Corresponding Author Address: Dr. Weixin Xu, Department of Atmospheric Sciences, Colorado. 28. State University, 3915 West Laporte Avenue, Fort Collins, ...
Journal of Climate
Convective Variability associated with the Boreal Summer Intraseasonal Oscillation in the South China Sea Region --Manuscript Draft--
Manuscript Number:
JCLI-D-18-0091
Full Title:
Convective Variability associated with the Boreal Summer Intraseasonal Oscillation in the South China Sea Region
Article Type:
Article
Corresponding Author:
Weixin Xu Colorado State University Fort Collins, UNITED STATES
Corresponding Author's Institution:
Colorado State University
First Author:
Weixin Xu
Order of Authors:
Weixin Xu Steven Rutledge
Abstract:
This study investigates the convective cloud population, precipitation microphysics, and lightning activity associated with the Boreal Summer Intraseasonal Oscillation (BSISO) over the South China Sea (SCS) and surrounding landmasses. SCS rainfall shows a marked 30-60d intraseasonal variability. This variability is less evident over land. The population of mesoscale convective systems (MCSs) and the stratiform rain fraction over the SCS, Philippines, and Indochina increase remarkably after the onset of BSISO. Convection over the SCS during inactive periods exhibits a trimodal population including shallow cumulus, congestus and deep convection, mirroring the situation over tropical open oceans. The shallow mode is absent over land. Shallow cumulus clouds rapidly transition to congestus clouds over the SCS under active BSISO conditions. Over land, deep convection and lightning lead total rainfall and MCSs by 2-3 BSISO phases, while they are somewhat in phase over the SCS. Though convective instability over the SCS is larger during active periods compared to inactive periods, variability in convective intensity and precipitation microphysics is minimal, with active periods showing only higher frequency of moderate ice scattering and 30dBZ heights extending to -10 oC. Over the Philippines and Indochina, inactive phases exhibit substantially stronger ice scattering signatures, robust mixed-phase microphysics, and higher lightning flash rates, possibly due to greater convective instability and a stronger convective diurnal cycle. Total rainfall, convective environments, and convective structures over Borneo are all out of phase with that over Philippines and Indochina, while Southern China shows little BSISO variability on convective intensity and lightning frequency.
Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
1
Convective Variability associated with the Boreal Summer Intraseasonal Oscillation
2
in the South China Sea Region
3 4 5 6 7 8 9 10 11 12 13
Weixin Xu and Steven A. Rutledge
Department of Atmospheric Sciences, Colorado State University, Fort Collins, CO
14 15 16 17 18
Submitted to:
19
Journal of Climate
20 21
Submitted: February 26, 2018
22
Revised: June 7, 2018
23
Accepted: June 11, 2018
24 25 26 27
___________________________
28
Corresponding Author Address: Dr. Weixin Xu, Department of Atmospheric Sciences, Colorado
29
State University, 3915 West Laporte Avenue, Fort Collins, CO 80521.
30
ABSTRACT
31 32 33
This study investigates the convective cloud population, precipitation microphysics, and
34
lightning activity associated with the Boreal Summer Intraseasonal Oscillation (BSISO) over the
35
South China Sea (SCS) and surrounding landmasses. SCS rainfall shows a marked 30-60d
36
intraseasonal variability. This variability is less evident over land. The population of mesoscale
37
convective systems (MCSs) and the stratiform rain fraction over the SCS, Philippines, and
38
Indochina increase remarkably after the onset of BSISO. Convection over the SCS during
39
inactive periods exhibits a trimodal population including shallow cumulus, congestus and deep
40
convection, mirroring the situation over tropical open oceans. The shallow mode is absent over
41
land. Shallow cumulus clouds rapidly transition to congestus clouds over the SCS under active
42
BSISO conditions. Over land, deep convection and lightning lead total rainfall and MCSs by 2-3
43
BSISO phases, while they are somewhat in phase over the SCS. Though convective instability
44
over the SCS is larger during active periods compared to inactive periods, variability in
45
convective intensity and precipitation microphysics is minimal, with active periods showing only
46
higher frequency of moderate ice scattering and 30-dBZ heights extending to -10 oC. Over the
47
Philippines and Indochina, inactive phases exhibit substantially stronger ice scattering signatures,
48
robust mixed-phase microphysics, and higher lightning flash rates, possibly due to greater
49
convective instability and a stronger convective diurnal cycle. Total rainfall, convective
50
environments, and convective structures over Borneo are all out of phase with that over
51
Philippines and Indochina, while Southern China shows little BSISO variability on convective
52
intensity and lightning frequency.
53 54
2
55
1. Introduction
56 57
The Asian summer monsoon (ASM) is one of the strongest components of the global
58
monsoon system (Chang et al. 2011). The strong ASM circulation impacts a large region of the
59
tropics and subtropics in the eastern hemisphere. The ASM is thought to be driven by large-scale
60
thermal contrast between the Eurasian continent and Indo-Pacific Ocean (Yanai et al. 1992; Li
61
and Yanai 1996; Wu and Zhang 1998). The initial onset of the ASM occurs in the Bay of Bengal
62
(BOB) and the South China Sea (SCS) during early-to-mid May (Tao and Chen 1987; Wu et al.
