NOTES AND CORRESPONDENCE. Diurnal Convection Peaks over the Eastern Indian Ocean o Sumatra during Di erent MJO Phases

Journal of the Meteorological Society of Japan, Vol. 89A, pp. 317--330, 2011. 317 DOI:10.2151/jmsj.2011-A22 NOTES AND CORRESPONDENCE Diurnal Convection Peaks over the Eastern Indian Ocean o Sumatra during Di erent MJO Phases Mikiko FUJITA, Kunio YONEYAMA, Shuichi MORI, Tomoe NASUNO Research Institute for Global Change (RIGC), Japan Agency for Marine-Earth Science and Technology (JAMSTEC), Kanagawa, Japan and Masaki SATOH Research Institute for Global Change (RIGC), Japan Agency for Marine-Earth Science and Technology (JAMSTEC), Kanagawa, Japan Center for Climate System Research, The University of Tokyo (Manuscript received 31 May 2010, in final form 4 November 2010) Abstract The authors investigated diurnal convection peak characteristics over the eastern Indian Ocean o the island of Sumatra during di erent phases of the Madden Julian oscillation (MJO). During MJO phases 2 to 3 (P2 and P3) defined by Wheeler and Hendon (2004), prominent diurnal variation in convection was observed by satellites when moderate low-level westerly winds were dominant over the eastern Indian Ocean. The diurnal convection peaks were prominent over the island of Sumatra in the evening, while migrations of the convection toward the Indian Ocean were observed in the early morning. By using the Global Positioning System around the western region o shore of Sumatra, a significant reduction in water vapor was observed from evening until midnight, compensating for the upward motion over the island. During midnight to early morning, the water vapor increased in the western o shore region as the convections migrated from the island. This prominent diurnal variation confirmed the result from a numerical experiment by Miura et al. (2007) using the Nonhydrostatic ICosahedral Atmospheric Model (NICAM). During P2 to P3, the atmosphere over the eastern Indian Ocean contains abundant water vapor, while the Maritime Continent is fairly well heated by solar radiation under calm conditions. This situation should be favorable for the development of two diurnal convection peaks: the evening convection over the land induced by solar radiative heating and the midnight convection over the ocean triggered by convergence of the low-level westerly wind and the land breeze. Corresponding author address: Mikiko Fujita, Research Institute for Global Change (RIGC), Japan Agency for Marine-Earth Science and Technology (JAMSTEC), 2-15 Natsushima-cho, Yokosuka, Kanagawa 237-0061, Japan. E-mail: fmiki@jamstec.go.jp 6 2011, Meteorological Society of Japan 1. Introduction The maritime continent is one of the most active regions for the development of deep convection. Using satellite images, many studies have indicated that this convective activity has a notable diurnal cycle in most parts of the maritime continent (e.g., Nitta and Sekine 1994; Ohsawa et al. 2001). Diurnal migration of clouds toward the Indian Ocean is frequently observed, in particular around the island

318 Journal of the Meteorological Society of Japan Vol. 89A of Sumatra. Mori et al. (2004) analyzed the precipitation radar (PR) data obtained by the Tropical Rainfall Measuring Mission (TRMM) satellite to investigate the regional characteristics of precipitation around Sumatra. They stated that the precipitation in this region follows a clear diurnal cycle. The precipitation system is first generated in the evening around the southwestern mountains near the coastline of Sumatra, and then it migrates toward both the inland and coastal regions in the morning. Sakurai et al. (2005) analyzed the propagation of convection around Sumatra using the equivalent black-body temperature (T bb ) data obtained by the geostationary meteorological satellite; they found that the convection around the mountains propagated leeward. The convection propagates mostly westward throughout the year, except during the summer monsoon period when it propagates eastward under the westerly wind. Ichikawa and Yasunari (2006) further suggested that the propagating signal of the diurnal cycle strongly depends on the low-level zonal wind. They investigated the time-space characteristics of the diurnal rainfall over the island of Borneo and its surrounding seas using the TRMM PR data. From midnight to morning, the precipitation system propagates westward when the low-level easterly wind prevails around the island and eastward when the westerly wind prevails around the island. In the tropics, the Madden Julian Oscillation (MJO; Madden and Julian 1971, 1972, 1994) is a dominant mode of variability with characteristic periods of 30 60 days. The MJO convection usually originates over the equatorial Indian Ocean. The convective area of the MJO propagates eastward across the Maritime Continent into the western Pacific, accompanied by deep convective activity in the eastern hemisphere. Some previous studies have suggested an interaction between the diurnal cycle and the MJO. Many studies have indicated that the diurnal cycles of tropical sea-surface temperature (SST) and deep convection are strongly modulated by the MJO (e.g., Sui and Lau 1992; Johnson et al. 1999). Other studies have shown that the diurnal cycle of convection can influence the MJO. For example, Tian et al. (2006a) found that the diurnal cycle of tropical deep convection is enhanced over both land and ocean, while it is reduced during the convectively suppressed phase of the MJO. Ichikawa and Yasunari (2007) investigated diurnal disturbances over the Maritime Continent, which were embedded in the MJO. The diurnal cycle becomes pronounced during