On 2 October 2005, a record-breaking rainfall event with 152 mm of rainfall in an hour occurred as Typhoon Longwang approached Fujian Province, China. The severe rainfall was unexpected and significantly underpredicted by the local weather forecasters and caused a total of 96 deaths. Because of the severe damage over Taiwan and mainland China, the name of Longwang, which means a dragon in charge of rainfall in Chinese, was removed from the name list for future typhoons.

EVOLUTION AND BASIC FEATURES OF THE RAINBAND.

The formation and evolution of the rainband associated with the rainfall event was captured by the radar mosaic produced by the Central Weather Bureau (CWB) of Taiwan (Fig. 1). Longwang had frequent rainband activity as it made landfall and passed over Taiwan Island (Yu and Tsai 2010, 2013). Rainbands during this period were more transient and had the features of squall-line dynamics (Yu and Tsai 2013). As Typhoon Longwang approached the coast of Fujian Province at 0800 UTC, one type of this transient rainband in the northeast sector started to weaken and dissipate (Fig. 1b). At the same time, the eyewall underwent an asymmetry transformation accompanied by a bended convection pattern in the north (Fig. 1b). The bended convection transformed into a strong convective band along the eyewall to the north and moved outward relative to the storm center (Fig. 1c). The convective band continued to intensify with a sharp inner edge (Fig. 1d). An hour later, the convective band achieved its maximum intensity with a large area of stratiform precipitation outward and downstream (Fig. 1e). At this time, cloud brightness temperatures as low as −80°C were measured by a Geostationary Operational Environmental Satellite (GOES; see Fig. ES1 in the online supplement to this article: https://doi.org/10.1175/BAMS-D-17-0122.2) and hourly precipitation of 152 mm was measured at Changle under this band. As the convective band detached from the original eyewall and propagated outward relative to the storm center, a new weak eyewall started to form and gradually intensified as the storm center moved over the Taiwan Strait (Figs. 1e–h). Before Longwang made landfall near Quanzhou around 1500 UTC (Fig. 1i), the rainband persisted with strong intensity. Finally, after the second landfall of Longwang over mainland China, the rainband slowly weakened (Figs. 1j–o), but still had a striking line form with large reflectivities until it dissipated completely after 2300 UTC 2 October. The rainband that produced the heavy rainfall lasted for at least 13 h. The triggering and maintenance of this long-duration strong rainband are worthy of a detailed investigation.

Fig. 1.

Doppler radar reflectivity mosaic images at 1-h interval (unit: dBZ) from the CWB, from 0700 to 2100 UTC 2 Oct 2005. The rainband of interest is denoted using a purple arrow.

Fig. 1.

Doppler radar reflectivity mosaic images at 1-h interval (unit: dBZ) from the CWB, from 0700 to 2100 UTC 2 Oct 2005. The rainband of interest is denoted using a purple arrow.

The Doppler radar at Changle captured a more detailed view of the evolution of the rainband (Fig. 2). The sharp inner edge of the band, which was sustained and fueled by the strong low-level convergence, was distinct during the whole lifespan of the rainband. This can be inferred from the radial wind pattern recorded by the radar (Figs. 2e–h). There is a sharp gradient of cross-band winds along the inner edge of the band with low-level outward flow inside of the band and inward flow outside of the band, especially toward the later times. This is similar to the secondary circulation in a hurricane, but converged along the band instead of the eyewall (Figs. 2e,f). The strong convergence was responsible for the sustained vertical motion extending to the upper troposphere, as indicated by reflectivities of up to 40 dBZ at an altitude of ∼9 km MSL as measured by the radar at Xiamen (Fig. ES2). Another feature to note is the very sharp tip or tail of the band (Figs. 2a–d), which was preferentially located near the coast during the whole life duration of the band. The inner edge was very sharp along a narrow zone with an arc-shaped radar echo over 50 dBZ (Figs. 2a,b). This feature is in contrast to that of typical squall lines, which generally have a broad area of stratiform precipitation located behind the leading-edge convection (Yu and Tsai 2013; Meng and Zhang 2012). Instead, some typical inner rainband-like characteristics are evident. For example, the band propagated outward (Willoughby et al. 1984), with an along-band jet (>30 m s–1; Figs. 2e,f) on the radially outward side of the convective cells (Houze 2010). The band continued to move outward and approached Changle radar station at 1139 UTC (Fig. 2b). Although the Doppler radar at Changle was shut down for protection from 1200 to 1400 UTC, it still captured the evolution and fine structures of the rainband. Reflectivities up to 50 dBZ were measured along a ∼200-km-long band extending from the ocean toward the inland area of Fujian Province. No hail was recorded at the surface during the passage of the band. Thus, reflectivities of 50 dBZ implied a very large number concentration of raindrops and/or large sizes of raindrops.

