Interannual variability of tropical cyclone activity in the southern South China Sea



[1] A study of tropical storm activity in the southern South China Sea region was carried out for the period of 1960 to 2006 using data obtained from the UNISYS website archive, which was provided to them from the Joint Typhoon Warning Center (JWTC) best track data. This study was motivated by two particularly costly storms that impacted Malaysia during the 1996–2001 period. This study demonstrated that November and December were the most active months for tropical cyclone activity in this region. A majority of these storms attained tropical storm intensity. Also, a majority of the tropical cyclones originated within the study area near Malaysia as opposed to moving into the area. The long-term trend showed that there has been a slight increase in tropical cyclone activity in the region, but the trend was not statistically significant. A study of the interannual variability revealed that there was more (less) tropical cyclone activity in the region during La Niña (El Niño) years. Longer-term variability, such as that related to the Pacific Decadal Oscillation, was not found in the analysis here. Using spectral methods confirms that there was significant El Niño–related variability in climatological quantities such as monthly sea surface temperatures or pressures. Finally, the background climatological state was examined in order to determine whether or not the atmosphere in the region was more conducive to tropical cyclone formation or maintenance during active years. It was found that the most active years were associated with warmer SSTs in the study region, relatively weak 200–850 hPa wind shear, a warm-core structure, more water vapor, and more cyclonic low-level relative vorticity, and these were all La Niña–type years. Nonactive years were associated with weaker wind shear, less water vapor, and a more anticyclonic (vorticity) background, regardless of whether the SSTs were warmer or cooler, and most of these were El Niño–type years.

1. Introduction

[2] In recent years, there have been deadly tropical cyclones that have impacted the southern South China Sea (SSCS) region, including Malaysia. Malaysia suffered direct strikes from tropical storm Greg on 25 December 1996 and tropical storm Vamei on 27 December 2001 [Chang et al., 2003]. Furthermore, Vamei had reached typhoon intensity before it reached Malaysia near 1° to 2° north latitude, which made such an occurrence unusual [Chang et al., 2003]. The previously recorded lowest latitude for a typhoon was Sarah (3.3°N) in 1956 [Fortner, 1958]. Additionally, Greg caused 238 deaths and economic losses of $52 M in US dollars [Ooi and Ling, 1996]. These storms have caused concern in Malaysia as to whether or not this was part of a long-term trend toward more frequent tropical cyclone strikes.

[3] Tropical cyclones are among the most deadly and destructive of natural disasters in terms of loss of human life and economic destruction [Bengtsson et al., 1996]. The 2004 and 2005 seasons in the Atlantic Ocean basin region were two of the most severe in recent memory for the nations bordering the region. Thus, interest in research on tropical cyclones has increased substantially in recent years [e.g., Emanuel, 2005]. This has been especially true regarding the study of intraseasonal and interannual variations in the intensity and occurrence of tropical cyclones, as associated with such phenomena as the Madden Julian Oscillation (MJO [e.g., Maloney and Hartmann, 2001; Chang et al., 2005b]), the Quasi-biennial Oscillation (QBO [e.g., Gray, 1984a, 1984b; Chan, 1995]), the El Niño and Southern Oscillation (ENSO), and the Pacific Decadal Oscillation (PDO)/North Atlantic Oscillation (NAO) [e.g., Gray, 1984a, 1984b; Chan, 1985; Wu and Lau, 1992; Lander, 1994; O'Brien et al., 1996; Bove et al., 1998; Landsea et al., 1999; Pielke and Landsea, 1999; Lupo and Johnston, 2000, hereinafter referred to as LJ00; Houghton et al., 2001].

[4] Within the Atlantic region, the occurrence and intensity of tropical cyclones have been shown to vary quite strongly with ENSO in general (see references above). These studies have shown that during El Niño (La Niña) years, tropical cyclone activity is suppressed (enhanced) and they tend to be weaker (stronger). This does not preclude the occurrence of strong storms during relatively quiet years, such as hurricane Andrew in 1992.

[5] LJ00 not only found results similar to those discussed above, but their study of hurricane activity in the Atlantic region from 1944 to 1999 demonstrated that there was an interdecadal signal in the behavior of the ENSO response. This variability could have been associated with the PDO. Specifically, that there were more hurricane events per year during one phase of the PDO (e.g., 1947–1976), but little or no statistically significant ENSO variability. During the opposite phase of the PDO (1977–1998), there was strong ENSO-related variability in the intensity and number of Atlantic storms.

