Asymmetric Distribution of Convection in Tropical Cyclones over the Western North Pacific Ocean

In recent years, while forecasts of tropical cyclone (TC) tracks have improved significantly, there are still many uncertainties in forecasts of TC intensity and quantitative precipitation (Demaria et al., 2007). This is mainly due to the fact that the strength and structure of TCs show a complicated spatiotemporal pattern and are affected by various factors. Among these, asymmetric convection plays an important role. Mass convergence, downdrafts, strong rainfall, and latent heat from energy release, which are caused by the convective activities of TCs, can significantly alter the strength and structure of TCs to a large degree (Merrill, 1988; Chen and Meng, 2001).

In a mature TC, convective activities are mostly concentrated in the eyewall, inner spiral rainbands, and outer spiral rainbands. The eyewall is the area of strongest convection, comprising cumulonimbus clouds embedded with hot vortex towers. Inner spiral rainbands are the MCSs located within the inner core of a TC, which is the zone defined by the triple radius of maximum wind (Wang, 2009). In most cases, inner spiral rainbands cannot be observed from satellite cloud images, mainly because they are hidden by overlying cirrus clouds. In contrast to inner spiral rainbands, outer spiral rainbands usually possess broad stratiform precipitation, and occur alongside various organized convection cells (Barnes et al., 1983, 1991; May, 1996). All of these convection systems show distinct asymmetric structures.

Observational and numerical studies have been conducted on the asymmetric structures in TC convections. (Miller, 1958) observed a maximum of rainfall in the front-right quadrant (viewing in the direction of motion) from a 16-storm composite around Florida, USA. (Rodgers et al., 1994) obtained a similar conclusion by observing fast-moving TCs through satellites over the North Atlantic Ocean. However, other studies (Burpee and Black, 1989; Corbosiero and Molinari, 2003) have shown that convective activities are scattered. (Burpee and Black, 1989) constructed radial profiles of rainfall and observed that the precipitation maxima of Hurricane Allen (1980) appeared in the front-right quadrant in the core (30 km <r<100 km, r means the radius from a TC center), whereas the precipitation maxima of Hurricane Frederick (1979) and Hurricane Alicia (1983) were more prominent to the left of the TC motion. This discrepancy suggests that there is a significant difference regarding the asymmetric distribution of different storms. Owing to a shortage of data, most studies of TCs over the western North Pacific Ocean have investigated the convection distribution near the TC's landfall. All of these studies agreed with the results obtained for TCs over the North Atlantic, with maximal precipitation tending to appear in the front-right quadrant of a TC. However, there are significant differences between the environmental conditions of the two oceans. So, are the previously determined results applicable to all TCs over the western North Pacific? And what is the typical convection distribution in the ocean away from the coasts? A general investigation of the convection distribution is required. Ground-based radar data have a limited spatial survey range; data obtained from polar orbiting satellites have a poor temporal resolution; and numerical modeling results show little universality. Therefore, in order to meet the statistical requirements for such an investigation, data obtained from geostationary satellites, with high temporal and high spatial resolution, are used in the present study.

Previous studies have linked the asymmetric structures in TC convections to many factors, including environmental vertical wind shear (Wingo and Cecil, 2010), TC motion (Shapiro, 1983), and the underlying surface (Wu et al., 2002). In a moving storm, friction-induced asymmetric boundary layer convergence results in stronger winds in the lower troposphere to the right of the storm path (Shapiro, 1983; Rodgers et al., 1994; Corbosiero and Molinari, 2003; Lonfat et al., 2004). Accordingly, in the inner core, the front-right quadrant is more favorable for the formation of convection. Environmental vertical wind shear is also an important factor affecting the asymmetric structure of convection in the inner core. Several studies have indicated that a strong convective system is predominantly located in the downshear-left quadrant (Frank and Ritchie, 1999, 2001; Blank et al., 2002; Corbosiero and Molinari, 2002), consistent with the largest vertical motion. However, only a small number of studies have been conducted on convection in the outer region of a TC (outside the inner core), and the two regions have notably different asymmetric characteristics. (Corbosiero and Molinari, 2002) observed a maximum occurrence of lightning in the downshear-right quadrant of the outer region, on the basis of data from the NLDN for Atlantic hurricanes and considered environmental shear as a dominant factor controlling convection activities. The same conclusion was also found in the research of (Chen et al., 2006).

