2022 Volume 100 Issue 2 Pages 387-414
Typhoon Jongdari (2018) took an unusual track along the circumference of an upper-tropospheric cold low (UTCL) before making landfall in Japan on 29 July. To investigate the effects of atmosphere–ocean interactions and interactions between the UTCL and Jongdari on the storm's track, numerical simulations were conducted with a 3-km-mesh nonhydrostatic atmosphere model and an atmosphere–wave–ocean coupled model, using different initial conditions created by adopting different start times of numerical integration. The UTCL was characterized by high potential vorticity (PV), low pressure, and low relative humidity on the 355-K isotherm surface. While the UTCL moved southwestward north of Jongdari from 25 to 27 July, simulation results indicate that Jongdari traveled counterclockwise along the circumference of the UTCL. After Jongdari moved westward, the coupled model clearly simulated sea surface cooling along the track. Jongdari weakened after making landfall while the UTCL also weakened south of Japan. In particular, latent heat flux from the sea and the resulting humidification of the upper troposphere through the convection affected the UTCL. When Jongdari redeveloped over the ocean south of Kyushu, some simulations showed that Jongdari merged with the UTCL there as a result of high PV in Jongdari and relatively low upper-tropospheric PV near the UTCL. Ocean coupling helped sustain the uppertropospheric PV near the UTCL and weakened the column of elevated PV associated with Jongdari, which affected the location of the tropopause folding transformed from the UTCL by lowering the PV column of Jongdari and weakening the upper-tropospheric outflow from the center. Because the steering flow of Jongdari was affected by the geostrophic-balanced cyclonic circulation created by the UTCL, a larger difference of the atmospheric initial conditions between the initial times had a stronger influence on track and intensity simulations of both Jongdari and UTCL than ocean coupling.
Typhoon Jongdari originated as a tropical depression that evolved to a tropical storm around 19.7°N, 136.7°E at 12 UTC on 24 July 2018, according to the Regional Specialized Meteorological Center Tokyo best track data set (https://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-pub-eg/besttrack.html, accessed on 16 October 2021). It then followed an unusual course. After moving northward in the early intensification phase from 12 UTC on 24 to 06 UTC on 25 July, Jongdari moved cyclonically in a wide arc corresponding to a circle of approximately 600 km radius, reaching a central pressure of 960 hPa at 00 UTC on 27 July (Figs. 1a–c). Jongdari made landfall in Shima City, Mie Prefecture (the asterisk in Fig. 1c) at 15 UTC on 28 July and moved over the Japanese archipelago (Fig. 1d). It entered the East China Sea at 21 UTC on 29 July and moved cyclonically in a circular loop approximately 150 km in radius over the ocean south of Kyushu until 31 July (Figs. 1e, f); it then moved westward over the East China Sea. Elucidating the mechanism of Jongdari's irregular and unusual track is a scientifically interesting topic.
During its early intensification phase, Jongdari had to its north a cold vortex in the upper troposphere over the ocean east of Japan and west of Typhoon Wukong (2018) (Fig. 1a). The cold vortex was an upper-tropospheric cold low (UTCL) or upper-tropospheric cyclonic low, a low-pressure system that has become completely detached from the basic westerly flow in the jet stream (Nieto et al. 2005, 2008). These form more often in summer than in winter (Wei et al. 2016), are usually advected slowly toward the equatorial side of the mid-latitude westerlies, and often stay over the same region for several days (Gimeno et al. 2007). Their four-stage life cycle is divided into upper-level trough, tear-off, cut-off, and final stages. At the time Jongdari was intensifying on 25 July, the cold vortex was already cut off from the mid-latitude westerlies. From 25 to 28 July, the UTCL was in the cut-off stage and moved slowly southwestward while weakening. As Jongdari approached Japan, it moved cyclonically along the circumference of this UTCL.
Previous studies have suggested that UTCLs originate in a tropical upper-tropospheric trough (TUTT) and lead directly to the formation and development of tropical cyclones (TCs) because of the presence of the cold air mass aloft or by acting as additional outflow channels to reinforce TC development (Sadler 1976, 1978). The outflow channels are connected to large-scale westerlies through the enhancement of upper-level divergence, which creates favorable conditions for interactions of upper-level troughs with TCs (Sadler 1976). An investigation of the relationship between TCs and UTCLs in the western North Pacific from 2000 to 2012 by Wei et al. (2016) found that 83 % of UTCLs were part of a TUTT and the rest were cut off from westerlies; it also reported that 73 % of TCs coexisted with a UTCL and 44 % of TCs interacted with a UTCL, whereas only 21 % of UTCLs coexisting with TCs were within an initial cut-off distance of 15° from each other.
Wei et al. (2016) also showed that UTCLs can affect the track of TCs, depending on their relative distance and orientation. TCs in the southern half of a UTCL are more likely to intensify and those in the northern half are more likely to weaken. Moreover, TCs in a UTCL's northeastern quadrant tend to weaken more slowly than the others in the western North Pacific in a climatological view. Patla et al. (2009) proposed a graphical representation of the conceptual model of the influence of a TUTT cell or UTCL on TC motions as guidance for operational use at the Joint Typhoon Warning Center. Yan et al. (2021) reported that the removal of the observed UTCL from their numerical simulations of Jongdari substantially changed the simulated motion and intensity of Jongdari.
In addition to the above-mentioned statistical studies, the dynamics between a UTCL and a TC have never been studied. Molinari et al. (1998) noted that the dynamics between a UTCL and a TC are affected by three advective transports related to vertical wind shear, diabatic heating, and vortex interaction as follows: the advection of the upper-tropospheric low potential vorticity (PV) anomaly away from the TC toward the vertical wind shear vector (Shapiro 1992; Wu and Emanuel 1993), the mutual advection between the positive and negative PV anomalies at 350 K, and the mutual advection between 350-K anomalies and anomalies above and below them (Jones 1995). These dynamics, however, do not include the effect of multiscale interactions such as those between the two vortices and mid-latitude westerlies and between the atmosphere and the ocean. In that sense, the understanding of the dynamics of the interactions between a UTCL and a TC is still not complete. Numerical simulation of a TC using a sophisticated atmosphere–wave–ocean-coupled model (Wada et al. 2010, 2018) is one of the effective methods for studying multiscale interactions. The dynamics of vortex interactions as well as the behavior of a UTCL moving toward low latitudes should be clarified in the coupled atmosphere–ocean model framework to better understand the role of the UTCL in Jongdari's unusual track and to improve the accuracy of track predictions.
This study sought to understand the mechanism by which the UTCL influenced Jongdari's track and intensity evolution and to understand the role of atmosphere–ocean interactions and initial conditions in that mechanism. For this purpose, we conducted numerical simulations with a 3-km-mesh nonhydrostatic atmosphere model (NHM) and an atmosphere–wave–ocean coupled model (CPL), using different atmospheric and oceanic initial conditions generated by adopting different initial times. The initial time means a start time of numerical integration.
This paper is structured as follows. Section 2 explains the experimental design of our numerical simulations. Section 3 consists of five subsections and describes the results obtained from simulations using NHM and CPL with different initial times. Section 4 presents the discussion and summary.
Our simulations using NHM and CPL required initial conditions for the atmosphere and ocean and boundary conditions for the atmosphere; the ocean–surface–wave model included in CPL assumed a motionless initial condition.
