1 Introduction

China is one of a few countries in the world that are severely affected by typhoons, with an average of seven to eight landfalling typhoons a year. These landfalling typhoons originate mainly from the western North Pacific (WNP). Therefore, to accurately forecast their tracks, it is important to understand the physical mechanisms that control typhoon movement in the WNP.

Previous researches have suggested that various factors may influence tropical cyclone (TC) motion, some of which (e.g., Kasahara, 1957, 1960; Anthes, 1982) indicate that the most important factor is the environmental steering flow. Kasahara(1957, 1960) found that the advection of a storm by the environment was consistent with the mass-weighted averaged horizontal wind field in the storm. Chan and Gray (1982) suggested that the average horizontal wind in the mid-troposphere (500–700 hPa) within a 5–7° latitude radius from the cyclone center is a good indicator of the direction of typhoon movement. Holland (1984) found that the vertically averaged horizontal wind field between 850 and 300 hPa can represent the steering flow of a typhoon. After removing smaller-scale perturbations from the mid-tropospheric (500-hPa) horizontal wind, Brand et al. (1981) found that typhoons tend to move toward the left of the steering flow. A stronger TC tends to move in a more westward direction (Fiorino and Elsberry, 1989). Wang et al. (1998) illustrated the influence of beta gyre circulation caused by the interaction between the circulation of a TC and the planetary vorticity gradient on TC movement. Li and Zhu (1991) suggested that an asymmetric TC structure can also contribute to its movement, in addition to the environmental steering flow and the beta effect. Chen and Meng (2001) summarized five factors that impact TC tracks, including large-scale steering flow, interactions between large-scale and smaller-scale circulation systems, TC dynamic and thermodynamic structure asymmetry, and interactions between mid and low latitude systems and topography. Ren et al. (2007) separated WNP typhoons into three types: westward- and northwestward-moving typhoons, recurving typhoons, and northward-moving typhoons, and found that different subtropical-high patterns over the western Pacific are responsible for the different types. Recent modeling studies (e.g., Fovell et al., 2016, Yamada et al., 2016) have emphasized the effects of cloud microphysics, radiation, and TC vertical structure on TC tracks.

Atmospheric intraseasonal oscillation (ISO) is very active in the WNP in boreal summer (Li and Wang, 2005; Li, 2010, 2014) and has a considerable impact on the genesis (Liebmann et al., 1994; Maloney and Hartmann, 1998; Fu et al., 2007; Li, 2012; Li and Zhou, 2013a; Xu et al., 2013, 2014), intensification rate (Hsu et al., 2011; Cao et al., 2014), and track (Li et al., 2012; Yoshida et al., 2014; Bi et al., 2015; Yang et al., 2015) of TCs. Tian et al. (2010) examined the characteristics of the lower- and upper-tropospheric intraseasonal circulation, and found that ISO may affect TC motion through a change in the intensity and location of the subtropical high and the monsoon trough. Tao et al. (2012) suggested that there are more frequent recurving typhoons east of 140°E during the easterly phase of ISO. Li and Zhou (2013b) found that there is an increasing tendency of westward and northwestward (recurving) TC tracks during the convective (suppressed) phase of ISO in the WNP.

Typical typhoon tracks in the WNP are westward, northwestward, and northeastward-recurving. Straight northward tracks with small zonal displacement are unusual. These typhoons often form near the 10°–20°N band south of the subtropical high, where there is pronounced climatological easterly steering flow. Despite the climatological zonal steering, these typhoons move northward, which makes predicting their tracks difficult (Xu et al., 2006; Gao et al., 2017). Given the strong ISO variability in the region, it is likely that intraseasonal flows may affect TC tracks. However, as shown by a recent study (Jiang et al., 2015), only a quarter of the current state-of-the-art general circulation models worldwide can simulate a realistic eastward propagation of the Madden–Julian Oscillation. Midlatitude ISOs are even more difficult to predict. One of the low-frequency circulations in the WNP is the monsoon gyre (MG) (Lander, 1994). Carr and Elsberry (1995) showed that the interaction between a TC-like vortex and a MG can lead to a sudden northward-moving track. Bi et al. (2015) investigated the cause of the sudden track change of Typhoon Megi (2010) and found that it was primarily caused by its interaction with a 10–60-day low-frequency MG circulation system. In their idealized simulation, Typhoon Megi continued moving westward, if the low-frequency MG was removed. This again demonstrates the importance of accurately predicting the intraseasonal background flow, to correctly predict the typhoon track.