The majority of previous observational and modeling studies have focused on convection in the core of TCs near the coast (e.g., Chan et al., 2004). In this study, the generality and complexity of convection asymmetry in the outer region of TCs over both the open sea and the land in the western North Pacific region is investigated, for the period 2005-2012. Black body temperature (TBB) data from satellites are used to determine the location, intensity and geometry morphology of convections. Convection distributions from the entire database, based on radial and azimuthal variations for various TC intensities, are discussed in sections 3 and 4 from the perspective of convection occurrence frequency and intensity, respectively. In section 5, quasi-circular and elongated MCSs are classified, and the spatial distribution of different categories of convective cells is analyzed. A summary and conclusions are provided in section 6.

2.1.1. Satellite data

TBB data were obtained from infrared images taken by the FY 2series Geostationary Meteorological Satellites C and E (http://satellite.nsmc.org.cn/portalsite/default.aspx), with a temporal interval of 1 h and a spatial resolution of 0.1^°× 0.1^°, and used as a measure of convection. In general, a lower TBB value implies a higher cloud-top height, and as cumulonimbus clouds rise high in strong convections, in a sense the TBB data can directly reflect the strength of convection (Zheng et al., 2008). When comparing satellite images with TMI rainfall data, results showed that convective precipitation (>10 mm h^-1) matched well with low TBB values (<-52^°C). TRMM data with a high spatial resolution can acquire a TC's 3D structure, but TRMM data have poor spatiotemporal continuity. However, in one orbit, TRMM cannot always catch a whole TC system. Therefore, TBB data have been widely applied to statistical studies on convection.

A "threshold method", in which a suitable maximum TBB threshold is empirically selected to identify convective clouds, is typically adopted. In their study into the formation of TCs, (Lee et al., 2008) defined areas where the TBB was lower than the thresholds of -32^°C, -60^°C and -75^°C as moderate cumulus convection (MC), deep convection (DC), and extreme convection (EC), respectively, in vertical hot towers. In this paper, we use Lee's definition in order to classify convections. Given that high cirrus clouds mainly occur in the inner core (Li and Duan, 2013), we focus on convections in the outer region, defined by a circle with radius ranging between 100 and 1000 km, and whose center lies at the center of the TC.

2.1.2. Best-track data

The best-track dataset for TCs over the western North Pacific used in this study was compiled by the Shanghai Typhoon Institute (www.typhoon.gov.cn). The six-hourly track and intensity analyses were recorded. In order to match the temporal resolution of the TBB data, linear interpolations of two closest best-track records between standard hours are calculated. In this paper, TCs are classified into six levels follows the national standard of China ICS 07. 060 (Appendix A) , in which levels 1 to 6 respectively represent: tropical depression (TD), tropical storm (TS), severe tropical storm (STS), typhoon (TY), severe typhoon (STY), and super typhoon (Super TY). Excluding the cases that are outside the range of the basin, as well as twin typhoons, a total of 136 TCs were recorded during 2005-2012.

2.1.3. Reanalysis data

NCEP-NCAR (2005-2012) 1^°× 1^° global reanalysis data with six-hourly resolution are used in this study. Vertical wind shear is defined as the average difference in horizontal wind vectors between 200 hPa and 850 hPa within a 1000-km radius from the center of a TC.

In this section, we discuss the spatial distribution of convection from the perspectives of occurrence frequency, intensity, and the morphology of convection systems.

