Journal of Geophysical Research: Atmospheres

The role of a convective burst in the genesis of typhoon Hagupit (2008)

Authors

  • Akihiko Murata

    Corresponding author
    1. Atmospheric Environment and Applied Meteorology Research Department, Meteorological Research Institute, Tsukuba, Japan
    • Corresponding author: A. Murata, Atmospheric Environment and Applied Meteorology Research Department, Meteorological Research Institute, Nagamine 1-1, Tsukuba, Ibaraki 305-0052, Japan. (amurata@mri-jma.go.jp)

    Search for more papers by this author

Abstract

[1] This study explores the effects of the mesoscale descent associated with a convective burst during the genesis of typhoon Hagupit (2008), based on a high-resolution cloud-resolving numerical simulation. The simulation result captures the synoptic-scale circulation surrounding the pre-Hagupit depression and the evolution of the storm. A burst of intense deep convection occurs about 1 day before the genesis of Hagupit. After the convective burst, temperature deviation near the center of the depression increases in the lower troposphere. This warming contributes to a drop in the central pressure of the depression and hence to the beginning of the so-called system-scale intensification. In addition, the low-level warming tends to inhibit vertical motion by acting as a lid. Horizontal flow is therefore dominant in the boundary layer; thereby, the air can efficiently gain energy from the sea surface. Increased energy in the boundary layer air feeds intense deep convection near the center of the depression just before the genesis time. These results are consistent with a previous observational study. Tangential momentum budget analysis demonstrates that, just before the genesis time, actual tendency of tangential velocity has larger values throughout the depth of the troposphere, indicating the importance of the deep-layer spin-up of the depression. These large values are attributed to the upward transport of tangential momentum by intense deep convection. In contrast, when the convective burst occurs about 1 day before the genesis time, positive actual tendency is confined to the lower troposphere because of smaller upward transport of tangential momentum.

1 Introduction

[2] The mechanisms that control tropical cyclone (TC) genesis are not yet fully understood, despite the efforts of many previous studies. Early investigations were based on climatological studies that dealt with synoptic or larger-scale environments on the seasonal time scale. Gray [1975] demonstrated that the distribution of TC genesis is related to six environmental factors: three dynamic factors (planetary vorticity, relative vorticity, and vertical shear of horizontal wind) and three thermodynamic factors (sea-surface temperature, vertically conditional instability, and relative humidity). Although the seasonal distribution of TC genesis locations and the factors responsible for the distribution are well understood, the time and location of the genesis of a TC on shorter timescales (i.e., the time and place of TC genesis) remain difficult to predict.

[3] Relatively little is known of the mesoscale processes of TC genesis, although two theories have been proposed: the top-down theory [e.g., Bister and Emanuel, 1997; Ritchie and Holland, 1997; Simpson et al., 1997] and the bottom-up theory [e.g., Hendricks et al., 2004; Montgomery et al., 2006]. Each theory is associated with key mesoscale phenomena: mid-level mesoscale vortices for the former and vortical hot towers for the latter. According to the top-down theory, low-level winds are strengthened due to either downward vorticity advection associated with a mid-level mesoscale convective vortex (MCV) or a merging of several MCVs. The bottom-up theory, on the other hand, suggests that a low-level vortex with deep convection is enhanced within mesoscale convective systems. The low-level vortex is strengthened through mergers and axisymmetrization of small-scale positive vorticity associated with vortical hot towers. The vorticity concentration by the low-level convergence results in the so-called system-scale intensification, leading to TC genesis.

[4] Bursts of deep convection, which are also a mesoscale phenomenon, appear to be a key factor that controls TC genesis [e.g., Zehr, 1992; Lee et al., 2008]. Using satellite data for a large number of individual cases of tropical cyclones, Zehr [1992] showed that bursts of deep convection occurred in a disturbance that developed into a tropical storm. He demonstrated that with the burst of convection that occurred about 1 day before TC genesis the sea level pressure of the disturbance decreased to a level at which the surface winds responded to the pressure gradient. Davidson et al. [1990] observed phenomena similar to those in Zehr [1992]. Utilizing the Australian Monsoon Experiment (AMEX) observational dataset, they examined the genesis of two tropical cyclones in the northern Australian basin and found a burst of deep convection within the inner core of the depression about 1 day before TC genesis. Also observed were large increases in low-level convergence and intensification of the low-level inner circulation. Recently, based on remote sensing data for a large sample of tropical cyclone formations, Lee et al. [2008] has also found that bursts of deep convection appear to be associated with subsequent near-surface enhancement in relative vorticity. They identified five characteristic temporal evolutions of deep convection and found several common routes in the evolution of convective events for TC genesis cases related to the monsoon. The common routes they found were the temporal evolutions of two or three convection events before tropical cyclone genesis related to the monsoonal flow. Tory et al. [2006], who also noted the importance of bursts of deep convection, showed that upright vortex cores develop from convergence/stretching and from vertical advection of absolute vorticity within the deep convective updrafts.

