factor of 0.95 (long dashed line) and 0.90 (short dashed line).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
(a) Minimum sea level pressure (hPa) for simulations that used the Purdue–Lin microphysics scheme in which the diabatic heating rate was multiplied by a factor ranging from 0.85 to 1.1. The minimum sea level pressure from the ECMWF analysis (black) is shown for reference. (b) Vertical profiles of the average diabatic heating rate (K h−1) over the boxed region in Fig. 8 for the period 1800 UTC 27 Aug–0600 UTC 28 Aug for the Lin simulation (red line), the WSM3 simulation (purple line), and for simulations that used the Purdue–Lin microphysics scheme in which the diabatic heating rates were multiplied by a factor of 0.95 (long dashed line) and 0.90 (short dashed line).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
(a) Minimum sea level pressure (hPa) for simulations that used the Purdue–Lin microphysics scheme in which the diabatic heating rate was multiplied by a factor ranging from 0.85 to 1.1. The minimum sea level pressure from the ECMWF analysis (black) is shown for reference. (b) Vertical profiles of the average diabatic heating rate (K h−1) over the boxed region in Fig. 8 for the period 1800 UTC 27 Aug–0600 UTC 28 Aug for the Lin simulation (red line), the WSM3 simulation (purple line), and for simulations that used the Purdue–Lin microphysics scheme in which the diabatic heating rates were multiplied by a factor of 0.95 (long dashed line) and 0.90 (short dashed line).
Citation: Monthly Weather Review 144, 10; 10.1175/MWR-D-15-0259.1
The large differences in outcome between the simulations with diabatic heating rates reduced by 10% (factor of 0.9) and 5% (factor of 0.95) (Fig. 15a) is perhaps surprising, and might suggest the diabatic heating rate adjustments resulted in unrealistic heating profiles. However, an examination of the average diabatic heating profiles within the boxed region in Fig. 8 from 1800 UTC 27 August to 0600 UTC 28 August (Fig. 15b) reveals that the diabatic heating rates below 500 hPa were within the range of values exhibited by the Lin simulation (factor of 1.0) and the WSM3 simulation. Above 500 hPa, the 0.95 profile was very similar to the WSM3 profile, whereas the diabatic heating rates for the 0.90 case were notably smaller. However, Park et al. (2013) suggested that low-level diabatic heating is most effective at inducing low-level convergence and storm spinup, which suggests that the low-level diabatic heating rates are perhaps most important to consider.
Park and Elsberry (2013) examined Electra Doppler Radar (ELDORA)-derived latent heating rates and found that the maximum latent heating and cooling rates were more uniformly distributed over a deeper layer for TCS025, indicating there was a mixture of stratiform (mid- to upper-level heating maxima) and convective (low-level heating maxima) rain areas. Using a model configuration similar to that of the WSM6 1-km simulation (see Fig. 1), Park et al. (2013) found that compared to the ELDORA observations of TCS025, the modeled latent heating rates were too large and that the modeled latent cooling rates were too small. Given the similarity in the minimum sea level pressure between the WSM6 1-km simulation and the Lin simulation, the diabatic heating rates in the Lin simulation were likely exaggerated as well. In addition to the convective-scale differences identified during the first three hours of the simulations, deep convection tended to dominate the Lin simulation following the initial period of convective activity, while stratiform precipitation was more prevalent in the WSM3 simulation. Interactions with other physical processes (e.g., stabilization of the thermodynamic profile and cloud–radiative feedbacks) may have contributed to the continued development of convection in the Lin simulation, while acting to inhibit strong convection in the WSM3 simulation. This suggests that in addition to accurately representing the convective characteristics (i.e., intensity, location, duration etc.), the proportion of stratiform and convective rain areas is also an important aspect for numerical models to correctly represent for forecasting tropical cyclogenesis.
It is interesting that the WSM3 simulation produced the most accurate result with respect to the intensity evolution of TCS025, despite the use of a less-sophisticated microphysics scheme. Although the explicit representation of graupel and mixed-phase processes is generally considered important for realistic high-resolution simulations, the additional complexity of the six-class schemes tended to degrade the solution. The overdevelopment when using more sophisticated microphysics schemes seems to agree with McFarquhar et al. (2006), who found that the use of microphysics schemes with additional hydrometeor species and mixed-phase processes resulted in more intense tropical cyclones compared to simulations that used less-sophisticated schemes. It is also important to consider that the physical processes responsible for the nondevelopment of TCS025 in the WSM3 simulation may have differed from those in reality. Similarly, the representation of convective processes in the Lin simulation that led to overdevelopment may have been a more correct response to thermodynamic conditions that resulted from other deficiencies in the model and/or erroneous initial conditions. Before this can be concluded, however, additional observation-based validation of the microphysical processes represented in numerical models is needed, especially for nondeveloping systems. It is possible that focusing solely on developing systems might mask model deficiencies that are more readily apparent upon careful examination of systems near the threshold of development.