In this study, we examine the cause of the straight northward-moving typhoon tracks. We focus on the influence of the 10–60-day low-frequency circulation. In addition, we investigate the year-to-year variability of these northward-moving typhoon tracks. The paper is organized as follows:

The data and methodology are described in Section 2, including the definition of “straight” northward-moving typhoon tracks and the determination of active and inactive years of northward-moving typhoon tracks. In Section 3, the cause of the interannual variability of the straight northward-moving typhoon tracks is investigated. In Section 4, we examine how the different types of intraseasonal background flows interact with the typhoons to cause northward movement. In Section 5, we compare the relative contributions of the seasonal mean steering flow (that includes an interannual component) and the intraseasonal steering flow to the northward-moving typhoon tracks. Finally, a summary is presented in Section 6.

2 Data and methodology 2.1 Data

The primary datasets used in this study are from the NCEP Final Operational Global Analysis (available online at http://dss.ucar.edu/), with an 1° × 1° grid resolution at 6-h intervals for the period 2001–14. The major meteorological fields used for our analysis include multi-level horizontal wind fields ranging from 850 to 300 hPa and the mid-tropospheric (500-hPa) geopotential height field. These fields are used to compute the steering flow, vorticity, and stream function fields, as well as the horizontal pattern of the subtropical high. The typhoon dataset is from the Joint Typhoon Warning Center (JTWC), at a 6-h temporal resolution. The JTWC best-track data are used to determine the typhoon genesis time and location in the WNP. The typhoon genesis time is defined as when its maximum tangential wind reaches 34 m s^–1 (65 knots). As most WNP typhoons form from June to October, the current analysis is confined to the active typhoon season (June–October).

2.2 Methodology

We focus on typhoons over the region 10°–30°N, 120°–160°E. The dominant northward track is defined as when the ratio of zonal displacement of a typhoon to its meridional displacement during its lifetime is less than half. That is, the meridional moving distance is at least twice as large as the zonal moving distance ( $\displaystyle \frac{{\Delta y}}{{\Delta x}} > 2$ ; Fig. 1).

Figure 2 displays all the tracks of the selected northward-moving typhoons during 2001–14. These typhoons formed in the area between 10°N and 25°N. Approximately 43% of the typhoons formed between 15°N and 20°N. The northward-moving typhoons can in general reach approximately 30°N.

Table 1 lists the distribution of the number of northward-moving typhoons each year. There are a total of 23 northward-moving typhoons during 2001–14, which account for 17% of all typhoons over the region 10°–30°N, 120°–160°E. There is a clear interannual variation in the northward-moving typhoons, with a maximum of 4 in 2001 and 0 in 2008 and 2014. The maximum typhoon ratio (33%) occurs in 2012, when there are 3 northward-moving typhoons compared with the total typhoon number of 9.

Based on the percentage of northward-moving typhoons each year, we form two groups. An active northward-moving typhoon year is defined when this percentage exceeds a standard deviation of 0.7, and an inactive year when it exceeds a standard deviation of –0.7. Based on these criteria, 3 active and 3 inactive years are identified during 2001–14. The active years are 2001, 2010, and 2012. The inactive years are 2008, 2013, and 2014.

To analyze the influence of the background mean flow on typhoon tracks, we adopt the TC removal method proposed by Kurihara et al. (1993), which removes the typhoon circulation and retains only the large-scale environmental fields. To examine the structure and evolution of the intraseasonal circulation, a 10–60-day Lanczos filter (Duchon, 1979) is applied to the original data to extract the ISO component. By separating the ISO flow and a TC, we can examine the impact of the intraseasonal steering flow on the TC track.