[5] However, the influence of convective bursts on TC genesis, particularly the role of downdrafts associated with convective bursts, remains unknown. Downdrafts in cumulus convection, induced by falling precipitation and subsequent evaporation of precipitation, spread laterally upon reaching the surface [e.g., Houze, 1993; Markowski and Richardson, 2010]. On the other hand, mesoscale downdrafts associated with stratiform clouds can warm the air below the clouds by descending at close to the dry adiabatic lapse rate [e.g., Dolling and Barnes, 2012b]. Only recently, the effects of downward motion associated with convective activity in pre-TC Humberto were investigated using observational data [Dolling and Barnes, 2012b]. They showed a possible path from a mesoscale convective system to a warm-core vortex of Tropical Storm Humberto as follows: Mesoscale downdrafts below an anvil, caused by the exhaust from deep convection, induce adiabatic warming in the lower troposphere, leading to the drop in the surface pressure and the boundary layer capping. Under the cap, the air in the boundary layer, flowing into the nascent eye, gains energy from the sea surface and feeds intense deep convection, which becomes the nascent eyewall. No studies on this topic using a numerical model have appeared in the literature. In addition, in the western North Pacific, very few modeling studies have been reported on the influence of convective bursts on TC genesis.

[6] The present study aims to clarify the influence of convective bursts on TC genesis, particularly to evaluate the role of downdrafts associated with convective bursts. The results show that bursts of deep convection are a key factor that governs TC genesis, based on data obtained from a high-resolution cloud-resolving numerical simulation of the genesis of typhoon Hagupit, which occurred in the western North Pacific in September 2008.

[7] The remainder of this paper is structured as follows. In section 2, the methods for numerical experiments using a cloud-resolving model are described. Section 3 provides a brief overview of typhoon Hagupit and its synoptic background. Section 4 presents the results obtained from the simulation. In section 5, the role of downdrafts associated with a burst of intense deep convection is discussed. In section 6, a tangential velocity momentum budget analysis is performed to investigate the source of the increase in cyclonic circulation. Finally, the main results of this study are summarized in section 7.

2 Numerical Model and Experimental Design

[8] The numerical model employed here is the Japan Meteorological Agency Nonhydrostatic Model (JMANHM) [Saito et al., 2006, 2007], which is used operationally by the JMA. The model has been upgraded from an anelastic version [Ikawa et al., 1991] and now possesses fully compressible Euler equations. The bulk cloud microphysics in the model predicts the mixing ratios of six water species (water vapor, cloud water, rain, cloud ice, snow, and graupel) and the number concentrations of cloud ice, snow, and graupel. The size distributions of the water substances are assumed to be inverse exponential for rain, snow, and graupel, and mono-disperse for cloud water and cloud ice [Lin et al., 1983; Murakami, 1990; Murakami et al., 1994]. The Mellor-Yamada-Nakanishi-Niino Level 3 scheme [Nakanishi and Niino, 2004] is used for the planetary boundary layer. A clear-sky radiation scheme [Yabu et al., 2005] and a cloud radiation scheme [Kitagawa, 2000] are employed. JMANHM has performed well in simulating TCs [e.g., Murata et al., 2003; Murata, 2009a; Mashiko et al., 2009] and cumulus convection [e.g., Murata and Ueno, 2005; Murata, 2009b] under realistic conditions.

[9] In the present study, JMANHM with a horizontal grid spacing of 2.5 km (681 × 681 horizontal grid points; herein referred to as 2.5 km-NHM) is used as a cloud-resolving model (CRM) for a numerical simulation of the genesis of typhoon Hagupit, which occurred in a tropical region of the western North Pacific (around 14°N, 133°E) on 19 September 2008. The official time, stated by Regional Specialized Meteorological Center (RSMC)-Tokyo, of the genesis of Hagupit (i.e., 1200 UTC on 19 September) is used as the genesis time for the present study. This is because the simulation performs reasonably well in capturing the storm's evolution as shown later. A one-way grid-nesting strategy is adopted for the lateral boundary conditions (Figure 1). The nesting procedure is as follows. The 2.5 km-NHM simulation is initialized at 0600 UTC on 18 September 2008 and integrated for 30 h. The initial and lateral boundary data for 2.5 km-NHM are obtained from forecasts produced by JMANHM with a horizontal grid spacing of 5 km (721 × 577 horizontal grid points; herein referred to as 5 km-NHM). The simulation using 5 km-NHM is initialized at 0000 UTC on 18 September 2008. The initial and lateral boundary data for 5 km-NHM are obtained from the JMA global analysis data produced with a four-dimensional variational assimilation technique [JMA, 2007]. High-resolution daily sea-surface temperatures in the global ocean on a grid of 0.25° × 0.25°, objectively analyzed by JMA [JMA, 2007], are also used for the boundary conditions. Ocean-atmosphere interaction in the simulation is one-way. The time step intervals are 5 s for 2.5 km-NHM and 15 s for 5 km-NHM. The vertical coordinates of the 5 km- and 2.5 km-NHMs are terrain-following and comprise 50 levels. The vertical grid increment is 40 m at the surface, gradually increasing to 1240 m at the highest model level (29 km). The depth of the Rayleigh friction layer is 10 km. The slightly deep layer is due to an increase in the depth of a model layer with increasing height. The Rayleigh friction layer is different from the planetary boundary layer: The former is introduced to prevent the false reflection of internal gravity waves from the upper boundary, whereas the latter is based on a turbulent closure model to represent the effects of atmospheric turbulence. The Kain-Fritsch convection scheme [Kain and Fritsch, 1990; Kain, 2004] is included in the 5 km-NHM in addition to a bulk cloud microphysical scheme.