63
1998). As the monsoon trough over the SCS deepens and extends northward, the ASM marches
64
northward and establishes itself over southern China and the northern SCS, marking the
65
beginning of the Meiyu/Baiu rainy season in East Asia (Wu et al. 1998; Chen 2004; Ding and
66
Chan 2005). After the ASM onset, active and break monsoon periods are largely determined by
67
the boreal summer intraseasonal oscillation (BSISO), a specific mode of the tropical
68
intraseasonal oscillation (ISO) prevailing in boreal summer (Wang and Xie 1997). While the ISO
69
in boreal winter is of course the Madden-Julian Oscillation (MJO) dominated by eastward
70
propagation of deep convection in the tropics, the BSISO exhibits an additional northward
71
propagating feature due to effects of monsoonal vertical wind shear (Wang and Xie 1997; Jiang
72
et al. 2004) and air-sea coupling (Kemball-Cook and Wang 2001, Fu et al. 2003).
73
The BSISO consists of two major modes, the 30-60 day mode and the 10-20 day
74
oscillation (Wang et al. 2005; Waliser 2006; Kikuchi et al. 2012; Chu et al. 2012; Lee et al.
75
2013). The dominant 30-60 day mode (similar time scales to the MJO) features prominent
76
eastward and northward/northeastward propagation, and is identified by a northwest-
77
southeastward tilted rain band between BOB and the SCS (Wang and Xie 1997). The 10-20 day
78
oscillation mainly shows northward/northwestward progression and a southwest-northeast
3
79
elongated frontal-like rain band (Lee et al. 2013). The BSISO has broad impacts on the ASM
80
including monsoon onset (Wang and Xie 1997; Kang et al. 1999), active and break monsoon
81
periods (Goswami 2005; Hoyos and Webster 2007; Ding and Wang 2009), seasonal mean
82
rainfall (Krishnamurthy and Shukla 2007, 2008), and precipitation extremes (Hsu et al. 2016;
83
Gao et al. 2016; Chen and Zhai 2017; Lee et al. 2017). Virts and Houze (2016) showed that
84
mesoscale organization, vertical cloud structure, and lightning activity within convection over
85
and around the BOB all vary significantly across the BSISO cycle (particularly the 30-60 day
86
mode). The BSISO also influences the midlatitude weather and climate by interacting with the
87
midlatitude circulation directly or indirectly (Ding and Wang 2005, 2007; Lee et al. 2011; Moon
88
et al. 2013). Unfortunately, our ability to simulate and predict the BSISO is severely limited
89
(Waliser et al. 2003; Kim and Kang 2008; Sabeerali et al. 2013), due to model misrepresentation
90
of processes key to the BSISO. Model improvement requires a more complete understanding that
91
can only be obtained by extensive observations of the BSISO, e.g., evolution and structure of
92
cloud and precipitation processes, vertical profiles of tropospheric moistening and heating,
93
surface fluxes, atmospheric boundary-layer processes, upper-ocean mixing and air-sea
94
interaction. Indeed these outstanding problems have motivated the Propagation of the Intra-
95
seasonal Tropical Oscillation (PISTON) field experiment to be held in late summer-early fall
96
2018 (http://onrpiston.colostate.edu).
97
The South China Sea Monsoon Experiment (SCSMEX) held in May-June 1998 collected
98
a rich dataset of radar, atmospheric sounding, and surface observations for studying convection
99
and convective environments associated with the SCS monsoon onset (Lau et al. 2000).
100
SCSMEX observations have been extensively used to investigate the organization and
101
propagation characteristics of convection, convective structures and precipitation microphysics,
4
102
regional variability of convection (southern versus northern SCS), convective heating profiles,
103
and the diurnal cycle (Johnson and Ciesielski 2002; Wang 2004; Johnson et al. 2005; Wang and
104
Carey 2005; Ciesielski and Johnson 2006; Aves and Johnson 2008). SCSMEX studies suggested
105
that SCS convection generally resembled other tropical oceanic systems in terms of convective
106
organization, e.g., shear-parallel convective lines (Wang 2004; Johnson et al. 2005), convective
107
intensity (Wang and Carey 2005), and ensemble precipitation microphysics (Wang and Carey
108
2005). However, some unique features of SCS convection were also identified. Two unique
109
convective organizational modes were found for the SCS convection possibly owing to its
110
interaction with subtropical frontal systems propagating from southern China (Johnson et al.
111
2005). The stratiform rain fraction within precipitation systems over the northern SCS (25%) was
112
substantially smaller compared to general tropical oceanic convection (40-50%), possibly due to
113
reduced atmospheric instability, low sea surface temperatures (SSTs), and relatively dry upper
114
tropospheric conditions during SCSMEX (Johnson et al. 2005; Wang and Carey 2005). While
115
SCSMEX focused on convection occurring before and during the onset of the SCS monsoon in
116
May-June, SCSMEX did not observe mid-to-late summer convection over the SCS nor consider
117
the BSISO cycles (or active-break monsoon cycles).
118
Many studies have showed contrasting convective intensities and structures between
119
active and break periods of various monsoons (e.g., Williams et al. 1992; Petersen and Rutledge
120
2001; Cifelli and Rutledge 1998; Xu and Zipser 2012), and between the low-level westerly and
121
easterly wind regimes in the Amazon (Petersen et al. 2002; Cifelli et al. 2002; Williams et al.
