First simultaneous measurement of vertical air velocity, particle fall velocity, and hydrometeor sphericity in stratiform precipitation: Results from 47 MHz wind-profiling radar and 532 nm polarization lidar observations

Authors


Abstract

[1] Results from simultaneous measurements of vertical air velocity (W), particle fall velocity, and hydrometeor sphericity in stratiform precipitation are reported for the first time. Cases of stratiform precipitation on 8 (case A) and 16 December 2008 (case B) observed at Sumatra, Indonesia (0.2°S, 100.32°E), are described. A 47 MHz wind-profiling radar measuredWand reflectivity-weighted particle fall velocity relative to the air (VZ) simultaneously. Upward W above ∼6.0 km altitude in case B (>0.2 m s−1) was greater than in case A (<0.1 m s−1). VZ at 300 m above the 0°C altitude in case B (1.8 m s−1) was greater than in case A (1.3 m s−1). The thickness of melting layer (ML) in case B (900 m) was greater than in case A (300 m). Because the large-sized aggregates contribute to produce greaterVZ and thicker ML, it is likely that entangled growth of dendritic crystals under the presence of significant upward Wand enhanced aggregation occurrence by the well-developed dendritic crystals produced the large-sized aggregates. Lidar measured an increase of linear depolarization ratio (δ) and lidar dark band in the ML. Volume δ of raindrops was 0.08–0.10 in case B and close to zero in case A. Stronger multiple scattering in case B is likely a cause that produced the greater δ. In case B, a dip of δ was measured at the bottom of ML. The decrease of hydrometeor nonsphericity at the final stage of melting explains the dip.

1. Introduction

[2] Precipitation is generally considered to be of two clearly distinguished types: stratiform and convective. Stratiform precipitation falls from nimbostratus clouds, while convective precipitation falls from cumulus and cumulonimbus clouds. Nimbostratus reaches well above the 0°C altitude, and hence growth and phase change of particles characterize microphysical properties of stratiform precipitation. In the upper part of stratiform precipitation, ice particles grow by vapor deposition and riming, and aggregation concentrates the condensates into large ice particles. When ice particles fall down and reach just below the 0°C altitude, they melt into rapidly falling raindrops [Houze, 1993].

[3] In situ observations using aircrafts and videosondes have been used for measuring particle characteristics in precipitating clouds [e.g., McFarquhar et al., 2007; Takahashi and Keenan, 2004]. Remote sensing tool is also a useful means to measure precipitation. Weather radars have been proved to be useful for measuring characteristics of precipitation [Bringi and Chandrasekar, 2001; Doviak and Zrnić, 1993]. Lidars which detect echoes scattered by hydrometeors are also a promising tool to measure microphysical properties of precipitation. Especially, linear depolarization ratio (hereafter δ) measured by lidars has been proved to be useful for differentiating the phase, shape, and orientation of hydrometeors [e.g., Bissonnette et al., 2001; Kolev et al., 2005; Noel et al., 2002; Sassen, 1991].

[4] In stratiform precipitation, though vertical air velocity (hereafter W) is small compared to fall velocity of ice crystals, upward air motion above the 0°C altitude plays a crucial role in maintaining supersaturation by condensing vapor, which is deposited onto ice particles [Houze, 1993]. Therefore, observations to quantitatively evaluate W in stratiform precipitation were carried out [e.g., Houze, 1989]. However, in order to measure W by radiosondes or weather radars, an assumption of W = 0 at lower or higher boundary of W computation is necessary because altitude profile of Wis computed from vertical integration of wind divergence. On the other hand, clear air Doppler radars, also referred to as wind-profiling radars, can measureW without such assumption because they directly measure wind velocities using a Doppler shift of clear air echoes scattered by refractive index irregularities [Fukao, 2007]. Especially, wind-profiling radars operated near 50 MHz frequency (i.e., 6 m wavelength; hereafter 50 MHz wind-profiling radars) have been proved to be useful for measuringW in stratiform precipitation region, because their low frequency enables them to detect clear air echoes and precipitation echoes separately [e.g., Balsley et al., 1988; Cifelli and Rutledge, 1994; Nishi et al., 2007]. Though frequencies used by wind-profiling radars (i.e., less than 2–3 GHz) are lower than those of weather radars, they have sufficient sensitivity to measure particle fall velocity relative to the ground (hereafterVZ+air) using echoes scattered by hydrometeors [e.g., Rao et al., 2008; Tabata et al., 2011; Williams et al., 1995]. Further, Wmeasured by wind-profiling radars contributes to measure reflectivity-weighted particle fall velocity relative to the air (hereafterVZ) correctly, because VZ can be retrieved by subtracting W from VZ+air [e.g., Schafer et al., 2002; Luce et al., 2010a; Yamamoto et al., 2008].