Following Chan and Gray (1982) and Holland (1984), we define the typhoon steering flow as the vertically integrated horizontal wind field between 850 and 400 hPa within a radius of 500 km. Table 2 shows the average moving speed and the calculated steering flow for each of the 23 typhoons when their intensity reaches or exceeds the typhoon strength. By separating the typhoon moving direction into eight categories, including north, northeast, east, southeast, south, southwest, west, and northwest, we find that approximately 60% of the typhoons are in accordance with the direction of the steering flow, and 95% of the typhoons have a directional error of less than 45°. The estimated averaged steering flow speed (4.2 m s^–1) is close to the averaged typhoon moving speed (4.9 m s^–1). This result indicates that the definition used for calculating the steering flow is reasonable.

Composite analysis is used to analyze the differences in the large-scale circulation fields between the active and inactive northward-moving typhoon years. In addition, we examine the structure and evolution characteristics of the low-frequency (10–60-day) horizontal wind fields and their relationship with the northward-moving typhoons. Typhoon centers are considered as reference points for the intraseasonal flow composite. Because the location coordinates of typhoon centers provided by the JTWC are floating point values, they are interpolated to the nearest grid points in the reanalysis data for the flow composite.

The 10–60-day low-frequency cyclonic circulation center is determined from a stream function field based on a least-squares fitting method. A quadratic equation is assumed:

$\psi (x,y) = a{x^2} + b{y^2} + cx + dy + f,$

and the stream function values at grids near the center of the cyclonic flow are used to determine the position of the minimum value of the stream function field. Using this method, we can calculate the relative location of a TC and the associated 10–60-day low-frequency circulation center (Bi et al., 2015).

3 Factors controlling the interannual variability of the frequency of straight northward-moving typhoons

Composite analysis is carried out to identify significant differences in the large-scale circulation between inactive and active northward-moving typhoon years. The vertically integrated horizontal wind from 850 to 400 hPa, with the typhoon circulation filtered out, is composited to analyze the extent to which the northward-moving typhoon frequency depends on the interannual change in the large-scale steering flow.

Figures 3a and 3b show the seasonal mean steering flow during the typhoon season (June–October) of inactive and active northward-moving typhoon track years. This composite steering flow consists of both the climatological seasonal mean component and an interannual component (the composite is calculated based on longitude and latitude grids). The subtropical high in both the active and inactive years presents a generally zonal-oriented distribution. However, the subtropical high, represented by the 5872-gpm contour in the inactive years, shifts further westward. Westward steering flows south of the subtropical high are clearly stronger in the inactive years compared with the active years. In the vicinity of (23°N, 100°–120°E), southerly steering flows are stronger in the inactive years than in the active years, whereas the relative strength of the steering flows becomes opposite over 120°–140°E.

The westward extension of the subtropical high during inactive years causes a strengthening (weakening) of the southerly steering flow west (east) of 120°E. The strengthened meridional gradient of the geopotential height contours south of the subtropical high during the inactive years is consistent with a strengthened westerly steering flow in the region. As major typhoons form in the WNP monsoon trough south of the subtropical high, the change in the steering flow from active to inactive years in the region is consistent with the marked contrast in the northward-moving typhoon numbers.

Figure 3c shows the difference in the steering flow between the inactive and active northward-moving typhoon track years. In the difference field, an anomalous anticyclonic steering flow occurs near Taiwan, while an anomalous cyclonic flow appears over the region 0°–20°N, 150°–180°E. A zonally oriented large-scale anomalous high between 10°N and 30°N (Fig. 3c) over the entire WNP implies that the subtropical high has strengthened in general. The significant zonal and meridional gradient differences at 500 hPa are associated with the strengthened easterly south of the subtropical high and weakened southerly near (20°–30°N, 130°E) during the inactive years, which is consistent with the less frequent northward-moving typhoon tracks.

The above analysis is based on the composite fields. To examine whether the steering flow changes are statistically significant, we have carried out significance tests. Figure 4 illustrates the zonal steering flows south of the subtropical high (averaged between 12°N and 18°N) for individual active and inactive years. East of 130°E, easterly steering flows are fairly strong in all inactive years (Fig. 4a), compared with those in the active years (Fig. 4b). The averaged zonal steering flow over 12°–18°N, 120°E–180° in the inactive years is approximately 1 m s^–1 greater than that in the active years, suggesting that typhoons that form in the monsoon trough region are more likely to move westward in inactive years. The difference field (Fig. 4c) clearly shows that the westward steering flow is statistically significantly larger (exceeding the 90% confidence level) south of the subtropical high (along the latitudinal zone of 12°–18°N between 120°E and 160°E) in the inactive years compared with the active years.

