a. Palau region

Figure 3 plots the time series of a 5-day running mean for 850-hPa zonal winds in 1992. Westerly winds prevailed from June to November. The other period indicated easterly winds. We used a threshold of 5 m s⁻¹ to define the westerly wind regime. The first day that the 5-day running mean zonal wind exceeded 5 m s⁻¹ was 20 June 1992. The last day that the zonal wind exceeded 5 m s⁻¹ was 26 November 1992. We defined the westerly wind regime as the period between these two days in 1992. Zonal-wind fluctuations of less than a seasonal cycle are intraseasonal oscillations.

We expanded this monsoon definition to include the radiosonde data collected from 1973 to 2003. Figure 4 illustrates the westerly wind regime at Palau. Onset of the westerly wind regime occurred between May and July, and the withdrawal dates appear between September and December. The westerly winds did not exceed the threshold of 5 m s⁻¹ until August in 1973, 1988, 1995, and 1998. Onset of the westerly wind regime was vague and undefined during these years. Westerly winds exceeded the threshold before May in 9 yr. However, considering previous studies of the seasonal march, we did not take these dates into account for the onset of the westerly wind regime. Tanaka (1997) suggested that the onset date of the western North Pacific monsoon delayed in the El Niño years and appeared early in the La Niña years. There were three El Niño years in which the westerly winds did not exceed the threshold. There were four La Niña years in which the westerly winds exceeded the threshold before May. The monsoon onset defined by low-level westerly wind at Palau was also affected by ENSO events as suggested by previous studies (Tanaka 1997; Wu and Wang 2000). We used the term of El Niño and La Niña years defined by the Japan Meteorological Agency (more information available online at http://okdk.kishou.go.jp/products/elnino). Two major concentrations of the onset of the westerly wind regime occurred in the middle of May and the middle of June. There may be a phase lock of intraseasonal oscillation during these periods. Nakazawa (1992) and Tanaka (1992) studied this June phase lock over the western Pacific. The dates of the onset of the westerly wind regime may be affected by the intraseasonal oscillation.

We analyzed the seasonal variations of moisture the same way as for zonal wind. The maximum variance was observed at 700-hPa height. Therefore, we chose 700 hPa and plotted the time series for moisture as in Fig. 5. Large amounts of moisture appeared during the westerly wind regime. Moisture increased earlier than the onset of the westerly wind regime and lasted beyond the withdrawal dates. During easterly wind regimes, moisture frequently increased due to the activity of convections (see section 4b). Therefore the dry season was not as clear over Palau. However, a strong El Niño event occurred during 1983 and 1998 and a dry period continued from January to April in these years. Next we expand the monsoon definition to the western Pacific region using QuikSCAT data.

b. Western Pacific region

Figure 6 depicts the times series of QuikSCAT surface zonal winds in 2001 in the same region as Koror (7°N, 134°E) and data superimposed by Koror 850-hPa zonal wind in the same year. The amplitude of these two values was different, but their fluctuations were very similar. The onset date of the westerly wind regime and the withdrawal date using QuikSCAT surface zonal winds were consistent with Koror 850-hPa radiosonde data. Both datasets determined onset on 18 May and withdrawal on 26 November 2001. Therefore QuikSCAT surface zonal winds help us expand the definition of the onset of the westerly wind regime. Initially we created the first harmonic component of 1 yr of QuikSCAT surface-zonal-wind data. Figure 7 shows the amplitude of this first harmonic component. High amplitudes represented the apparent seasonal variation of surface zonal winds. High amplitudes appeared on both sides of the Philippines, including Palau. This was consistent with the northwestern Pacific summer monsoon region as defined by Murakami and Matsumoto (1994) and also appeared over the adjacent Sea of Indonesia (Ramage 1971). We analyzed the onset dates of the westerly wind regime over the region where the amplitude exceeded 3 m s⁻¹. Figure 8 shows the onset dates of the westerly wind regime defined by the same threshold of greater than 5 m s⁻¹ of a 5-day running mean zonal wind using QuikSCAT surface-zonal-wind data in 2001. This year 2001 was not an El Niño or La Niña year. The rapid-onset regions were in the southern part of the South China Sea in early May and then appeared in the southwestern part of the western Pacific, including Palau, in the middle of May. The onset region’s spread to the north and east was more or less consistent with previous studies (Lau and Yang 1997; Wang and LinHo 2002). However, the onset region did not spread continuously. In nearly all of the regions in the western Pacific and the South China Sea, the westerly wind regime lasted from May to August. In Southern Hemisphere regions, around the adjacent sea of Indonesia, the westerly wind regime occurred from October to December and the onset region spread from west to east. Figure 9 shows the withdrawal dates of the westerly wind regime based on QuikSCAT surface zonal-wind data for the year 2001. The withdrawal region spread southward; however the progress was not continuous. The southwestern part of the western Pacific, including Palau, was the region with the latest withdrawal of the westerly wind regime, and this occurred at the end of November. Around the adjacent Sea of Indonesia, withdrawal dates of the westerlies occurred from March to April and spread from west to east. The seasonal march of the monsoon onset and withdrawal in this study was consistent with previous studies. Zonal-wind direction was useful for defining the westerly wind regime over the western Pacific region. Palau had the earliest start and latest end of the regions in the western Pacific.