Fig. 2.

The 1.45°-elevation plan position indicator (PPI) scans of the (a)–(d) reflectivities (dBZ) and (e)–(h) dealiased radial velocities (m s–1) from the Changle Doppler radar at four representative times. The four selected times approximately correspond to the passage of the rainband over Putian (PT), Changle (CL), Luoyuan (LY), and Ningde (ND), as denoted using the black dot. The black star marks the position of the radar. The geometric definitions of the Va and Vc wind components are shown in (c), with arrows denoting the positive direction.

Fig. 2.

The 1.45°-elevation plan position indicator (PPI) scans of the (a)–(d) reflectivities (dBZ) and (e)–(h) dealiased radial velocities (m s–1) from the Changle Doppler radar at four representative times. The four selected times approximately correspond to the passage of the rainband over Putian (PT), Changle (CL), Luoyuan (LY), and Ningde (ND), as denoted using the black dot. The black star marks the position of the radar. The geometric definitions of the Va and Vc wind components are shown in (c), with arrows denoting the positive direction.

Unprecedented heavy rainfall due to this band occurred over the coastal area of Fujian Province before and during the landfall of Typhoon Longwang. More than 10 stations recorded hourly precipitation greater than 40 mm in a short period during 2–3 October (Fig. 6a), with the maximum precipitation of 152 mm h–1 (minute rainfall up to 5.5 mm) recorded at Changle between 1100 and 1200 UTC 2 October (Fig. 3). The 3-h accumulated rainfall is over 300 mm over a wide area swept by the rainband (not shown).

Fig. 3.

Time series of minute precipitation rate (PR), temperature (T), surface pressure (P), and Va and Vc wind components at 1-min temporal resolution at the six coastal sites (as shown in Fig. 6a) from 0800 to 1800 UTC 2 Oct 2005. The translucent rectangles denote the arrival of the rainband based on the start of precipitation associated with the rainband at each station, showing the propagation of the rainband.

Fig. 3.

Time series of minute precipitation rate (PR), temperature (T), surface pressure (P), and Va and Vc wind components at 1-min temporal resolution at the six coastal sites (as shown in Fig. 6a) from 0800 to 1800 UTC 2 Oct 2005. The translucent rectangles denote the arrival of the rainband based on the start of precipitation associated with the rainband at each station, showing the propagation of the rainband.

The temporal evolution of rainfall at a few representative stations (Fig. 3) clearly showed the propagation of rainfall associated with the movement of the rainband. At each station, heavy rainfall lasted for 1–2 h. The maximum precipitation rates at the six representative stations were 77, 152, 111, 84, 108, and 43 mm h–1, and occurred at 0900–1000, 1100–1300, 1200–1400, 1200–1400, 1500–1700, and 1600–1800 UTC, respectively (Fig. 3). Associated with the passage of the rainband over these stations, the surface pressure experienced a ∼3-hPa increase accompanied by a temperature drop of ∼3 K (Fig. 3). Another prominent feature is the abrupt change of winds following the passage of the rainband. We decomposed the 10-m winds into their along- (Va) and cross-band (Vc) components. There is a distinct peak of Va with a sharp reduction of Vc associated with the passage of the rainband (Fig. 3). For example, Va increased from 0 to ∼12 m s–1 with the Vc being reduced from 15 m s–1 to near 0 when the band passed over Changle (Fig. 3). There was also a clear reversal of Vc before and after the passage of the band with positive Vc (inward flow) before the passage and negative Vc (outward flow) after the passage as the rainband propagated farther away from the center of the storm (e.g., Luoyuan in Fig. 3). This implied that the band was actually induced and accompanied by strong low-level cross-band convergence.

POSSIBLE CAUSES FOR THE RAINBAND.