122
2002). Convection during monsoon breaks and easterly regime in the Amazon is more intense,
123
containing strong mixed-phase processes (owing to higher CAPE and stronger updrafts) resulting
124
in frequent lightning activity (Mohr and Zipser, 1996; Petersen and Rutledge 2001; Williams and
5
125
Stanfill 2002; Xu and Zipser 2012). Monsoon (or the Amazon westerly regime) convection
126
resembles maritime convection characterized by weak to moderate intensity, significant “warm-
127
rain” processes, and only weak mixed-phase precipitation growth (Petersen and Rutledge 2001;
128
Williams and Stanfill 2002; Xu and Zipser 2012). Lightning is markedly reduced during this
129
phase. These regime-based variations in convective intensity and mixed-phase microphysics are
130
attributed to thermodynamic variability (Williams et al. 1992; Rosenfeld and Lensky 1998;
131
Petersen and Rutledge 2001) or aerosol loading (Rosenfeld and Lensky 1998; Williams and
132
Stanfill 2002), or both processes working in concert (Williams and Stanfill 2002; Stolz et al.
133
2015). Over the SCS, convective intensity is stronger before the monsoon onset than after, as
134
indicated by larger flash rates (Yuan and Qie 2008). Ho et al. (2008) showed that inactive ISO
135
periods (easterly conditions) exhibited more intense radar structures compared to the active ISO
136
periods (westerly conditions). Furthermore they showed that convection over the SCS during the
137
inactive ISO was more land-based, i.e., peak convection was mainly initiated over land in the
138
afternoon and migrated offshore in the evening. Similarly, convection over the SCS was stronger
139
than open ocean convection (i.e., Philippine Sea) as indicated by more intense radar reflectivity
140
structures and microwave ice scattering signatures (Park et al. 2007). Convection over the SCS
141
might be best described by a combination of continental and oceanic regimes, possibly impacted
142
by aerosols advected from adjacent continents/islands or the migration of continental convective
143
systems from land (Takayabu 2006; Park et al. 2007; Ho et al. 2008). Nevertheless, the full
144
spectrum of the convective cloud population and microphysical structures associated with the
145
BSISO cycles over the SCS and surrounding landmasses warrants further study.
146
This study provides a large-scale climatological context of the convective cloud
147
populations and microphysical characteristics throughout the BSISO cycles over the SCS and
6
148
adjacent land areas. Sixteen years of measurements from the Tropical Rainfall Measuring
149
Mission (TRMM) are used to quantify convective structures as a function of BSISO phase (30-
150
60 day mode), specifically, convective organization, precipitation type, cold cloud top
151
characteristics, vertically integrated ice content, vertical precipitation structures, and lightning
152
flash rates. These TRMM-based convective parameters are also related to environmental
153
variables derived from reanalysis data to determine how major environmental factors may
154
modulate convective structures throughout the 30-60 day cycles. We also examine variability in
155
precipitating cloud populations and precipitation microphysics between regions along the active
156
BSISO rain band (i.e., the SCS, Philippines, Indochina Peninsula) and areas outside the rain band
157
(Borneo and Southern China).
158 159
2. Data and Methodology
160
Data used in this study include a multi-satellite rainfall product, large-scale reanalysis,
161
TRMM precipitation features (PFs), and TRMM Lightning Imaging Sensor (LIS) orbital
162
lightning observations during BSISO peak periods (June-September) from 1998 to 2013. These
163
data are stratified by BSISO phase defined by the BSISO index (Lee et al. 2013). This study
164
focuses on the key BSISO region over the SCS and surrounding landmasses [5S-30N; 100-
165
130E], including the Indochina peninsula (INDO), Philippines (PHIL), Borneo (BORN), and
166
Southern China (SCH) as marked in Fig. 1.
167
a. Large-scale rainfall maps and environmental variables
168
To study the large-scale rainfall patterns as a function of BSISO phase, we use the TRMM
169
Multi-satellite Precipitation Analysis surface rainfall product (version 7) called 3B42 (Huffman
170
et al. 2007). The 3B42 rainfall data have a 3 h temporal resolution and a 0.25o spatial resolution, 7
171
available from 50 S to 50 N during 1998-2014. This dataset mainly uses passive-microwave
172
measurements from low-earth orbit satellites, which were first calibrated by the TRMM PR.
173
This study examines a set of environmental variables including SST, near surface (2 m)
174
temperature, low-level (850 hPa) winds, mid-level (500 hPa) geopotential height, relative
175
humidity at these two levels, vertical wind shear (150-700 hPa), and convective available
176
potential energy (CAPE). SST data are taken from the TRMM Microwave Imager and Advanced
177
Microwave Scanning Radiometer for EOS (Gentemann et al. 2010). SST data are on daily time
178
scales, the same scale as the BSISO index data. The atmospheric environmental variables are
179
derived from the ancillary data of TRMM PF database (Liu et al. 2008). The ancillary data are
180
based on European Centre for Medium-Range Weather Forecasts re-analysis Interim (ERA-
181
Interim) reanalysis data (Dee et al. 2009), which have a horizontal spatial resolution of 1o by 1o,
182
temporal resolution of 6 h and mean vertical resolution of 25 hPa. Reanalysis data were
183
interpolated into locations of PFs using linear interpolation methodology (Liu et al. 2008).
184
Specifically, atmospheric variables including near surface air temperature at 2 m, convective
185
available potential energy (CAPE), relative humidity (RH), and deep vertical wind shear are
186
analyzed. CAPE is accumulated from the near-surface level (1000 hPa over ocean and 925 hPa
187
over land) up to the level of neutral buoyancy. RH is averaged over lower troposphere (850-700
188
hPa) and mid-troposphere (500-300 hPa) levels. Deep vertical wind shear is defined as the wind
189
shear between 150-700 hPa (V150 – V700). Deep shear was thought to be an important factor for
190
convective vigor and development of stratiform precipitation during the upscale growth of
191
convective systems (Saxen and Rutledge 2000).