the passage of the MJO, and the eastward propagating diurnal disturbances dominate as part of the internal structure of the large-scale convection system of the MJO. In many previous studies, general circulation models (GCMs) have been adopted to simulate the MJO (e.g., Hendon 2000). A majority of GCMs have di culty in simulating a realistic MJO. One of the key problems with GCMs that one could think of is a convection scheme (e.g., Lin et al. 2006 and Sato, N. et al. 2009). On the other hand, to simulate the diurnal variation of convection induced by local circulation, numerical models with fine spatial resolution, su cient to resolve topography and thermally induced local circulations, are required. Hara et al. (2009) compared the diurnal variation of tropical convection simulated by a GCM and regional model. The GCM was able to simulate the convection system; however, it did not simulate well the spatial distributions of the total amount and structures of the convection systems over large islands. The abovementioned strong diurnal cycle that shows convection migrating toward the western region o shore of Sumatra was also simulated by Hara et al. (2009) using a fine regional model without cumulus parameterization. The global cloud-resolving Nonhydrostatic ICosahedral Atmospheric Model (NICAM) has been developed jointly by the Japan Agency for Marine- Earth Science and Technology (JAMSTEC) and the Center for Climate System Research (now Atmosphere and Ocean Research Institute) at the University of Tokyo (Tomita and Satoh 2004; Satoh et al. 2008). NICAM has a great ability to simulate convective systems explicitly in the tropics, where mesoscale convection is fundamental to the organization of multiscale convective systems. Miura et al. (2007) successfully simulated an eastward propagating MJO event that occurred in December 2006. Masunaga et al. (2008) and Liu et al. (2009) also analyzed the output of Miura s experiment by comparing it with the observations; they found that the three-dimensional structure of the MJO was realistically simulated by NICAM. This model output was also analyzed by Sato, T. et al. (2009) by focusing on the diurnal cycle of precipitation simulated by NICAM. They stated that NICAM successfully simulates the diurnal cycle of precipitation associated with the land sea breeze and the thermally induced topographic circulations as well as the horizontal propagation of the diurnal cycle signals.

February 2011 M. FUJITA et al. 319 The diurnal variations around the Maritime Continent, shown in the previous studies, are likely to be a ected by the MJO. In this study, we investigate the characteristics of the diurnal convection peaks around the island of Sumatra during di erent MJO phases using satellite data, reanalysis data, and the numerical model output of NICAM that adequately reproduces both the diurnal cumulus convection and large-scale dynamics such as the MJO. 2. Data and numerical simulation We use three-hourly TRMM PR data (3B42 product version 6) and hourly T bb data merged from major satellites, which are provided from the NASA as Global Infrared Radiation (IR) data, to analyze the diurnal propagation of convections during 2000 2008. The Global Positioning System (GPS) data covers the period of 2005 2008. The GPS stations were installed around the western o shore region of Sumatra (see Fig. 1) for tectonic surveillance after the severe damage by the western Sumatra earthquake in 2004. The Scripps Orbit and Permanent Array Center provides the raw GPS data. The GPS data from 24 sites are used to calculate precipitable water vapor (PWV) at three-hour intervals. The PWV data is converted from the atmospheric delays of the GPS signals. The atmospheric delays caused by water vapor are estimated Fig. 1. Research area map with topography (shaded; [m]). The GPS stations are marked by dots. The dashed line indicates the location of a cross-section used for analysis. by the analytic software RTnet (e.g., Rocken et al. 2005, Fujita et al. 2008). The atmospheric delay is converted to PWV using surface temperature and pressure data. The meteorological data is taken from the nearest grid point data of the NCEP final analysis data interpolated to three-hour intervals. In this paper, we analyze the NICAM output of the same experiment as in Miura et al. (2007), with an emphasis on the diurnal cycle of convections around Sumatra. The NICAM adopts the icosahedral horizontal grid system that covers the globe quasi-homogeneously. Its governing equations describe the non-hydrostatic, fully compressible atmosphere and permit acoustic waves. The model has 40 vertical layers using a terrain-following grid system; its model top is at 38 km, and its lowest layer thickness is 162 m. The cloud microphysics scheme of Grabowski (1998) is used with no cumulus parameterization. Over land, a simple bucket hydrological model is used, in which the initial soil moisture is given by the NCEP final analysis data on 1.08 grids. More details about the model configuration can be found in Satoh et al. (2008). The numerical experiment covered one month starting on 15 December 2006. Initial atmospheric conditions were obtained from the NCEP tropospheric analysis (with a 1:08 1:08 grid). The spatial and temporal variations of SST were interpolated from the weekly mean data of Reynolds SST (Reynolds and Smith 1994). The experiment, with a horizontal mesh size of 7 km, was integrated over 32 days, which corresponds to MJO phases P2 to P6, defined by Wheeler and Hendon (2004) as this experiment was aimed at the onset of the MJO convection over the Indian Ocean. Six-hourly snapshots of 3-dimentional variables and 1.5 hourly averaged 2-dimentional variables from the NICAM integration were used in the analyses. 3. Results 3.1 MJO phenomenon In order to show the climatological diurnal cycle of convection, we first define the Real-Time Multivariate MJO (RMM) index (Wheeler and Hendon 2004) using multiple-variable empirical orthogonal functions (EOFs) of band-averaged (15 Sto15 N) 850 hpa and 200 hpa daily zonal winds from the NCEP, and satellite-observed outgoing longwave radiation (OLR) data for the NOAA Interpolated OLR dataset of 1999 2009. The RMM index is composed of a pair of spectral properties of the temporal coe cients of the EOFs (RMM1 and