The atmospheric initial and boundary conditions were created from the Japan Meteorological Agency (JMA)'s 6-hourly global atmospheric analysis data with a horizontal grid spacing of approximately 20 km. The initial times ranged from 12 UTC on 25 July to 12 UTC on 28 July 2018 at 6-h intervals. The atmospheric boundary condition at each initial time was created every 6 h during the integration time. The oceanic initial condition, other than sea surface temperature (SST), was created from the JMA daily North Pacific oceanic analysis data, which include water temperature, salinity, currents, and sea surface height anomaly data at a horizontal resolution of 0.5° in longitude and latitude. The oceanic initial condition was created at 6-h intervals from 12 UTC on 25 July to 12 UTC on 28 July 2018, depending on the initial integration day because the oceanic analysis data are a daily product, and the same data are imposed throughout a day.
The SST initial condition was derived from the Microwave Optimally Interpolated SST daily product (obtained from the Remote Sensing Systems site, http://www.remss.com/, accessed on 16 October 2021). This is a daily merged satellite dataset that combines measurements by the WindSat (Gaiser et al. 2004), the Advanced Microwave Scanning Radiometers 2, and the Global Precipitation Measurement Microwave Imager. The dataset covers the global ocean with a 0.25° horizontal spacing at a depth of approximately 1 m. We used brightness temperatures from the Himawari-8 infrared imager to monitor the behavior of the UTCL and hourly atmospheric motion vectors above 350 hPa in height, derived from the Himawari-8 brightness temperatures, to monitor the upper-tropospheric flow.
2.2 ModelsThe NHM and CPL used in this study have been used in many other studies to analyze the role of atmosphere–ocean interactions on TC simulations (e.g., Wada et al. 2018; Wada and Oyama 2018; Oyama and Wada 2019; Wada 2021). NHM was introduced in Saito (2012). The components of the coupled model (Wada et al. 2010) include the Meteorological Research Institute (MRI) third-generation ocean-surface wave model MRI-III (Japan Meteorological Agency 2013) and a multilayer ocean model developed by MRI based on Bender et al. (1993). The ocean model includes a diurnally varying SST scheme based on Schiller and Godfrey (2005) with the short-wave absorption/penetration formulation of Ohlmann and Siegel (2000). Details of the MRI models and the exchange processes between the atmosphere, oceansurface waves, and ocean are described in Wada et al. (2018).
The physical processes in our models are important for determining the accuracy of simulation results. We relied on an explicit three-ice bulk microphysics scheme (Ikawa and Saito 1991; Lin et al. 1983). Air–sea momentum fluxes and sensible and latent heat fluxes with exchange coefficients for air–sea momentum and enthalpy transfers over the sea were based either on bulk formulas (Kondo 1975) or, when the ocean-wave model was coupled (Wada et al. 2010), on the roughness lengths proposed by Taylor and Yelland (2001), a sea spray formulation (Wada et al. 2018), a turbulent closure model (Klemp and Wilhelmson 1978; Deardorff 1980), and a radiation scheme (Sugi et al. 1990). These are also the same as those specified in Wada et al. (2018) and Wada (2021).
2.3 Configuration of numerical simulationsThe horizontal resolution of NHM and CPL was 3 km. The computational domain was single and approximately 2,760 × 2,760 km, centered at 30°N, 140°E. All simulations used 55 levels in vertical coordinates, with intervals ranging from 40 m for the near-surface layer to 1,013 m for the uppermost layer. The top height was approximately 27 km. The lowermost end index for upper Rayleigh damping layer was 44 (approximately 17.5 km). The width of lateral boundary relaxation sponge layers was 150 km in all experiments. The time steps were 6 s for NHM, 36 s for the ocean model, and 6 min for MRI-III. The physical components were exchanged between NHM and the ocean model every 36 s and between those models and MRI-III every 6 min. The integration time was 144 h in all experiments. Simulations by each model incorporated 13 initial conditions at 6-h intervals, for a total of 26 simulations. It should be noted that the difference in the width of lateral boundary relaxation sponge layers could affect the track and intensity of the simulated Jongdari; therefore, the simulation results were validated with RSMC best track data.
The passage of Jongdari during its intensification phase induced sea surface cooling on the right side (outside) of the storm track along its counterclockwise track (Fig. 2). This cooling is known as the negative feedback effect on TCs and helped suppress the intensification of Jongdari (e.g., Bender et al. 1993; Wada et al. 2018). However, SST was relatively high at 28–29°C from 25 to 28 July over the ocean south of Kyushu, where Jongdari passed after 21 UTC on 29 July. The warm surface water south of Kyushu was a favorable condition for Jongdari's second intensification (Kuo et al. 2018; Wu et al. 2008).
JMA weather maps at 00 UTC on (a) 26 July, (b) 27 July, (c) 28 July, (d) 29 July, (e) 30 July, and (f) 31 July 2018. Contour intervals are 4 hPa for solid isobars and 2 hPa for dashed isobars. TD, tropical depression; TS, tropical storm; L (H), area where sea-level pressure is lower (higher) than in the surroundings.
Daily SST (colors in the right-hand color bar) from 25 July to 1 August 2018 with best track positions of Jongdari every 6 h (circles). Colors of the circles (left-hand color bar) indicate the best track central pressure. The large circle indicates the position of Jongdari at the time of the plot.
The brightness temperature and atmospheric motion vectors clearly show the dry area around the UTCL and the clockwise flow circulation centered on the continental high at 12 UTC on 25 July (Fig. 3). Northeasterly winds were relatively strong southwest of the UTCL at that time (“C” in Fig. 3). While the UTCL moved southwestward, Jongdari, represented by a cluster of low brightness temperature areas, moved cyclonically along its circumference. As Jongdari approached the UTCL after making landfall in Japan, the dry area in the UTCL gradually became subdued, an indication that the UTCL was weakening. After Jongdari moved over the ocean south of Kyushu, the TC appeared to occupy the same position as the UTCL. Our simulations of the interactions between the UTCL and Jongdari sought to investigate the role of Jongdari, a marginal TC (Molinari et al. 1998), in the UTCL transition and vice versa.
Brightness temperatures (Band 13, gray scale) and atmospheric motion vectors (arrows; red arrow represents 30 m s−1) between 250 hPa and 350 hPa heights at 12 UTC from 25 July to 1 August. Red circles indicate the position of Jongdari, and the color indicates the minimum central pressure. The letter “C” from 12 UTC on 25 July to 12 UTC on 27 July indicates the location of the UTCL.
Numerical simulations of TCs are strongly affected by the uncertainty of atmospheric and oceanic initial conditions (e.g., Wada and Kunii 2017; Wang and Wu 2004). Even if a simulation at a given initial condition fits the observations and the best track analysis of a TC, it does not mean that the numerical system used can be successfully repeated with another initial condition at another start time of numerical integration. We therefore, used ensemble simulation results from NHM and CPL to obtain more accurate simulated Jongdari's irregular track without affecting the uncertainty included in the atmospheric initial conditions.