The characteristics of convection are herein described in storm-relative coordinates. Azimuthal variation is depicted in the clockwise direction and divided into 36 equal parts from 0^° to 360^°, in which 0^° represents the direction of storm motion. Radial variation is divided into 10 equal parts outward from the storm center, up to a 1000-km radius. Thus, the TC is segmented into 36× 10 angular rings. As each angular ring differs in size for different radius values, the conditional probability flux density functions (PFDFs) of convection occurrence are calculated to describe the occurrence of MC, DC and EC. The number of grids whose values are below the TBB threshold divided by the total number of grids denotes the PFDF within each angular ring, averaged over time. Next, in order to be more intuitive, the PDFs of convection occurrence are determined by considering only azimuthal variation, and the asymmetric distributions of varying grades of TCs for different radius values are compared.

The distribution of convections based on their intensity, given by TBB values, is discussed. Azimuthal asymmetry is discussed in section 3; thus, we calculate the PDFs of TBB values, representing the intensity of convection, considering only radial distance. In order to determine the asymmetry of TBB values, first-order Fourier coefficients are computed using the data from satellite grids (Boyd, 2001; Lonfat et al., 2004): \begin{eqnarray} \label{eq1} a_1&=&\sum_i{T_i\cos\theta_i} ,(1)\\ \label{eq2} b_1&=&\sum_i{T_i\sin\theta_i} , (2)\end{eqnarray} where T_i is the TBB value of each individual and θ_i is the phase angle of the value relative to storm motion. The spatial structure of the first-order asymmetry (M₁) can be represented by \begin{equation} \label{eq3} M_1=(a_1\cos\theta+b_1\sin\theta)/T , (3)\end{equation} where T is the mean TBB calculated over the entire annulus.

The methods for identifying MCSs from infrared satellite data are the same as those employed in (Yang et al., 2015). First, a predetermined TBB value and a cloud area threshold need to be provided (Lee et al., 2012), for which a detailed description is given in section 5.1. Secondly, all convection clouds are identified in a single image. Then, the "maximum spatial correlation tracking technique" developed by (Carvalho and Jones, 2001) is used to track target systems between consecutive images. Once an MCS is identified, several features of the MCS can be computed, including mean and minimum cloud shield temperatures, the area of the MCS, the number of cold clusters embedded in the cloud systems, and its eccentricity, displacement, and duration. Finally, after automated identification, a visual check is performed to ensure the accuracy of the results.

Figure 1 illustrates the spatial distribution of convection occurrence frequency (PFDF) over a span of eight years. The left column represents MC (TBB <-32^°C), the middle column represents DC (TBB <-60^°C), and the right column represents EC (TBB <-75^°C). In general, DCs are distributed within a 500-km radius around the center of a TC, while ECs are distributed within a 300 km-radius from the center. It is apparent that MC occurs more frequently than DC. Irrespective of the intensity of convection, when a TC is weak, the front-left quadrant of the TC is more favorable for the formation of convections. With the development of the TC, the frequency of convection occurrence increases; the high-value zone rotates anticlockwise from the front-left quadrant to the rear-left quadrant and the distribution becomes more asymmetrical. When TC transforms into Super TY, convection is concentrated towards the rear. In contrast, the front-right quadrant of the TC shows an inhibition of convection. It should be noted that when a TC matures into an STY, a closed low-value zone appears near the core of the TC in the front-right quadrant, which is equivalent to the average location of the typhoon eye.

Figure .  Spatial distribution of the integral ratio of convection occurrence. Left column represents moderate convection, middle column represents deep convection, and right column represents extreme convection. In essence, this ratio is the conditional probability flux density function. The colored shading indicates the frequency of convection occurring at each grid. The ordinate unit is km.