[6] While the most extensive studies have been performed in the Atlantic Region, there have been studies demonstrating ENSO variability in Pacific region tropical cyclone occurrence as well [e.g., Chan, 1985; Wu and Lau, 1992]. These studies showed similar results for the Pacific Ocean basin as a whole. However, Ramage and Hori [1981] and Lander [1994] found no ENSO variability in the overall tropical cyclone numbers throughout the North Pacific as a whole, but significant variability in the primary tropical cyclone genesis regions. Since then, studies have shown that the focus of tropical storm activity moves southeastward (northwestward) during El Niño (La Niña) years [e.g., Chan, 1985; Wu and Lau, 1992; Lander, 1994; Wang and Chan, 2002].

[7] Recently, there has been concern that tropical cyclone frequencies and intensities may be increasing because of climatic (whether attributed naturally forced or anthropogenic) warming, and several studies have investigated the issue [Henderson-Sellers et al., 1998; Houghton et al., 2001; Pielke et al., 2005; Emanuel, 2005; Webster et al., 2005]. These studies, however, have shown inconsistent results. Malaysia, which is located close to the equator, was thought to be generally immune to the threat of tropical cyclones. Gray [1968] found that for tropical cyclone development, the disturbance should be approximately a minimum of 5° latitude from the equator, generally.

[8] There has also been some disagreement in the literature regarding the relationship between the QBO and tropical cyclone activity over the western north Pacific. Chan [1995] found that the westerly phase of the QBO was positively correlated with increased tropical cyclone activity over the western north Pacific, which is similar to the results of Gray [1984a] for the Atlantic Ocean basin. However, Lander and Guard [1998] showed that the global tropical cyclone activity was not significantly correlated to the QBO.

[9] Thus, this study will have three objectives and they are (1) to develop a detailed long-term climatology of tropical cyclone activity in the SSCS region, (2) to examine the character of the long-term trends and interannual variations in tropical cyclone activity, and (3) to examine variations in the background atmospheric and surface environment in the SSCS (especially in the active season, November and December). In order to accomplish the second objective, we examined whether or not there were significant ENSO or PDO related variations in the SSCS tropical cyclone climatology. Other recent studies have examined the regional climatological character of tropical cyclones and the background climatological state in which they occur [e.g., Tonkin et al., 2000; Dare and Davidson, 2004; Paterson et al., 2005]. One outcome of this study will be to produce information that may provide useful guidance for the long-range prediction of tropical cyclone occurrence for tropical weather forecasters in the SSCS region (Malaysia and adjacent nations). This will include information for advising policy makers regarding future tropical cyclone occurrences.

2. Methods and Analyses

2.1. Analyses

[10] The data set used here was acquired from the UNISYS website ( All tropical cyclones (tropical depressions, tropical storms, and typhoons) from 1960 to 2006 were listed on this website and included in this data set. These data were provided to UNISYS through the Joint Typhoon Warning Center (JTWC) best track data set. The data set contains tropical cyclone categorization, name, life span, maximum wind speed, and track information. As stated by Lander [1994], tropical cyclones in the western North Pacific Ocean have been monitored by the United States Navy since 1945 and by the United States Air Force since 1947 using aircraft reconnaissance. In 1959, these entities were combined under one command, the JTWC, to track tropical cyclones over the western North Pacific. Weather satellite observations became routine in 1965, and these observations contributed to the data set as well.

[11] The monthly sea surface temperature (SST) information and atmospheric variables were provided by the National Center for Environmental Prediction (NCEP) and National Center for Atmospheric Research (NCAR) gridded reanalyses [Kalnay et al., 1996; Kistler et al., 2001]. These data are archived at NCAR and obtained from the National Oceanic and Atmospheric Administration (NOAA) Climate Diagnostics Center (CDC) website ( These reanalyses are displayed on a 2.5° by 2.5° latitude-longitude grid available on 17 mandatory levels from 1000 to 10 hPa at 6-h intervals. These data included standard atmospheric variables geopotential height, temperature, relative humidity, vertical motion, u and v wind components, and surface information.

2.2. Methods

[12] The region of study was the SSCS, and it is bounded here by 0° to 10°N and 100°E to 120°E (Figure 1). The monthly tropical cyclone counts include those that develop locally, as well as those that enter the region from the western North Pacific. The intensity was given as the maximum intensity attained by the tropical cyclone following the methodology of LJ00. Briefly, central pressure data were not always available for each storm in the earlier years in the period described above and were not used. In the Atlantic, maximum wind data have been shown to contain biases (e.g., 1944–1969). Landsea [1993] discusses the 1944–1969 bias in maximum wind speeds (as much as 5 kt or 2.5 m s−1) at some length. To reduce the influences of these biases in maximum wind speed, hurricane intensity was assigned on the basis of the maximum Saffir-Simpson intensity attained by the storm. Thus, only storms which reported maximum wind speeds close to the limits of a particular Saffir-Simpson category would have been vulnerable to being improperly classified. The statistical analysis of trends and distributions were performed using the standard statistical tests found in any standard text [e.g., Neter et al., 1988] and also follows those used in LJ00.