Our two sets of 13 simulations successfully tracked the center position and central pressure of the simulated Jongdari from each initial time and in each model used. The simulated center position of Jongdari was determined as the grid point with the lowest pressure at sea level. The center of the simulated UTCL was the grid point with the lowest pressure at an altitude of 12,000 m, consistent with previous studies (e.g., Wei et al. 2016). Both positions were tracked from 12 UTC on 25 July to 00 UTC on 2 August.
Regarding the predictability of Jongdari's track, the error tendencies in predictions carried out by major numerical prediction centers such as the European Centre for Medium-Range Weather Forecasts (ECMWF), the Meteorological Service of Canada, and Deutsche Wetterdienst showed westward deflection in the first intensification phase and northward deflection in the mature, landfalling, and weakening phases when the initial time of the prediction was 12 UTC on 25 July (not shown). The ECMWF ensemble forecasts showed a westward shift in the track, whereas those of the National Centers for Environmental Prediction showed an eastward shift (Lei et al. 2020). One factor in this difference in track forecasts is the poor simulation of intensity due to the model's coarse horizontal resolution (Fierro et al. 2009; Kanada and Wada 2016).
Figures 4a, b show the tracks of Jongdari and the UTCL simulated by NHM (hereafter, the noncoupled-model simulation) (Fig. 4a) and by CPL (hereafter, the coupled-model simulation) (Fig. 4b). Both of these simulations reduced the northward deflection in the mature, landfalling, and weakening phases to some extent compared with the previously mentioned track forecasts, but they did not reduce the westward deflection in the intensification phase. In addition, all simulations failed to reproduce the looping feature south of Kyushu. The smaller loop in the simulations was different from the larger loop of the best track analysis. The simulated UTCL first appeared east of Japan around 35°N, 148°E, moved southwestward over the ocean, and then stayed around 30°N, 136°E, which is consistent with the observed behavior of the UTCL shown in Fig. 3. Figures 4c, d show the positions of simulated UTCL relative to the simulated TC positions in the noncoupled- (Fig. 4c) and coupled-model simulation (Fig. 4d). The simulated UTCL approached the TC center counterclockwise from 12 UTC on 25 July to 00 UTC on 29 July and then stayed on the east side of the TC center. The noncoupled- and coupled-model simulations had no notable difference in their TC tracks, which suggests that the sea surface cooling induced by Jongdari had little or no effect on the track simulations. This result is consistent with previous studies (Bender et al. 1993; Ito et al. 2015; Mogensen et al. 2017; Wada et al. 2018). In addition, the sea surface cooling induced by Jongdari had little or no effect on the relative position of the simulated UTCL to the simulated TC. In other words, the difference in the simulated TC track and movement of the UTCL was mainly caused by the difference in the initial conditions.
Simulations by NHM (a, c) and CPL (b, d). (a, b) Blue thick line with circles (colors indicate the central pressure) and error bars (one standard deviation) shows the ensemble mean TC tracks at 12 h intervals. Black thick line with diamonds shows best track TC positions at 12 h intervals. Red thick line with stars (colors indicate the temperature) and error bars (one standard deviation) shows locations of the UTCL at 12 h intervals. “A” (12 UTC on 25 July), “B” (06 UTC on 27 July), “C” (12 UTC on 27 July), “D,” (18 UTC on 27 July), and “E” (00 UTC on 28 July) with thin lines indicate the simulated TC track, whereas “a” (12 UTC on 25 July), “b” (06 UTC on 27 July), “c” (12 UTC on 27 July), “d,” (18 UTC on 27 July), and “e” (00 UTC on 28 July) with thin lines indicate the simulated locations of UTCL at each initial time. (c, d) Relative position evolutions between TC (origin, red mark) and UTCL (the end of the arrow) from 12 UTC on 25 July to 12 UTC on 31 July. Each two digits depicted in each panel as four digits mean day and time (UTC).
Figure 5a shows the time series of the best track central pressure, mean central pressure averaged from 13 simulations of the noncoupled- and coupled-model simulations, respectively, with the time series of central pressure at five different initial times (12 UTC on 25, 09 UTC on 27, 12 UTC on 27, 18 UTC on 27 and 00 UTC on 28 July). These show underdevelopment from 12 UTC on 25 July to 00 UTC on 29 July and then overdevelopment after 00 UTC on 29 July. At the time when the real Jongdari was over land at 16 UTC on 28 July, the simulated Jongdari was moving over the ocean to the south of the best track, which was favorable for the development of Jongdari. The ocean coupling effect tended to alleviate the overdevelopment of the simulated Jongdari due to sea surface cooling (Bender et al. 1993; Ito et al. 2015; Mogensen et al. 2017; Wada et al. 2018; Wada 2021). Figure 5b shows the time series of the mean radius of the maximum wind speed at 20-m altitude averaged from 13 simulations of the noncoupled- and coupled-model simulations, respectively. The range of the evolution of the radius of the maximum winds varied from 30 km to 80 km from 00 UTC on 26 July to 12 UTC on 31 July, an indication that the size of the simulated TC was relatively small compared with the distance between the TC and the UTCL (Figs. 4c, d). When the difference in the average radius of the maximum winds between the noncoupled- and coupled-model simulations was tested, no significant difference was found at the 95 % confidence level based on t-test. Therefore, the effect of ocean coupling on the simulated TC size is not significant in the present study.
(a) Time series showing the best track central pressure (gray circles), ensemble mean central pressures with the error bar (one standard deviation) simulated by NHM (red circles with the vertical line) and the atmosphere–wave–ocean coupled model (light blue circles with the vertical line) and simulated central pressures with atmospheric initial conditions at 12 UTC on 25 July, 06, 12, 18 UTC on 27 July and 00 UTC on 28 July. See Table 1 for the identity of the diff erent simulations. (b) Time series showing ensemble mean radius of the maximum winds at 20-m height with the error bar (one standard deviation) simulated by NHM (red circles with the vertical line) and CPL (light blue circles with the vertical line).
Table 1 lists the simulation results (position, central pressure, and maximum wind speed at 20 m), the best track data for 144 h for each initial time and their averages. The average position from the coupled-model simulations (30.13°N, 133.00°E) was southeast of the average position from the noncoupled-model simulations (30.24°N, 132.97°E) and slightly closer to the average position obtained from the best track data (30.65°N, 135.37°E). The same was true for the average central pressures as follows: 978.52 hPa from the coupled-model simulations, 973.90 hPa from the noncoupled-model simulations, and 980.19 hPa from the best track data. The average maximum wind speed was 28.47 m s−1 in the coupled-model simulations, which was weaker than the noncoupled-model results (33.74 m s−1) and closer to the best track result (25.54 m s−1). These results suggest that the use of CPL slightly improved the track and intensity predictions for Jongdari. However, the prediction error due to the diff erence in initial times was much greater than the eff ect of ocean coupling (Table 1).
Figure 6 shows the time series of the average diff erences in the track and central pressure between the noncoupled- and coupled-model simulations. CPL reduced the error in the track simulations after 60 h compared with NHM. The simulated intensity was closer to the best track for the noncoupled-model simulations up to 60 h and for the coupled-model simulations after 84 h. The improvement in central pressure simulations in the latter part of the integration time is an expectable effect of ocean coupling (Bender et al. 1993; Ito et al. 2015; Mogensen et al. 2017; Wada et al. 2010, 2018; Wada 2021). The question is how ocean coupling affected the TC vortex itself and the UTCL surrounding the TC. The next section discusses how Jongdari and the UTCL behaved in the simulation results.