a. Zonal wind and cloud amount

Figure 10 illustrates the time series of the 5-day running mean equivalent cloud amount and Koror 850-hPa zonal wind from 28 June 2001 to 30 June 2002. Equivalent cloud amount was calculated from ceilometer data (see section 2). The correlation coefficient between equivalent cloud amount and zonal wind was 0.71 during the westerly wind regime of 28 June to 25 November 2001. When westerly winds intensified with the active phase of intraseasonal oscillation, the equivalent cloud amount increased, indicating that convective activity intensified over Palau. This result was consistent with the study of intraseasonal oscillation in the off-equatorial regions (Kemball-Cook and Weare 2001). However, during the easterly wind regime, equivalent cloud amount did not decrease as zonal-wind direction changed to easterlies. There was no correlation between equivalent cloud amount and zonal wind during the easterly wind regime. After the westerly wind regime vanished, the northeasterly trade wind prevailed over Palau. The core latitude of the westerly winds observed within the intraseasonal oscillation shifted nearly to the equator during the Northern Hemisphere winter (Lau and Chan 1985). The convective activity and westerly wind were correlated over the equatorial region associated with intraseasonal oscillation during the Northern Hemisphere winter (Nitta and Motoki 1987). In contrast, a cyclonic circulation was produced over the off-equatorial region by the intraseasonal oscillation (Matsuno 1966; Gill 1980). Therefore, westerly winds did not blow, even when the active phase of intraseasonal oscillation passed over the western Pacific. These factors presumably weaken the correlation between equivalent cloud amount and zonal winds. The next westerly wind regime began on 18 May 2002. When westerly winds intensified, equivalent cloud amount increased as well. Zonal wind and equivalent cloud amount are correlated once again, after the westerly wind regime began.

b. Precipitation and precipitable water

Figure 11 plots the time series of daily precipitation and 5-day running mean precipitable water at Peleliu Island. Precipitable water was observed by GPS at Peleliu Island and by SSM/I in the same region (7°N, 134°E) and is superimposed in the figure. The difference in precipitation between westerly and easterly wind regimes was small compared to that of the Indochina Peninsula land region (Matsumoto 1997), and average precipitation amount was 8.8 and 8.2 mm day⁻¹, respectively (see Table 1 and section 4g). During the observation period, heavy rains exceeding 80 mm per day were observed mostly in the easterly wind regime. Rainy days persisted for several days in the westerly wind regime and were scattered during the easterly wind regime. Figure 11 shows precipitable water identified by GPS and SSM/I. SSM/I precipitable water data compensated the limited Peleliu GPS data. Averaged daily GPS precipitable water data exceeded 2.1 mm compared to data from SSM/I. Precipitable water reached nearly 65 mm during the westerly wind regime with only small fluctuations, probably because the amount of precipitable water was nearly saturated. During the easterly wind regime, the dry period appeared and precipitable water decreased below 40 mm at the end of January and early April. Precipitation was suppressed during these days. Figure 12 presents a latitudinal cross section of precipitable water at 134°E obtained from SSM/I. High precipitable water values (greater than 55 mm) spread northward beyond 20°N during the boreal summer and southward below the equator during the winter. Meridional propagation of precipitable water seemed to be associated with seasonal variations of the intertropical convergence zone (ITCZ). During the easterly wind regime, the dry region spread from north to south. At the end of January and early April, the dry air below 45 mm propagated southward, reaching 3°N. Over Palau, the atmosphere was affected by dry subtropical air, and when dry air was propagated from the subtropics, precipitation was suppressed during the easterly wind regime.