192 193
8
194
b. TRMM observations
195
Two different levels of TRMM data are used in this study, specifically, the TRMM PF
196
database (Liu et al. 2008), and orbital lightning observations from TRMM LIS (both version 7).
197
LIS flash data (from all orbits) are used to calculate the lightning flash density over the entire
198
study region [5S-30N; 100-130E]. During each TRMM overpass, LIS detects total lightning
199
(both intracloud and cloud-to-ground lightning) events (radiance-based), employing a 5 km
200
horizontal resolution and a temporal resolution of 2 ms (Christian 1999). LIS lightning events are
201
grouped into flashes (a flash is identified as events within 6 km of one another and occurring
202
within a 330 ms interval). The flash detection efficiency of LIS has been established to be 70%–
203
90% (Boccippio et al. 2002; Christian et al. 2003).
204
The PF database utilizes various instruments on TRMM (Kummerow et al. 1998)
205
including the PR, LIS, TRMM Microwave Imager (TMI), and Visible and IR Scanner (VIRS).
206
These multiple sensor measurements were grouped into PFs as described by Liu et al. (2008).
207
PFs are first defined as clusters of PR raining pixels (5 km horizontal res.) at the near surface
208
level, with the centroid (based on elliptical fitting) of the cluster as the location of each PF. PR
209
vertical profiles (250 m vertical res.), as well as measurements of various resolutions from VIRS,
210
TMI, and LIS, are then collocated into the PR pixels of each PF. Therefore, a PF contains
211
quantities such as precipitation area, fraction of convective/stratiform rain, radar echo top height,
212
frequency of radar reflectivity values as a function of altitude, microwave brightness temperature
213
(Tb), lightning flash rate, etc. Horizontally and vertically polarized Tb at 85 GHz from TMI is
214
converted into the polarization corrected temperature (PCT). PCT at 85 GHz (PCT85) is a good
215
proxy for ice scattering (Spencer et al. 1989) or vertically integrated ice content (Vivekanandan
216
et al. 1991). Frequencies of precipitation (PR pixels) relative to PF size, stratiform rain fraction,
9
217
20-dBZ echo-top height, infrared (IR) Tb, and PCT85, as well as 30-dBZ occurrence in the
218
mixed-phase region and PDF of flash rates, are all derived from the PF database. Temperatures
219
at PR vertical range heights (above sea level) are calculated through the interpolation of
220
pressure-level temperatures and geopotential heights from ERA-Interim reanalysis data.
221 222
c. BSISO indices
223
A primary goal of this study is to examine convection as a function of BSISO phase in
224
the 30-60 day mode. For this purpose, we use the BSISO indices developed by Lee et al. (2013)
225
(hereafter,
226
(http://iprc.soest.hawaii.edu/users/jylee/bsiso). The Lee index is derived from the Multivariate
227
empirical orthogonal functions (MV-EOFs) of 850-hPa zonal wind and outgoing long-wave
228
radiation over the entire ASM region [10S-40N, 40-160E]. The 30-60 day mode is represented
229
by the first and second EOF principal components (PC). The Lee index better captures BSISO
230
features in the ASM region compared to previous BSISO indices (e.g., Waliser et al. 2004;
231
Annamalai and Sperber 2005; Kikuchi et al. 2012) and the MJO index (Wheeler and Hendon
232
2004). BSISO index is on a daily timeframe and has eight phases, indicating various states of the
233
ISO over the ASM region. In order to focus on major BSISO conditions, BSISO data are
234
inclusive to days with significant amplitude, i.e., (PC12 + PC22)1/2 > 1 during June-September.
235
The total number of days, TRMM overpasses, and PFs in specific BSISO phases is listed
236
in Table 1. Obviously, there are a large number of data samples in each BSISO phase (e.g., 100-
237
200 sample days with over 1000 TRMM overpasses in each phase). These statistics provide
238
confidence for the phase composites used in this study.
Lee
index),
which
is
available
239 240 241 10
from
1981
to
present
242 243 244
3. Overview of 30-60 Days Variability a. Large-scale rainfall and synoptic conditions
245
Figure 2 shows the composites of rainfall and 850-hPa winds as a function of BSISO
246
phase during June-September. Generally, precipitation and low-level winds are suppressed over
247
the SCS during phases 1-3 (Figs. 2a-c), whereas phases 5-7 feature bands of heavy rain and
248
accompanying strong westerlies over the BOB, SCS, and Philippine Sea (Figs. 2e-g). In contrast,
249
BORN and SCH, to the south and north of the SCS, show a nearly opposite rainfall pattern to the
250
SCS (also shown in Fig. 3). Specifically, phases 1-2 mark the most suppressed conditions of
251
precipitation (< 5 mm d-1) and winds (southwesterlies at 850 hPa < 5 m s-1) over the SCS. During
252
these same phases, heavy rainfall is located over the eastern Indian Ocean, or immediately west
253
of Sumatra, as well as western BORN (Figs. 2a-b). The BSISO convective envelope begins
254
propagating eastward and northward (to northern BOB) during phase 3 (Fig. 2c), and forms a
255
northwest-southeastward oriented rain band between the BOB and the tropical western Pacific by
256
phase 4 (Fig. 2d). A southwest-northeast oriented rain band accompanied by strong low-level
257
southwesterlies (> 10 m s-1) develops along the northern edge of an anticyclonic circulation (the
258
western Pacific subtropical high) between southern China and Japan (phases 3-4). Heavy
259
precipitation accompanied by strong low-level westerly winds (~15 m s-1) over the southern SCS
260
and western Pacific propagates further northward during phases 5-7, and produces peak rainfall
261
over the central and northern SCS areas. During phases 5-7, precipitation over BORN and SCH
262
(outside the BSISO rain band) is generally suppressed possibly owing to enhanced subsidence
263
adjacent to the intense rain band. The heavy BSISO rain band (phases 5-7) tends to be
264
interrupted by landmasses, i.e., the INDO and the islands of the PHIL. Figure 3 clearly shows
265
that both maximum rainfall and intraseasonal variability of precipitation are reduced over land
11
266
(INDO and PHIL) compared to over ocean (SCS). Previous studies suggested that the low
267
thermal inertial of the land surface could suppress intraseasonal surface fluxes that help to
268
promote the intraseasonal oscillation (Maloney and Sobel 2004; Sobel et al. 2010). Furthermore,
269
the strong diurnal cycle over land could release the atmospheric instability on the diurnal
270
timescale and compete with the intraseasonal timescale (BSISO) for instability (Neale and Slingo
271
2003; Sobel et al. 2010).