(Left) Mean track errors with respect to the best track position and (right) central pressure errors with respect to the mean best track central pressure in the noncoupled-model simulations (red circles and lines), and coupledmodel simulations (blue circles and lines). The error bars indicate one standard deviation.
This section presents Jongdari and the UTCL in the noncoupled-model simulations. Figure 7 maps pressure at an altitude of approximately 10 km along with sea-level pressure and Ertel's PV (Davis and Emanuel 1991) (1 PV unit = 10−6 m2 s−1 K kg−1) on the 355-K isotherm surface and shows vertical profiles of PV, potential temperature, and horizontal–vertical wind vectors parallel to the cross section along the line connecting the centers of Jongdari and the UTCL at the integration times of 0, 36, 72, and 108 h after the initial time of 12 UTC on 25 July. It should be noted that the height of 10 km is lower than that of the UTCL (12 km) used in Figs. 3 and 4, although this difference does not affect our conclusions. The horizontal distributions in all cases in this paper are averaged over the neighboring 16-grid cells (approximately 50 km), except the distribution of hourly precipitation.
Simulations by NHM with initial conditions at 12 UTC on 25 July at integration times of (a–c) 0 h, (d–f) 36 h, (g–i) 72 h, and (j–l) 108 h. The left panels show the distribution of pressure at an altitude of approximately 10 km (colors), sea-level pressure (black lines, contour interval 8 hPa), and wind vectors at an altitude of approximately 10 km. The center panels show the distribution of PV (colors) and wind vectors on the 355-K isotherm surface. The right panels show a vertical profile of PV (colors), potential temperature (black lines, interval 10 K), and horizontal–vertical wind vectors parallel to the cross section along the line between the centers of Jongdari and the UTCL (see Section 3.2), shown as dashed lines in the left and central panels. “A–H” indicate the location of the start and end points of the cross section. “T” and “V” in Fig. 7l are explained in the text.
In the intensification phase (0–36 h), the TC vortex was relatively small (Figs. 7a, b), resembling that of a marginal tropical storm (Molinari et al. 1998) or a midget TC (Lander 1994). However, a tower of positive PV (> 1 PV unit) associated with the TC (hereafter referred to as the PV tower) was clearly depicted extending from the lower to the middle troposphere (Fig. 7c). On the other hand, the UTCL was characterized by low pressure around 10 km and high PV in the upper troposphere (Fig. 7b). As Jongdari moved north–northeastward in the simulation and the UTCL moved southwestward (Figs. 7d, e), southerly winds around the eastern edge of the UTCL strengthened. The simulated TC moved along the northeastern edge of the cyclonic circulation, and the value of PV on the 355-K isotherm surface became small. Jongdari was intensifying in the simulation when the outflow from the PV tower became evident (Fig. 7f).
As the simulated Jongdari moved westward along the Japanese coast (Fig. 7g) and the high-PV area in the UTCL descended in altitude while the UTCL was moving southwestward (Fig. 7h), the PV tower approached the Japanese archipelago (Fig. 7i) and then weakened (Fig. 7l). The high-PV area on the 355-K isotherm surface was stretching and cyclonically folding (Fig. 7h), resembling the deformation of a fluid surface computed by a barotropic model in which the layers behave like a two-dimensional ideal fluid (Welander 1955). In the simulation, Jongdari continued to follow the geostrophic-balanced cyclonic circulation centered at the UTCL, which is one of the factors that affects the steering flow. When Jongdari moved over the Japanese archipelago, the central pressure in the UTCL increased. This indicates that the TC weakened at that time as a result of surface friction on land. Indeed, the PV tower weakened during the passage over land (not shown). When Jongdari moved over the ocean south of Kyushu in the simulation (Fig. 7j), the cyclonic circulation centered at the area within the UTCL shrank in size (Fig. 7k) and the PV tower intensified again. Even though the UTCL became weak, for convenience we continue to refer to it as the UTCL. At 00 UTC on 30 July in the simulation, the tilt of the upper-level high PV (> 1 PV unit) (Agusti-Panareda et al. 2004) or tropopause folding (Price and Vaughan 1993; Bosart 2003) had reversed from its direction at 36 h and 72 h (Fig. 7l), and the location of the TC appeared to be identical to that of the UTCL.
Geostrophic-balanced cyclonic circulation was induced below the UTCL at the initial time (Fig. 8a). As the UTCL moved southwestward at 36 h, the distance between the TC and the UTCL became closer than before, but the geostrophic circulation of Jongdari was still separated from the geostrophic-balanced cyclonic circulation centered within the UTCL (Fig. 8b). The geostrophic flows within the inner core of Jongdari were clearly found in the intensification phase, although, in general, the gradient wind balance is established rather than the geostrophic wind balance within the inner core of a TC (e.g., Miyamoto et al. 2014). At 72 h, geostrophic-balanced cyclonic circulation centered within the UTCL was located in the area centered around 29°N, 136°E (Fig. 8c). The TC moved westward along the northern edge of the UTCL-induced geostrophic-balanced cyclonic circulation. At 108 h, the geostrophic-balanced cyclonic circulation was not clear and became part of the inner-core structure of Jongdari (Fig. 8d). This suggests that the magnitude of the UTCL-induced geostrophic-balanced cyclonic circulation became weak. The gradient winds of the simulated TC may affect the steering flows. The modification of the steering flow due to the excessively simulated gradient winds of Jongdari possibly affect the difference in TC tracks between the simulations and the best track analysis.
Horizontal distribution of the magnitude of geostrophic flow on the 355-K isotherm surface (colors), sea-level pressure (purple lines, contour interval 8 hPa), and geostrophic flow vectors on the 355-K isotherm surface simulated by NHM with initial conditions at 12 UTC on 25 July at integration times of (a) 0 h, (b) 36 h, (c) 72 h, and (d) 108 h.
Next, we present the thermodynamic conditions of Jongdari and the UTCL in the simulation. Figure 9 maps the horizontal moisture flux (specific humidity multiplied by momentum per unit mass) at an altitude of approximately 10 km, the relative humidity at the height of the 355-K isotherm surface, and the relative humidity along with the potential temperature and horizontal–vertical wind vectors parallel to the cross section along the line between the centers of Jongdari and the UTCL at the integration times of 0, 36, 72, and 108 h after the initial time of 12 UTC on 25 July. Each panel in Fig. 9 is a counterpart to one in Fig. 7. The altitude of the horizontal moisture flux and relative humidity is set to 10 km to clearly show the difference between the dry area at the lower end of the UTCL and the convection area of the TC.
Simulations by NHM with initial conditions at 12 UTC on 25 July at integration times of (a–c) 0 h, (d–f) 36 h, (g–i) 72 h, and (j–l) 108 h. The left panels show the distributions of horizontal moisture flux (colors) at an altitude of approximately 10 km. The center panels show the relative humidity on the 355-K isotherm surface (colors). The right panels show a vertical profile of relative humidity (colors), potential temperature (black lines, interval 10 K), and horizontal–vertical wind vectors parallel to the cross section along the line between the centers of Jongdari and the UTCL shown as white lines in the left and center panels. “A–H” indicate the location of the start and end points of the cross section. “X” in Fig. 9i is explained in the text.