What is the cause of such a long-lasting and strong rainband? Radiosonde measurements at Fuzhou captured the thermodynamic and kinematic environmental characteristics during the evolution of the rainband (Fig. 4). At 0000 UTC before the passage of the band, the CAPE was only 131 J kg–1. At 1200 UTC around the time of the passage of the rainband over Fuzhou, the CAPE increased to about 561 J kg–1 with the level of free convection at about 1,450 m. Although convection and precipitation might have reduced the CAPE at 1200 UTC, it is very difficult to generate and sustain such strong convection and updrafts solely by the release of CAPE. This is part of the reason why we think there are other types of forcing, possibly wave forcing and frictional contrast across the coastline as well, rather than thermodynamic instability alone, for this band.

Fig. 4.

Skew T–logp diagram at Fuzhou at (a) 0000 and (b) 1200 UTC 2 Oct 2015.

Fig. 4.

Skew T–logp diagram at Fuzhou at (a) 0000 and (b) 1200 UTC 2 Oct 2015.

Similar to Yu and Tsai (2013), the possibility of cold pool dynamics was examined first. There was an obvious cold pool around 3 K associated with the passage of the rainband (Fig. 3). Based on the strength of the cold pool, we estimated the moving speed of the cold pool approximately as

 
formula

where g is the gravitational acceleration, H is the cold pool depth, θυ0 denotes initial the environmental virtual potential temperature, and ∆θυ is the virtual potential temperature deficit influenced by the cold pool. For the present case, ∆θυ is about 3.1 K and θυ0 is estimated to be about 305 K. Assuming the cold pool depths to be 1–3 km, this gave a speed of approximately 10–17 m s–1, which is larger than the cross-band wind shear of 6–8 m s–1 estimated over the same depths, as measured at Fuzhou (Fig. 4b). Clearly, the cold-pool-induced vorticity is larger than the cross-band wind shear and thus is in the so-called suboptimal state (Rotunno et al. 1988), which implied that the system tended to slope rearward instead of having an upright updraft. In addition, the weak shear does not favor the regeneration of strong cells along the outflow boundary and thus the maintenance of squall lines (Rotunno et al. 1988; Weisman and Rotunno 2004). However, this robust rainband lasted for more than 13 h with strong upright convection located in the inner edge of the rainband, in contrast to most relatively short-lived squall lines having the leading convective line (e.g., Meng and Zhang 2012; Yu and Tsai 2013). The small CAPE and suboptimal state implied that squall-line dynamics alone probably could not maintain such a long-lived rainband.

Regarding the possible wave dynamics for this rainband, we noted that the wave dynamics in this band is different from that in Yu and Tsai (2010, 2013). For example, the temperature and pressure perturbations were 90° different across the band studied in Yu and Tsai (2010, cf. their Fig. 12), which is consistent with the basic features of gravity waves. In contrast, the temperature and pressure perturbations were approximately out of phase (180° difference) at several stations we studied that were passed by the rainband (Fig. 5). This feature is consistent with Rossby-type wave characteristics. Note that caution should be paid since convection itself can significantly influence temperature and pressure perturbations.

Fig. 5.

Time series of pressure (red line) and temperature (green line) perturbations over a 2-h time period covering the passage of the rainband at the six stations labeled in Fig. 6a. A 2-h running average was used to compute the perturbations. The yellow line denoted the time of the maximum precipitation at each station.

Fig. 5.

Time series of pressure (red line) and temperature (green line) perturbations over a 2-h time period covering the passage of the rainband at the six stations labeled in Fig. 6a. A 2-h running average was used to compute the perturbations. The yellow line denoted the time of the maximum precipitation at each station.

