Possible Linkage Between Tropical Indian Ocean SST Anomalies and the Date of First and Last Tropical Cyclones Landfalling in the Chinese Mainland

The western North Pacific (WNP) is the most active area of tropical cyclone (TC) genesis, where approximately one third of all the TCs in the world form. On average, more than seven TCs make landfall over the Chinese mainland each year, exerting disastrous impacts on the life and economy of people living in the coastal regions (Wang and Chen ^[1]). Existing studies have indicated that the location for TC genesis is controlled by environmental conditions (Gray ^[2]; Cao et al. ^[3]; Zhou and Chen ^[4]; Huangfu et al. ^[5]) and the tracks of TCs are determined by large-scale mid-tropospheric flows related to the western Pacific subtropical high (e. g., Wu et al. ^[6]). Moreover, the sea surface temperature (SST) in different tropical regions could play an essential role in the TC landfall behavior by modulating the TC genesis location and mid-tropospheric steering flows (Fudeyasu et al. ^[7]; Chen and Tam ^[8]). Generally, the boreal summer and autumn are the main seasons for TCs making landfall in the Chinese mainland with the annually averaged first landfall appearing in late-June and the last occurring in early-October. A better understanding of the season of landfalling TCs and its influencing factors seems to be of great scientific and practical importance. El Niño-Southern Oscillation (ENSO) events have profound effects on the landfalling TCs in China with the annually first landfalling TC later and last one earlier when El Niño persists from July to September (Wang and Song ^[9]). Recent work by Li et al. ^[10] has showed that the relationship between ENSO and the date of the first-TC-landfall is not stationary and could be modulated by different phases of the Pacific Decadal Oscillation (PDO), which is closely associated with anomalous anticyclone/cyclone over the tropical WNP forced by the tropical Indian Ocean SST variations.

As the first EOF mode of tropical Indian Ocean SST on the interannual time-scale, Indian Ocean basin (IOB) mode peaks in late winter and persists into the following spring and summer (Klein et al. ^[11]; Saji et al. ^[12]; Yang et al. ^[13]; Guo et al. ^[14]). The IOB warming is in close relation with El Niño event in preceding winter but the maintenance of IOB warming in the spring and summer of El Niño decaying year could exert its impacts over the WNP region after El Niño itself has dissipated (Yuan et al. ^[15]; Xie et al. ^[16]; Zhang et al. ^[17]). Previous work has demonstrated that under El Niño/La Niña conditions, the thermal state of the tropical Indian Ocean could affect the WNP TC activity, including TC genesis, intensity, landfall behavior, etc. (Du et al. ^[18]; Zhan et al. ^{[19, 20]}; Tao et al. ^[21]; Yao et al. ^[22]; Huangfu et al. ^[23]; Li et al. ^[10]). Thus, there is a need to investigate the role of IOB in the variability of the first-TC-landfall date (FLD) in China with and without the ENSO signals systematically.

Indian Ocean Dipole (IOD) mode is the second empirical orthogonal function (EOF) mode of tropical Indian Ocean SST anomalies, characterized by a zonal tropical SST oscillation pattern that usually develops in boreal summer, matures in autumn, and decays in winter (Saji et al. ^[24]). The IOD event usually occurs with the ENSO but some IOD cases are independent of ENSO and mainly determined by intrinsic processes within the Indian Ocean region (Friedrich et al. ^[25]; Zhang et al. ^[26]; Liu et al. ^[27]; Wang et al. ^[28]; Sun et al. ^[29]; Dong et al. ^[30]; Wang et al. ^[31]; Fan et al. ^[32]; Stuecker et al. ^[33]). Zhou et al. ^[34] have examined the impacts of the IOD on the interannual variations in prevailing tracks and landfalling areas of TCs over the WNP during boreal autumn. However, the specific relationship between the IOD and the variations of the last-TC-landfall date (LLD) over the Chinese mainland seems to have received less attention.

The prediction of first and last TCs making landfall in the Chinese mainland is of great significance for disaster prevention and reduction in the coastal areas (Liu and Chan ^[35]). Although the remote Indian Ocean SST anomalies have been regarded as the influencing factors of WNP TC activity, their role in the interannual variations of the FLD/LLD have not been well analyzed until now, as seen from the aforementioned studies. Therefore, the objective of this study is to conduct a detailed diagnosis of the linkage between the FLD/LLD and tropical Indian Ocean SST anomaly modes and explore the related physical processes. The rest of this paper is organized as follows. The data and analysis methods adopted in the present study are described in Section 2. The interannual variation of the LLD and its linkage with the autumn IOD, the possible association between the FLD and the spring IOB are investigated in Section 3. Finally, the major findings of this study are summarized in Section 4.