In the adjoint sensitivity experiments by Doyle et al. (2012), wind, temperature, and moisture perturbations to the model initial state failed to yield significant differences in the development outcome for the TCS025 disturbance. As a result, Doyle et al. (2012) concluded that TCS025 was not close to the critical development threshold, at least from an initial condition perspective. In this study, however, the multiphysics ensemble and the additional simulations in which the diabatic heating rate was artificially adjusted reveal a stronger sensitivity from a model physics perspective. It is also important to note that the developmental differences for TCS025 were almost as large between two simulations (WSM6 and MYJ) that used different boundary layer schemes. Bao et al. (2012) also found that different boundary layer parameterization schemes led to large differences in the intensity evolution and structural characteristics of a simulated tropical cyclone. These findings indicate that the modeled development of TCS025 was more sensitive to the representation of microphysical and boundary layer processes than initial condition uncertainties. This is perhaps due to the fact that these changes alter the nature of convection throughout the duration of the simulation, and serves to highlight the impact of model uncertainty or error when integrated over time. The development of TCS025 in these simulations was sensitive to the nature of convection, such that the cumulative effect of small changes in the latent heating rate had a large impact on the development pathway and outcome.
Acknowledgments
This research was funded by National Science Foundation Grant ATM-0736003 and by the Office of Naval Research Marine Meteorology Grants N0001413WX20824 and N0001414WX20029. Support for JDD was provided by the Chief of Naval Research through the NRL Base Program and the Office of Naval Research’s Program Element 0601153N and 0602435N.
REFERENCES
Bao, J.-W., S. G. Gopalakrishnan, S. A. Michelson, F. D. Marks, and M. T. Montgomery, 2012: Impact of physics representations in the HWRFX on simulated hurricane structure and pressure–wind relationships. Mon. Wea. Rev., 140, 3278–3299, doi:10.1175/MWR-D-11-00332.1.
Bister, M., and K. A. Emanuel, 1997: The genesis of Hurricane Guillermo: TEXMEX analyses and a modeling study. Mon. Wea. Rev., 125, 2662–2682, doi:10.1175/1520-0493(1997)125<2662:TGOHGT>2.0.CO;2.
Chen, S.-H., and W.-Y. Sun, 2002: A one-dimensional time dependent cloud model. J. Meteor. Soc. Japan, 80, 99–118, doi:10.2151/jmsj.80.99.
Davis, C. A., and T. J. Galarneau, 2009: The vertical structure of mesoscale convective vortices. J. Atmos. Sci., 66, 686–704, doi:10.1175/2008JAS2819.1.
Doyle, J. D., C. A. Reynolds, C. Amerault, and J. Moskaitis, 2012: Adjoint sensitivity and predictability of tropical cyclogenesis. J. Atmos. Sci., 69, 3535–3557, doi:10.1175/JAS-D-12-0110.1.
Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 3077–3107, doi:10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2.
Dudhia, J., 1996: A multi-layer soil temperature model for MM5. Preprints, Sixth PSU/NCAR Mesoscale Model Users’ Workshop, Boulder, CO, NCAR, 3 pp. [Available online at http://www.mmm.ucar.edu/mm5/lsm/soil.pdf.]
Elsberry, R. L., and P. A. Harr, 2008: Tropical cyclone structure (TCS08) field experiment science basis, observational platforms, and strategy. Asia-Pac. J. Atmos. Sci., 44, 209–231.
Elsberry, R. L., T. D. B. Lambert, and M. A. Boothe, 2007: Accuracy of Atlantic and eastern North Pacific tropical cyclone intensity forecast guidance. Wea. Forecasting, 22, 747–762, doi:10.1175/WAF1015.1.
Emanuel, K. A., 1989: The finite-amplitude nature of tropical cyclogenesis. J. Atmos. Sci., 46, 3431–3456, doi:10.1175/1520-0469(1989)046<3431:TFANOT>2.0.CO;2.
Fritsch, J. M., and J. S. Kain, 1993: Convective parameterization for mesoscale models: The Fritsch–Chappell scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 46, Amer. Meteor. Soc., 159–164.
Fritz, C., and Z. Wang, 2013: A numerical study of the impacts of dry air on tropical cyclone formation: A development case and a nondevelopment case. J. Atmos. Sci., 70, 91–111, doi:10.1175/JAS-D-12-018.1.
Hendricks, E. A., M. T. Montgomery, and C. A. Davis, 2004: The role of “vortical” hot towers in the formation of Tropical Cyclone Diana (1984). J. Atmos. Sci., 61, 1209–1232, doi:10.1175/1520-0469(2004)061<1209:TROVHT>2.0.CO;2.
Hong, S.-Y., J. Dudhia, and S.-H. Chen, 2004: A revised approach to ice microphysical processes for the bulk parameterization of clouds and precipitation. Mon. Wea. Rev., 132, 103–120, doi:10.1175/1520-0493(2004)132<0103:ARATIM>2.0.CO;2.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, doi:10.1175/MWR3199.1.
Janjić, Z. I., 1990: The step-mountain coordinate: Physical package. Mon. Wea. Rev., 118, 1429–1443, doi:10.1175/1520-0493(1990)118<1429:TSMCPP>2.0.CO;2.
Janjić, Z. I., 1996: The surface layer in the NCEP Eta Model. Preprints, 11th Conf. on Numerical Weather Prediction, Norfolk, VA, Amer. Meteor. Soc., 354–355.