272
Analysis of the geopotential height at 500 hPa (Fig. 4) shows that precipitation-
273
suppressed conditions (phases 1-3) over the SCS is dominated by the western Pacific subtropical
274
high (Figs. 4a-c), whereas the active BSISO rain band (phases 5-7) over the BOB, SCS, and
275
tropical western Pacific is associated with a train of low-pressure circulations (Figs. 4e-g). The
276
subtropical high has maximum westward extent and is the strongest during phase 3, prior to
277
retreating eastward in phase 4. A low-pressure circulation initiates over the central SCS in phase
278
5, accompanied by strong low-level westerlies. The strongest westerlies during active phases (5-
279
7) are to the south of the low-pressure systems, along the zone of large pressure gradients. The
280
northward propagation of the active BSISO rainfall maxima (phase 5 to phase 7) over the SCS
281
and the western Pacific region corresponds closely to the northward movement of monsoon lows.
282
Previous studies suggested that land surface heat fluxes into the boundary layer could destabilize
283
the atmosphere ahead of the ascending zone and cause a northward shift of convective activity
284
(Webster, 1983; Srinivasan et al., 1993). Similarly, positive SST anomalies to the north of the
285
BSISO convective center could induce boundary layer convergence and promote northward
286
propagation of the intraseasonal disturbance (Back and Bretherton 2009; Hsu and Li 2012). Of
287
course, many other propagation mechanisms for BSISO have been proposed (summary in
288
DeMott et al. 2012).
12
289 290 291
b. Precipitation features and lightning activity
292
6, the quantitative PF population (density) as a function of BSISO phase is shown. In general,
293
phases 1-2 feature the most suppressed convective condition (minimum population of PFs by all
294
metrics) over the SCS (Figs. 5a-b and Fig. 6), while phases 5-6 are characterized by maximum
295
number of MCSs and cold clouds over the SCS (Figs. 5e-f and Figs. 6b-c). Behaviors of the PF
296
population, especially mesoscale convective systems (MCSs) with PF areas > 2000 km2 (Nesbitt
297
et al. 2000; Liu et al. 2008), are consistent with rainfall patterns (shown in Fig. 2) over both the
298
SCS and surrounding landmasses. During phases 1-3, PF frequency is relatively small and MCSs
299
are infrequent over the SCS, INDO, and the PHIL, marking suppressed conditions in these areas
300
(Figs. 5a-c). However, during these same periods MCSs occur more frequently over BORN and
301
the most southern portions of the SCS. Many of the MCS over the southern SCS west of BORN
302
may originate from convection over Borneo, given the pronounced nocturnal offshore
303
propagation from BORN (Ichikawa and Yasunari 2006). In the wake of active BSISO conditions
304
(phase 5, Fig. 5e), MCSs increase substantially over INDO, the SCS, and PHIL, and exhibit a
305
northward propagating pattern (phases 6-7, Figs. 5f-g). As will be shown in Sec. 4a, MCSs
306
contribute over 80% of the total rainfall during active periods, therefore, the propagation of these
307
MCSs constitutes the main BSISO rain band. During these phases (5-7), BORN and SCH
308
become convectively suppressed, marked by the minimum population of all types of PFs in these
309
regions (Figs. 5e-f and Figs. 6b-c).
Locations of PFs categorized by PF size in each BSISO phase are shown in Fig. 5. In Fig.
310
Populations of deep convection (min IR Tb < -60 oC, Fig. 6c) and intense convection (20-
311
dBZ echo-tops exceeding 12 km, Fig. 6d) reveal a different pattern compared to MCSs (Fig. 6b),
312
especially over land. The deepest convection over PHIL and INDO leads the frequent MCS
13
313
period by 1-2 phases, whereas deep convection and MCSs are more in phase over SCS and
314
regions outside the BSISO rain band (BORN and SCH). For example, deep/intense convection
315
frequently develops over PHIL and INDO during phases 2-3 (Figs. 6c-d), whereas MCSs are less
316
frequent during these same periods (Fig. 6b). Following onset of active periods associated with
317
BSISO over the SCS (phase 5), deep convective clouds increase over the SCS (Fig. 6c), but
318
decrease over land including INDO, PHIL, and BORN. However, intense convection (e.g., 20-
319
dBZ echo-top > 12 km) over SCS only increases slightly during active periods of the BSISO,
320
possibly because deep convection (producing cold IR Tb) over ocean have only moderate updraft
321
that is not able to lift precipitation size particles (20-dBZ) to high altitudes (12 km).