The most significant difference in the meridional steering flow between the active and inactive years occurs in the region west of 120°E and east of 100°E (Fig. 5c). All individual events in the inactive years display a stronger northward steering flow in the region, with an average speed of 1.5 m s^–1 (Fig. 5a), compared with those in the active years, which have an average speed of 1 m s^–1 (Fig. 5b). The meridional steering flow exhibits a peak near 130°E in all active years, with an average greater than 2 m s^–1 (Fig. 5b). The meridional steering flow is much weaker in the inactive years in the same longitudinal zone, with an average speed of 1 m s^–1. In addition, the significant differences also occur near 140°E and 160°E (Fig. 5c). The greater northward steering flows east of 120°E in the active years favor more frequent straight northward-moving typhoon tracks.

To summarize, the interannual variation of the large-scale steering flow has a significant impact on the frequency of northward-moving typhoons. The change in strength of the steering flow is closely related to the change in intensity of the subtropical high. A strengthening and westward extension of the subtropical high leads to a strengthened westward steering flow south of the subtropical high, which causes typhoons to move westward and hinders northward movement. Conversely, a weakening and eastward withdrawal of the subtropical high leads to a northward steering flow in the WNP and favors more frequent northward tracks.

4 Influence of ISO flows on the northward tracks

In the previous section we analyzed the general impact of the seasonal mean steering flow on northward-moving typhoons. However, for a given year—for example, an active year—not all typhoons move northward. This means that typhoon tracks are not solely controlled by the interannual variation of the summer mean steering flow, but are also influenced by the intraseasonal variation of the background steering flow. In this section, based on a detailed analysis of the individual northward-moving typhoon tracks, we study the role of ISO flows in affecting the northward movement of typhoons.

The intraseasonal variability in the WNP is dominated by the 10–60-day timescale (Li and Wang, 2005; Bi et al., 2015). Therefore, the 10–60-day low-frequency steering flow is obtained by the vertical integration of the 10–60-day filtered wind fields from 850 to 400 hPa. By analyzing the structure and evolution of the 10–60-day low-frequency steering flows and their relationship with the northward-moving typhoons, we separate all cases into three categories, including Type-A (10–60-day low-frequency MG pattern), Type-B [10–60-day low-frequency wave train (WT) pattern], and Type-C (10–60-day low-frequency trough pattern). Although subjective, this method can identify the main 10–60-day low-frequency circulation systems accompanied by the straight northward-moving typhoons. This analysis follows Fu et al. (2007), who identified three types of synoptic-scale precursory signals for typhoon genesis in the WNP during 2000–01, which provided valuable information to understand and predict TC genesis in the WNP.

After analyzing the structure and evolution characteristics of the 10–60-day low-frequency steering flows for the 23 northward-moving typhoons, we identify 9 Type-A, 4 Type-B, and 4 Type-C cases. The sum of the three types accounts for 74% of all the northward-moving typhoon cases. Among the three types, Type-A constitutes nearly 53%, indicating that this type represents the major 10–60-day low-frequency flow pattern of the northward-moving typhoons. To present the evolution of the composite low-frequency steering flow for each type, we further separate each type into three stages. The first stage describes the initial period when a typhoon is still located south of the 10–60-day low-frequency circulation system. The second stage describes the main evolution period when the typhoon is in phase with the low-frequency system and moves together with the low-frequency system. The third stage describes the final period after a typhoon moves out of the low-frequency system.

4.1 Type-A

This first type involves the 10–60-day low-frequency MG system. For this type, a typhoon that is initially located over the southeast of the gyre moves gradually into the center of the gyre and then moves northward together with the gyre. Figure 6 shows the different stages of the composite 10–60-day low-frequency flow and vorticity fields for the selected nine Type-A typhoons. The composite is made using the typhoon center (0, 0) as a reference at multiple time levels (6-h intervals) for the selected nine cases. The typhoon is always co-located with a large-scale cyclonic circulation system, i.e., the MG (Lander, 1994). The MG has a domain of 3000 km × 3000 km. In Fig. 6, (0, 0) denotes the center of the typhoon, so a gradual southward shift of the 10–60-day low-frequency MG center from the top to the bottom panel of Fig. 6 implies northward movement of the typhoon. The evolution of the Type-A typhoons is as follows.