The monthly mean atmospheric data used in this study are the NCEP/NCAR reanalysis products with a 2.5º × 2.5º horizontal resolution at multiple vertical pressure levels from 1000 hPa to 10 hPa. The SST data are available from the HadISST dataset, constructed on a 1º × 1º grid. The TC best track dataset over the WNP region is taken from the Tropical Cyclone Data Center, China Meteorology Administration (CMA) (Lu et al. ^[36]). The TC genesis is defined as the first record of TC best track data when the maximum sustained wind speed reaches 18m s^-1. We concentrate our analysis on the TCs over the WNP region (north of the equator, west of the dateline) from 1960 to 2019 since the earlier reanalysis data over the East Asian-WNP region may have a larger degree of uncertainty (Wu et al. ^[37]).

The FLD and LLD indices are defined as the annual date of first and last TC making landfall in the Chinese mainland, respectively. The autumn mean IOD index is defined as the strength of the SST differences between the western (10°S-10°N, 50°-70°E) and eastern (10° S-0°, 90° E-110° E) regions of the tropical Indian Ocean because boreal autumn (September-November) is usually the mature phase of IOD (Jiang et al. ^[38]). Similarly, the spring mean SST anomaly averaged over the tropical Indian Ocean (20° S-20° N, 40º - 110ºE) is employed as the IOB index since boreal spring (March-May) is usually its mature phase. These two indices are obtained from the CMA website (http://cmdp.ncc-cma.net/Monitoring/cn_nino_index.php?product=cn_nino_index_iobw). The seasonal mean SST anomaly averaged over the Niño 3 region (5°S-5°N, 150°-90°W) is taken as an index (Niño 3 index) to denote the thermal state of the equatorial eastern Pacific. Correlation and composite analysis are adopted throughout this study and the statistical significance is assessed using Student' s two-sided t-tests.

Figure 1a illustrates the normalized time series of the LLD index during the study period of 1960-2019 with obvious interannual variations. The LLD-related monthly atmospheric circulation anomalies over the WNP have shown to be much more enhanced in October and November than those in September (figures not shown). Hence, we focus on the responses of the tropospheric circulation during October-November to the LLD index which can be seen in Figs. 1b and c. When the LLD is later, an evident cyclonic anomaly at 850hPa appears around the Philippines, favorable for the occurrence of TCs there, and below-normal values of 500-hPa geopotential height dominate the tropical WNP, indicating the northward shift of western Pacific subtropical high (WPSH). The resulting mid-tropospheric southeasterly anomalies over (10° - 25° N, 110° - 140° E) could steer more TCs towards southeast China. On the contrary, the convective activities around the Philippines are suppressed accompanied by the anomalous westerly steering flows that prevent the TCs from approaching the Chinese mainland during October-November, in coincidence with the earlier LLD. The tropical SST anomalies in relation to the LLD show the typical negative IOD and prominent La Niña signals during boreal autumn. That is, the warming SST anomalies cover the tropical eastern Indian Ocean while cooling domains appear in the tropical western Indian Ocean and equatorial eastern Pacific Ocean, suggesting that there may be impacts of tropical SST forcings on the LLD variability (Fig. 2).

Figure 1.  (a) Time series of the normalized LLD index during 1960-2019. Regression of (b) 850-hPa winds (vectors: m s^-1), (c) 500-hPa winds (vectors: m s^-1) and geopotential height (contours: gpm) on the LLD index during October-November. Dark (light) shadings in (b) and (c) indicate the regions of 850-hPa winds and 500-hPa geopotential height with positive (negative) values significant at the 99% (95%) confidence level.

Figure 2.  Correlations between the autumn SST and the LLD index. Contour interval is 0.15. Dark (light) shadings indicate the regions of positive (negative) values significant at the 99% (95%) confidence level.