320 Journal of the Meteorological Society of Japan Vol. 89A P8) in this paper refer to RMM Index phases, and daily phase data for the period from January 1, 2000, to December 31, 2008, is used. Figure 2 shows 10 100 day band-pass-filtered composite maps of MJO phenomena in each phase. Over the central Indian Ocean, an easterly wind is dominant during P7 to P8, as well as P1, while a westerly wind is seen during P3 to P5. During P2 and P6, the low-level wind is relatively weak over the central Indian Ocean. From P2 to P3, convections are activated over the central Indian Ocean and propagate toward the east during P4 to P8. The westerly wind is relatively strong around the negative OLR area. Meanwhile, around the eastern Indian Ocean o Sumatra, calm wind conditions appear in P2 P3 and P6 P7. An easterly wind is dominant in P8 and P1, while a westerly wind is seen during P4 and P5. Fig. 2. Composite horizontal distribution of OLR (color; [W m 2 ]), and horizontal wind (vector; [m s 2 ]) in 850 hpa level in each MJO phase, after being band-pass filtered around 10 100 days. The numbers indicate the MJO phase and counts of applied days. RMM2). These two EOFs describe features of the MJO, such as the eastward propagation of convection anomalies in the Eastern Hemisphere, the out-of-phase relationship between the lower- and upper-tropospheric wind anomalies, and the predominance of the lower-tropospheric westerly anomalies near and to the west of the enhanced convection. We selected days with strong MJO activity, defined by ðrmm1 2 þ RMM2 2 Þ > 1, for this analysis. All references to Phases 1 to 8 (P1 to 3.2 Observed diurnal cycle around the west of Sumatra The diurnal variation of convection over the eastern Indian Ocean and the island of Sumatra was calculated for the respective MJO phases. Figure 3 shows horizontal maps of diurnal convection and precipitation observed by IR satellites and TRMM at local times (LT) of 1600 (left), 2200 (middle), and 0400 (right). During P2 to P3, the amplitudes of the diurnal cycle of convection were large, while a clear contrast between the diurnal peaks over the land and ocean was observed. In the evening, convection and precipitation were prominent around the mountains of Sumatra. The evening land precipitation was formed by an upslope wind induced by earlier daytime solar radiation (Mori et al. 2004; Wu et al. 2009). In the early morning of the following day, convection and precipitation peaks appeared around the western region o shore of the island. The convections remained over the ocean during the daytime. On the other hand, during P4 to P6, the amplitude of the diurnal cycle became progressively smaller. The evening convection over Sumatra increased gradually in P7 P8 and in P1. Figure 4 shows the time-longitude cross-section of averaged diurnal variation of convection along the dashed line shown in Fig. 1. As shown in Fig. 3, the diurnal cycles were distinctly di erent from one MJO phase to the next. During all phases, the land precipitation appeared around 1700 and 2300 LT. Around midnight, the convection had begun to migrate to the west, o shore of the island. The most promi-

February 2011 M. FUJITA et al. 321 Fig. 3. Averaged diurnal convective migration around the Sumatra Island at (a) 1600, (b) 2200, and (c) 0400 LT, observed by IR (color; T bb [K]) and TRMM (contour; precipitation rate [mm h 1 ]) satellites drawn in each MJO phase. The precipitation rate was illustrated with a 0.5 mm h 1 interval. nent amplitude of the diurnal cycle appeared in P2 and P3. During P4 to P5, when the westerly wind was strong over the Maritime Continent corresponding to the MJO convective area, weak diurnal migration was also observed. In the easterly wind phases of P7 to P8, although the diurnal amplitudes were small both over the land and to the west of the island, the contrast of the diurnal peaks between land and sea had a clear diurnal variation. The diurnal amplitude gradually became larger during P7 to P8 and during P1. The diurnal variation of GPS-PWV over the smaller islands in the western region o shore of Sumatra (see Fig. 1; the Mentawai Islands) was averaged for each MJO phase (Fig. 5). The diurnal variation is shown as a deviation from its daily