To be more precise, the distribution of convection is discussed by considering only the azimuthal variation. Figure 2a shows the PDF curves of convection occurrence for all TCs. Irrespective of the intensity of convection, the PDFs are approximately sine curves and display a wavenumber-1 pattern. Fitting the experimental data using the least-squares method reveals that all of the results pass the significance tests. According to Fig. 2a, stronger convection accompanies larger amplitude, which implies a higher asymmetry. Convection dominates in the rear-left quadrant (180^°-300^°) of a TC, compared to the front-right quadrant (40^°-90^°). Each curve has two maxima, at 200^° and at 280^°, but only one minimum.

The PDFs of different levels of TC are shown in Figs. 2b-g. In the TD period, the distribution of convection varies over a smaller range. Two peaks are observed at 180^° and 330^°, and two troughs exist in the rear-left (230^°) and front-right quadrants (70^°) of the TC. At the beginning of TC formation, multiple convections merge from all directions, and the curves vary within a small range. With the development of a TC, the convection asymmetry changes are significantly more dramatic. Two peaks merge into one; meanwhile, the location of the peak moves from the left-rear to the rear. By comparing the trends of the four functions, it can be determined that the more severe the convective activities are, the more sensitive they are to larger amplitudes of curves and stronger asymmetry. Thus, more severe convections are more sensitive to the azimuth around the core of a TC. In Fig. 2, the yellow, green and blue curves show similar tendencies, in contrast to that for MC. Before a TC matures into an STY, MC is more likely to form in the front-left quadrant of a TC, while DC and EC are more likely to form in the left quadrant. When TCs mature into STY or Super TY, all forms of convective activity converge toward the direction opposite to TC movement.

Figure 2.  PDF curves of convection occurrence based on azimuthal variations, in which (a) shows the convection occurrence PDF curve with azimuth for all TCs. PDFs of different grades of TCs are shown in (b)-(g).

Figure 3.  (a) Shear direction distribution relative to TC motion over the western North Pacific. (b) Sketched relationship between shear and convection-prone areas.

Figure 4.  PDF curves of convection occurrence based on azimuthal variations and PDFs of convection occurrence by azimuth for varying radial distances from the center of a TC. Left column represents moderate convection, and right column represents deep convection. The blue, green, red, purple, yellow and black curves represent PDFs within 100 km, 100-200 km, 200-300 km, 300-400 km, 400-500 km, 500-600 km and 600-700 km from the center, respectively.

As mentioned in the introduction, the effects of asymmetric friction suggest that the front-right quadrant is more favorable for the formation of convection in the inner core. However, this is not supported by our results, indicating that the movement of TCs is not a crucial factor influencing the distribution of convection.

One possible explanation for the observed distribution is vertical wind shear. We present in Fig. 3a the average shear direction relative to TC motion in the western North Pacific, and comparison with Fig. 2a reveals that the convection-prone zone is located to the average downshear left. In particular, a stronger convection-prone zone is located closer to the direction of shear. When TCs became stronger with counterclockwise circulation enhancing, convections are more likely to be located in the average downshear left. Therefore, the maximum of convection incidence shifts from the left-front to the rear of TCs. The specific influence of shear on convection will be discussed in a future study.

Figure 4 illustrates the PDF curves of convection occurrence based on azimuthal variations for different radius values. The left column represents MC, and the right column represents DC. As only a few convections lie in the periphery, for this analysis only convections within a 700-km radius of the center of a TC are considered for simplicity. As can be seen from the images in Fig. 4, the larger the radius, the greater the change in amplitude, which means that the distribution of convections is more asymmetrical. As the radius increases, the peak of the convection occurrence rate first rotates clockwise at a radius of 100 km, before rotating anticlockwise between radii of 200 km and 700 km from the center. However, after a TC matures into an STY, the peak rotates in an opposite direction to the previous developmental stages. The transformation of a cyclonic outflow in the core of the TC into an anticyclone outflow is a particular phenomenon of a mature TC. This change is consistent with the movement of the convection peak when a TC matures. In summary, the environmental flow significantly influences the asymmetric structure of convection.