The vortex Rossby wave (VRW) has been proposed to explain the formation of typhoon rainbands (Montgomery and Kallenbach 1997). To determine the propagation of the rainband, the isochrones of the area of strong convection with radar reflectivity greater than 45 dBZ in the band were tracked from 0900 to 1756 UTC (Fig. 6a). Surprisingly, the radial speed of the rainband relative to the typhoon center is rather steady at about 5.78 m s–1 (Fig. 6b). Relatively small radial propagation speeds of VRWs have been found in both numerical modeling [e.g., 4–5 m s–1 for wavenumber-1 VRWs in Wang (2002)] and observational studies [e.g., 5.2 m s–1 in Corbosiero et al. (2006)]. The small radial propagation speed of this band is thus consistent with VRWs rather than the high-frequency gravity waves [e.g., greater than 15 m s–1 in Diercks and Anthes (1976)]. However, the azimuthal propagation speed determined based on the direct tracking of the strong convective sector is rather steady with a speed of about 4.67 m s–1 anticyclonically (Fig. 6b), which is inconsistent with the cyclonically propagating speed noted for VRWs (e.g., Wang 2002; Corbosiero et al. 2006). A careful wave decomposition analysis following Corbosiero et al. (2006) using the CWB radar images indicated that the convective segment used for the previous tracking is actually only a part of a longer rainband, more specifically, the tail of a long spiral rainband (Fig. 7). The evolution of the wavenumber-2 component corresponded well with that of the major body of the northwestern rainband (cf. Fig. 1). Note that wavenumber-2 analysis is fairly robust and does not strongly depend on the position of the storm center (Fig. ES3; Reasor et al. 2000; Corbosiero et al. 2006). The major body of the rainband identified by the wave analysis actually propagated cyclonically with an azimuthal phase speed of about 12 m s–1 (Fig. 7g). Note that the ambient 850- and 700-hPa tangential wind speeds near this location of the rainband are about 16.5 and 27.7 m s–1, respectively, based on the sounding from Fuzhou at 1200 UTC, implying that this wave propagated against the tangential flow of the typhoon. The azimuthal propagation against the tangential flow together with the slow outward radial propagation from the typhoon center is consistent with the characteristics of VRWs documented in previous studies (Wang 2002; Corbosiero et al. 2006). For example, the mean azimuthal speed was 68% of the local tangential mean flow (Corbosiero et al. 2006). Similar orientation between the rainband and cyclonic tangential flow also implies that the rainband was strongly affected by the dynamics of the typhoon. Note that although the majority of the band propagated cyclonically, the convective sector of interest was actually moving anticyclonically relative to the storm center and was always located near the coast (Fig. 1). One possible reason for the strong convection near the coast is mainly due to the strong low-level moisture convergence enhanced by surface friction over land, as we can see in the European Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis (ERA-Interim; not shown). In contrast, the rainband is much weaker in the western part of the storm, likely because of unfavorable moisture supply although the wave signal is still present.

Fig. 6.

(a) The isochrones of the rainband and the track of the typhoon with a mean time interval of 1 h (represented by different colors), and the position of the typhoon center at 1-h intervals was interpolated from the International Best Track Archive for Climate Stewardship (IBTrACS) dataset. The rainband is portrayed as the connection of the average position of the points on each longitude with a reflectivity greater than 45 dBZ from the 1.45°-elevation PPI scans of the Changle Doppler radar. The red dots are those surface stations that measured an instantaneous precipitation rate greater than 40 mm h–1, which are marked from 1 to 6 for PT, CL, Fuzhou (FZ), Minhou (MH), LY, and ND, respectively. (b) The solid and dashed lines denote the temporal variations of the radial and tangential distances (calculated as aR, in which a is the slant angle of the rainband relative to the storm center as shown in Fig. 2c and R is the average radial distance of the rainband to the center of the storm) of the rainband to the typhoon center, respectively.

Fig. 6.

(a) The isochrones of the rainband and the track of the typhoon with a mean time interval of 1 h (represented by different colors), and the position of the typhoon center at 1-h intervals was interpolated from the International Best Track Archive for Climate Stewardship (IBTrACS) dataset. The rainband is portrayed as the connection of the average position of the points on each longitude with a reflectivity greater than 45 dBZ from the 1.45°-elevation PPI scans of the Changle Doppler radar. The red dots are those surface stations that measured an instantaneous precipitation rate greater than 40 mm h–1, which are marked from 1 to 6 for PT, CL, Fuzhou (FZ), Minhou (MH), LY, and ND, respectively. (b) The solid and dashed lines denote the temporal variations of the radial and tangential distances (calculated as aR, in which a is the slant angle of the rainband relative to the storm center as shown in Fig. 2c and R is the average radial distance of the rainband to the center of the storm) of the rainband to the typhoon center, respectively.