At an altitude of approximately 10 km, the horizontal moisture fluxes were relatively high on the western side of the UTCL and higher than the moisture fluxes around Jongdari at the initial time, 12 UTC on 25 July (Fig. 9a). This high-moisture area corresponds to the area where the relative humidity was higher than 50 % on the 355-K isotherm surface, while the relative humidity at the center of the UTCL was close to zero (Fig. 9b). The area of > 50 % relative humidity was spread zonally around Jongdari. The cross section between Jongdari and the UTCL shows that the tropopause, where the vertical temperature gradient is steeper than that within the troposphere, dropped to an altitude of approximately 12 km around 36°N, 147°E, while the air with relatively high relative humidity (> 70 %) at an altitude of around 8–10 km near 32.5°N, 145°E was carried upward to the upper troposphere (Fig. 9c).
At 36 h integration time, the moisture flux was relatively high in the arc-shaped area from north to east of UTCL (Fig. 9d). High-moisture fluxes around Jongdari were on the southeastern side of the UTCL and joined the arc-shaped area. The area of low relative humidity (< 20 %) stretched horizontally to the southwest and then folded around 30°N, 140°E (Fig. 9e). The shape of the arc of the dry area was opposite to the arc-shaped area of the moisture fluxes. The relative humidity around Jongdari increased on the 355-K isotherm surface during the intensification phase of TC, and the distance between the UTCL and Jongdari decreased (Figs. 4, 9f). The folding of the tropopause around the UTCL was deflected toward Jongdari. Immediately below the area of folding, the relative humidity was locally higher than 70 % around 31°N, 139°E, and at an altitude of 8–10 km.
At 72 h, the moisture flux was highest around Jongdari just before landfall, and relatively high to the north and northwest of the UTCL. On the 355-K isotherm surface, an arc of relatively dry air south of Jongdari formed as the dry area of the UTCL combined with another body of dry air, a slot in the middle-to-upper troposphere that flowed cyclonically from the continent (Fig. 9h). The flow of this dry slot was captured by the atmospheric motion vectors above 350 hPa (Fig. 3). The UTCL gained moisture while it was moving southwestward and the distance between the UTCL and Jongdari further decreased (Figs. 4, 9i). Since the TCs simulated by NHM and CPL are affected by the lateral boundary conditions updated every 6 hours (e.g., Wada 2017), the influence of the dry slot on the interactions between the UTCL and Jongdari may be affected by the setting of the computational domain and the width of the lateral boundary, as explained in Section 2.3. The effect of setting the width of the lateral boundary on the simulation of the dry slot is beyond the scope of this study.
At 108 h, high-moisture flux was confined to an area around Jongdari (Fig. 9j). On the 355-K isotherm surface, an area of > 50 % relative humidity likewise surrounded Jongdari (Fig. 9k), and the area of > 70 % relative humidity around Jongdari had become reduced in altitude from its height at 72 h (Fig. 9l). The UTCL structure was no longer visible in the cross section. During these movements of the UTCL, the dry air within it was gradually humidified and the tropopause around it rose, and it became obscured as it approached and then coalesced with Jongdari.
This humidification process below the UTCL should degrade the capacity of the UTCL to sustain its low pressure and dry condition in the upper troposphere due to increases in specific humidity at an altitude of approximately 10 km from 0.1 g kg−1 at the initial time to 0.6 g kg−1 at 108 h at the center of the UTCL (Fig. 4a). This means that the low pressure of the UTCL was hardly sustained due to the humidification process and thereby increases in specific humidity in the UTCL. The increases in specific humidity (humidification) were considered to be caused by cumulus convection over the warm ocean and its associated diabatic heating since the interaction between the TC and the ocean plays a crucial role in supplying heat and moisture from the ocean to the atmosphere and in transporting the heat and moisture upward by cumulus convection around the TC and the edge of the UTCL. Given that the TC intensity produced by NHM was stronger than the best track TC intensity (Table 1), the following questions arise: how ocean coupling processes affect the interaction between Jongdari and the UTCL and how simulation results can better incorporate ocean coupling.
3.4 Effect of ocean coupling on simulated Jongdari and UTCLSea surface cooling such as that induced by Jongdari along its track, shown in Fig. 2, is mainly caused by vertical turbulent mixing and upwelling in the upper ocean (Price 1981). This fact suggests that CPL or at least the atmosphere–ocean coupled model is required to reflect the dynamic and thermodynamic processes in simulations of Jongdari. An accurate simulation of sea surface cooling requires an accurate atmospheric forcing to be applied to CPL as well as an accurate oceanic initial condition, particularly the stratification in the upper ocean. In addition, CPL needs to simulate surface wind speeds realistically. Here we investigate the results simulated by CPL in detail. The ocean waves simulated by CPL affect the roughness length over the ocean and thereby change the wind stress or frictional velocity between the atmosphere and the ocean as well as the vertical turbulent mixing caused by breaking waves (Wada et al. 2010).
Figure 10 maps the SST simulated by CPL. The SST initial condition at 12 UTC on 25 July successfully matches the observations shown in Fig. 2, an indication that the SST initial condition was well created by interpolations in the simulations. However, the simulated sea surface cooling induced by Jongdari was relatively weak because the simulated intensity of Jongdari was weaker than the best track intensity even when NHM was used (Table 1). The reason why the sea surface cooling was small as the TC moved rapidly westward is that the vertical turbulent mixing beneath the TC had weakened owing to the weakening of the atmospheric forcing.
Horizontal distribution of SST simulated by CPL from 25 July to 1 August with initial conditions at 12 UTC on 25 July and simulated positions of Jongdari every 6 h (circles). The large circle indicates the simulated position of Jongdari at the time of the plot. The colors of the circles indicate the simulated central pressure.
Figure 11 maps hourly precipitation in simulations by NHM and CPL at the integration times of 36, 72, and 108 h. NHM simulated heavy rainfall around the TC center at 36 h (Fig. 11a), and another area of precipitation was centered around 29°N, 145°E, where the relative humidity on the 355-K isotherm surface was relatively high (see Fig. 9e). As Jongdari approached land at 72 h, its center was an area of heavy rainfall, and narrow spiral rainbands trailed it on its southeastern side (Fig. 11b). At 108 h, a concentric rainfall pattern surrounded Jongdari as the TC redeveloped south of Kyushu (Fig. 11c). The simulation by CPL showed a small effect of ocean coupling on the distribution of hourly precipitation at 36 h (Fig. 11d). At 72 and 108 h, however, the area of heavy precipitation became smaller than in the noncoupledmodel simulations (Figs. 11e, f). The presence of narrow spiral rainbands below the UTCL during the integration reveals that local convection and associated diabatic heating occurred below the UTCL.
Distributions of hourly precipitation (colors) and sea-level pressure (contours, interval 8 hPa) in the noncoupled-model simulation at integration times of (a) 36 h, (b) 72 h, and (c) 108 h and in the coupled-model simulation at integration times of (d) 36 h, (e) 72 h, and (f) 108 h.