c. Diurnal variation of precipitation

Previous studies of diurnal variation of rainfall over the western equatorial Pacific demonstrated that the nocturnal maximum occurred when convection was active. During the inactive phase, the diurnal variation was weaker and had a weak afternoon maximum (Sui et al. 1997; Chen and Houze 1997; Kubota and Nitta 2001). Therefore, the diurnal variation was changed with the existence of large-scale disturbances. In this study, we also divided the analyses periods into the active phase of convection and the inactive phase based on equivalent cloud amount data. We determined a threshold of 5-day running mean equivalent cloud amount greater than 4 to be the active phase, and less than 3 to be the inactive phase (see Fig. 10). We further compared the diurnal variation of precipitation during the westerly wind regime to easterly winds. We observed 30% of the active-phase and 33% of the inactive-phase days during the westerly wind regime, and 20% of the active-phase and 44% of the inactive-phase days during the easterly wind regime. Figure 13 depicts the diurnal variation of precipitation during the westerly wind regime, which was divided into active and inactive phases. Precipitation increased from nighttime to early morning and decreased during the afternoon in the active phase. In the inactive phase, diurnal variation was weak and had a lesser afternoon maximum. The diurnal variation of precipitation during the westerly wind regime was consistent with previous work over the western equatorial Pacific. We demonstrated that during the westerly wind regime at Palau, the equatorial region atmosphere affected convection. Figure 14 depicts the diurnal variation of precipitation during the easterly wind regime. We did not see diurnal variation of precipitation in either the active phase or the inactive phase.

d. Duration and intensity of rain event

We examined precipitation properties further to determine the cause of the differences between the easterly wind regime and the westerly wind regime. We defined a rain event using hourly precipitation data at Peleliu Island. Figure 15 depicts the time series of hourly precipitation from 0900 local time (LT) 28 June to 0800 LT 30 June 2001. Continuous rain is defined as a rain event, and a rain event was divided into different events if the rain intermittence was 3 h or more. Figure 15 illustrates three rain events observed using this definition. Figure 16 presents the frequencies and intensities of rain events of both westerly and easterly wind regimes. The maximum frequency of a rain event appeared at 1-h duration in both westerly and easterly wind regimes and the frequencies decreased as duration increased except for 3 h. The most apparent difference of rain event frequency between westerly and easterly wind regimes is found in the duration of rain events exceeding 13 h. Rain event frequency during the easterly wind regime was over twice that of the westerly wind so the long-rain event was more evident in the easterly than in the westerly wind. The intensity of rain increased as duration of rain event became longer. Durations of almost all rain events demonstrated that rain events were more intense during the westerly wind regime. Lau et al. (1998) suggested that rain intensity increased during the Asian monsoon active period; however they did not mention about the difference of the duration of rainfall. Tagami (1990) suggested that the most important factors for the apparent diurnal variation of precipitation over Japan were rain events of short duration and strong intensity. Unlike in the easterly wind regime, our observation of rain events during the westerly wind regime also indicated short duration and strong intensity consistent with Tagami (1990).

e. Atmospheric stability

We examined the difference of vertical atmospheric environment between the westerly wind regime and easterly wind regime. Convective available potential energy (CAPE) is an index for representing the atmospheric stability, given by

View Expanded

where θ_υ and are the virtual potential temperatures of the air parcel and environment. The level of free convection (LFC) is the height at which the air parcel becomes warmer than the environment, enabling the air parcel to rise above this level without force. The level of neutral buoyancy (LNB) is the height at which the air parcel temperature is again equal to that of the environment.