At the beginning (Fig. 6a), the typhoon is located south of the MG center, and then moves gradually into the gyre center (Fig. 6b). Next, it moves to the north of the MG (Fig. 6c) and may eventually move away from the 10–60-day low-frequency MG after reaching the mid-latitude westerly zone. During most of its northward journey, the typhoon is inside the MG and is surrounded by the gyre circulation, as both the typhoon and the gyre move northward. Therefore, the main feature of Type-A is that the typhoon moves northward under the control of the northward-propagating 10–60-day low-frequency MG. This scenario resembles the TC–MG interaction proposed by Carr and Elsberry (1995) and Bi et al. (2015).

Typhoon Francisco (2001) is an example of a Type-A typhoon. It formed at 1200 UTC 21 September 2001. The distribution of the typhoon’s location, the 10–60-day low-frequency steering flow, and the vorticity fields around the typhoon center from 0000 UTC 22–25 September are displayed in Fig. 7. At the time of formation, the typhoon was located to the east of the large-scale MG. As the typhoon moved westward, it entered the MG circulation area. On 22 September, it was located to the south of the MG center. During the subsequent days, the typhoon moved northward with the MG. The typhoon moved to the northern part of the MG center on 25 September. During the three-day (22–25 September) journey, it moved northward by approximately 15° of latitude, while the 10–60-day low-frequency MG also moved northward at a slower speed. The average speed of the MG was 2° of latitude per day, or 6° of latitude during the three days (e.g., from 25°N on 22 September to 31°N on 25 September).

The actual locations of Typhoon Francisco’s center and the center of the 10–60-day low-frequency MG (denoted by a minimum center of the stream function field) are displayed in Fig. 8 at a 6-h temporal resolution. During the period of interest, Typhoon Francisco moved northward from 18°N to 35°N, while the 10–60-day low-frequency MG moved northward from 21°N to 28°N. Figure 8 clearly illustrates that both the typhoon and the MG circulation moved northward together during the typhoon’s life cycle. Initially, the typhoon was located south of the MG and the distance between them was relatively long (approximately 400 km). As the typhoon caught up with the MG, the distance between them decreased and reached a minimum at 1800 UTC 22 September. Under the impact of the northward-moving MG, the typhoon also moved northward. The specific mechanism by which the MG influences the typhoon track is not currently clear. However, it is likely that the MG may impact the typhoon through the following physical processes.

First, the cyclonic vorticity associated with the MG may provide favorable environmental dynamic conditions for the typhoon. Second, the Ekman pumping-induced ascending motion may moisten the column within the MG region, providing favorable environmental thermodynamic conditions for typhoon development. Third, the overlap of the MG cyclonic flow and the typhoon flow may enhance the beta effect, leading to a northward track. Fourth, a generalized Fujiwhara binary effect, as proposed by Bi et al. (2015), may operate, leading to a northward track.

4.2 Type-B

The second type involves a 10–60-day low-frequency WT system, where a typhoon is steered primarily by the southerly flow of the WT with cyclonic (anticyclonic) vorticity to its west (east). Figure 9 demonstrates the different stages of the composite 10–60-day wind and vorticity fields for four selected Type-B typhoons. Different from Type-A, the composite 10–60-day low-frequency circulation field exhibits a dominant zonal dipole pattern, with a cyclonic circulation to the west and an anticyclonic circulation to the east of the typhoon center at each stage. The typhoon is initially located south of the WT, gradually catches up with the WT, and is then steered by the southerly flow of the WT. As the composite 10–60-day low-frequency circulation is based on vertical integration from 850 to 400 hPa, it represents deep steering flow for the typhoon. The superposition of the southerly steering flow associated with cyclonic vorticity to the west and anticyclonic vorticity to the east strengthens the northward advection effect.