Figure 1.

A map of the South China Sea and part of the west North Pacific Ocean. A map of the study area and Malaysia are indicated in the inset.

[13] Many studies, such as Gray [1984a], have examined the variability in the climatological background state for atmospheric quantities that have been associated with tropical cyclone development and intensification. Such variables used in other studies [e.g., Tonkin et al., 2000; Chang et al., 2005b] and included here were the area-averaged monthly (1) SSTs and sea level pressure (SLP), (2) 850–200 hPa wind shear, (3) low-level divergence, and (4) low-level vorticity.

2.2.1. Analysis of SST and SLP

[14] The method of cycles as described by Mokhov et al. [2004] and modified by Lupo et al. [2007], was used to extract interannual and interdecadal variability from a one-dimensional time series in this study. It is based on the analysis of simple phase plots of the first derivative of the time series versus the time series itself. For a function that is represented by a cyclic time series, or one that is sinusoidal or nearly sinusoidal, we can use;

equation image

where X(t) represents the time series of some variable, A(t) is the amplitude, ω is the frequency, and ϕ (t) is the initial phase. Then, X(t) represents a general solution for differential equations of form;

equation image

[15] The time series in this study were also filtered to extract periodicities on the interannual and interdecadal timescale, or timescales of 2 years or longer following the procedure given by Mokhov et al. [2004]. However, their publication used a 2-year running mean in order to filter out the higher-frequency noise. Lupo et al. [2007] modified this technique by using a Shapiro [1970] filter. This technique has the advantage of retaining the original length of the data set and can be applied a successive number of times in order to control the retention of signal versus noise. This filter also (1) preserves some of the annual cycle, (2) does not result in a phase shift in the low-order variability, and (3) does not introduce significant aliasing error. Finally, Fast Fourier Transforms were used in order to extract significant interannual and interdecadal variability. Unfortunately, the methods used here filter out the signal induced by such phenomena as the MJO. Phenomena on these smaller timescales, however, are beyond the scope of our analysis here and are influenced likely by phenomena such as ENSO and the QBO.

2.2.2. Analysis Using Quantities at the 850 hPa and 200 hPa Geopotential Height Level

[16] Winds at the 850 hPa and 200 hPa geopotential height surfaces were used to generate the quantities described below. The monthly mean vertical wind shear (850–200 hPa), the 850 hPa vorticity and divergence were calculated using the observed winds. The wind shear was calculated using the formula:

equation image

where equation image is the observed horizontal wind vector. Wind shear is a measure of baroclinicity in the atmosphere, and many researchers and forecasters have shown that tropical cyclone development or intensification is not favored in a strong wind shear environment [e.g., Fitzpatrick, 1997]. Wind shear was used, for example, in the Statistical Hurricane Intensity Prediction Scheme (SHIPS) [e.g., DeMaria et al., 2005]. SHIPS is an operational model using environmental conditions in order to forecast hurricane intensity. The 850 hPa divergence and relative vorticity were calculated using;

equation image
equation image

where positive (negative) values represent divergence (convergence) in 4a, and ∇ is the three-dimensional del operator applied to a similar wind vector. The horizontal component of the divergence and the equation image-component of the vorticity were then used here, as has been done commonly in standard meteorological analysis. Also, these calculated quantities were averaged over the period of 1 November to 31 December (the active SSCS tropical season), as well as area-averaged over the study region (see Figure 1). These variables were chosen since tropical cyclone formation generally requires preexisting low-level cyclonic disturbance for formation. These variables were also used in the study of Chang et al. [2005b] as they studied convective activity over the SSCS. We used these months because such a large fraction of the tropical cyclone activity in the SSCS occurred in these 2 months. While the time compositing was performed using the CDC website, the area averaging was carried out by downloading the data and integrating over the area using mean value theorem [see Burkhardt and Lupo, 2005].