Figure 12 maps latent heat fluxes from the ocean to the atmosphere simulated by NHM and CPL at the integration times of 36, 72, and 108 h. The latent heat flux was relatively high around the edge of the cyclonic circulation and exceeded 400 W m−2 around the simulated TC and along the south coast of Japan around 34.5°N, 139°E (Fig. 12a). At 72 h, when the simulated TC approached the Japanese archipelago, the latent heat flux exceeded 400 W m−2 along the south coast of Japan around 34.5°N, 137°E, while the latent heat flux around the edge of cyclonic circulation became smaller than before (Fig. 12b). At 108 h, the area of latent heat flux exceeding 400 W m−2 was clearly found only within the inner core of simulated TC (Fig. 12c). The difference in latent heat fluxes caused by ocean coupling was found within the inner core of the simulated TC at 36 h. The latent heat flux also decreased below the UTCL around 32°N, 139°E, although the amount of the decreases was relatively small compared to that around the TC. The decrease in latent heat fluxes was found not only around the TC but also below the UTCL around 30°N, 135°E (Figs. 12e, f). The area of the decreases in latent heat fluxes extended from the limited TC area to the entire area below the UTCL as the integration time proceeded. Hereinafter, it will be shown that the reduction in latent heat fluxes below the UTCL due to ocean coupling helped suppress the convection and associated diabatic heating there.
Distributions of latent heat flux (colors) and sea-level pressure (contours, interval 8 hPa) in the noncoupled-model simulation at integration times of (a) 36 h, (b) 72 h, and (c) 108 h and in the coupled-model simulation at integration times of (d) 36 h, (e) 72 h, and (f) 108 h.
Figure 13 corresponds to Fig. 7, except the results simulated by CPL. At 36 h, there was no significant difference between the noncoupled- and coupled-model simulations of the distribution of pressure at an altitude of approximately 10 km or PV on the 355-K isotherm surface (compare between Figs. 13a, b and Figs. 7d, e); however, the effect of ocean coupling appeared in a decrease in the height of the PV tower of Jongdari and its magnitude (compare Fig. 13c and Fig. 7f). In addition, ocean coupling decreased the upper-tropospheric outflow by more than 10 m s−1 oriented from the top of the PV tower at 14–16 km and thus, modified the locations of low-PV areas formed in the upper troposphere.
At 72 h, the pressure at 10 km altitude at the center of the UTCL was lower in the coupled-model simulation (Fig. 13d) than in the noncoupled-model simulation (Fig. 7g). The PV surrounding the UTCL was higher in the coupled-model simulation (Fig. 13e) than in the noncoupled-model simulation (Fig. 7h). The height of the PV tower was shorter in the coupled-model simulation (Fig. 13f) than in the noncoupled-model simulation (Fig. 7i). These differences due to ocean coupling were more apparent at 108 h (compare between Figs. 13g, h and Figs. 7j, k). They caused a delay in the coalescence of the UTCL and Jongdari; in the noncoupled-model simulation, the PV tower extended from the surface to the tropopause (“V” in Fig. 7l) and the tropopause folding tilted away from the PV tower (“T” in Fig. 7l), whereas in the coupled-model simulation (Fig. 13i), the tropopause folding (“V” in Fig. 13i) still tilted toward the PV tower (“T” in Fig. 13i).
Figure 14 shows the difference in geostrophic flows at an altitude of approximately 10 km between the noncoupled- and coupled-model simulations with wind vectors indicating geostrophic flows in the coupled-model simulation. The map of the difference in geostrophic flows at 36 h (Fig. 14a) shows that the magnitude of the geostrophic flows was almost the same as that in the noncoupled-model simulation (Fig. 8b), although the magnitude around the TC was ∼ 25 m s−1 smaller in the coupled-model simulation (Fig. 14a) than that in the noncoupled-model simulation (Fig. 8b). These features were also found at 72 h (Fig. 14b). However, the geostrophic-balanced cyclonic circulation was ∼ 20 m s−1 stronger in the coupledsimulations (Fig. 14b). At 108 h, the difference in geostrophic flows exceeded 20 m s−1 only around the TC (Fig. 14c). Even though the locations of the TC and the UTCL significantly differed between the noncoupled- and coupled-model simulations, particularly at the latter integration time (72 h and 108 h), the UTCL-induced geostrophic-balanced cyclonic circulation was not clearly found east of the simulated TC at 108 h and thus, it is hard to determine the direct impact of ocean coupling on the geostrophic-balanced cyclonic circulation.
Horizontal distributions in the difference in magnitude (colors) and vector (arrows) of geostrophic flows between the noncoupled-model and coupled-model simulations (CPL minus NHM) with initial conditions at 12 UTC on 25 July at integration times of (a) 36 h, (b) 72 h, and (c) 108 h. The contours indicate sea-level pressure simulated by CPL. The interval is 8 hPa.
Figure 15 corresponds to Fig. 9, except the results simulated by CPL. At 36 h, ocean coupling had produced no significant difference in the maps (compare Figs. 15a–c and Figs. 9d–f) between the noncoupled- and coupled-model simulations, except regarding the area around the PV tower, where relative humidity was relatively high in the noncoupled-model simulation.
At 72 h, the area where moisture flux exceeded 2 g m−2 s−1 at an altitude of 10 km around the circumference of the UTCL (Fig. 15d) was larger than the area in the noncoupled-model simulation (Fig. 9g). The area with less than 10 % relative humidity on the 355-K isotherm surface around the UTCL was smaller (Fig. 15e) than in the noncoupled-model simulation (Figs. 9g, h), but the downward intrusion of dry air from the UTCL toward Jongdari was stronger above an altitude of 12 km in the coupled-model simulation (“X” in Fig. 15f) than that in the noncoupled-model simulation (“X” in Fig. 9i). At 108 h, the moisture flux at an altitude of 10 km near Jongdari was more widespread in the coupled-model simulation (compare Fig. 15g and Fig. 9j). Unlike the result at 72 h (Fig. 15e), an area with < 10 % relative humidity on the 355-K isotherm surface was apparent on the north side of Jongdari (Fig. 15h). In addition, the downward intrusion of dry air from the UTCL toward Jongdari was still apparent (Fig. 15i). Around the PV tower of Jongdari, the area with > 70 % relative humidity was higher than in the noncoupled-model simulation, exceeding an altitude of 10 km (compare Fig. 15i and Fig. 9l). This may partly result from the difference between the noncoupled- and coupled-model simulations in the structure of the PV tower and the nearby low-PV area in the area of upper-tropospheric outflow.
Our results demonstrate that adding ocean coupling to the atmosphere model helps reduce the PV around the PV tower of Jongdari. In addition, ocean coupling helps suppress warming of the air below the UTCL and thereby helps suppress the spread of decreased PV around the UTCL. The reduction due to ocean coupling in hourly precipitation below the UTCL also reduces upper-tropospheric warming around the UTCL through processes such as reduction in latent heat fluxes from the ocean to the atmosphere; weakening convection and associated diabatic heating, particularly around the eyewall of the TC that moved along the circumference of the UTCL; suppressing the production of areas of low PV in the upper troposphere there; and environmental effects that may help maintain high upper-level PV around the UTCL. The reduction in PV around the outflow area of Jongdari reduced upper-troposphere warming around the UTCL, and the strengthened intrusion of dry air from the UTCL to the vicinity of Jongdari due to relatively strong geostrophic-balanced cyclonic circulation were all factors in delaying the coalescence of the UTCL and Jongdari. The delay in the coalescence due to ocean coupling, interacting with the enlarged and strengthened geostrophic-balanced cyclone circulation induced by the relatively strong UTCL, resulted in a difference in the simulated track of Jongdari.