Fig. 7.

(a)–(f) Wavenumber-2 asymmetry of radar reflectivity of Longwang from CWB from 0900 to 1400 UTC 2 Oct 2005, with the purple dashed line showing the +0.5-dBZ perturbation overlaid with green contours showing the rainband of interest with radar reflectivities greater than 30 dBZ. Black circles are the 50-, 150-, and 250-km radii from the typhoon center. (g) Azimuth–time Hovmöller diagram of the wavenumber-2 asymmetry at a radius of 170 km denoted by the black dashed circles in (a)–(f).

Fig. 7.

(a)–(f) Wavenumber-2 asymmetry of radar reflectivity of Longwang from CWB from 0900 to 1400 UTC 2 Oct 2005, with the purple dashed line showing the +0.5-dBZ perturbation overlaid with green contours showing the rainband of interest with radar reflectivities greater than 30 dBZ. Black circles are the 50-, 150-, and 250-km radii from the typhoon center. (g) Azimuth–time Hovmöller diagram of the wavenumber-2 asymmetry at a radius of 170 km denoted by the black dashed circles in (a)–(f).

In addition, we have performed a Weather Research and Forecasting (WRF) Model simulation of the storm with 2-km grid spacing and found a similar rainband in the simulation (Fig. 8). Similar to Chen and Yau (2001), we found that the model reflectivity corresponded well with the wavenumber-2 potential vorticity and vertical motion (Fig. 8). Of course, the wave mechanism noted in the simulation may not be the same as that in the observed rainband. Nevertheless, the results indicate the possible contributions by VRWs to the triggering and maintenance of the rainband. Surprisingly, the strong convection tends to move with the main body of the wave farther inland in another WRF simulation with the terrain height reduced by half (Fig. ES4) instead of remaining stationary near the coast in the control simulation. This indicated that the strong convection is probably triggered and maintained by the wave dynamics, but reinforced by the differential friction along the coast. Overall, it is suggested that the interplay between VRWs and the coastal convergence is the most likely reason for this long-lasting rainband. More details about the simulation results will be reported upon in a separate study.

Fig. 8.

(a)–(c) Model 850-hPa reflectivity (dBZ, color shading) overlaid with 850-hPa wavenumber-2 asymmetry of potential vorticity (solid line for 0.1 PVU and dashed line for 1 PVU, where 1 PVU = 10−6 K kg–1 m2 s–1) at 0700, 0800, and 0900 UTC 2 Oct 2005. (d)–(f) As in (a)–(c), but overlaid with the 850-hPa wavenumber-2 vertical velocity (solid line, 0.05 m s–1; dashed line, 0.2 m s–1). Purple circles are the 50-, 100-, 150-, and 200-km radii from the typhoon center marked using the typhoon symbol.

Fig. 8.

(a)–(c) Model 850-hPa reflectivity (dBZ, color shading) overlaid with 850-hPa wavenumber-2 asymmetry of potential vorticity (solid line for 0.1 PVU and dashed line for 1 PVU, where 1 PVU = 10−6 K kg–1 m2 s–1) at 0700, 0800, and 0900 UTC 2 Oct 2005. (d)–(f) As in (a)–(c), but overlaid with the 850-hPa wavenumber-2 vertical velocity (solid line, 0.05 m s–1; dashed line, 0.2 m s–1). Purple circles are the 50-, 100-, 150-, and 200-km radii from the typhoon center marked using the typhoon symbol.

Considering the small radial propagation speed and other features noted above, we think that this rainband is probably related to VRW dynamics although other mechanisms cannot be completely ruled out at this stage. Further investigations, including numerical simulations, are needed for a better understanding of the trigger and maintenance of this rainband.

ACKNOWLEDGMENTS

The authors thank the two anonymous reviewers for their thoughtful comments. The radar images used in this study were provided by the Taiwan Central Weather Bureau. This work was supported by the outreach project of the State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences (2016LASW-B02). Y. Wang was supported in part by the National Basic Research and Development Project (973 program) of China under Contract 2015CB452805, and in part by the National Natural Science Foundation of China under Grants 41675044 and 41730960 and the Basic Research Fund of CAMS 2016Z003 and 2017Y013.