Figure 1.

Map showing the domains of 5 km-NHM and 2.5 km-NHM. Divergence [10−6 s−1] (shading) and wind vectors (arrows) at 850 hPa at 0000 UTC on 18 September are also shown.

3 Typhoon Hagupit (2008) and Synoptic Background

[10] Typhoon Hagupit, assigned the serial number 14 by RSMC-Tokyo, developed out of cloud clusters to the west of the Mariana Islands on 19 September 2008. Hagupit tracked westward as it gradually intensified. The typhoon moved farther westward in the South China Sea and reached its lowest minimum sea level pressure (MSLP) of 935 hPa on 23 September. The storm made landfall over China on 24 September and dissipated on 25 September.

[11] The JMA global analysis data at 1200 UTC on 18 September reveal a low-level circulation (LLC; 15°N, 140°E) between the Philippines and the Mariana Islands (Figure 2). An easterly (westerly) flow exists to the east (west) of the LLC. The easterly flow exists in the trade wind region and is located to the south of the subtropical high-pressure system. The westerly flow is associated with the Asian monsoon and originates from an easterly flow in the Southern Hemisphere, which crosses the equator and turns eastward in the Northern Hemisphere. The easterly and westerly flows encounter each other in the northwestern and southeastern parts of the LLC.

Figure 2.

Horizontal wind vectors (arrows) and zonal wind speed (shading) at 850 hPa from (a, c, e) JMA global analysis data and from (b, d, f) 5 km-NHM results at 1200 UTC on 18 September (t = 6 h) (in Figures 2a and 2b), 0000 UTC (t = 18 h) (in Figures 2c and 2d), and 1200 UTC on 19 September (t = 30 h) (in Figures 2e and f). Positive (negative) values indicate westerly (easterly) flows.

4 Simulation Results

[12] Lower tropospheric synoptic-scale circulation is reasonably well reproduced by 5 km-NHM (Figure 2). The simulated large-scale pattern is similar to that in the JMA global analysis. For example, at the center of the domain there exists an LLC located between the Philippines and the Mariana Islands. The model also shows an easterly (westerly) flow to the east (west) of the LLC. This pattern is responsible for horizontal confluence and convergence to the west of the pre-Hagupit depression, as in the analysis field (Figure 1).

[13] Mesoscale convective features are reasonably well reproduced by 2.5 km-NHM. Figure 3 shows 3 h accumulated rainfall amounts obtained from the simulation and observations. The observational data used are precipitation analyses based on satellite microwave observations with Climate Prediction Center morphing technique (CMORPH) [Joyce et al., 2004] provided by the National Oceanic and Atmospheric Administration (NOAA). The comparison shows that heavy precipitation to the south of the pre-Hagupit depression around 1800 UTC (t = 12 h) on 18 September is observed by the satellite (Figures 3a and 3b). It should be noted that Figure 3b shows precipitation averaged over a 25 km square, equivalent to the resolution of CMORPH, and that Figure 3c shows raw precipitation values (i.e., 2.5 km resolution). The comparison reveals that spatial patterns in precipitation is reasonably well reproduced by the model: areas of heavy precipitation just south of the depression center and to the southeast of the depression center. Simulated precipitation on the north side of the depression center may be overestimated. However, it is possible that the observations could not capture precipitation on the northern area owing to weak signals due to relatively light rain. As for precipitation near the depression center, the observation may not capture the rain-fee area because of its coarser resolution compared with the model. Figure 3d shows a vertical cross-section of the sum of mixing ratios of total water and ice (i.e., cloud water, cloud ice, rain water, snow, and graupel). The cross-section reveals that heavy precipitation to the south of the depression is associated with intense deep convection.

Figure 3.

Three-hour accumulated rainfall amount [mm 3h−1] at 1800 UTC on 18 September (t = 12 h) from (a) CMORPH data, (b) 2.5 km-NHM results (averaged over surrounding 11 × 11 grids), (c) 2.5 km-NHM results (not averaged). Vertical cross-section of total water and ice [g kg−1] along the line W-E in (Figure 3c) is also shown in (Figure 3d).