2.3. ENSO and PDO Definitions

[17] ENSO years were defined using the Japan Meteorology Agency (JMA) ENSO index to classify these as El Niño, La Niña, or neutral years. A complete description of this index can be found at the Center for Ocean and Atmospheric Prediction Studies (COAPS) website ( Briefly, this index classified each year on the basis of a 5 month running mean of spatially averaged SST anomalies over the Niño 3.4 region in tropical Pacific Ocean. This value should be equal to or exceed (be less than) 0.5°C (−0.5°C) for 6 straight months to be classified as an El Niño (La Niña) year. When the SST anomalies were close to 0°C, these years were considered neutral years. Table 1 provides a list of these years, and, for example, the ENSO year 1970 begins 1 October 1970 and ends 30 September 1971. This definition of El Niño has been used by several studies which have examined the interannual variability of atmospheric phenomenon [e.g., Bove et al., 1998; LJ00; Smith and O'Brien, 2001; Wiedenmann et al., 2002], and is similar to other definitions used by other investigators [e.g., Pielke and Landsea, 1999].

Table 1. A List of Years Examined in This Study Separated by ENSO Phase
 El NiñoLa NiñaNeutral
Years examined1963, 1965, 1969, 1972, 1976, 1982, 1986–1987, 1991, 1997, 2002, 20061964, 1967, 1970–1971, 1973–1975, 1988, 1998–19991960–1962, 1966, 1968, 1977–1981, 1983–1985, 1989–1990, 1992–1996, 2000–2001, 2003–2005
Total years121025

[18] Finally, the Pacific Decadal Oscillation is a 50 to 70 year basinwide variation in Pacific Region SSTs. This phenomenon is described by, for example, Gershanov and Barnett [1998] in more detail, but the positive (negative) phase of the PDO represents the presence of relatively warm (cool) SSTs persisting throughout the eastern portion of the Pacific Ocean Basin. The positive (negative) phase of the PDO (see also the Joint Institute for the Study of the Ocean and Atmosphere at the University of Washington ( (Table 2) persisted during the period from 1977 to 1998 (1947 to 1976), and in 1999 it was believed to have switched phases again [e.g., LJ00; Houghton et al., 2001; Lupo et al., 2007].

Table 2. Phase of the PDO
PDO PhasePeriod
Phase 2 (negative)1947–1976
Phase 1 (positive)1977–1998
Phase 2 (negative)1999–2006

3. SSCs Climatological Analysis

[19] A total of 50 tropical cyclones occurred within the SSCS region over the 47 year period of study (Table 3). This sample included 11 typhoons (TY, ∼20%), 29 tropical storms (TS, ∼60%), and ten tropical depressions (TD, ∼20%). While these events constituted a very small fraction of the total Pacific Ocean basin activity (3%), two of these tropical cyclones have had recently a major impact on the nation of Malaysia. Of the 50 tropical cyclones observed over the period, 27 of these (54%) were of local origin (Table 3). Local origin here refers to storms developing within the study area shown in Figure 1.

Table 3. Total Number of Tropical Cyclones and Annual Averages Over the SSCS Region From 1960 to 2006
 TDTSTYTotalLocal Origin
Annual average0.210.620.231.060.57

[20] Table 4 demonstrated that the active tropical season for Malaysia persists from October to April. However, most of the activity (78%) occurred during the months of November (42%) and December (36%). This represented a sharp peak in the active season and likely is related to the annual migration of the Intertropical Convergence Zone (ITCZ), and the associated convection, through the SSCS [e.g., Hurrell et al., 1995; Chang et al., 2005a]. This seasonal peak seemed to exist regardless of whether the tropical cyclones were of local origin or they migrated into the area from the adjacent area to the east. While tropical cyclones have occurred in all months in the greater western North Pacific in general, the poleward portion of the greater Pacific region is climatologically more active from July to October [Yumoto and Matsuura, 2001]. Thus, the peak of the tropical season for the SSCS followed that of the greater western North Pacific.

Table 4. Total Number and Percentage of Tropical Cyclone Occurrence in the SSCS by Month for the Period 1960–2006, Following the ENSO Year Definition Used Here (Beginning With October)

[21] Tropical cyclones in the SSCS region tended to move from east to west and more of these tended to move across the poleward portion of our study domain. Tropical cyclone Vamei is an example of this type of behavior (Figure 2b). Vamei developed rapidly off the coast of Malaysia and was classified as a typhoon early on 27 December 2001 ( [see also Chang et al., 2003]. The storm then moved inland and decayed quickly over land, but not before causing widespread flooding in Malaysia and neighboring nations. However, some storms did move west to east (Figure 2a) and tropical cyclone Greg is an example of this type of behavior. Greg developed out of a tropical depression on 24 December 1996 and became a tropical storm just before striking northern Malaysia and then moving out of the SSCS. Following the move out into the SSCS, Greg was downgraded to a tropical depression again. Additionally, locally developing tropical cyclones tended also to have a short track, while most of the typhoons were storms that have migrated into the region.