The composite 10–60-day low-frequency flow exhibits a dominant WT pattern, oriented from west to east. Under the Type-B scenario, a typhoon typically enters the 10–60-day low-frequency WT pattern from the south. Because the low-frequency WT moves slowly, a typhoon from the south can catch up with the WT. Under the impact of the southerly wind associated with the WT, the typhoon is steered along the narrow southerly band. Eventually, the strong southerly flow pushes the typhoon northward out of the low-frequency WT region. Therefore, the 10–60-day low-frequency southerly associated with the WT acts as a channel, moving the typhoon along the channel, resulting in a northward-moving typhoon track.

Typhoon Jelawat (2012) is an example of the Type-B scenario. During 23–30 September, a 10–60-day low-frequency WT pattern occurred to the north of the Philippines (Fig. 10). The WT had positive cyclonic vorticity to the west and negative anticyclonic vorticity to the east. The low-frequency WT moved slowly northward, from 18°N on 23 September to 22°N on 30 September. Initially, Typhoon Jelawat was located south of the WT. Then, the typhoon gradually moved into the 10–60-day low-frequency WT system. In the subsequent days, the typhoon was advected by strong southerly steering flow associated with the WT. Under the impact of the intraseasonal steering flow, the typhoon moved northward from the initial latitudinal location at 10°N to approximately 30°N within 7 days.

4.3 Type-C

The third type involves a 10–60-day low-frequency trough system. In this type, a typhoon is located mainly in the low-frequency trough, and moves along with the trough during its northward journey. Figure 11 demonstrates the different evolution stages of the composite 10–60-day filtered wind and vorticity fields for four selected Type-C typhoons. The typhoon is initially located at the bottom of a 10–60-day low-frequency trough, and then moves northward towards the trough center. Eventually, it moves to the front of the trough and may move away from the trough area under the control of a strong northward steering flow. Different from Type-A, which has closed circulation, the 10–60-day low-frequency flow in Type-C shows an open, trough-like pattern (Fig. 11b). This trough has a large meridional scale, connecting to the midlatitude trough (MT). Most of the time, a typhoon is located near the maximum vorticity region of the trough. This scenario is referred to as the 10–60-day low-frequency MT pattern. The relationship between the typhoon and the 10–60-day low-frequency circulation in Type-C is similar to Type-A, where both the typhoon and the 10–60-day low-frequency MT system move northward.

Typhoon Vamco (2009) is a Type-C case. Initially, the typhoon formed in a tropical 10–60-day low-frequency cyclonic circulation system, and was located at (16°N, 158°E) on 19 August. A few days later, this tropical circulation merged with an MT, and together they evolved into a deep trough with a large meridional extent ranging from north of 40°N to south of 20°N (Figs. 12c, d). The typhoon moved northward, along the maximum positive vorticity zone within the trough (red color in Figs. 12d–f). As the 10–60-day low-frequency MT moved slowly northward, Typhoon Vamco moved along with it, and reached 26°N on 23 August and 30°N on 24 August. In the final stage of the typhoon, it shifted to the front of the trough and moved away from the trough region due to strong northward steering flow (Fig. 12g).

The relative locations of the typhoon center and the 10–60-day low-frequency maximum vorticity center associated with the MT are shown in Fig. 13. As Typhoon Vamco moved northward from 17°N to 38°N during 19–25 August (red curve in Fig. 13), the 10–60-day low-frequency MT system also moved northward at a slower speed. It is likely that the 10–60-day low-frequency MT system has an impact on the northward-moving typhoon track by providing favorable vorticity and moisture conditions.

5 Relative contributions of seasonal mean and intraseasonal steering flows to northward-moving typhoon tracks

In the previous sections, it was found that both the seasonal mean and intraseasonal steering flows can contribute to the northward movement of a typhoon. But what are the relative roles of these two types of steering flow? To address this question, we calculate and compare the amplitudes of the seasonal mean (including the interannual component) steering flow and the intra-seasonal steering flow for each of the northward-moving typhoon cases. Here, the seasonal mean steering flow is calculated based on a 90-day low-pass filtered vertically integrated wind field. The intraseasonal steering flow is calculated based on a 10–60-day band-pass filtered vertically integrated wind field. To test the sensitivity of the result to different vertical and horizontal domains, we compare the result using 700–300 hPa vertical inte-gration to that using 850–300 hPa, as well as the result using a horizontal average radius of 500 km to that using a radius of 700 km.