By calculating the correlation coefficient, we found that the LLD is significantly and negatively correlated with the autumn IOD index during the study period (i.e., - 0.37). Due to the close relation between the IOD and ENSO fluctuations (Fig. 3a), the partial correlation method is adopted here in order to isolate the influence of IOD from the ENSO signals (Feng et al. ^[39]; Zhou et al. ^[34]). As can be seen in Fig. 3b, the IOD-SST partial correlation distribution with respect to the autumn Niño 3 index shows apparent west-east dipole in the tropical Indian Ocean and diminished SST anomalies over the equatorial Pacific. In addition, the partial correlation coefficient between the LLD and IOD index is - 0.27, also significant above the 95% confidence level. Thus, although the IOD's climate effects may be modulated by ENSO variations, the IOD seems to exert great impacts on the LLD separately without the ENSO. We select independent IOD years of 1961, 1994, 2011, 2012, 2019 (1960, 1974, 1990, 1992, 1996, 2016) with IOD index above + 0.5 (below - 0.5) standard deviation while autumn Niño 3 index within ±0.5 standard deviation to represent the typically positive (negative) IOD cases alone. These typical years are basically consistent with the results of previous studies (e. g., Zhang et al. ^[40]; Zhou et al. ^[34]). The composite SST anomalies between the two different IOD categories highlight the dipole mode over the Indian Ocean with nearly no responses over the tropical Pacific, manifesting the effectiveness of the chosen cases for further analysis (Fig. 3c).

Figure 3.  (a) Time series of the normalized IOD index (red line) and Niño 3 index (grey shadings) in the autumn during 1960-2019. (b) Partial correlations between autumn SST (contours: 0.2) and IOD index. (c) Composite autumn SST anomalies (contours: 0.25℃) between independently positive and negative IOD cases. Dark (light) red and blue shadings represent the areas of positive (negative) values significant at the 99% (95%) confidence level.

Previous work has shown that dynamic and thermodynamic environmental conditions could influence the formation of the WNP TCs (Wu et al. ^[41]; Huang et al. ^[42]). Fig. 4 exhibits the composited differences of the 850-hPa vorticity, 200-hPa divergence, 600-hPa relative humidity, and vertical zonal wind shear (between 200hPa and 850hPa) during October-November between different IOD phases. It is notable that the stronger vertical wind shear, reduced mid-tropospheric moisture, weaker lower-level convergence and upper-level divergence are all dominating the western parts of the tropical WNP region. The environmental fields over the western WNP are unfavorable for TC genesis in the positive IOD phase while conductive to TC genesis in the negative IOD phase (Figs. 5a and b). At 500hPa, there exists an anomalous anticyclone over the tropical WNP modulating the large-scale mid-tropospheric flows and determining the pathways of the WNP TCs (Fig. 5c). Specifically, the WPSH splits into two separate cells around 130° E with the south flank of its western part more equatorward during October-November under positive IOD conditions, advantageous for the TCs to take recurved tracks or westward pathways to the southern South China Sea. By contrast, the WPSH presents itself as a contiguous entity placing more poleward especially in the west of 130°E under negative IOD conditions, giving rise to an increasing number of TCs moving northwestward and making landfall in the Chinese mainland (Figs. 5a and b). Therefore, the changes in the regions of TCs genesis and mid-tropospheric steering flows could well explain the prevailing tracks away from (towards) southeast China during October-November, resulting in the earlier (later) LLD in the year of positive (negative) IOD phase.

Figure 4.  Composite anomalies of (a) 850-hPa vorticity (contours: 2×10^-6 s^-1), (b) 200-hPa divergence (contours: 0.75×10^-6 s^-1), (c) vertical zonal wind shear between 200hPa and 850hPa (contours: 1.5 m s^-1), and (d) 600-hPa relative humidity (contours: 3%) during October-November between positive and negative IOD phases. Dark (light) red and blue shadings represent the areas of positive (negative) values significant at the 99% (95%) confidence level, respectively.

Figure 5.  500-hPa western Pacific subtropical high (contours: gpm) and streamlines, TCs genesis positions (black dots) during October-November for (a) positive and (b) negative IOD cases, respectively. (c) Composite anomalies of 500-hPa winds (vectors: m s^-1) between different IOD phases. Dark (light) shadings in (c) indicate the regions significant at the 99% (95%) confidence level.

Figure 6b displays the anomalous divergent winds and velocity potential at 200hPa between different IOD cases for better understanding atmospheric responses over WNP regions. The variations of Walker circulation and the related large-scale divergent motion triggered by the tropical SST anomalies associated with a positive IOD event are characterized by the evident rising branch in the western Indian Ocean and sinking branch in the eastern Indo-western Pacific region. The resulting robust lower-tropospheric easterly anomalies over the tropical Indian Ocean could further produce a pair of anticyclones symmetrically around the equator. Along the flank of the northern anticyclone, the easterly flows turn northward towards the Indian continent and then merge into westerlies over northern India, ultimately entering the east of southeast China (Fig. 6a). Consequently, an anticyclone anomaly is formed around the Philippines by these westerlies near 20° N and easterlies near the equator over the western WNP. On the contrary, a pronounced Philippines cyclonic anomaly is established during the negative IOD phase. The Philippines anticyclone/cyclone in lower-troposphere has been shown to play an important role in the WNP TC activity, including the intensity of the convection near the Philippines and the prevailing wind direction in the mid-troposphere over southeast China.