[14] The simulation performs reasonably well in capturing the evolution of typhoon Hagupit. The trend in MSLP agrees well with observations, although the simulated deepening rate of the storm is slightly lower than that observed (Figure 4). Tangential velocity at the surface, averaged within a radius of 100 km from the storm center, is also shown in Figure 4. The tangential velocity increases with time although abrupt changes are seen when bursts of convection occur between 1200 (t = 6 h) and 1800 UTC (t = 12 h) on 18 September. It is possible that the model intensity could be underestimated, thus delaying the time of genesis relative to the observed one. However, even if the genesis time in the model delays, the period when the analyses were made falls into a period before the genesis time. Therefore, there is no effect on our results in essence. As mentioned earlier, the official time, stated by RSMC-Tokyo, of the genesis of Hagupit (i.e., 1200 UTC on 19 September) is used as the genesis time for the present analyses. This approach is reasonable given that the simulation performs reasonably well in capturing the storm's evolution.

Figure 4.

Time series of the minimum sea level pressure [hPa] of the pre-Hagupit depression. Closed squares and solid line indicate the observed and simulated minimum sea level pressure, respectively. Simulated tangential velocity near the surface (snapshot in time) averaged within a radius of 100 km from the center of the depression is also shown by the dashed line.

[15] Figure 5 shows the temporal evolution of tangential wind, averaged within a radius of 100 km from the center of the simulated storm. The storm center is determined at 1 h intervals, following Braun [2002]; however, this approach employs the tangential wind field at the lowest model level (rather than pressure) in the TC genesis simulation in the case that the pressure minimum is not apparent. This method is a kind of variational approach that adjusts the location of the center until the azimuthal variance of the tangential wind field at all radii between the center and the 125 km radius is minimized. Using 125 km radius is reasonable because radii of 100 km and 150 km are mainly used for the calculations of area-averaged quantities, although quantities averaged over a radius of 150 km are not shown in the text. In the lower troposphere, the area-averaged tangential velocity shows a gradual increase during the period approaching the TC genesis time (Figure 5). The increase in velocity is more pronounced at altitudes below 1 km. A rapid increase in the tangential velocity is also apparent at around 1800 UTC on 18 September (t = 12 h).

Figure 5.

Time-height diagram of vertical velocity [10−2 m s−1] (shading, positive is upward) averaged within a radius of 100 km from the center of the simulated storm, and tangential velocity [m s−1] (contours) averaged within a radius of 100 km from the center. Solid (dashed) lines indicate positive (negative) values. The period of the convective burst is marked by the double-headed arrow.

[16] Figure 5 also shows the vertical velocity averaged within a radius of 100 km from the center of the simulated storm. Deep updrafts are found between 1200 (t = 6 h) and 1800 UTC (t = 12 h) on 18 September, about 1 day before the TC genesis. These remarkable updrafts correspond to a convective surge, as reported by Zehr [1992], accompanying heavy precipitation around the storm center during this period (not shown). The timing of the event (i.e., about 1 day before the TC genesis time) is consistent with the results of Zehr [1992]. He used the same definition of the genesis as that in the present study except that the genesis time was derived from the Joint Typhoon Warning Center best track data instead of the RSMC-Tokyo best track data. A deep updraft at about 0600 UTC on 18 September (t = 0 h) is not examined because the updraft is affected by model spin-up.

[17] The bursts of intense deep convection can contribute to the thermodynamic structure, such as a warm core, of the pre-Hagupit depression. Figure 6 shows the temporal evolution of temperature deviation, averaged within a radius of 50 km from the center of the simulated storm. Temperature deviation is defined as temperature subtracted from the reference value, which is temperature averaged within a radius of 500 km from the center of the simulated storm and varies with time and height. It is found from the figure that, after the convective burst occurred at 1800 UTC (t = 12 h) on 18 September, temperature deviation at an altitude of 2.5 km increases rapidly and remains nearly constant after that. This large value at the 2.5 km altitude indicates that a warm core forms at a lower height relative to that in a mature tropical cyclone [e.g., Hawkins and Rubsam, 1968; Hawkins and Imbembo, 1976], consistent with previous studies on a pre-TC or an early-stage TC [e.g., Leary and Thompson, 1976; Heymsfield et al., 2006; Dolling and Barnes, 2012b]. Of note, the convective burst occurs just before the rapid increases in tangential velocity (see above) and in temperature deviation, indicating that the convective burst influences the subsequent intensification of the pre-Hagupit depression. Guided by this result, the structure of the convective burst is investigated in detail in the next section.

Figure 6.

Time-height diagram of temperature deviation [K] averaged within a radius of 50 km from the center of the simulated storm. Temperature deviation is defined, for each time and height, as temperature subtracted from temperature averaged within a radius of 500 km from the center of the simulated storm. The period of the convective burst is marked by the double-headed arrow.