3.5 Initial conditions and predictabilityThe dry areas surrounding both the UTCL and Jongdari included the continental high, the dry slot from the continental high, and another UTCL over the ocean east of Japan (Fig. 3) that appeared at 108 h integration time (Fig. 9k) when the initial time was 12 UTC on 25 July. The atmospheric environments at the initial time and those provided as lateral boundary conditions differ depending on the initial time of integration, influencing the simulated track and intensity of Jongdari. In fact, the simulated location (Fig. 4), intensity (Fig. 5a) and size (Fig. 5b) of Jongdari differed greatly depending on the initial time of integration; the resulting difference in track simulations was much greater than that caused by ocean coupling. In this section we compare the noncoupled-model and coupledmodel simulations under initial conditions based on four different initial times.
Figure 16 shows the distribution of relative humidity and PV on the 355-K isotherm surface and the vertical cross section of PV on the line between the centers of Jongdari and the UTCL based on the JMA data at the four following times: 06, 12, and 18 UTC on 27 July and 00 UTC on 28 July. These were selected as initial times to provide an integration time of less than 72 h before 00 UTC on 30 July, the time at which Jongdari redeveloped south of Kyushu and coalesced with the UTCL (Fig. 7i). An integration time of less than 72 h was chosen to reduce, to some extent, the effect of ocean coupling on the simulations and to focus on the effect of the difference in the atmospheric initial conditions.
Atmospheric conditions based on JMA data at four different times: (a–c) 06 UTC on 27 July, (d–f) 12 UTC on 27 July, (g–i) 18 UTC on 27 July, and (j–l) 00 UTC on 28 July. These were used as initial conditions for simulations ending at 00 UTC on 30 July (see text). The left panels show the distribution of the relative humidity on the 355-K isotherm surface (colors), wind vectors at that altitude, and sea-level pressure (black lines, contour interval 8 hPa). The center panels show the distribution of PV (colors) and sea-level pressure (black lines, contour interval 8 hPa). The right panels show a vertical profile of PV (colors), potential temperature (black lines, interval 10 K), and horizontal–vertical wind vectors parallel to the cross section along the line between the centers of Jongdari and the UTCL (see Section 3.2), shown as dashed lines in the left and center panels. “A–H” indicate the location of the start and end points of the cross section.
At 06 UTC on 27 July, the UTCL with low relative humidity (Fig. 16a) and high PV (Fig. 16b) lay south of Japan. At that time, Jongdari was located southeast of the UTCL. The PV tower of Jongdari and the tropopause folding (> 1 PV unit) from the UTCL lay along the vertical cross section, and they were approximately 500 km apart at an altitude of 8 km (Fig. 16c). At 12 UTC on 27 July, the UTCL had moved southward and Jongdari had moved cyclonically around the circumference of the UTCL compared to their positions 6 h earlier (Figs. 16d, e). The centers of Jongdari and UTCL were less than 500 km apart at an altitude of 10 km (Fig. 16f). At 18 UTC on 27 July, the dry area (Fig. 16g) with high PV was oriented northwest–southeast rather than north–south (Fig. 16h). Although the height of the PV tower was unchanged from the simulation starting 6 h earlier (Fig. 16i), the area with high relative humidity on the 355-K isotherm surface had become concentrated near the center of Jongdari. At 00 UTC on 28 July, Jongdari's motion had changed to west–northwestward while its maximum intensity matched the best track central pressure (Fig. 5). The center of the UTCL coincided with the dry area at the eastern edge of high-PV area, and Jongdari had moved along the circumference of the UTCL (Figs. 16j, k). The PV tower and the tropopause folding extending from the UTCL were approximately 200 km apart at an altitude of approximately 11 km (Fig. 16l). The upward motion was clear from the lower troposphere to the top of the PV tower, and the outflow from the PV tower went to the southeast. In contrast, the winds from the tropospheric folding to the PV tower were northwesterly. Overall, the behavior of the UTCL and Jongdari as analyzed from different atmospheric initial conditions was continuous. It appears unfeasible to detect a difference in the simulations that can be attributed to the atmospheric initial conditions.
Figure 17 shows maps of relative humidity (Fig. 17a) and potential vorticity (Fig. 17b) at the height of the 355-K isotherm surface and the vertical cross section of PV on the line AB shown in Figs. 17a, b (Fig. 17c) based on JMA 6-hourly global atmospheric analysis data at 00 UTC on 30 July. The relative humidity was high west of the TC within the inner core, whereas PV was high east of the TC. The height of the TC tower was approximately 8 km, and the PV in the UTCL was relatively high, from 12 km to 14 km, east of the TC.
Atmospheric conditions based on JMA 6-hourly global atmospheric analysis data at 00 UTC on 30 July. (a) Distribution of the relative humidity on the 355-K isotherm surface (colors), wind vectors at that altitude, and sea-level pressure (purple lines, contour interval 8 hPa). (b) Distribution of PV (colors) and sea-level pressure (black lines, contour interval 8 hPa). (c) A vertical profile of PV (colors), potential temperature (black lines, interval 10 K), and horizontal–vertical wind vectors parallel to the cross section along the line between the centers of Jongdari and the UTCL (see Section 3.2), shown as dashed lines in the left and center panels.
Figure 18 corresponds to Fig. 16 except results of simulations by NHM starting at four different initial times (i.e., at integration times of 48, 54, 60, and 66 h). All four simulations featured a relatively dry area east–southeast of Jongdari (Fig. 18a) and high PV on the 355-K isotherm surface (Fig. 18b), although the locations of high-PV area relative to the TC center were different from the global analysis, particularly around the analyzed high-PV area at the height of the 355-K isotherm surface (Fig. 17b). The cross-section line in the noncoupled-model simulation that started at 18 UTC on 27 July (Figs. 18g, h) differed from the others in its orientation owing to the difference in the relative positions of Jongdari and the UTCL due to the track error. In fact, the simulated track that started at 18 UTC on 27 July corresponds to a track passing through Jeju Island in Fig. 4, which is approximately 227 km northwest of the ensemble mean position. The relative positions of the tropopause folding and the PV tower differed with the initial time of the integration (Figs. 18c, f, i, l). The shorter the integration time, the closer the relative position to the analysis distribution. Note that the heights of the PV tower in these four simulations were lower than that in the noncoupled-model simulation with the initial time of 12 UTC on 25 July (Fig. 7l), which implies that the intensity of Jongdari was overestimated in the noncoupled-model simulation with the earlier initial time. However, the heights of the PV tower in all four simulations (Figs. 18c, f, i, l) were still higher (∼ 10 km or higher) than that of the analyzed PV tower (∼ 8 km) (Fig. 17c). This suggests that the coalescence of Jongdari and the UTCL did not actually occur and thus, the coalescence in the noncoupled-model simulation starting at 12 UTC on 25 July was unrealistic.
Results of simulations by NHM at 00 UTC on 30 July from the initial conditions in Fig. 16.