322
Figure 7 shows the distribution of lightning flash density (unconditional) in each BSISO
323
phase. Lightning activity over land is most robust during suppressed conditions (phases 2-3, figs.
324
7a-c), corresponding to the presence of intense convection (Figs. 6d). However, during inactive
325
BSISO periods, lightning activity is highly suppressed over the SCS, although there are indeed a
326
significant population of deep convective clouds (Figs. 6c-d). This is due to the fact that
327
lightning production is more related to the mixed-phase physics (e.g., 30-dBZ echo-top height)
328
rather than the cloud-top height (Zipser and Lutz 1994; Petersen et al. 1996, 1999). [Note: the
329
frequency of 30 dBZ above the freezing level is very low during the inactive periods over the
330
SCS as will be shown in Sec. 4c] The increase of lightning activity over the SCS during active
331
periods may be partially attributed to larger CAPE (will be shown in Sec. 3c), which may in fact
332
be enhanced by stronger heat and moisture fluxes from the ocean (discussed in Sec. 3c). The
333
increase in MCS organization over SCS during active periods may also promote lightning (Bang
334
and Zipser 2015; Bang and Zipser 2016). As shown previously, deep convective systems over
335
the SCS during inactive BSISO periods are relatively isolated, therefore, buoyant updrafts in the
14
336
convective cores of these systems are more likely to be diluted by ambient air (Bang and Zipser
337
2016). After the onset of the BSISO (phase 5), flash density increases over the SCS
338
corresponding to an increase in the MCS population. During active BSISO phases, flash density
339
over land (i.e., PHIL and INDO) decreases substantially, consistent with reduced CAPE (will be
340
shown in Sec. 3c).
341
c. Composites of “inactive” versus “active” BSISO periods
342
Figure 8 shows composites of rainfall, 850-hPa winds, and lightning density during
343
“inactive” (phases 1-3) and “active” periods (phases 5-7). Obviously, inactive periods are
344
characterized by minimum rainfall (Fig. 8a) and infrequent lightning activity (Fig. 8c) over the
345
SCS, while a remarkable west-eastward rain band spanning over INDO, SCS, and PHIL marks
346
the active phases (Fig. 8b). However, BORN and SCH show a rainfall pattern opposite to INDO,
347
SCS, and PHIL, as these locations are outside the BSISO rain band region.
348
Figure 9 shows mean values of several environmental variables during inactive and
349
active BSISO periods as a function of location (marked in Fig. 1). SST (SCS only) and surface
350
air temperature in the BSISO rain band regions (SCS, INDO, and PHIL) are reduced (by 0.5-1
351
o
352
precipitation cooling, and stronger upper-ocean mixing owing to enhanced low level winds (Fig.
353
9a). As a result, the inactive BSISO induces larger CAPE relative to the active phases,
354
particularly for the INDO and PHIL regions (Fig. 9b), likely due to stronger surface heating over
355
land. It is intriguing that SCS shows higher CAPE during active periods even though SST/air
356
temperature is lower compared to inactive periods. A possible mechanism is that higher sea
357
surface fluxes (due to stronger surface winds) and stronger low-level moisture convergence
358
during active BSISO phases (as in active MJO phases, Johnson and Ciesielski 2013) lead to
C) during active periods compared to inactive phases, likely owing to cloud shading,
15
359
larger CAPE during these periods. Petersen et al. (1996) found that CAPE was highly correlated
360
to the mixing ratio (wm) of the lowest 50 hPa (i.e., CAPE = 747.7*wm – b) over the tropical
361
western Pacific Ocean, suggesting that CAPE over warm oceans is largely moisture driven. Over
362
SCS, mixing ratio in the lowest 50 hPa is 17.6 g kg-1 during inactive periods and 18.2 g kg-1
363
during active periods. Using the relationship reported by Petersen et al. (1996), the derived
364
CAPE differences are ~400 J kg-1 between active and inactive periods, consistent to the
365
difference shown in Fig. 9b.
366
Relative humidity in the lower-to-mid troposphere and vertical wind shear over SCS,
367
INDO, and PHIL are all enhanced during active periods (Figs. 9c-d) with influences from rain
368
bands, low-pressure circulations, and strong low-level westerly winds. Convective environments
369
over BORN also vary significantly between inactive and active periods, showing patterns
370
opposite to INDO and PHIL, e.g., higher air temperature and larger CAPE during active BSISO
371
periods. Note that active BSISO periods correspond to dry periods over BORN (and SCH), vice
372
versa. So it is not surprising the CAPE over BORN is high during active periods as there is little
373
convection to remove the CAPE. Deep shear over BORN during active phases becomes
374
exceptionally large (25 m s-1), as the South-Asia upper-level anticyclonic circulation is enhanced
375
by active BSISO deep convection leading to strengthened upper-level easterlies/northeasterlies
376
over BORN. In contrast, SCH shows the minimum BSISO-related variability in most of the
377
environmental variables. Compared to other regions, SCH shows lower rainfall maxima and less
378
variability in total precipitation between active and inactive periods (Fig. 3 and Figs. 8a-b),
379
therefore minimal variability in cloud shading, precipitation cooling, and condensational heating
380
across the BSISO cycle.