[18] The effects of large-scale systems are also examined. Figure 7 shows a time-height cross-section of large-scale vorticity and vertical velocity. There is a high vorticity region at upper levels (centered around 13 km) after 2100 UTC on 18 September (t = 15 h). Upward motion is relatively strong (weak) throughout the troposphere before (after) 2100 UTC on 18 September (t = 15 h) except for just before the genesis time. These structures of vorticity and vertical velocity are consistent with that of so-called tropical-depression-type (TD-type) disturbances, sometimes called easterly waves [e.g., Reed and Recker, 1971; Lau and Lau, 1990; Takayabu and Nitta, 1993]. This disturbance affects some parts of the structures near the pre-Hagupit depression. For example, between 0000 UTC (t = 18 h) and 0600 UTC (t = 24 h) on 19 September, vertical velocity sometimes shows downward motion at levels between 4 and 12 km (Figure 5). The large-scale vertical velocity field also shows that upward motion is weak at these levels (Figure 7). The TD-type disturbance therefore provides conditions favorable for mesoscale descent near the depression center. It should be noted that the upper-level warming (Figure 6) at levels between 14 and 15 km, corresponding to strong upward motion (Figure 5), is attributed to diabatic warming induced by latent heat of freezing. In fact, a time-height cross-section of mixing ratio of cloud ice reveals that cloud ice concentrates on these levels (not shown). A close inspection reveals that the increased cloud ice is attributed to exhaust from deep convection associated with the depression, whereas there is no evidence suggesting a connection between the cloud ice and the TD-type disturbance.

Figure 7.

Time-height diagram of vertical velocity [10−2 m s−1] (shading, positive is upward) averaged within a radius of 500 km from the center of the simulated storm, and relative vorticity [10−5 s−1] (contours) averaged within a radius of 500 km from the center. Solid (dashed) lines indicate positive (negative) values. The period of the convective burst is marked by the double-headed arrow.

5 Role of Downdrafts Associated With Convective Burst

[19] Figure 8 shows the horizontal distribution of the vertical component of relative vorticity and horizontal wind vectors near the surface, at the time of the burst of deep intense convection. At lower levels, environmental flow induces horizontal convergence to the west of the pre-Hagupit depression in advance of the burst of deep convection (Figure 1). This convergence produces repeated deep convective events. In Figure 8, there exists an area of high vorticity (the area enclosed by the ellipse). In the area, horizontal confluence results in enhanced horizontal wind. The strong wind at the center of the deep convective area is evident in the wind vectors. There exists a relative vorticity couplet: positive (negative) vorticity is located to the northeast (southwest) of the axis of the zone of intensified horizontal wind. It should be noted that Tory and Montgomery [2008] pointed out that a dipole of vorticity anomalies are generated by the tilting term associated with vertical motion.

Figure 8.

Relative vorticity [10−4 s−1] (shading) and horizontal wind vectors (arrows) at the lowest model level (z = 20 m) at (a) 1700 UTC (t = 11 h), (b) 1800 UTC (t = 12 h), (c) 1900 UTC (t = 13 h), and (d) 2000 UTC (t = 14 h) on 18 September. For each panel, vorticity mentioned in the text is enclosed by ellipse. The star indicates the location of the storm center.

[20] The burst of intense deep convection is also evident in vertical velocity fields. Figure 9 shows the horizontal distribution of vertical velocity at a height of 2.3 km. The area of intense deep convection is apparent at the location of the vorticity couplet mentioned above (the area enclosed by the ellipse). The group of deep convective cumulonimbus, corresponding to a mesoscale convective system as has been observed in observational studies, seems to be at its mature and decaying stages because the area of downdrafts within the system increases with time. The system approaches the depression center with time and is located near the center by 2000 UTC (t = 14 h) on 18 September (Figure 9d). At this time, the area of mesoscale downdrafts associated with convective burst is located just to the southeast of the depression center. This area of mesoscale downdrafts corresponds with stratiform clouds associated with intense deep convection. Figure 10 shows a vertical cross-section of the sum of mixing ratios of cloud water and cloud ice. In the southern area, there are deep updrafts where mixing ratio is relatively large. In contrast, in the middle area where warming occurs, areas of large mixing ratio are confined to the middle troposphere, indicating that there are stratiform clouds caused by exhaust from deep convection. Below the levels, weak downward motion is dominant. These mesoscale downdrafts do not reach the surface, unlike cool convective downdrafts. The results indicate that weak mesoscale downdrafts associated with stratiform clouds are responsible for warming in the lower troposphere, a mechanism similar to that in Dolling and Barnes [2012b].

Figure 9.

As for Figure 8, but vertical velocity [m s−1] at an altitude of 2.32 km.

Figure 10.

Vertical cross-section of the sum of mixing ration of cloud water and cloud ice [g kg−1] (shading), and vertical velocity [m s−1] (contours) along the line S-N in Figure 9d. Quantities are averaged over the east-west direction, where the east and west boundaries are indicated by dashed lines in Figure 9d.

[21] The mesoscale descent contributes to warming in the lower troposphere near the center of the pre-Hagupit depression. Figure 11 shows the horizontal distribution of temperature at an altitude of 2.5 km. The area enclosed by each ellipse is the same as that in Figure 8 where the ellipse is determined based on vorticity. Temperature within the enclosed area is relatively low compared with its surroundings from 1700 UTC (t = 11 h) to 1900 UTC (t = 13 h) on 18 September (Figures 11a–11c). In contrast, at 2000 UTC (t = 14 h) on 18 September (Figure 11d), temperature increases and has higher values compared with its surroundings. This warming corresponds to the rapid increase in temperature deviation at an altitude of 2.5 km previously shown in Figure 6.