322 Journal of the Meteorological Society of Japan Vol. 89A Fig. 4. Time-longitude cross-sections of averaged diurnal variation of convection (shaded; T bb [K]), precipitation (contour; precipitation rate [mm h 1 ]) in each MJO phase, along the dashed line in Fig. 1. The precipitation rate was illustrated with a 0.5 mm h 1 interval. The Y-axis indicates the local time; 24-hour data sets are shown. The blue line indicates the western coastline of Sumatra. mean. During all phases, the water vapor around the Mentawai Islands had a clear diurnal cycle. In the evening, water vapor converged over the islands due to daytime solar heating. In contrast, at night, a reduction of water vapor was observed. The most prominent reduction in water vapor was observed in P2 and P3. After that, GPS-PWV increased again in the early morning for all phases. Figure 6 shows a composite section of the averaged diurnal cycle of GPS-PWV over the Mentawai Islands in each MJO phase. As shown in Fig. 5, clear diurnal variations were observed; moreover, the diurnal cycle itself varied in terms of its peak from one MJO phase to the next. During P2 and P3, the water vapor decreased between evening and midnight. After that, a clear increase in water vapor was observed in the early morning. The daytime increment in water vapor was likely to be formed by solar radiative heating (Wu et al. 2009). In P4 and P5, the start time of the weak evening reduction was later than that in P2 and P3. A rise in water vapor was observed during 0300 and 0900 LT. Small amplitudes were observed in P6 and P7, while the nighttime reduction was indistinct. After the insignificant reduction, the water vapor increased relatively early at 0200 LT. Similar varia-

February 2011 M. FUJITA et al. 323 Fig. 6. Time and MJO phase cross-section of GPS-PWV. Area-averaged GPS-PWV (color; [mm]) over the Mentawai Islands at the o the west coast of Sumatra is shown as the deviation from the daily mean of each phase. tions were observed in P8 and P1, which had larger amplitudes than those of P6 and P7. The water vapor diurnal cycle is summarized as follows. First, the water vapor reduction between evening and midnight indicated a downward trend to compensate for the upward motion around the island of Sumatra. The reduction amplitude is an indication of the strength of subsidence flow induced by the land heating; a sharp water vapor reduction would indicate a strong convection forming over Sumatra due to the upslope wind. Next, we consider the water vapor increment from midnight to early morning. This increment is caused by the migration of convections from the island toward the western region o shore. This diurnal migration is also seen in Fig. 4; especially, the modulation of the GPS-PWV diurnal peaks was consistent with satellite observations. The diurnal cycle of water vapor observed by GPS at one station on Siberut Island was reported in Wu et al. (2008). Their results indicated a clear diurnal water vapor transport induced by the local circulation, consistent with what we found here in P2 and P3. Fig. 5. Diurnal variation of GPS-PWV (color; [mm]) averaged for each MJO phase during 2005 2008, shown as the deviation from the daily mean of each phase (a) 1600, (b) 2200, and (c) 0400 LT.

324 Journal of the Meteorological Society of Japan Vol. 89A 3.3 Simulated diurnal variation by NICAM Considering the observational results, the diurnal cycle around the west coast of Sumatra was clearly di erent in each MJO phase, with a prominent westward migration of convection tending to appear during P2 and P3. In this section, we study these characteristics of diurnal variation using the NICAM output reported by Miura et al. (2007). Fig. 7. Time-longitude plots of the OLR (color; [W m 2 ]), precipitable water (contour; [mm]), and horizontal wind at 850 hpa (vector; [m s 2 ]) averaged between 5 S and 5 N simulated by NICAM from December 16, 2006, to January 15, 2007. The precipitable water data larger than 50 mm was illustrated with a 10-mm interval. The bold line indicates the precipitable water of 60 mm. Figure 7 shows the time-longitude cross-section from December 16, 2006, to January 16, 2007, which corresponds to MJO phases P2 through P6, because the experiment was conducted at the onset of MJO convection over the Indian Ocean. In the observations and numerical simulation, significant convective clouds of MJO passed through the island of Sumatra around January 1 2, 2007, (Miura et al. 2007; Nasuno et al. 2009). Liu et al. (2009) determined the MJO characteristics using the same NICAM output. They indicated that the MJO has a realistic evolution in amplitude pattern, geographical locations, eastward propagation, and baroclinic- and westward-tilted structures. However, in the simulated event, the MJO grows faster in phases 2 and 3, and peaks with a 30% higher amplitude than that observed, although the 7-km version shows slight improvement with respect to the biases. Based on the results of Liu et al. (2009), in the present study, the following results of the MJO simulated by NICAM were focused on the geographical location of the MJO and its characteristics in di erent MJO phases. Figure 8 shows simulated horizontal diurnal distribution of convection and surface wind in P2 (December 17, 2006), P4 (January 3, 2007), and P6 (January 13, 2007). The diurnal variations of the convections were simulated clearly. The diurnal migration toward the Indian Ocean appeared in P2, when calm surface winds were simulated. In the evening, convections appeared over Sumatra, due to the convergence induced by the land being heated by solar radiation. Around midnight, the convections over the mountain region shifted toward the Indian Ocean. Near the coastal region of Sumatra, a strong land breeze was simulated. In P4 and P6, however, the diurnal amplitude of convection was unclear when both land and sea were covered with clouds throughout the day. Westerly surface winds prevailed in P4 and P6, and the nighttime land breeze was not simulated. The observational results of the corresponding cases are illustrated in Fig. 9. Especially, the diurnal variation of convection at the geophysical location were quite similar to the simulated results in Fig. 8. In P2, the convections converged over the mountains of Sumatra at night and were located above the western o shore region of Sumatra after midnight. The convection variations around the western o shore area were not observed in the other phases. Figure 10 shows the time-longitude cross-sections of precipitable water, OLR, and surface wind simu-