When CAPE was greater the environment was more suitable for developing the convection (Thompson et al. 1979). The averaged CAPE during the westerly wind regime and easterly wind regime was 2824 and 1879 (m s⁻¹)², respectively (see Table 1). CAPE also differed between westerly and easterly, having a greater value during the westerly wind regime. This result indicated that the generation of deep convections was not suitable condition during the easterly wind regime. Furthermore we examined the composite of a relative humidity profile during the active phase defined by using a threshold of a 5-day running mean equivalent cloud amount data and divided the westerly wind regime from the easterly wind in Fig. 17 (see Fig. 10). During the easterly wind regime, middle layers around 250 to 700 hPa were dryer than the westerly wind regime even at the active phase.

f. Spatial distribution of precipitable water

A composite of the horizontal distribution of SSM/I precipitable water during the active phase of convection of the westerly wind regime and easterly wind regime is depicted in Fig. 18. A moist region of more than 55 mm of precipitable water was widely spread from the South China Sea to the western Pacific. This large-scale environment was suitable for developing convection during the westerly wind regime. However, most of the moist region of more than 55 mm was located south of 5°N, except the Palau region, during the easterly wind regime. The expansion of this wetness environment that promotes convection was limited during the easterly wind regime.

g. Summary

Table 1 lists the averaged characteristics of precipitation properties and the environment during the westerly wind and easterly wind regimes. The values of zonal wind at 850-hPa height were nearly identical in the opposite direction. The difference of precipitation and equivalent cloud amount between westerly and easterly wind regimes is within 10%. Precipitation and cloud amount do not decrease even during the easterly wind regime. These features differ from the Indochina Peninsula land region and the South China Sea (Matsumoto 1997; Lau et al. 1998). The diurnal variation of precipitation had a nocturnal maximum during the active phase in the westerly wind regime but was not clear during the easterly wind regime. The precipitation was intense and brief during the westerly wind regime, and weak intensity and long duration characterized the easterly wind regime. CAPE also differed between westerly and easterly, having a greater value during the westerly wind regime. Different environments appeared during the active phase of convection between the westerly wind regime and easterly wind regime. The middle layers were dryer during the easterly wind regime than in the westerly wind regime. A moist region expanded widely over the western Pacific during the westerly wind regime. In contrast, this region appeared only around Palau during the easterly wind regime. The horizontal and vertical environments were more suitable for developing convection during the westerly wind regime.

a. The environment for diurnal variation of precipitation

We divided the monsoon season into the westerly wind regime and the easterly wind regime for Palau, the off-equatorial region of the western Pacific. We used the threshold of a 5-day running mean of an 850-hPa zonal wind. The onset of the westerly wind regime occurred between May and July, and the withdrawal dates were between September and December. The definitions of monsoon onset and withdrawal were more or less consistent with previous studies using different OLR or rainfall data (Lau and Yang 1997; Wang and LinHo 2002). However, the difference in precipitation and equivalent cloud amount between the westerly and easterly wind regimes was small. Precipitation and cloud amount did not decrease, even during the easterly wind regime. These features differ from other Asian monsoon regions, such as the Indochina Peninsula land region and the South China Sea. The precipitation properties differed between regimes despite the minimal difference in precipitation between the westerly wind regime and easterly wind regime. The diurnal variation of precipitation had a nocturnal maximum during the active phase in the westerly wind regime but was not clear during the easterly wind regime. The precipitation was intense and brief during the westerly wind regime but exhibited weak intensity and long duration during the easterly wind regime.