Janjić, Z. I., 2002: Nonsingular implementation of the Mellor–Yamada level 2.5 scheme in the NCEP Meso model. NCEP Office Note 437, NCEP, 61 pp.
Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170–181, doi:10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.
Kerns, B. W., and S. S. Chen, 2013: Cloud clusters and tropical cyclogenesis: Developing and nondeveloping systems and their large-scale environment. Mon. Wea. Rev., 141, 192–210, doi:10.1175/MWR-D-11-00239.1.
Lin, Y.-L., R. D. Farley, and H. D. Orville, 1983: Bulk parameterization of the snow field in a cloud model. J. Climate Appl. Meteor., 22, 1065–1092, doi:10.1175/1520-0450(1983)022<1065:BPOTSF>2.0.CO;2.
Mapes, B. E., and R. A. Houze Jr., 1995: Diabatic divergence profiles in western Pacific mesoscale convective systems. J. Atmos. Sci., 52, 1807–1828, doi:10.1175/1520-0469(1995)052<1807:DDPIWP>2.0.CO;2.
McFarquhar, G. M., H. Zhang, G. Heymsfield, J. B. Halverson, R. Hood, J. Dudhia, and F. Marks, 2006: Factors affecting the evolution of Hurricane Erin (2001) and the distributions of hydrometeors: Role of microphysical processes. J. Atmos. Sci., 63, 127–150, doi:10.1175/JAS3590.1.
Mellor, G. L., and T. Yamada, 1982: Development of a turbulence closure model for geophysical fluid problems. Rev. Geophys. Space Phys., 20, 851–875, doi:10.1029/RG020i004p00851.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682, doi:10.1029/97JD00237.
Montgomery, M. T., M. E. Nicholls, T. A. Cram, and A. B. Saunders, 2006: A vortical hot tower route to tropical cyclogenesis. J. Atmos. Sci., 63, 355–386, doi:10.1175/JAS3604.1.
Nicholls, M. E., and M. T. Montgomery, 2013: An examination of two pathways to tropical cyclogenesis occurring in idealized simulations with a cloud-resolving numerical model. Atmos. Chem. Phys., 13, 5999–6022, doi:10.5194/acp-13-5999-2013.
Nolan, D. S., 2007: What is the trigger for tropical cyclogenesis? Aust. Meteor. Mag., 56, 241–266.
Park, M.-S., and R. L. Elsberry, 2013: Latent heating and cooling rates in developing and nondeveloping tropical disturbances during TCS-08: TRMM PR versus ELDORA retrievals. J. Atmos. Sci., 70, 15–35, doi:10.1175/JAS-D-12-083.1.
Park, M.-S., A. B. Penny, R. L. Elsberry, B. J. Billings, and J. D. Doyle, 2013: Latent heating and cooling rates in developing and nondeveloping tropical disturbances during TCS-08: Radar-equivalent retrievals from mesoscale numerical models and ELDORA. J. Atmos. Sci., 70, 37–55, doi:10.1175/JAS-D-11-0311.1.
Penny, A. B., P. A. Harr, and M. M. Bell, 2015: Observations of a nondeveloping tropical disturbance in the western North Pacific during TCS-08 (2008). Mon. Wea. Rev., 143, 2459–2484, doi:10.1175/MWR-D-14-00163.1.
Penny, A. B., J. P. Hacker, and P. A. Harr, 2016: Analysis of tropical storm formation based on ensemble data assimilation and high-resolution numerical simulations of a nondeveloping disturbance. Mon. Wea. Rev., 144, 3631–3649, doi:10.1175/MWR-D-16-0100.1.
Rogers, R. F., M. L. Black, S. S. Chen, and R. A. Black, 2007: An evaluation of microphysics fields from mesoscale model simulations of tropical cyclones. Part I: Comparisons with observations. J. Atmos. Sci., 64, 1811–1834, doi:10.1175/JAS3932.1.
Sippel, J. A., and F. Zhang, 2008: A probabilistic analysis of the dynamics and predictability of tropical cyclogenesis. J. Atmos. Sci., 65, 3440–3459, doi:10.1175/2008JAS2597.1.
Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., doi:10.5065/D68S4MVH.
Waliser, D. E., and Coauthors, 2012: The “year” of tropical convection (May 2008–April 2010): Climate variability and weather highlights. Bull. Amer. Meteor. Soc., 93, 1189–1218, doi:10.1175/2011BAMS3095.1.
Wang, Z., 2012: Thermodynamic aspects of tropical cyclone formation. J. Atmos. Sci., 69, 2433–2451, doi:10.1175/JAS-D-11-0298.1.
Zhang, F., and J. A. Sippel, 2009: Effects of moist convection on hurricane predictability. J. Atmos. Sci., 66, 1944–1961, doi:10.1175/2009JAS2824.1.
Zhu, T., and D.-L. Zhang, 2006: Numerical simulation of Hurricane Bonnie (1998). Part II: Sensitivity to varying cloud microphysical processes. J. Atmos. Sci., 63, 109–126, doi:10.1175/JAS3599.1.