February 2011 M. FUJITA et al. 325 Fig. 8. Simulated diurnal variation of convections around Sumatra (color; OLR [W m 2 ]) including surface wind (vector; [m s 2 ]) at (a) 0830, (b) 1430, (c) 2030, and (d) 0230 LT. lated by NICAM in P2 (December 17 21, 2006), P4 (January 2 6, 2007), and P6 (January 12 16, 2007). During P2, the westward migration of convection was well simulated. The diurnal variation of water vapor corresponding to the convective migration was also simulated in the early morning, while the water vapor showed a decrease at nighttime around the western region o shore. The surface wind around the coastal area varied diurnally; the surface wind blowing from Sumatra toward the Indian Ocean tended to appear before the migration of the convection. This land breeze might converge with the westerly wind from the Indian Ocean; after that, the convection should migrate toward the o shore region west of Sumatra. This migration mechanism was speculated in Mori et al. (2004), in their introduction section. They mentioned that the convection was formed around the west coast of Sumatra, and was caused by the convergence between the low-level westerly wind and the land breeze from the island of Sumatra. On the other hand, during P4 and P6, these diurnal migrations were no longer clear in the simulation, which is consistent with the climatological observational results. In P4, both the land and ocean around the island were continuously covered with convection and moist air masses, because the MJO convection developed around the Maritime Continent. The westerly surface wind prevailed, while a weak topography-induced wind was simulated. In the MJO inactive phase of P6, weaker convections were simulated sparsely. The atmosphere was relatively dry during this period, while the westerly surface wind prevailed. These simulated features are basically consistent with the migration of convection observed by satellites and the diurnal cycle of water vapor derived from the GPS measurements. 4. Discussion The early morning o shore convection usually migrated from Sumatra in P2 P3, as shown in section 3. Notably in P2 P3, when the land convections had developed around midnight, a strong land breeze was simulated. The results indicate that a strong land breeze should be associated with the evening convection over the land induced by

326 Journal of the Meteorological Society of Japan Vol. 89A Fig. 9. Similar to Fig. 8, but for observation by IR satellite (color; T bb [K]) and TRMM (contour; precipitation rate [mm h 1 ]) at (a) 0800, (b) 1400, (c) 2000, and (d) 0200 LT. The TRMM data is illustrated with an interval of 3 mm h 1. daytime heating. In this section, we discuss this strong land breeze because we consider it to be a key di erence of the diurnal variation in the MJO phases. Figure 11 shows the simulated evening-tomidnight vertical structure of water vapor and wind over the west coast of Sumatra in P2 (December 17, 2006), P4 (January 3, 2007), and P6 (January 13, 2007). During P2, clear topography-induced local circulation was simulated. Significant land-sea local circulation prevailed in the evening around the west coast of the island. During the evening of P2, the western o shore area (98 99 E) is assumed to be a downward area for a compensating subsidence of the upward motion and the anabatic wind. Thus, the water vapor around the o shore area decreased during the night. In contrast, the water vapor over the island converged around the mountainous area in an up-sloping motion due to daytime solar radiation. Around midnight, a high water vapor area existed o shore. The strong land breeze was also simulated with a cold air mass over the coastal area. On the other hand, the diurnal cycle in P4 and P6 was not simulated clearly; this agrees with the observational results. The simulated lower atmosphere had abundant water vapor in P4, but was relatively dry in P6. The water vapor converged over the island of Sumatra; however, the local circulation induced by land heating was unclear in the strong westerly wind. The cold land breeze blowing from Sumatra toward the Indian Ocean was reported as a topography-induced gusty o shore flow in Wu et al. (2009). They observed the sudden o shore wind accompanied by an abrupt drop in surface temperature in the late afternoon and evening. Fujita et al. (2010) investigated the diurnal convection over the Strait of Malacca using a regional numerical model. The model results in Fujita et al. (2010) indicated that the morning precipitation peak in the strait was induced by the convergence of two cold outflows that had been produced by the precipitation systems in the previous evening over Sumatra and the Malay Peninsula. The cold outflow blowing toward the

February 2011 M. FUJITA et al. 327 Fig. 10. Simulated time-longitude cross-sections of precipitable water (shaded; [mm]), OLR (contour; [W m 2 ]), and surface horizontal wind (vector; [m s 2 ]) averaged between 2 S and 2 N. (a) P2, (b) P4, and (c) P6. The OLR data smaller than 200 W m 2 is illustrated with an interval of 25 W m 2. The bold line indicates the OLR of 200 W m 2. Fig. 11. Simulated vertical structures of potential temperature (shaded; [K]), water vapor mixing ratio (contour; [g kg 2 ]), and wind (vector; horizontal [m s 2 ], vertical [cm s 2 ]), averaged between 0.5 S and 0.5 N. (a) 1900 and (b) 0100 LT. The mixing ratio data larger than 15 g kg 1 is illustrated with an interval of 1gkg 1. The bold line indicates the mixing ratio of 16 g kg 1. The ground level is illustrated by a dashed line.