Figure 11.

As for Figure 8, but temperature [K] at an altitude of 2.49 km.

[22] The mesoscale descent associated with the burst of intense deep convection is also evident in moisture fields. Figure 12 shows the horizontal distribution of relative humidity at an altitude of 2.5 km. The area of intense deep convection is enclosed by the ellipse as in the previous figure. Relative humidity within the enclosed area is relatively high compared with its surroundings from 1700 UTC (t = 11 h) to 1900 UTC (t = 13 h) on 18 September (Figures 12a–12c). In contrast, at 2000 UTC (t = 14 h) on 18 September (Figure 12d), relative humidity decreases and has lower values compared with its surroundings. This drying corresponds to the rapid increase in temperature at an altitude of 2.5 km (Figure 11d) and hence in temperature deviation at the same level (Figure 6).

Figure 12.

As for Figure 8, but relative humidity [%] at an altitude of 2.49 km.

[23] Previous studies have reported the importance of mesoscale descent in warming in the lower troposphere. For example, Dolling and Barnes [2012b] investigated the thermodynamic structures in Tropical Storm Humberto (2001) during the early stage of the formation of an eye using observational data from global positioning system dropwindsondes, radar, and in situ aircraft measurements. In an area of an anvil accompanying a mesoscale convective system contained in Humberto, there was mesoscale descent that could induce dry adiabatic warming in the lower troposphere.

[24] The temporal evolution of MSLP is consistent with the results of the previous study mentioned above. The time series of MSLP, shown in Figure 4, exhibits a decrease with time after the burst of intense deep convection around 1800 UTC (t = 12 h) on 18 September. This drop in MSLP is probably caused by the low-level adiabatic warming centered on 2.5 km altitude mentioned above. In fact, in Dolling and Barnes [2012b], dry adiabatic warming in the lower troposphere resulted in a drop in pressure at the surface. Another factor that can cause the pressure decrease is diabatic heating associated with the intense deep convection around 1800 UTC (t = 12 h) on 18 September. At this time, warming near the depression center occurs at levels between 4 and 12 km (Figure 6), corresponding to intense updrafts (Figure 5).

[25] The temporal evolution of moist static energy in the boundary layer air is also consistent with the results of Dolling and Barnes [2012a]. Figure 13a shows the time series of moist static energy near the depression center at the surface. Moist static energy, h, is defined as follows:

display math(1)

where g is gravity acceleration, z is geopotential height, cp is specific heat capacity, T is temperature, L is latent heat release constant from water vapor to liquid water, and q is specific humidity. There is a sudden drop in moist static energy around 1800 UTC (t = 12 h) on 18 September. This time corresponds to that of the occurrence of the convective burst. Cold downdrafts associated with the intense deep convection affect moist static energy near the surface. Moist static energy shows an increase after 2100 UTC on 18 September (t = 15 h) and maintains high values after 0200 UTC on 19 September (t = 20 h). This temporal evolution does not fully correspond to that of water vapor flux from the surface, although the flux shows high values after 2000 UTC (t = 14 h) on 18 September (Figure 13b). This result can be explained as follows: During the period between 2100 UTC on 18 September (t = 15 h) and 0200 UTC on 19 September (t = 20 h), moist static energy near the surface increases because of high water vapor flux and suppressed convection. After this period, the consumption of moist static energy by convection is responsible for the inhibition of an increase of moist static energy. In Dolling and Barnes [2012a], the energy content in the boundary layer air increased because the warm air just above acted to cap the boundary layer, and thereby the boundary layer air could efficiently gain energy from the sea surface. A similar mechanism can work in the present case considering warming in lower troposphere. Dolling and Barnes [2012b] pointed out that the increased energy in the boundary layer air could feed vigorous deep convection for the nascent eyewall of Tropical Storm Humberto. In the present case, intense deep convection appears around 0900 UTC (t = 27 h) on 19 September (Figure 5), just before the genesis time of Hagupit. This is consistent with the results of Dolling and Barnes [2012b], although their case is a tropical storm whereas the present case is a tropical depression. This deep convection can gain high moist static energy from the boundary layer air (Figure 13). In fact, moist static energy upstream of deep convection near the depression center is relatively high (Figures 14a and 14b). A vertical cross-section clearly shows that an updraft core for this deep convection contains high moist static energy sourced from the boundary layer (Figure 14c). Convection occurring after 0900 UTC (t = 27 h) on 19 September has a direct influence on the genesis of Hagupit, as will be discussed in the next section.

Figure 13.

Time series of (a) moist static energy [103 m2 s−2] and (b) water vapor flux from the surface [10−5 kg kg−1 m s−1], at the lowest model level (z = 20 m) averaged within a radius of 100 km from the center of the simulated storm. The period of the convective burst is marked by the double-headed arrow.