The results of the coupled-model simulations at the four different initial times (Fig. 19) differed somewhat from those of the noncoupled-model simulations (Fig. 16). The direction of the cross-section line did not change due to ocean coupling, while the line clearly differs among the four atmospheric initial conditions (compare Figs. 18a, d, g, j and Figs. 19a, d, g, j). This indicates that the atmospheric initial conditions determined the arrangement of simulated Jongdari and UTCL and their evolutions. The high-PV area on the 355-K isotherm surface southeast of Jongdari was slightly larger (from 10,000 km2 to 46,000 km2) than that in the noncoupled-model simulations due to ocean coupling (Figs. 19b, e, h, k), except the simulated area that started at 18 UTC on 27 July (Fig. 19h). The location of the high-PV area relative to the TC center in the coupled-model simulations (Figs. 19b, e, h, k) was closer to the location of the analyzed high PV area (Fig. 17b) than those in the noncoupled-model simulations (Figs. 18b, e, h, k). The reduction in the amplitude of positive PV in the PV tower was approximately 1.2 PVU (Fig. 19c), 0.2 PVU (Figs. 19f, i), and 1.4 PVU (Fig. 19l) in the coupled-model simulations compared to those in the noncoupled-model simulations. The height of the simulated PV tower (Figs. 19c, f, I, l) became close to the analyzed PV tower (Fig. 17c) due to the reduction in the amplitude of positive PV in the PV tower caused by ocean coupling.
Results of simulations by CPL at 00 UTC on 30 July from the initial conditions in Fig. 16.
Although the geostrophic-balanced cyclonic circulation may dominate the TC movement according to the analysis in Fig. 17, the simulated TC intensity was still overdeveloped compared to the best track analysis even in the coupled-model simulations. The excessively strong gradient winds of the overdeveloped TC may be factors that a large loop south of Kyusyu analyzed in the best track data shrank in the noncoupled- and coupled-model simulations.
We conducted numerical simulations of Typhoon Jongdari (2018) with the 3-km mesh NHM and CPL using different initial conditions from different start times of numerical integration to investigate the effects of the ocean and the atmospheric environment on the storm's irregular and unusual track. We also investigated the interactions between Jongdari and the UTCL in detail to understand the relation of their intensities to their respective tracks.
In the early intensification phase, Jongdari lay on the south–southeastern edge of the UTCL, represented by high PV on the 355-K isotherm surface. While the UTCL moved southwestward, Jongdari moved north–northeastward and then moved cyclonically along the western edge of the UTCL. When the UTCL slowed as it stayed south of Japan, Jongdari moved across the northern edge of the geostrophic-balanced cyclonic circulation induced below the UTCL, where midtropospheric relative humidity was high. In Jongdari's mature phase, the storm induced sea surface cooling along its track. Jongdari weakened after landfall in Japan as the UTCL also weakened, staying south of Japan. After Jongdari moved west over the Japanese archipelago, it redeveloped over the ocean south of Kyushu.
The NHM simulation with the earliest initial time (12 UTC on 25 July) showed that the simulated intensity of TC tended to be overdeveloped compared to the best track analysis. During the southwestward movement of UTCL, the cyclonic loop of the TC track was controlled by the UTCL-induced geostrophic-balanced cyclonic circulation, while the UTCL was weakening by humidification caused by cumulus convection over the warm ocean and associated diabatic heating. The PV tower of simulated Jongdari appeared to merge with the UTCL south of Kyushu and thus, TC motion simulated by NHM may be controlled by excessively strong gradient winds of the overdeveloped TC rather than by the UTCL-induced geostrophic-balanced cyclonic circulation.
The corresponding simulation by CPL showed an increased PV in the tropopause around the UTCL and a decreased PV at the PV tower. The increased PV in the upper troposphere represented by PV on the 355-K isotherm surface could be accounted for by reduced hourly precipitation below the UTCL due to ocean coupling, which resulted in less warming in the upper troposphere around the UTCL, because of (1) reduced latent heat flux, (2) weakened cumulus convection over the cooled ocean and reduced diabatic heating particularly around the eyewall of TC and partly below the UTCL, (3) suppressed production of the low-PV area in the upper troposphere around the UTCL, and (4) environmental effects that may help maintain high upper-level PV around the UTCL. Ocean coupling suppressed the upper-tropospheric outflow at the top of the PV tower, which affected the location of low upper-tropospheric PV and thus, helped sustain the amplitude of the PV in the tropopause around the UTCL. Finally, ocean coupling led to reduction in PV within the PV tower, which helped avoid the coalescence of Jongdari and the UTCL. Thus, ocean coupling affected the simulation of Jongdari's track through modifications of the intensities of both Jongdari and UTCL, although the TC intensity simulated by CPL was still overdeveloped compared to the best track analysis, and thus, a large loop south of Kyusyu analyzed in the best track data shrank in the coupled-model simulations due to relatively strong gradient winds. A new finding of this study is that the intensity of not only the TC but also the UTCL is affected by ocean coupling. This study also showed that this modification of intensities altered the interactions between the TC and the UTCL to affect the simulated TC track.
Our simulations show that differences in the initial conditions lead to errors in the simulated Jongdari track that originate in differences in the synoptic environments such as the UTCL, the continental high, the dry slot from the continental high, and another UTCL over the ocean east of Japan. When the effects of ocean coupling are added to these, the track simulation of Jongdari is further perturbed by affecting the interactions between Jongdari and the UTCL through the change in their respective intensities. Figure 20 summarizes the characteristics of the atmospheric and oceanic environments, the interactions between the TC and UTCL, and the effect of ocean coupling on the TC and UTCL as a schematic diagram. The addition of ocean coupling improves Jongdari simulations by weakening the PV tower and increasing the PV in the UTCL. On the other hand, our results suggest that the error in simulating the intensity of Jongdari, attributed to the error of the synoptic environment at the initial integration time, may in turn affect the simulation of the atmospheric environment itself and lead to unexpected errors in TC simulations. Our results suggest that the TC simulations are much more sensitive to atmospheric initial environments than to ocean coupling. Although the effect of ocean coupling on the behavior of the UTCL may be important for improving the accuracy of TC simulations, further systematic research is needed on the effects of atmospheric initial conditions on TC simulations while at the same time making improvements in atmospheric analysis.
Schematic diagrams depicting the interactions between Jongdari and the UTCL (hatched solid boxes) addressed in this study. Factors associated with uncertainty of atmospheric environments and ocean coupling are shown. Solid boxes show factors addressed in this study. SST and difference in atmospheric initial conditions are shaded. Solid ellipses indicate the comparison of geostrophic and gradient winds that affect simulated TC tracks between the noncoupled- and coupled-model simulations. Dashed boxes show the comparison regarding the factor in each solid box. Humidification indicated in large arrows is directly affected by diabatic heating and the effect is accumulated in the atmospheric environments, resulting in the impact on UTCL. Difference in atmospheric environments at the initial time indicated in another large arrow is also one of the atmospheric environment factors that affects the simulations of the TC and UTCL.
The author appreciates Prof. T. Sato and two anonymous reviewers for comments that helped improve the first manuscript. Generic Mapping Tools software (https://gmt.soest.hawaii.edu/) was used to draw the figures. This study was supported by Grants-in-Aid for Scientific Research (KAKENHI) Numbers JP19H 01973 and JP19H05696 from the Japan Society for the Promotion of Science.