328 Journal of the Meteorological Society of Japan Vol. 89A warmer sea surface could be a trigger for new convection, and the upward flow would decrease the moist static stability. As some previous studies have indicated (e.g., Wu et al. 2009; Fujita et al. 2010), the strong midnight land breeze was induced by the convections over the land during the evening. This cold air mass was probably produced by evaporation of raindrops from land precipitation. The early morning convections in the o shore area could be formed by convergence, which might be triggered by this strong land breeze and low-level westerly wind. Therefore, convections over Sumatra from evening till midnight would be one of the important processes of cloud migration. To form the evening convection over Sumatra, topography-induced circulations driven by radiative heating and abundant water vapor are essential. Calm environmental conditions are also favorable for creating local circulation. These situations during P2 P3 are shown in the present study, namely the low-level wind was relatively weak (Fig. 2), while the atmosphere over the eastern Indian Ocean contained much water vapor (Fig. 7). Moreover, the SST environmental field might have some relationship with the diurnal variation. The atmospheric response to the tropical intra-seasonal SST anomalies has been addressed in Woolnough et al. (2001). When the eastward phase speed of the imposed SST anomaly was similar to that of the MJO, a significant atmospheric response shows a convection lagging the positive SST anomaly, consistent with the observations of the MJO. The maximum SST anomaly tends to appear when the MJO convection is activating over the Indian Ocean (P1 P3). When the cold midnight land breeze blows toward the Indian Ocean, a warmer SST creates a larger temperature di erence in the low-level atmosphere. This temperature di erence will produce stronger instability around the eastern Indian Ocean. The di erence of diurnal convection peaks in each MJO phase is summarized as follows. The environmental atmospheric fields of the MJO around the eastern Indian Ocean produced a characteristic diurnal variation of convection. In P2 and P3, the atmosphere over the eastern Indian Ocean was damp, while the Maritime Continent was fairly well heated by solar radiation under calm conditions. This situation should be favorable not only for the evening land convection caused by radiation heating but also for the midnight o shore convection triggered by the cold strong land breeze. The maritime land was also heated in P6 and P7; however, the diurnal variation was weak because the convectively suppressed field prevailed and the atmosphere was drier. During P4 to P5, the strong westerly wind around the Maritime Continent, which corresponds to the MJO convection passing through, weakened the local circulation induced by the diurnal heating contrast between land and ocean. Therefore, the presence of the MJO clearly has an impact on the diurnal peaks of convection around Sumatra. This characteristic diurnal variation enhanced by the MJO should also play an important role in moisture transportation in the tropics. This diurnal convection peak in di erent MJO phases was reported in Tian et al. (2006a). For the region around Sumatra, the modulation of the diurnal amplitude in the present study is consistent with their results. However, the relationship of the convection induced by local circulation was not stated in Tian et al. (2006a). Additionally, the results of the present study can be extended to a discussion about the onset and propagation of the MJO itself. Tian (2006b) discussed the low-level moist thermodynamic preconditioning of MJO observed from atmospheric profiling satellite data. In their results, clearly preconditioned lower-tropospheric water vapor anomaly corresponding to the convection anomaly was observed over the Indian Ocean and western Pacific; it was quite consistent with predictions by the frictional Kelvin Rossby wave conditional instability of the second kind (wave-cisk) for the MJO (e.g., Seo and Kim 2003). In Tian et al. (2006b), the low-level moisture, especially over the eastern Indian Ocean, tended to be preconditioned when the MJO was activated over the central Indian Ocean. The diurnal water vapor transport in the present study probably enhances the low-level preconditioning of MJO, or interacts with it. 5. Conclusion Using observational data and numerical model output, we have investigated the diurnal convective peak characteristics over the eastern Indian Ocean o Sumatra during the di erent MJO phases defined by Wheeler and Hendon (2004). During P2 to P3, a prominent diurnal variation of convection was observed by satellites when moderate westerly winds were dominant. In the evening, convection and precipitation were prominent around the mountains on Sumatra. The evening