Figure 2.

Tracking information for (a) Tropical Storm Greg (24–31 December 1996) and (b) Typhoon Vamei (27–28 December 2001). The mauve, dark blue, and light blue colors indicate a tropical depression, storm, and typhoon, respectively. This figure is adapted from the archive available through the UNISYS website.

[22] Figure 3 shows a time series of tropical cyclone activity over the 47 year period. With the exception of the quiet period from 1976 to 1980, there has been steady activity in the SSCS during the period. A regression analysis demonstrated that there was a slight upward trend in the number of tropical cyclones during the period of record overall (Figure 3a). However, when this trend was tested for significance, it was found that the trend was not statistically significant at even the 90% confidence level by using the F test and assuming that there should be no trend a priori. This agreed with the findings of many other studies of the active tropical regions [e.g., Free et al., 2004]. There was a prolonged period of inactivity for tropical cyclones developing within the SSCS from 1974 to 1994 (Figure 3b). The relatively active period since 1995 has been the result of an increase in tropical depressions developing in the SSCS. The observations for the period during the 1974–1994 would coincide with the fact that there were more and stronger El Niño events. The observations also support the assertion of the studies cited in the reference section, mainly, that during El Niño years, the focus of tropical storm activity was located further east. Regardless, this would favor the migration of tropical cyclones into the region if there were no locally developing storms during years where one storm was observed. Alternatively, the noted increase in the climatological trend noted above could be the result of better satellite identification techniques. However, more observations (and/or reanalysis of previous records) would be needed to confirm this hypothesis, and the satellite record may also provide better insight into trends in tropical cyclone intensity.

Figure 3.

The time series of (a) all tropical cyclones and (b) tropical cyclones of local origin occurring within the SSCS region from 1960 to 2006. In Figure 3a the dashed line is the linear regression line.

[23] In order to detect interannual variations in tropical cyclone occurrences, the whole SSCS data set was partitioned into El Niño, La Niña, and neutral years. Table 5 shows that La Niña (El Niño) years were more (less) active than other years over the SSCS, and this result was significant at the 95% confidence level when testing the means of these years against the mean of the entire sample using the z-score test [Neter et al., 1988] using the null hypothesis. It should be noted that since the data set is small, statistical testing was only carried out on the complete data set and not subsets of the data, even though the distribution of tropical storms and events of local origin were similar to the distribution of the total sample. These results, however, were similar to the results of other studies for the entire western Pacific Region as well [e.g., Wu and Lau, 1992]. Finally, an examination of the variability with respect to the predominant phase of the PDO (not shown) demonstrates that there was little or no difference in the relative annual distribution of tropical cyclones. Additionally, there was no change in the interdecadal character of the ENSO related variability as found, for example, in the Atlantic by LJ00. It was not clear whether this result was a function of the small area included in the study, the small data set, or other real factors.

Table 5. Average Annual Tropical Cyclone ACTIVITY Separated by El Niño Years for All Tropical Cyclones (All), Typhoons (TY), Tropical Storms, Tropical Depressions (TD), and Tropical Cyclones of Local Origin
 AllTYTSTDLocal Origin
El Niño0.
La Niña1.
All years1.

4. Interannual Variation in the SSCs Background State

[24] One goal of this study was to examine whether the tropical cyclone activity in active and nonactive years could also be associated or correlated with variability in meteorological parameters related to tropical cyclone formation in the climatological background of the SSCS region. In order to accomplish this, the time series of the November/December mean variables were analyzed. These included SSCS area-averaged monthly mean (1) SSTs, (2) sea level pressure (SLP), (3) 850–200 hPa wind shear, (4) divergence (calculated using equation (4a)), (5) 850 hPa relative vorticity (equation (4b)), (6) 500 hPa heights, (7) precipitable water, and (8) lifted index. The time series of variables 1–5 were then correlated with one another. We examined the background climatological state of the SSCS as a study of the general circulation of a much wider region is beyond the scope of this study. There are many available studies in the literature on this topic [e.g., Hurrell et al., 1995; Houghton et al., 2001].