The distributions of different convection intensities, given by TBB values, corresponding to different radial variations are discussed in this section. The radial variation is divided into 20 km-wide annuli outward from the center of a TC, up to a 1000-km radius. This method is identical to that employed by (Lonfat et al., 2004). Figure 5 illustrates the unconditional azimuthal mean TBB values as a function of the radial distance from the center of a TC. The averaged TBB values reach a minimum at a radial distance of 100 km, which is the average position of the eyewall.

The change in the PDFs of convection intensity as a function of radial distance for different grades of TC is shown in Fig. 6. It is worth noting that a bright band with a TBB value of -95^°C is observed near the core of the TC in every image. This bright band may represent high-altitude cirrus clouds, and a weak TC is likely to be influenced by cirrus clouds. (Li and Duan, 2013) pointed out that a TC's inner core is more easily affected by cirrus, but the outer region is basic insusceptible, relatively speaking. Moreover, cirrus's lifespan is short (about one hour), and on a small scale. After time-averaging, the influence of cirrus can be neglected, as in our longstanding and large-scale survey samples.

In Fig. 6, it can be seen that the area within a radius of 300 km is essentially a deep convective zone, whereas outside of this zone is dominated by MC. When TCs develop into STY (Fig. 6e), low TBB values appear at a radius of 100 km and, in particular, a closed low-value center is observed in Super TYs (Fig. 6f). (Lonfat et al., 2004) suggested that the width of the PDF is a good indication of the degree of convection symmetry. Thus, from the images it appears that, although convections are weak, their distribution broadens with elevated asymmetry at large distances from the center (r>300 km).

A first-order Fourier transformation was calculated according to Eqs. (2)-(4), in order to analyze the asymmetric spatial distribution of convection intensity. As shown in Fig. 7, positive values indicate a highly asymmetric distribution while negative values indicate a relatively low asymmetric distribution. In Fig. 7, it is evident that convection intensity is more uniform near the center of the TC, as the convection encircles the entire storm eye. Maximum asymmetry is observed in the rear-right quadrant, while minimum asymmetry is observed in the front-left, within a radius of 300-500 km. As evident from Fig. 3a, a stronger convection is more likely to deflect away from the direction of downshear. These results are in accordance with the distribution convection occurrence described in the preceding paragraphs. Generally, with increasing radius, TCs have the greatest asymmetry in the front-left quadrant, as far as 500 km from the center.

Figure 5.  Radial profile of azimuthally averaged TBB values.

Figure 6.  PDFs of TBB values with varying radial distance. The PDFs of different grades of TCs are shown in (a)-(f).

Figure 7.  Normalized phase maximum of first-order convection asymmetry.

Under different circumstances, MCSs take on a variety of morphologies, such as a quasi-circular shape, a persistent elongated shape, or a convective line (e.g., Carbone et al., 2002; Johnson et al., 2005). Several previous studies have used multiple satellite observations to identify the various MCS morphologies over land, and their results indicate that the spatiotemporal properties and characteristics of individual MCSs differ significantly (Houze et al., 2007; Luo et al., 2011; Yang et al., 2015). Determining the variations in form and associated distribution regularities in outer-MCSs of TCs is applicable to the aims of this study.