Figure 6.  Composite anomalies of (a) 850-hPa winds (vectors: m s^-1), (b) 200-hPa velocity potential (contours: 1×10⁶ m² s^-1), and divergent winds (vectors: m s^-1) between positive and negative IOD phases. Dark (light) shadings indicate the regions significant at the 99% (95%) confidence level in (a) and the areas of positive (negative) values significant at the 95% confidence level in (b).

More efforts were made to investigate the FLD variability (Fig. 7a) and its influencing factors in a similar way. The SST anomalies in association with the FLD have remarkable above-normal region in tropical Indian Ocean basin and equatorial eastern Pacific Ocean from proceeding winter to spring (Figs. 7c and d). The detrended-IOB index (with linear-trend removed) instead of the original value is used to calculate its correlation coefficient with the FLD (i. e., 0.33, above the 95% confidence level) due to the evident warming trends over Indian Ocean basin during past few decades (e. g., Saji et al. ^[12]). Additionally, as the IOB warming/cooling in the spring closely coincide with the El Niño/La Niña episode in the proceeding winter (Fig. 7b), we calculate the partial correlation coefficient between the FLD and detrended-IOB index with respect to wintertime Niño 3 index, which decreases sharply and becomes insignificant (i. e., 0.12), suggesting the combined climate effects of ENSO and IOB on the FLD fluctuations. Hence, the independent IOB events and the combined IOB and ENSO cases are selected for further comparison. The combined positive IOB with El Niño (negative IOB with La Niña) are based on the criterion of both the spring detrended-IOB index and winter Niño 3 index higher than + 0.8 (lower than - 0.8) standard deviation, whereas the independent IOB cases are chosen when only the springtime detrended-IOB index reaches the criterion. The selected years during the study period of 1960-2019 are shown in Table 1.

Figure 7.  Time series of the normalized (a) FLD index, and (b) detrended-IOB index (red line) in the spring and Niño 3 index (grey shadings) in preceding winter during 1960-2019. Correlations between the SST (contours: 0.15) and the FLD index in (c) spring and (d) preceding winter, respectively. Dark (light) shadings in (c), and (d) indicate the regions significant at the 95% (90%) confidence level.

Diagnosis analysis has indicated that the FLD-related atmospheric circulation anomalies are much stronger during May-June than those in other months (figures not shown). The most striking feature is an anticyclone over the western WNP in the lower-troposphere, which can affect the TCs genesis positions directly and modulate the large-scale steering flows indirectly (Fig. 8a). Compared with that in later FLD years, more TCs occur in the western parts of WNP and move towards southeast China led by the easterly steering flows during May-June in earlier FLD year. Figs. 8b and d plot the impacts of ENSO-IOB co-occurrence on the atmospheric circulation throughout the troposphere during May-June. It is found that the lower-level Philippines anticyclone may be induced by the robust descending motion of the anomalous Walker circulation with the ascending branches in the eastern Pacific and western Indian Ocean, which is forced by the SST anomalies related to the warm IOB during El Niño decaying spring. The corresponding WPSH tends to be strengthened and expanded prominently, preventing the TCs from landfalling in the Chinese mainland because of the mid-tropospheric westerlies over southeast China. By contrast, the convective activity around the Philippines is enhanced and the WPSH seems to be shrunk with the southeasterly covering southeast China during May-June in cold IOB and La Niña decaying spring, leading to that the TCs formation concentrates near the Philippines and an increase number of TCs take northwestward paths to approach the coast of China. The circulation patterns are disadvantageous (advantageous) for WNP TCs landfall in the Chinese mainland during May-June and fit particularly well with the later (earlier) FLD in the warming (cooling) phase of the IOB and ENSO. Moreover, the involved anomalous lower-level Philippines anticyclone (cyclone) is found to play a key bridge role of in the linkage between tropical SST modes and WNP TCs activity, which is in agreement with the former studies (e. g., Li et al. ^[10]). However, when the IOB cases occur independently during boreal spring, as shown in Fig. 8c and e, the atmospheric circulation responses to the Indian Ocean SST anomalies throughout the troposphere are suppressed and nearly non-existent over the WNP region, resulting in the poor linkage between IOB and WNP TC activity.