[25] As expected, there were strong correlations between the SSTs and the SLP, low-level divergence, vorticity, and the wind shear (Table 6). Some of these correlations were significant at the 95% confidence level, and only the SLP correlated with the wind shear and SST correlated with vorticity showed weak correlations. Then some of these time series were filtered following the procedures in section 2 and analyzed using FFTs (e.g., SSTs were representative of the other variables and shown in Figures 4 and 5) . A significant peak in the periodograms was found in the 3–7 year range (peak near two cycles per decade in Figure 5) which is a time period consistent with ENSO variability. A cross-spectral analysis was then performed using, for example the SST and SLP time series (not shown), and the results were consistent with the observations in Figure 5. Thus, it is likely that variability in SSCS background climatological state was real, and may have been due to (or at least associated with) the ENSO. Thus, the next step would be to demonstrate that during active years, the atmospheric background state must be more conducive to tropical cyclone occurrence than during nonactive years. It must also be shown that the active (nonactive) years were associated with La Niña (El Niño) years. Additionally, an analysis of long-term trends in the SSTs (Figure 4) and SLP (not shown) revealed that there was no statistically significant secular trend in the SSCS region.

Figure 4.

The raw mean monthly SSTs (°C) for the SSCS region from 1963 to 2002.

Figure 5.

The power spectra for the filtered mean SSTs in the SSCS region. The filter was set in order to remove the signal for time periods smaller than 2 years. The abscissa is cycles per decade; thus 10 is 10 cycles per decade or the annual cycle. The dashed line represents the 99% confidence interval for accepting the red noise spectrum null hypothesis.

Table 6. Correlation Between SSTs, SLP, Geostrophic Wind Shear, Wind Divergence, and Vorticitya
  • a

    The asterisk denotes a correlation at a 95% confidence interval. Div, divergence; vor, vorticity.


[26] Then, in order to test further the hypothesis (objectives, section 1 and elucidated above) and determine which primary and calculated variable(s) may have been encouraging or inhibiting tropical cyclone formation and maintenance in the SSCS, the SSTs, 200–800 hPa wind shear, and the 850 hPa divergence and relative vorticity were examined, and these were calculated as discussed in section 2. In order to sample these data objectively, the five most active years were chosen (Table 7) and then compared to two sets of nonactive years (Tables 8 and 9, no occurrences). The variables displayed in Figures 68 are the mean 1 November to 31 December SLP, 500 hPa height, precipitable water, and lifted index (stability) in the SSCS region. These data were composites of the conditions within the samples shown in Tables 79 and these groups overall. The two sets of nonactive years were compared by using the 5 years with the warmest SSTs and the 5 years with the coldest SSTs in order to generate objective samples of years with no activity and to determine the possible impact of SSTs in the SSCS. The mean SSTs for the warm (cold) nonactive years was about one standard deviation (0.45°C) greater (less) than the 45 year mean (28.14°C).

Figure 6.

The mean (a) sea level pressure (hPa, interval 0.1 hPa), (b) 500 hPa heights (m, interval 0.5 m), (c) precipitable water (kg m−2, interval 0.5 kg m−2), and (d) sea level pressure anomaly (hPa, interval 0.05 hPa) for November and December for the years shown in Table 7 in the SSCS region.

Figure 7.

(a–d) As in Figure 6 except for the years in Table 8.

Figure 8.

(a–d) As in Figure 7 except for the years in Table 9.

Table 7. Area-Averaged November and December Means for the Five Most Active Tropical Seasons in the SSCS
YearTotal TCsENSO YearSSTs (°C)Wind Shear (ms−1)Divergence (×10−6 s−1)Relative Vorticity (×10−6 s−1)
19983La Niña28.112.4−2.4910.00
19703La Niña28.112.0−2.22−2.35
Means  28.110.1−1.883.42
Table 8. As in Table 7 Except for Five Nonactive Years With the Warmest SSTs
YearTotal TCsENSO YearSSTs (°C)Wind Shear (ms−1)Divergence (×10−6 s−1)Relative Vorticity (×10−6 s−1)
19870El Niño29.09.2−1.950.90
20020El Niño28.82.7−1.31−0.90
19630El Niño28.47.5−0.08−4.95
19860El Niño28.35.8−2.941.55
Means  28.68.5−1.78−1.13
Table 9. As in Table 8 Except for Five Nonactive Years With the Coldest SSTs
YearTotal TCsENSO YearSSTs (°C)Wind Shear (ms−1)Divergence (×10−6 s−1)Relative Vorticity (×10−6 s−1)
19910El Niño27.92.9−0.02−0.68
Means  27.75.2−0.26−0.63

[27] The active sample contained two La Niña and three neutral years, but these three neutral years were “cold” neutral, that is SST anomalies close to La Niña conditions during the October–January period. The nonactive warm SST years were predominantly El Niño, but 1 year was a cold neutral year. The nonactive cold years included one El Niño year and two “warm neutral” and 1 year with SST anomalies very close to zero. However, when examining all 18 years in which there was no tropical cyclone activity, 12 years were El Niño or warm neutral and two were years with SST anomalies very close to zero. The three groups described above together represent almost one third of the total number of years.