The criteria for identifying MCSs using geostationary satellite images have long been contentious. A diversity of research objectives has resulted in non-uniform classification criteria for MCSs. However, all criteria developed to date have been based on the mesoscale convective complex (MCC) survey compiled by (Maddox, 1980) (Yang et al., 2015). In his research, MCCs are defined with specific shapes, durations, area thresholds, and the lowest limit of TBB. In subsequent studies, scholars have found that Maddox's definitions were too rigorous, and could not be meaningfully applied to other cases. Therefore, they assumed -52^°C as the TBB threshold, and discovered that the centroid of the -52^°C cold-cloud-shield is closely related to precipitation (Kane et al., 1987; Augustine and Howard, 1988). Other studies found that MCCs represent only a part of the MCSs; (Anderson and Arritt, 1998) classified another type of MCS, termed persistent elongated convective systems (PECSs), as the major convective system occurring over the North American continent. Further to this, considering the scale of convections, (Jirak et al., 2003) classified MCSs into four types: MCCs, PECSs, MβCCSs (meso-β circular convective systems), and MβECSs (meso-β elongated convective systems). They noted significant differences between various morphologies of MCSs in terms of both spatiotemporal distribution and environmental conditions. This classification has been widely used in census studies of MCSs over land; however, such studies into TCs remain relatively limited.

Figure 8.  Identified MCSs in satellite images of typhoon Matsa from 1200 to 1400 UTC and 1900 to 2100 UTC 12 August 2005. The shaded region is the area with an IR brightness temperature less than $-32^\circ$C.

Figure 9.  (a) Distribution of MCS occurrence. (b) Distribution of MCS areal extent (units: 100 km$^2$). (c) Averaged TBB of MCSs (units: $^\circ$C). (d) Lifespan of MCSs.

Figure 10.  Scatterplot showing the locations of MCSs. Each black point represents a single MCS. The ordinate units are 1 km.

A TC is an MCS in itself, and the rainbands and vortices inside the TC cannot be identified if the -52^°C threshold is adopted. Therefore, the identification threshold must be revised in order to be successfully applied in our study. Through comparative analysis and multi-group tests, the thresholds of TBB and minimum area were changed to -62^°C and 12 000 km², respectively, which proved to be the optimal results from the tests in order to identify MCSs from a TC. Furthermore, we defined the range of outer-MCSs within a radius of 1000 km from the center. A part of the results of these analyses is presented in Fig. 8, which demonstrates that the algorithm has a preferable automatic identification result, in that it can isolate the rainbands and vortices from a TC.

As the thresholds have been modified from previously published classification criteria, it is also imperative that the various MCSs are redefined. In view of this, the sizes of the MCSs must be scaled down, as shown in Table 1.

In order to focus only on the outer-MCSs, the MCSs found embedded within the core of a TC are first eliminated. Figures 9a and b separately show the distribution of outer-MCSs' occurrence and average area. Owing to the larger radius, both the numbers and sizes of MCS are greater in the outer region than at the center of the TC. Large MCSs lie in the rear-left quadrant of the TC according to these images. Figure 9c shows the averaged TBB values of MCSs. In general, the rear-left region is favorable for lower TBB, as well as strong MCSs. However, both the maximum and minimum average TBB appear near the center of the TC, within 300 km of the center, implying that the intensity of MCSs is more uniform in the outer region. Additionally, it can be noted from Figs. 9a-c that the maximum value alternates with the minimum value in each image, which may be due to the fact that, in the convection of a TC, the ascending and descending motions change discontinuously. Figure 9d shows the duration of the MCSs. From this, it can be seen that most of the identified MCSs survive for only one hour, and that only a few MCSs meet the previously published time criteria.

The proportions of different categories of MCSs listed in Table 1 are computed. The results reveal that the outer-MCSs are dominated by MβECSs (46%), followed by PECSs (24%), while MCCs make up the smallest number (11%). Nearly 70% of the identified MCSs fall into the elongated categories, and 65% of the MCSs are small-scale. This general pattern is consistent with the results observed from MCSs over land (Yang et al., 2015).