FOR FURTHER READING

FOR FURTHER READING
Chen
,
Y.
, and
M. K.
Yau
,
2001
:
Spiral bands in a simulated hurricane. Part I: Vortex Rossby wave verification
.
J. Atmos. Sci.
,
58
,
2128
2145
, https://doi.org/10.1175/1520-0469(2001)058<2128:SBIASH>2.0.CO;2.
Corbosiero
,
K. L.
,
J.
Molinari
,
A. R.
Aiyyer
, and
M. L.
Black
,
2006
:
The structure and evolution of Hurricane Elena (1985). Part II: Convective asymmetries and evidence for vortex Rossby waves
.
Mon. Wea. Rev.
,
134
,
3073
3091
, https://doi.org/10.1175/MWR3250.1.
Diercks
,
J. W.
, and
R. A.
Anthes
,
1976
:
Diagnostic studies of spiral rainbands in a nonlinear hurricane model
.
J. Atmos. Sci.
,
33
,
959
975
, https://doi.org/10.1175/1520-0469(1976)033<0959:DSOSRI>2.0.CO;2.
Houze
,
R. A.
, Jr.
,
2010
:
Clouds in tropical cyclones
.
Mon. Wea. Rev.
,
138
,
293
344
, https://doi.org/10.1175/2009MWR2989.1.
Meng
,
Z.
, and
Y.
Zhang
,
2012
:
On the squall lines preceding landfalling tropical cyclones in China
.
Mon. Wea. Rev.
,
140
,
445
470
, https://doi.org/10.1175/MWR-D-10-05080.1.
Montgomery
,
M. T.
, and
R. J.
Kallenbach
,
1997
:
A theory for vortex Rossby-waves and its application to spiral bands and intensity changes in hurricanes
.
Quart. J. Roy. Meteor. Soc.
,
123
,
435
465
, https://doi.org/10.1002/qj.49712353810.
Reasor
,
P. D.
,
M. T.
Montgomery
,
F. D.
Marks
Jr.
, and
J. F.
Gamache
,
2000
:
Low-wavenumber structure and evolution of the hurricane inner core observed by airborne dual-Doppler radar
.
Mon. Wea. Rev.
,
128
,
1653
1680
, https://doi.org/10.1175/1520-0493(2000)128<1653:LWSAEO>2.0.CO;2.
Rotunno
,
R.
,
J. B.
Klemp
, and
M. L.
Weisman
,
1988
:
A theory for strong, long-lived squall lines
.
J. Atmos. Sci.
,
45
,
463
485
, https://doi.org/10.1175/1520-0469(1988)045<0463:ATFSLL>2.0.CO;2.
Wang
,
Y
.,
2002
:
Vortex Rossby waves in a numerically simulated tropical cyclone. Part I: Overall structure, potential vorticity, and kinetic energy budgets
.
J. Atmos. Sci.
,
59
,
1213
1238
, https://doi.org/10.1175/1520-0469(2002)059<1213:VRWIAN>2.0.CO;2.
Weisman
,
M. L.
, and
R.
Rotunno
,
2004
:
“A theory for long-lived squall lines” revisited
.
J. Atmos. Sci.
,
61
,
361
382
, https://doi.org/10.1175/1520-0469(2004)061<0361:ATFSLS>2.0.CO;2.
Willoughby
,
H. E.
,
F. D. J.
Marks
Jr.
, and
R. J.
Feinberg
,
1984
:
Stationary and moving convective bands in hurricanes
.
J. Atmos. Sci.
,
41
,
3189
3211
, https://doi.org/10.1175/1520-0469(1984)041<3189:SAMCBI>2.0.CO;2.
Yu
,
C. K.
, and
C. L.
Tsai
,
2010
:
Surface pressure features of landfalling typhoon rainbands and their possible causes
.
J. Atmos. Sci.
,
67
,
2893
2911
, https://doi.org/10.1175/2010JAS3312.1.
Yu
,
C. K.
, and
C. L.
Tsai
,
2013
:
Structural and surface features of arc-shaped radar echoes along an outer tropical cyclone rainband
.
J. Atmos. Sci.
,
70
,
56
72
, https://doi.org/10.1175/JAS-D-12-090.1.

Footnotes

A supplement to this article is available online (10.1175/BAMS-D-17-0122.2)

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Supplemental Material