Table 3 lists the composite northward-moving speeds for typhoon types A, B, and C, and their counterparts from the seasonal mean steering flow and the 10–60-day low-frequency steering flow. It is interesting to note that the northward-moving speed of the typhoons can to a large extent be explained by the combined seasonal mean and the 10–60-day low-frequency steering flow components. For types A, B, and C, the combined steering flows explain approximately 65%, 100%, and 52% of the total northward-moving speeds, respectively. The ratios are not very sensitive to the horizontal and vertical domains. While the seasonal mean steering flow component appears more important in Type-A cases, the intraseasonal steering flow component becomes dominant in Type-B and Type-C cases.

The overall contribution of the intraseasonal steering flow to the northward tracks for all three types is approximately 36%. A similar amount is contributed by the seasonal mean (including interannual) steering flow. This indicates that the straight northward-moving typhoon tracks are controlled equally by the intraseasonal and seasonal components. The overall percentage that explains the moving speed of the northward-moving typhoon by the combined seasonal mean steering flow and the intraseasonal steering flow is approximately 75% for all three types. This indicates that the seasonal mean steering flow and ISO are important contributors to the straight northward-moving typhoon tracks in the WNP.

6 Conclusions and discussion

This study examines the influence of intraseasonal (10–60-day) and interannual steering flows on straight northward-moving typhoons in the WNP. A straight northward-moving typhoon is defined as one where its northward displacement is at least two times greater than its zonal displacement. We find that the intraseasonal and interannual steering flows have significant impacts on these typhoons.

The interannual steering flow affects the northward-moving typhoons primarily through changes in the location and intensity of the WNP subtropical high, because the typhoons have a tendency to move around the edge of the subtropical high. The difference between the active and inactive northward-moving typhoon years indicates that in the latter (former) case the easterly steering flows south of the subtropical high are significantly strengthened (weakened). A strengthened easterly steering flow is conducive to guiding the typhoon westward and hinders its northward movement. In addition, during active years, the southerly steering flow east of 120°E is strengthened, as the subtropical high retreats eastward. This favors more frequent northward-moving typhoon tracks. Conversely, during inactive years, the southerly steering flow east of 120°E is weakened while the westerly steering flow is strengthened.

There are three types of ISO flows that accompany individual northward-moving typhoon tracks. The first type is a 10–60-day low-frequency MG pattern, in which the northward-moving typhoon is embedded in a closed cyclonic gyre circulation. The second type is a 10–60-day low-frequency WT pattern, where a low-frequency cyclonic (anticyclonic) vorticity circulation is located to the west (east) of the typhoon center, causing the typhoon to be advected by the southerly steering flow associated with the WT. The third type is a 10–60-day low-frequency MT pattern, in which the northward-moving typhoon center is located in the maximum cyclonic vorticity region of the trough, moving northward with the trough.

Comparison of the seasonal mean steering flow that includes an interannual component and the intraseasonal steering flow shows that both are important to steering the straight northward moving tracks. However, current operational (statistical or dynamical) models have difficulty in predicting the structure and evolution of the 10–60-day low-frequency circulation systems in the tropics and at midlatitudes (Zhu et al., 2015). This leads to forecast errors in the large-scale steering flows and, therefore, typhoon tracks—particularly when the ISO and interannual variabilities are strong. Therefore, to predict typhoon tracks accurately, an operational model must be able to capture the realistic ENSO teleconnection pattern and ISO evolution in the WNP.

While the current study identifies the observed 10–60-day low-frequency circulation features associated with straight northward-moving TC tracks, the specific physical mechanisms by which ISO flow interacts with typhoons to influence their tracks are still unclear. Idealized numerical model experiments that remove or retain the ISO flow or typhoon, following Bi et al. (2015), would be interesting as a future study to understand the roles of ISO and typhoons, and their interactive nature.

Acknowledgments. We greatly appreciate the constructive comments from the anonymous reviewers and Dr. Mingyu Bi.