Possible hypotheses for the characteristics of diurnal variation of precipitation differing during the active phase between the westerly wind regime and easterly wind regime were considered. One important factor that drives the diurnal variation of precipitation over the ocean is the existence of stratiform clouds. The spreading of stratiform clouds promotes radiative cooling at night and destabilizes the atmosphere (Xu and Randall 1995; Kubota et al. 2004). However, the middle layers were dryer during the active phase of the easterly wind regime than in the westerly wind regime (Fig. 17). We inferred that the environment was not favorable for the existence of stratiform clouds during the easterly wind regime. Expansion of the moist region is another factor. Previously diurnal variations of convection were observed during the active phase of large-scale disturbances that were associated with intraseasonal oscillations over the western Pacific equatorial region. The horizontal scale of a large-scale disturbance of intraseasonal oscillation observed over the equatorial region was more than 10 000 km (Nakazawa 1988). The active phase of intraseasonal oscillation continued for about a week during its passage (Lin and Johnson 1996). The moist region exceeded 55 mm of precipitable water spread over a large area of the western Pacific during the active phase of the westerly wind regime in our study (Fig. 18a). Palau was located in the southern region of the ITCZ during the westerly wind regime (Fig. 12). We can hypothesize that the environment was suitable for generating large-scale disturbances. This hypothesis may support the diurnal variation of precipitation during the active phase of the westerly wind regime. In contrast, the ITCZ moved south of Palau and subtropical dry air propagated around Palau during the easterly wind regime (Fig. 12). A moist region appeared in a limited area around Palau (Fig. 18b). This may diminish the horizontal scale of the disturbance. The seasonal variation of the sea surface temperature was less around the Palau region (not shown). Local-scale convection can be generated outside the ITCZ area, increasing water vapor during the easterly wind regime (Figs. 5, 15 and 18b). The rain event duration was longer during the easterly wind regime than the westerly (Fig. 16). These results may cause the precipitation properties to differ due to the seasonal variation and may minimize the difference in precipitation amounts between the westerly and easterly wind regimes.

A diurnal variation of precipitation was recently revealed over the tropical region using the Tropical Rainfall Measuring Mission (TRMM) satellite (Sorooshian et al. 2002). However it is necessary to average the data over a long duration because of the coarse resolution of the observations. The nighttime to early morning maximum in the active phase and the weak afternoon maximum in inactive phase were combined when we analyzed both active and inactive phases together. The amplitude of diurnal variation thus became smaller and we misinterpreted the diurnal variation. We demonstrated the difference in diurnal variation of precipitation in this study by separating the active phase of convection from the inactive and comparing the westerly wind regime to the easterly wind regime. Petty (1995) was resolute regarding strong regional variations in oceanic rainfall. In situ high-temporal-resolution observations are required to understand and explain the regional distribution of precipitation properties. We found different horizontal and vertical environments between the westerly wind regime and easterly wind regime, which may affect the difference in precipitation properties. However, the different convective activity structures and the precipitation clouds present in the westerly wind regime and easterly wind regime have not been discussed. We used 10 months of observation data for this study. In situ observation continues at Peleliu Island and Aimeliik state stations. We anticipate that further observation will help us to clarify these uncertainties.

b. Island effect in precipitation

In this study, we used Peleliu Island station data, which were less affected by the large island. The NWS station provided by NOAA is located at Koror (7.33°N, 134.48°E) and performs long-term meteorological observation. Gray and Jacobson (1977) showed diurnal variation of precipitation over western Pacific island stations including Koror from 1961 to 1973. They distinguished the island-affected locations from those less influenced by the island and analyzed the diurnal variation of precipitation. They demonstrated that island-affected locations including Koror have not only a nocturnal maximum but also increased daytime precipitation. We used Koror precipitation data for the same time period as for Peleliu Island precipitation data and compared the diurnal variation. Figure 19 shows the diurnal variation of precipitation at Koror during the active phase of the westerly wind regime. As in Gray and Jacobson (1977), both nighttime and daytime precipitation increased. This feature of diurnal variation differed from that of Peleliu Island (Fig. 13a). The large island of Bebeldaob affected Koror’s NWS station but Peleliu Island was less affected and represented an open-ocean region. Therefore our observational studies at Peleliu Island support the difference of precipitation properties between westerly and easterly wind regimes and indicated that the difference did not come from the island effect but from the oceanic environment.