381
16
382
4. BSISO Variations on Convective and Microphysical Structures
383 384
While Sec. 3 provided a large-scale overview of the variability in synoptic conditions,
385
convective environments, rainfall, precipitation features, and lightning activity over the BSISO
386
cycle, we now discuss the detailed convective and microphysical structures during inactive and
387
active BSISO periods. Specifically, we compare convective organization, radar echo-top heights,
388
IR Tb frequency, vertically integrated ice content, mixed-phase microphysical properties, and
389
lightning flash rates of convective clouds over various regions impacted by the BSISO including
390
SCS, INDO, PHIL, BORN, and SCH.
391
a. Convective organization
392
Convective organization is examined in terms of the size and convective/stratiform
393
proportions of precipitation systems. Figure 10 shows the frequency of precipitation occurrence
394
(PR raining pixels) as a function of precipitation area and areal fraction of stratiform
395
precipitation (FSP) of PFs. Generally, precipitation systems over SCS, PHIL, and INDO are
396
larger and consist of larger FSP during active phases compared to inactive periods, possibly due
397
to greater shear and more moist mid-tropospheric environments during active periods (Fig. 9).
398
Isolated systems contribute 20-30% of precipitation frequency (similar contribution from mostly
399
convective systems with FSP < 20%) over SCS, PHIL, and INDO during inactive periods, which
400
is twice that during active phases. This is consistent with higher CAPE, reduced deep shear, and
401
drier mid-tropospheric conditions over SCS, PHIL, and INDO during inactive phases (Fig. 9). In
402
these same regions, large systems with broad stratiform precipitation (e.g., PF size > 10,000 km2
403
and FSP > 50%) are responsible for nearly 40% of the precipitation frequency during active
404
phases, yet comprise only 15-25% of the precipitation frequency during inactive periods. The
405
precipitation contribution from large MCSs during active BSISO phases is similar to what has 17
406
been observed by shipborne radar during active MJO periods over the Indian Ocean (Xu and
407
Rutledge 2014) and the western Pacific warm pool (Rickenbach and Rutledge 1998). BORN and
408
SCH exhibit an opposite trend regarding precipitation area and FSP to the SCS, as the wet and
409
dry periods in these regions correspond to inactive and active BSISO phases over the SCS,
410
respectively.
411
b. Convective echo-top heights
412
Figure 11 shows the probability distribution function (PDF) of 20-dBZ echo-top heights
413
(ETH20) of convective cells (within rain areas identified as convective). Generally, ETH20
414
maximizes near the freezing level (5 km), corresponding to the “congestus mode” over the
415
tropical oceans (Johnson et al. 1999). The congestus mode is caused by the near 0o C stable
416
layer, owing to cooling and moistening effects associated with extensive regions of melting
417
within MCSs. SCS exhibits a broad peak (2-5 km) of ETH20 during inactive BSISO periods
418
(Fig. 11a). The lower-level (2-3 km) echo-top maximum represents the “shallow cumulus”
419
mode, as reported over the Pacific warm pool/western Atlantic (Johnson et al. 1999), Manus Is.
420
(Hollars et al. 2004), and the central Indian Ocean (Powell et al. 2015). It is not surprising that
421
this shallow cumulus mode disappears over land (Figs. 11b-e), given the strong surface heating
422
and greater instability over land. The shallow cumulus mode is better defined over more open
423
oceans during inactive periods, i.e., western tropical Pacific (WPC, Fig. 11f). An examination of
424
stability profiles shows that a low-level stable layer (850-750 hPa) associated with trade
425
inversion apparently exists over both SCS and WPC (Fig. 12). Of course, the mid-level stable
426
layer is also evident over these regions. The mid-level stable layer over the SCS is stronger
427
during active periods (Fig. 12a), due to the substantial increase of MCS population in active
428
BSISO thus enhanced cooling and subsidence within broad stratiform regions.
18
429
During active BSISO periods, the frequency of shallow cumulus over the SCS (e.g.,
430
MAXHT20 < 3 km) decreases (Fig. 11a), as the large-scale BSISO disturbance likely disrupts
431
the trade stable layer near 2 km (reduction of stability as shown in Fig. 12a), promoting the
432
growth of cumulus into congestus. The modulation of the shallow cumulus cloud population due
433
to active BSISO is even more evident over WPC (Fig. 11f). During inactive periods, convective
434
echo tops over WPC show a clear bimodal structure with maxima at 2-3 km and 5 km. After the
435
onset of BSISO, shallow cumulus decreases substantially and congestus clouds (with tops at 4-6
436
km) increase correspondingly. Hence the shallow cumulus mode virtually disappears.
437
There is little variability of shallow cumulus and congestus echo tops over land
438
between inactive and active periods (Figs. 11b-e). Over land, especially PHIL and INDO, the
439
frequency of deep convection (e.g., ETH20 > 8 km) is higher during inactive periods compared
440
to active periods. Nevertheless, ETH20 does not show a deep convective mode (cloud tops > 12
441
km, Johnson et al. 1999), possibly because ETH20 does not represent actual cloud tops (which
442
may be closer to 0 dBZ as in Johnson et al. 1999). Xu and Rutledge (2016) compared convective
443
top heights from 20 dBZ, 0 dBZ, and IR Tb using shipborne radar data and satellite imagery
444
collected during DYNAMO. They found that the discrepancies between these heights increase as
445
a function of altitude, e.g., these proxies are near in height for shallow to medium convection (
1.