Figure 14.

(a) Moist static energy [103 m2 s−2] (shading) and wind vectors (arrows) at the lowest model level (z = 20 m) at 1000 UTC on 19 September (t = 28 h). (b) As for Figure 14a, but vertical velocity [10−1 m s−1] (shading) and wind vectors (arrows) at an altitude of 400 m. The star indicates the location of the storm center. (c) Vertical cross-section of moist static energy [103 m2 s−2] (shading) along with the line W-E in Figures 14a and 14b, and wind vectors (arrows) projected on the cross-section.

6 Tangential Velocity Budget

[26] A tangential momentum budget analysis is performed to investigate the source of the increase in cyclonic circulation. An equation for the tendency of azimuthal means of tangential velocity is developed using cylindrical coordinates centered on the moving storm center, where the storm motion vector is subtracted from the wind, as follows:

display math(2)

where v is the tangential velocity, u is the radial velocity, w is the vertical velocity, η (= ζ + f; where ζ is the vertical relative vorticity and f is the Coriolis parameter) is the vertical absolute vorticity, overbars denote azimuthal averages (azimuthal bin size is 10° in this case) centered on the storm, and the primes denote deviations from the azimuthal average, representing eddies. The tangential and radial velocities are storm-relative quantities. The terms on the right-hand side are called (following Montgomery et al. [2006]) the mean radial flux of absolute mean vertical vorticity (MRAV), the eddy radial flux of eddy relative vorticity (ERRV), the mean vertical flux of mean tangential momentum, and the eddy vertical flux of eddy tangential momentum (EVTM). Each term, except for the residual term, is computed using the model output every 20 min. Figure 15 shows a time-height cross-section of the terms in equation (2) averaged within a radius of 100 km from the center of the simulated storm. It should be noted that the comparison between Figures 15f and 15g reveals that the right- and left-hand sides of equation (2) approximately balance. It should be also noted that small-scale features in each term are probably related to physical situations. For example, in Figures 15b, 15f, and 15g, positive and negative tendencies above 10 km height after 0000 UTC 19 September (t = 18 h) correspond to variations of outflow strength.

Figure 15.

Time-height diagrams of the components of an equation for the tendency of tangential velocity [10−5 m s−2] averaged over a circular area centered on the storm with a radius of 100 km: (a) mean radial flux of absolute mean vertical vorticity, (b) eddy radial flux of eddy relative vorticity, (c) mean vertical flux of mean tangential momentum, (d) eddy vertical flux of eddy tangential momentum, (e) the residual term, (f) the actual tendency derived from changes in tangential velocity, and (g) the sum of the terms (in Figures 14a–14d). The period of the convective burst is marked by the double-headed arrow.

[27] The actual tendency of tangential velocity has larger values throughout the depth of the troposphere during the period between 0900 UTC (t = 27 h) and 1200 UTC (t = 30 h) on 19 September (Figure 15f). This suggests that a deep-layer spin-up is important for the pre-Hagupit depression to attain tropical storm strength. The large values in the actual tendency are significantly contributed by EVTM (Figure 15d). This large EVTM is associated with intense deep convection during the period (Figure 5), indicating that each event of deep convection contributes to transporting tangential momentum upwards. Another important point is that MRAV is larger than ERRV at levels below 1 km during this period (Figures 15a and 15b). This indicates that the low-level spin-up is caused mainly by radial flux of vorticity on the system scale, so-called system-scale intensification process [e.g., Montgomery et al., 2006; Tory et al., 2006], not only by a single event of convective burst.

[28] In contrast, positive actual tendency is confined to the lower troposphere around 1800 UTC (t = 12 h) on 18 September, when the burst of intense deep convection occurs (Figure 15f). These large values in the actual tendency do not correspond to those in MRAV but those in ERRV (Figures 15a and 15b), indicating that the convective burst greatly affects the actual tendency at the time. However, the spin-up of the pre-Hagupit depression appears only in the lower troposphere because positive actual tendency is limited at the lower levels, which is related to weak EVTM around this time (Figure 15d). The low-level confined tendency can be affected by the TD-type disturbance mentioned earlier. When the large-scale vorticity at upper levels is higher (from 2100 UTC on 18 September; Figure 7), the large-scale upward motion tends to become weak over the troposphere, which is not favorable for deep convection. It should be noted that the large ERRV is influenced by the movement of the storm in addition to the distribution of vorticity. The calculations of vorticity and inflow (not shown) reveals that higher vertical absolute vorticity near the storm, combined with storm-relative inflow due to the westward movement of the storm, enhances ERRV.