[28] This analysis demonstrates that during the active years, the area averaged 200–850 hPa wind shear was of similar magnitude to that of the nonactive warm SST years (Tables 7 and 8) and close to the overall mean. A positive number indicated wind speeds increasing with height. The mean value for the five active years in Table 7 was smaller compared to midlatitude values of 15–30 ms−1 found by taking several random samples of mean November/December wind shear within the midlatitude region approximately 30° latitude poleward of the study area. The wind shear values in Tables 79 were lower than 15 ms−1, and these values are generally accepted as being supportive of tropical cyclone formation and maintenance. The relatively low tropospheric wind shear in the SSCS indicated a less baroclinic environment, especially when compared to the midlatitudes. Additionally, an examination of the low-level relative vorticity for the active seasons demonstrated that the lower troposphere was dominated by cyclonic relative vorticity (Table 7), but the value in Table 7 was only slightly greater than the sample mean. The active year composites in Figure 6 support the discussion above as it showed that there was relatively low SLP in the SSCS (Figures 6a and 6d), and there was also ridging in the midtroposphere (Figure 6b), especially over the eastern portion of the SSCS. This would suggest an SSCS atmospheric environment that was more cyclonic near the surface, and the troposphere was warmer than the samples from the other group. This atmosphere would also be supportive of warm core disturbances such as tropical cyclones. Additionally, the SLP was significantly lower (at the 95% confidence level, for the whole sample σ = 0.40 hPa) than the long-term mean across the SSCS, except along the equator.

[29] However, in the nonactive years with warm SSTs (Table 8), factors other than SSTs, such as the more anticyclonic environment seemed to differentiate the active year sample from the inactive years. The SSTs from this sample were actually warmer than those for active years, but not at any accepted level of statistical significance. The area averaged low-level vorticity was the more anticyclonic (lower than the mean at the 95% confidence level) for this group (Table 8) than the two groups in Tables 7 and 9. The low-level divergence was similar to that of the active year sample. The composites of the years shown in Table 8 are given in Figure 7. The near surface environment was clearly more anticylonic for the nonactive years (Figure 7a), but not different significantly from the long-term mean (Figure 7d). In the midtroposphere there were stronger height gradients over the region (Figure 7b), and the flow was nearly zonal or showed weak troughing over much of the domain. This atmospheric configuration is similar to results found by Chang et al. [2005b] who found that convection was suppressed over the SSCS when the near surface environment was more anticyclonic (or as stated in their work, the “Borneo vortex” was not present).

[30] Additionally, an examination of the precipitable water suggests that the SSCS environment in these nonactive years (Figure 7c) was clearly drier than for the active year sample (Figure 6c). Thus, it is suggested here that there was more low-level atmospheric water vapor available to potential tropical cyclones in the active years. Latent heat release has been identified as an important thermodynamic process in tropical cyclone intensification [e.g., Holland, 1997]. Since climatologies of the tropics [e.g., Hurrell et al., 1995] demonstrated that most atmospheric moisture is contained in the boundary layer, the precipitable water here was taken to represent low-level moisture available for possible ingest into convective systems.

[31] In the SSCS, it appears that even if the other background variables were similar (e.g., wind shear, low-level divergence, SSTs), the tropical cyclone activity in the region was inhibited during seasons when the relative vorticity was more anticyclonic (even though the low-level wind shear was favorable). The variables shown in Figure 6 (Figure 7) and Table 7 (Table 8) suggest a more (less) cyclonic and warm core background climatological state for the active (inactive) warm SST years. These results also suggest that in the SSCS, this set of El Niño years with warm SSTs may have been more hostile to tropical cyclone development as a result of the atmospheric conditions, and not necessarily because of cooler SSTs. Many studies have shown for different regions that, in addition to warm SSTs, the atmospheric background state was also a critical factor in contributing to tropical cyclone activity [e.g., Holland, 1997; Tonkin et al., 2000; Dare and Davidson, 2004; Paterson et al., 2005]. Free et al. [2004] also demonstrated that SST was not always associated with a higher potential intensity for tropical cyclones.