826 827 P1
P2
P3
P4
P5
P6
P7
P8
Days
152
192
157
112
182
136
139
165
Orbits
1416
1710
1446
1035
1680
1259
1212
1499
PFs
46535
59106
52602
37225
50184
36267
33485
45914
828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851
36
852 853 854 855 856
Figures
857 858 859
Fig. 1. Geography map of elevation (color shaded) overlaid with low-level winds at 925 hPa
860
(arrows) from June to September in the South China Sea and Indo-Pacific regions. Study regions
861
are marked by boxes, including South China Sea (SCS), Philippines (PHIL), Indochina (INDO),
862
Borneo (BORN), Southern China (SCH), and Western Pacific (WPC).
863 864 865 866 867 868 869 870 871 872 873 874
37
875 876
Fig. 2. Mean rainfall (mm d-1) from TRMM 3B42 (color shaded) overlaid with winds at 850 hPa
877
(arrows) derived from reanalyses as a function of BSISO phase during June-September 1998-
878
2013.
879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 38
897 898 899
900 901 902
Fig. 3. Mean rainfall (mm d-1) time series for five study regions derived from TRMM 3B42 as a
903
function of BSISO phase.
904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 39
922 923
924 925
Fig. 4. Geopotential heights at 500 hPa (contours) and winds at 850 hPa (arrows) as a function of
926
BSISO phase. Colorbar represents geopotential height.
927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944
40
945 946
947 948 949
Fig. 5. Distribution of Precipitation Features (PFs) categorized by PF area (color) as a function
950
of BSISO (30-60 days oscillation) phase during June-September, 1998-2013. Small PFs with
951
area < 100 km2 are not displayed.
952 953 954 955 956 957 958 959 960 961 962 963 964 965 966
41
967
968 969 970
Fig. 6. PF population as a function of BSISO phase: (a) all PFs, (b) PFs with area > 2000 km2,
971
(c) PFs with minimum IR Tb < -60 oC, and (d) PFs with 20dBZ echo-top height > 12 km. PF
972
numbers are normalized by area of study region and number of days in each BSISO phase.
973 974 975 976 977 978 979 980 981 982 983 984
42
985 986
987 988 989
Fig. 7. Distribution of lightning flash density (fl km-2 yr-1) derived from orbital TRMM LIS data
990
as a function of BSISO phase. Flash density is unconditional, including both raining and not
991
raining times sampled by LIS.
992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007
43
1008 1009 1010
Fig. 8. Mean rainfall overlaid with 850 hPa winds and lightning flash density during inactive
1011
(phases 1-3) and active (phases 5-7) BSISO periods: (a) rainfall during inactive, (b) rainfall
1012
during active, (c) lightning during inactive, and (d) lightning during active. White boxes in
1013
panels (a) and (b) mark the selected study regions.
1014 1015 1016 1017
44
1018 1019 1020
Fig. 9. Environmental variables during inactive (blue) and active (red) BSISO periods: (a) SST
1021
and 2-m air temperature, (b) convective available potential energy (CAPE), (c) relative humidity
1022
(RH) at lower-levels (850-700 hPa) and mid-to-upper levels (500-300 hPa), and (d) vertical wind
1023
shear between 700 and 150 hPa. Error bars represent 0.5 standard deviation.
1024 1025 1026 1027 1028 1029 1030 45
1031
1032 1033 1034
Fig. 10. Joint PDF of TRMM PR raining pixels as a function of PF area and PF stratiform rain
1035
fraction during (a)-(e), inactive BSISO, and (f)-(j), active BSISO periods. Number within each
1036
box represents the percentage of occurrence in the specific value ranges of PF area and stratiform
1037
rain fraction.
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054
46
1055 1056 1057
Fig. 11. PDF of 20-dBZ echo-top height over convective area of precipitation systems observed
1058
by TRMM PR during inactive (blue bars) and active (red bars) BSISO periods. Frequency is
1059
relative to convective rain pixels near the surface.
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073
47
1074 1075 1076 1077
1078 1079 1080
Fig. 12. Cumulative frequency (%) of stability (dT/dZ) greater than -5 oC km-1 over ocean for (a)
1081
SCS and (b) WPC. The frequency is calculated in every 50 hPa bin.
1082 1083 1084 1085 1086
48
1087 1088 1089
Fig. 13. Frequency of infrared brightness temperature during inactive (blue bars) BSISO and
1090
active (red bars) BSISO periods observed by TRMM VIRS during 1998-2013.
1091 1092 1093 1094 1095 1096 1097 1098
49
1099 1100
1101 1102 1103
Fig. 14. Frequency of TRMM TMI microwave polarization-corrected brightness temperature at
1104
85 GHz during inactive (blue bars) BSISO and active (red bars) BSISO periods. Frequency is
1105
relative to TMI raining pixels near the surface.
1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 50
1120 1121
1122 1123 1124
Fig. 15. Occurrence frequency of PR 30-dBZ radar reflectivity above the freezing altitude during
1125
inactive (blue bars) BSISO and active (red bars) BSISO periods. Frequency is relative to PR
1126
convective rain pixels at 2 km.
1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 51
1143 1144
1145 1146 1147
Fig. 16. Accumulated frequency of precipitation features as a function of PF’s lightning flash
1148
rates during inactive (blue bars) BSISO and active (red bars) BSISO periods. All PFs are
1149
included for the frequency calculation.
1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 52
1164 1165 1166
Fig. 17. Conceptual model (left panels) of convective structures and microphysical properties
1167
across the BSISO rainband region (INDO, SCS, and PHIL) during suppressed conditions and
1168
active BSISO periods, and corresponding diurnal variations (right panels) on 30-dBZ occurrence
1169
frequency above 6 km in convective precipitation areas.
1170 1171 1172 1173 1174
53