[29] During the period between these two convective events at 1800 UTC (t = 12 h) on 18 September and 0900 UTC (t = 27 h) on 19 September, MRAV near the surface tends to have positive values (Figure 15a). This is consistent with the temporal evolution of MSLP, which shows a decrease with time after 1800 UTC (t = 12 h) on 18 September (Figure 4), considering that absolute vorticity near the surface is transported inward by inflow induced by the drop of MSLP. The dominance of MRAV near the surface indicates the beginning of a larger-scale intensification process (system-scale intensification process). In fact, sustained convergence calculated below 1 km on the 2.5 km grid (not shown) appears at altitudes below 1 km during the period. In addition, MRAV has negative values in the upper troposphere, indicating the generation of relatively axisymmetric outflow. The evolution of the pre-Hagupit vortex is consistent with the above studies which reported system-scale intensification driven during the later stages of TC genesis.

[30] The merger of MCVs is also considered a key process in the genesis of TCs. For example, Kieu and Zhang [2008] examined tropical cyclogenesis from merging MCVs associated with the breakdown of the intertropical convergence zone over the eastern Pacific and demonstrated that the merger of two MCVs was critical for the genesis of a TC. In the present study, there is no MCV in the middle troposphere, although an MCV has been observed in many pre-TC storms [e.g., Simpson et al., 1997; Dolling and Barnes, 2012b].

[31] Differences exist between these previous studies and the present case. In the present study, a close inspection reveals that the vortex core moves toward the fixed area of ongoing deep convection, where vertical vorticity is relatively intense, thereby helping to strengthen the vortex core. In contrast, the scavenging of surrounding vorticity contributes to growth of the vortex core in the cases presented by Tory et al. [2006] and Fang and Zhang [2010]. These two studies, along with the present case, address the issue of one major vortex, whereas Kieu and Zhang [2008] considered two major vortices and their merger.

7 Summary and Conclusions

[32] The role of the mesoscale downdraft associated with the convective burst was investigated using a numerical simulation performed with a cloud-resolving model. The simulation performed reasonably well in reproducing synoptic-scale circulation surrounding the pre-Hagupit depression and in reproducing the evolution of the storm. In the simulation, intense deep updrafts occurred at about 1 day before the genesis of Hagupit, consistent with previous results. After the burst of intense deep convection, temperature deviation near the center of the depression increases rapidly and remains nearly constant after that in the lower troposphere.

[33] Detailed examinations of the convective burst reveal that mesoscale downward motion associated with stratiform clouds in the middle troposphere is responsible for the low-level warming near the center of the pre-Hagupit depression. This warming contributes to a drop in MSLP and hence to the beginning of the system-scale intensification. In addition, the low-level warming tends to inhibit vertical motion by acting as a lid. Therefore, horizontal flow is dominant in the boundary layer; thereby the air can efficiently gain energy from the sea surface. Increased energy in the boundary layer air feeds intense deep convection near the center of the depression just before the genesis time.

[34] To investigate the source of the increase in cyclonic circulation of the pre-Hagupit depression, a tangential momentum budget analysis was performed. Just before the genesis of Hagupit, actual tendency of tangential velocity has larger values throughout the depth of the troposphere, suggesting the importance of the deep-layer spin-up of the depression. These large values are attributed to the upward transport of tangential momentum by intense deep convection just before the genesis time. In contrast, positive actual tendency is confined to the lower troposphere when the burst of intense deep convection occurs about 1 day before the genesis time. This positive tendency is attributed to the horizontal eddy term, indicating that the convective burst greatly affects the actual tendency at the time. The spin-up of the depression, however, is confined in the lower troposphere at this time, because of smaller values in the vertical eddy term. During the period between these two convective events, the horizontal mean term near the surface tends to have positive values. This indicates that absolute vorticity near the surface is transported inward by inflow induced by the drop of MSLP. The dominance of the horizontal mean term suggests the beginning of the so-called system-scale intensification process. These results are consistent with those observed by Dolling and Barnes [2012b].

[35] One goal of future work must be to clarify the relationship between the genesis of Hagupit and its axisymmetrization before the genesis time. Previous studies reported steady intensification of a tropical depression after axisymmetrization [e.g., Montgomery et al., 2006; Tory et al., 2006]. In addition, previous studies on the process of axisymmetrization showed that the deformation of vorticity perturbation by the horizontal shear of symmetric flow causes upgradient momentum transport, leading to an increase in the kinetic energy of the symmetric flow [e.g., Carr and Williams, 1989; Montgomery and Kallenbach, 1997; Montgomery and Enagonio, 1998; Nolan and Farrell, 1999; Möller and Montgomery, 2000]. It is currently unclear whether this process operated in the present case. Further analysis of the process is required in a subsequent study.

Acknowledgments

[36] This study was conducted under the framework of the “Projection of the change in future weather extremes using super-high-resolution atmospheric models” supported by the KAKUSHIN Program of the Ministry of Education, Culture, Sports, Science, and Technology (MEXT). The simulations were performed with the NEC SX6 computer system at Meteorological Research Institute. The author would like to thank T. Kuroda, S. Hayashi, M. Kunii, and W. Mashiko for the use of their nesting tools. The author wishes to thank three anonymous reviewers whose valuable comments and suggestions greatly improved the quality of this paper.

Ancillary