The scatterplot in Fig. 10 presents the spatial distribution of the various MCSs, with each point marked in black indicating a single MCS. In order to quantitatively describe the asymmetrical distribution of MCSs, a TC is segmented into 8× 5 angular rings. With an increase in the radius of the TC, the area of angular rings becomes larger and the number of MCSs in the outer annulus increases. In view of this, we compute the density of MCSs in each sector. Taking the innermost zone as a benchmark, the magnitude of each kind of MCS in subsequent angular rings from the center is divided by 3, 5, 7 and 9, respectively (Fig. 11). As shown in Fig. 10, four types of MCSs have a homogeneous distribution. In general, meso-β MCSs (MβCCs, MβECSs) tend to cluster near the core of a TC, while α-mesoscale MCSs are scattered in the outer region. For semi-circular MCSs (MβCCSs, MCCs) with large eccentricity, the rear of the TC is a notably high incidence area. In contrast, elongated MCSs (MβECSs, PECSs) are more likely to be observed in the rear-right quadrant of a TC, within 400 km from the center. With increasing radius, the high incidence area for such convections rotates clockwise from the rear to the rear-left, which is consistent with the clockwise outward propagation of rainbands. The causes of these different distributions may be linked to environmental conditions, and the specific controlling mechanisms therefore require further research.

Figure 11.  Spatial distribution of the density of each type of MCS. Warmer colors indicate larger densities, and colder colors indicate smaller densities.

The aim of this study was to investigate the asymmetric distribution of convections in TC formations over the western North Pacific between 2005 and 2012. Infrared imagery from FY2 and the Shanghai Typhoon Institute best-track dataset were employed to determine the incidence, intensity and morphology of the convective activities in TCs.

The results of these investigations demonstrate that the convections in TCs have a clear asymmetric distribution. In general, the PDF curves of convection occurrence based on azimuthal variation for all TCs are approximately sinusoidal. The rear-left area relative to TC motion has the highest rate of convection occurrence, while the front-right area has the lowest. With the development of a TC, the asymmetry of convection increases slightly in the outer region, and the maximum incidence shifts anticlockwise from the front-left to the rear of the TC. In general, the convection asymmetry of a TC has larger amplitudes with respect to vertical wind shear than to TC motion. We found that the convection-prone zone is located in the downshear left region and, in particular, a stronger convection-prone zone is found closer to the direction of shear. When comparing PDFs at different TC radii, the results show that the distribution of asymmetric convections is proportional to the radius. With an increasing radius, the peak of the convection PDF first rotates in an anticlockwise direction in the inner core, and then in a clockwise direction outwards from the center when the TC is mature. This shift is consistent with the upper flow field.

The spatial distribution of convection intensity can be divided into an azimuthal average and a wavenumber-1 asymmetry. Weak convections are more widely observed in the front-left of a TC at large distances. In contrast, strong convections are more likely to appear to the rear-left of a TC, within 300 km of the center. The wavenumber-1 asymmetry amplitudes are much smaller near the center, and the maximum intensity of convection is observed in the rear-right quadrant, while the minimum intensity is observed to the front-left, within a radius of 300-500 km.

Using infrared satellite imagery, outer-MCSs were classified into four types on the basis of MCS activity and morphology: MCCs, PECSs, MβCCSs, and MβECSs. More than 70% of all MCSs examined here are elongated systems, and MβECSs are the most dominant type in the outer region of a TC. Smaller sized MCSs are more concentrated near the center of a TC, and semi-circular MCSs (MβCCSs, MCCs) with large eccentricity are concentrated to the rear of a TC. In contrast, elongated MCSs (MβECSs, PECSs) are more likely to appear to the rear-right of a TC, within a 400 km radius, and subsequently propagate outwards in the clockwise direction.

The present study provides new insights into the convection asymmetry in the outer region of TCs of the western North Pacific. However, this research does not attempt to investigate the factors influencing such a distribution and the dynamical processes involved. Further studies should therefore focus on the effects of environmental vertical shear, TC motion, and the underlying surface, on the characteristics of the convective activity. TCs should be classified according to their maturation time and landing time, and the environmental factors influencing the various types of outer-MCSs should be further investigated. In general, a better understanding of the distribution of convective activities within TCs will result in more accurate forecasts of heavy rainfall events.