Figure 8.  (a) Regression of 850-hPa winds (vectors: m s^-1) on the FLD index during May-June. (b) and (c) Composite anomalies of 850-hPa winds (vectors: m s^-1), (d) and (e) 200-hPa velocity potential (contours: 2×10⁶ m² s^-1) and divergent winds (vectors: m s^-1) during May-June between positive and negative phases of (b) and (d) both IOB and ENSO, (c) and (e) independent IOB. Red (Blue) 5880gpm contours in (b) and (c) indicate the western Pacific subtropical high for warming IOB of El Niño (cooling IOB of La Niña) decaying year and independent positive (negative) IOB year, respectively. Dark (light) shadings indicate the regions significant at the 99% (95%) confidence level.

The present analysis considers both individual influences of tropical Indian Ocean SST anomalies on the date of the first and last landfalling TCs in the Chinese mainland. It is revealed that when the IOD is positive (negative) in boreal autumn the LLD becomes earlier (later) and the impact of this IOD is independent of ENSO forcing. The environmental variables that contribute to TCs formation, such as the enhanced lower-level convergence and upper-level divergence, increased mid-level relative humidity and reduced vertical wind shear, are all present over the western WNP region in association with the anomalous cyclone around the Philippines during October and November in the negative IOD phase. The WPSH shifts northward and presents itself as a continuous entity, leading more TCs to move northwestward to southeast China and make landfall there, in accordance with the later LLD. On the contrary, a lower-tropospheric Philippines anticyclone could suppress TCs genesis over the western WNP and a separated weakened WPSH further results in the recurved pathways or westward movements into southern South China Sea of WNP TCs during October-November in the positive IOD phase. The dynamic and thermodynamic variations over the WNP could be attributed to the type of SST anomalies through the modulation of Walker circulation over tropical Indo-western Pacific regions.

Previous study has found that El Niño events could play an important role in the FLD over Continental East Asia (Li et al. ^[10]) through inducing the Indian Ocean warming. By dividing the springtime IOB cases into independent occurrences and co-occurrences with ENSO, our investigations have shown that the responses of the WNP TCs activity to the spring IOB depend on the ENSO event in the preceding winter. Under warming (cooling) IOB condition of El Niño (La Niña) decaying year, the FLD tends to be later (earlier) with fewer (more) landfalls of TCs over the Chinese mainland during May-June, resulting from the enhanced anticyclone (cyclone) near the Philippines and the expanded (shrunk) WPSH thermally motivated by the tropical SST signals related to co-occurrence of the IOB and ENSO. By comparison, the tropospheric circulation anomalies are much weaker over the WNP region in association with the anomalous SST pattern during independent IOB years, contributing to the insignificant connection between IOB and TC landfall behavior. According to previous work (Ren et al. ^[43]; Wang et al. ^[44]), the duration of the season of TC landfall in the Chinese mainland, defined as the period from the FLD to the LLD, is in close relation with summertime ENSO variability, also supported by the results in Fig. 9 which shows significant correlations between Niño3 from summer to autumn and the season of TC landfall in the Chinese mainland. In addition, although the lead/lag correlation analysis results with IOD index are discouraging and failed to pass the 95% confidence level, the correlation coefficients between IOB index from late-spring to early-summer (April-July) and season of TC landfall over China are negatively significant, suggesting that IOB may shed slight light on the seasonal prediction of TC landfall season. The present study reveals the possible linkage between the SST in the tropical Indian Ocean and the date of first and last TC making landfall over the Chinese coast and may provide indications to the short-term climate prediction of WNP TC activity. However, the tropical SST-TC activity relationship is unstable and varies with the phase of Pacific Decadal Oscillation (Li et al. ^[10]; Yang et al. ^[45]). Thus, more examinations are definitely needed in the future to comprehensively understand the interdecadal fluctuations of the relationship between the tropical remote SST anomalies and the season of landfalling TCs in the Chinese mainland.

Figure 9.  Lead/Lag correlation coefficients between three-month moving average Niño 3 (black solid line), IOB (red solid line), IOD (bule solid line), and season of TCs landfall in the Chinese mainland, respectively. Black dashed horizontal line indicates the values statistically significant at the 95% confidence level.