February 2011 M. FUJITA et al. 329 land precipitation was formed by an upslope wind induced by daytime solar radiation. In the early morning of the following day, the convection and precipitation peaks migrated toward the western region o shore of the island. The precipitable water vapor over the Mentawai Islands derived from GPS observations showed prominent diurnal water vapor transportation in P2 and P3. A reduction in water vapor between evening and midnight indicated a downward trend, compensating for the upward motion around the island. During midnight to early morning, a rise in water vapor was observed, which was caused by the westward migration of the convection from the island toward the western o shore region. These diurnal variations were captured by the NICAM output, obtained by Miura et al. (2007) with one-month integration from December 15, 2006. In P2, the prominent migrations of precipitation clouds were simulated over the western region o shore of Sumatra in the early morning. The cold land breeze blowing from the island toward the Indian Ocean tended to appear before the migration of the convection. Then, the early morning convection was simulated around the west coast of Sumatra. As mentioned in a previous study, this land breeze triggers the development of the convection in the o shore region when it converges with the westerly wind from the Indian Ocean. During P2 P3, the atmosphere over the eastern Indian Ocean was damp while the Maritime Continent was fairly well heated by solar radiation. This situation should be favorable not only for the evening land convection caused by the radiation heating but also for the midnight o shore convection triggered by the strong cold land breeze. On the other hand, during P4 to P5, the amplitude of the diurnal cycle became progressively smaller. In this phase, the strong westerly wind around the Maritime Continent, which corresponds to the MJO convection passing through, weakens the local circulation induced by the diurnal heating contrast between land and ocean. During P6 and P7, the diurnal variation was weak because a convectively suppressed field prevailed and the atmosphere was drier. Therefore, the presence of the MJO clearly has an impact on the diurnal peaks of convection around Sumatra. This characteristic diurnal variation enhanced by the MJO should also play an important role in moisture transportation in the tropics. Acknowledgments The authors thank Dr. Fujio Kimura of JAM- STEC for his helpful discussions and comments. Insightful comments and suggestions by reviewers helped us greatly improve the previous version of the manuscript. We thank all the members of the NICAM group for their e orts in developing and improving the model. References Fujita, M., F. Kimura, K. Yoneyama, and M. Yoshizaki, 2008: Verification of precipitable water vapor estimated from shipborne GPS measurements. Geophys. Res. Lett., 35, L13803, doi:10.1029/ 2008GL033764. Fujita, M., F. Kimura, and M. Yoshizaki, 2010: Morning precipitation peak over the Strait of Malacca under a calm condition. Mon. Wea. Rev., 138, 1474 1486. Grabowski, W. W., 1998: Toward cloud resolving modeling of large scale tropical circulations: A simple cloud microphysics parameterization. J. Atmos. Sci., 55, 3283 3298. Hara, M., T. Yoshikane, H. G. Takahashi, F. Kimura, A. Noda, and T. Tokioka, 2009: Assessment of the diurnal cycle of precipitation over the Maritime Continent simulated by a 20 km mesh GCM using TRMM PR data. J. Meteor. Soc. Japan, 87A, 413 424. Hendon, H. H., 2000: Impact of air sea coupling on the Madden Julian Oscillation in a general circulation model. J. Atmos. Sci., 57, 3939 3952. Ichikawa, H., and T. Yasunari, 2006: Time-space characteristics of diurnal rainfall over Borneo and surrounding oceans as observed by TRMM-PR. J. Climate, 19, 1238 1260. Ichikawa, H., and T. Yasunari, 2007: Propagating diurnal disturbances embedded in the Madden Julian Oscillation. Geophys. Res. Lett., 34, L18811, doi:10.1029/2007gl030480. Johnson, R. H., T. M. Rickenbach, S. A. Rutledge, P. E. Ciesielski, and W. H. Schubert, 1999: Trimodal characteristics of tropical convection. J. Climate, 12, 2397 2418. Lin, J. L. et al., 2006: Tropical intraseasonal variability in 14 IPCC AR4 climate models. Part I: Convective signals. J. Climate, 19, 2665 2690. Liu, P., M. Satoh, B. Wang, H. Fudeyasu, T. Nasuno, T. Li, H. Miura, H. Taniguchi, H. Masunaga, X. Fu, and H. Annamalai, 2009: An MJO simulated by the NICAM at 14- and 7-km resolutions. Mon. Wea. Rev., 137, 3254 3268. Madden, R. A., and P. R. Julian, 1971: Detection of a 40 50 day oscillation in the zonal wind in the tropical pacific. J. Atmos. Sci., 5, 702 708.