[32] During the warm SST years (Table 8) in which no storms occurred in the SSCS, it is clear that the wind shear was similar to that of the active years. During the cold SST nonactive years (Table 9), however, the low-level wind shear was even weaker than that of the samples in Tables 7 and 8. Nonetheless, the low-level background was also more anticyclonic (again, at the 95% confidence level) when comparing the low-level vorticities (anticyclonic relative vorticity) to those of the warm active years (Table 7). A composite map of the SLP (Figure 8a), however, showed that in spite of weak cyclonic conditions over the ocean, similar to the active years (Figure 6a), these values were higher than that of the long-term mean (but not at the 90% confidence level). But, there were strong height gradients and troughing over the much of the SSCS region and weak ridging over the east side (Figure 8b), which was more similar to the warm nonactive years (Figure 7b). This was also a season in which the tropospheric moisture was comparable to that of the active years (compare Figure 8c to Figure 6c), but more than that of the warm nonactive years (Figure 7c). Thus, in addition to the cooler SSTs, the nonactive years sampled in Table 9 presented a less favorable atmospheric background environment for tropical storm formation than that of the active years. Additionally, for Tables 8 and 9 (not shown), adding the next 3 years to these samples in which no tropical cyclone activity occurred did not change the interpretation of the results discussed here in section 4. Adding those years represented a total of 21 years sampled from the 47 year period.

5. Summary and Conclusions

[33] An analysis of the interannual variations and long-term trends in tropical cyclone activity from 1960 to 2006 in the SSCS was performed here in order to determine whether or not the threat of tropical cyclone activity to the nations of this region, particularly Malaysia, has increased in a detectable fashion, or whether the active period of the last decade was due to natural cycles. Malaysia was struck by two deadly tropical cyclones during the 1996 to 2001 period. There has been concern in Malaysia that tropical cyclone activity and intensity may increase in association with the expected climate change whether or not the climatic changes are naturally or anthropogenically forced. The tropical cyclone data was obtained from the UNISYS hurricane archive, and the NCEP-NCAR gridded reanalyses available via the worldwide web were used for the analysis of the climatological background conditions.

[34] The important findings resulting from this work were (1) the SSCS experiences about one tropical cyclone event per year, whether these develop locally within the SSCS (about 54%) or they propagate into the region; (2) of these, most attain tropical storm status, but relatively few become typhoons; (3) the most active part of the season was confined to November and December (78% of all activity) and was likely associated with the ITCZ as it moves through the region; and (4) there was a slight upward trend in the occurrence of these events, but this trend was not statistically significant.

[35] When examining the SSCS sample for interannual variability we found the following:

[36] 1. La Niña (El Niño) years were more (less) active than other years, and this result was significant at the 95% confidence level when examining the total sample. The variability of tropical storms and tropical cyclones of local origin was similar to that of the total sample.

[37] 2. There was no apparent climatic variability (statistically significant) in the SSCS that could be attributed to interdecadal variability such as the PDO,

[38] A spectral analysis of the filtered climatological background variables such as SST, SLP, 200–850 hPa wind shear, 850 hPa divergence and 850 hPa vorticity showed that there was significant variability found in the 3–7 year period, which is consistent with that of the ENSO period. Examining a subset of the most active years (all La Niña and “cold” neutral years) versus those years with no tropical cyclone activity for the 5 years of (1) warmest SSTs (predominantly El Niño years) and (2) coolest SSTs yielded some interesting results (mostly El Niño or “warm” neutral). First, during warm nonactive SST years, tropical cyclone activity was likely suppressed as the low-level relative vorticity was considerably more anticyclonic, even though SSTs were about one standard deviation warmer and wind shears were similar to those of active years. The SSCS atmospheric environment for warm SST nonactive years was drier than that of the active years, and did not exhibit a surface–500 hPa structure that would be as supportive of warm-core tropical cyclones. Most of these years were also ENSO years, and two thirds of all years with no activity were El Niño or warm neutral. These results may indicate that, in the SSCS region, it was atmospheric conditions that were unfavorable to tropical cyclone development during El Niño years with warm SSTs.

[39] During cool SST years, tropical cyclone activity was suppressed by a climatological background that was more anticyclonic at low levels than that of active years, even though the wind shear was even more favorable. The background climatological state was more similar to that of the warm SST nonactive years. These results should provide forecasters in the SSCS region with guidance regarding the type of background conditions that are favorable for the occurrence of tropical cyclones. The majority of these were also El Niño and “warm” neutral years.


[40] The authors would like to acknowledge the Government of Malaysia, which provided funding for the first author. We also thank Patrick Market and Neil Fox, University of Missouri Department of Soil, Environmental, and Atmospheric Science, and Chris Wikle, University of Missouri Department of Statistics, for their comments on earlier versions of this work. Finally, we thank the two anonymous reviewers whose comments strengthened this manuscript considerably.