330 Journal of the Meteorological Society of Japan Vol. 89A Madden, R. A., and P. R. Julian, 1972: Description of global-scale circulation cells in the tropics with a 40 50 day period. J. Atmos. Sci., 6, 1109 1123. Madden, R. A., and P. R. Julian, 1994: Observations of the 40 50-Day tropical oscillation A review. Mon. Wea. Rev., 122, 814 837. Masunaga, H., M. Satoh, and H. Miura, 2008: A joint satellite and global cloud-resolving model analysis of a Madden-Julian Oscillation event: Model diagnosis, J. Geophys. Res., 113, D17210, doi:10.1029/ 2008JD009986. Miura, H., M. Satoh, T. Nasuno, A. T. Noda, and K. Oouchi, 2007: A Madden-Julian Oscillation event simulated using a global cloud-resolving model. Science, 318, 1763 1765. Mori, S., J.-I. Hamada, Y. I. Tauhid, and M. D. Yamanaka, 2004: Diurnal land-sea rainfall peak migration over Sumatera Island, Indonesian maritime continent, observed by TRMM satellite and intensive rawinsonde soundings. Mon. Wea. Rev., 132, 2021 2039. Nasuno, T., H. Miura, M. Satoh, A. T. Noda, and K. Oouchi, 2009: Multi-scale organization of convection in a global numerical simulation of the December 2006 MJO event using explicit moist processes. J. Meteor. Soc. Japan, 87, 335 345. Nitta, T., and S. Sekine, 1994: Diurnal variation of convective activity over the tropical western pacific. J. Meteor. Soc. Japan, 72, 627 641. Ohsawa, T., H. Ueda, T. Hayashi, A. Watanabe, and J. Matsumoto, 2001: Diurnal variations of connective activity and rainfall in tropical Asia. J. Meteor. Soc. Japan, 79, 333 352. Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses using optimal interpolation. J. Climate, 7, 929 948. Rocken, C., J. Johnson, T. Van Hove, and T. Iwabuchi, 2005: Atmospheric water vapor and geoid measurements in the open ocean with GPS, Geophys. Res. Lett., 32, L12813, doi:10.1029/ 2005GL022573. Sakurai, N., F. Murata, M. D. Yamanaka, S. Mori, J.-I. Hamada, H. Hashiguchi, Y. I. Tauhid, T. Sribimawati, and B. Suhardi, 2005: Diurnal cycle of cloud system migration over Sumatra island. J. Meteor. Soc. Japan, 83, 835 850. Sato, N., C. Takahashi, A. Seiki, K. Yoneyama, R. Shirooka, and Y. N. Takayabu, 2009: An evaluation of the reproducibility of the Madden-Julian Oscillation in the CMIP3 multi-models. J. Meteor. Soc. Japan, 87, 791 805. Sato, T., H. Miura, M. Satoh, Y. N. Takayabu, and Y. Wang, 2009: Diurnal cycle of precipitation in the tropics simulated in a global cloud-resolving model. J. Climate, 22, 4809 4826. Satoh, M., T. Matsuno, H. Tomita, H. Miura, T. Nasuno, and S. Iga, 2008: Nonhydrostatic Icosahedral Atmospheric Model (NICAM) for global cloud resolving simulations. J. Comput. Phys., 227, 3486 3514, doi:10.1016/j.jcp.2007.02.006. Seo, K.-H., and K.-Y. Kim, 2003: Propagation and initiation mechanisms of the Madden-Julian oscillation, J. Geophys. Res., 108, 4384, doi:10.1029/ 2002JD002876. Sui, C. H., and K.-M. Lau, 1992: Multiscale phenomena in the tropical atmosphere over the western Pacific, Mon. Wea. Rev., 120, 407 430. Tian, B., D. E. Waliser, and E. J. Fetzer, 2006a: Modulation of the diurnal cycle of tropical deep convective clouds by the MJO. Geophys. Res. Lett., 33, L20704, doi:10.1029/2006gl027752. Tian, B., D. E. Waliser, E. J. Fetzer, B. H. Lambrigtsen, Y. L. Yung, and B. Wang, 2006b: Vertical moist thermodynamic structure and spatial-temporal evolution of the MJO in AIRS observations. J. Atmos. Sci., 10, 2462 2485. Tomita, H., and M. Satoh, 2004: A new dynamical framework of nonhydrostatic global model using the icosahedral grid, Fluid Dyn. Res., 34, 357 400. Wheeler, M. C., and H. H. Hendon, 2004: An all-season real-time multivariate MJO Index: Development of an index for monitoring and prediction. Mon. Wea. Rev., 132, 1917 1932. Woolnough, S. J., J. M. Slingo, and B. J. Hoskins, 2001: The organisation of tropical convection by intraseasonal sea surface temperature anomalies, Quart. J. Roy. Meteor. Soc., 127, 887 907. Wu, P., S. Mori, J.-I. Hamada, M. D. Yamanaka, J. Matsumoto, and F. Kimura, 2008: Diurnal variation of rainfall and precipitable water over Siberut Island o western coast of Sumatra Island. SOLA, 4, 125 128. Wu, P., M. Hara, J.-I. Hamada, M. D. Yamanaka, and F. Kimura, 2009: Why a large amount of rain falls over the sea in the vicinity of western Sumatra Island during nighttime. J. Appl. Meteor. Climate, 48, 1345 1361.