[5] Though measurements using a wind-profiling radar and lidar have been used to study wind and turbulence features in and around nonprecipitating clouds [Luce et al., 2010b; Yamamoto et al., 2009a, 2009b], measurement results in precipitating conditions have not been reported. W, VZ, and δprofiles in stratiform precipitation were measured simultaneously at the Equatorial Atmosphere Observatory at Kototabang, West Sumatra Indonesia (0.2°S, 100.32°E, 865 m height above the mean sea level (MSL); hereafter observatory) on 8 and 16 December 2008. Using the measurement results of a 47 MHz wind-profiling radar referred to as the Equatorial Atmosphere Radar (hereafter EAR) and 532 nm polarization lidar (hereafter lidar), characteristics ofW, VZ, and δ in the stratiform precipitation are described.

2. Data

2.1. Overview

[6] From 6 to 23 December 2008, continuous observation of the troposphere using the EAR and the lidar was carried out. The observation aims at clarifying W and turbulence motions associated with clouds and precipitation, and was named Cloud Observation Campaign by Lidar and the Equatorial Atmosphere Radar (hereafter CLEAR). Details of the EAR and lidar are described in sections 2.2 and 2.3, respectively. During CLEAR, the lidar measured δ up to the melting layer on 8 and 16 December 2008. Unfortunately the lidar could not measure δ up to the melting layer in other precipitation cases because of strong lidar signal attenuation by hydrometeors. Therefore two stratiform precipitation cases on 8 and 16 December 2008 were used for data analysis. Because this study focuses on characteristics observed above 2 km MSL, precipitation occurrence was defined by the presence of hydrometeors above 2.0 km MSL. Whether precipitation reached to the surface was not taken into account for defining the precipitation occurrence.

[7] Convective and stratiform precipitation can be defined in terms of their W scales. Upward W in convective precipitation is ∼1–10 m s−1, which equals or exceeds the typical fall speeds of ice crystals and snow (∼1–3 m s−1). On the contrary, stratiform precipitation is defined as precipitation processes in which upward W is small compared to the fall velocity of ice crystals and snow [Houze, 1993]. W profiles measured by the EAR were used in order to identify the two stratiform precipitation cases. W values in the altitude range 6–10 km MSL, where upward W is dominant both for convective and stratiform precipitation [Houze, 1989] were used to define whether precipitation is convective or stratiform. Because upward W in the altitude range 6–10 km was less than 1 m s−1 in both the two cases, it is concluded that the two precipitation cases were stratiform (later shown in section 3). It is noted that using W profiles is a straightforward way to define convective and stratiform precipitation. Because of the lack of means to measure W with high accuracy, horizontal distribution of radar reflectivity factor [Steiner et al., 1995], moments of Doppler spectra measured by vertically pointing Doppler radar [Rao et al., 2008; Williams et al., 1995], raindrop size distribution (hereafter DSD) measured by disdrometer [Tokay and Short, 1996] have been utilized in order to distinguish between convective and stratiform precipitation.

[8] During CLEAR, surface rainfall was collected by the optical rain gauge installed at the observatory. Further, 15 radiosondes launched from the observatory measured altitude profiles of temperature. Because radiosonde soundings were not carried out on 8 and 16 December, all the altitude profiles of temperature measured by radiosondes are averaged in order to produce Figure 1. The altitude of 0°C was ∼4.8–4.9 km MSL. 9.74 GHz weather radar installed at the observatory (hereafter weather radar) was used to observe radar reflectivity factor (hereafter Ze). The basic specification of the weather radar has been described by Konishi et al. [1998]. Because of the maximum elevation angle of 29.5°, the altitude profile of Ze at the vertical direction (i.e., above the EAR and lidar) was not measured. Therefore, Ze obtained with the 29.5° elevation angle was used to represent the values around the EAR and lidar. Time and range interval of data collection was ∼8 min and 500 m, respectively. Ze attenuation by raindrops was not corrected because data collected only within 10 km range were used for the data analysis. Because the weather radar was not operated on 8 December, measurement results only on 16 December are presented.

Figure 1.

Altitude profile of temperature collected by radiosondes. Temperature values were calculated by averaging all the available data during the Cloud Observation Campaign by Lidar and the Equatorial Atmosphere Radar (CLEAR). Each temperature profile was averaged every 50 m. Horizontal lines indicate standard deviations. The dashed line indicates 0°C.

2.2. The 47 MHz Wind-Profiling Radar

[9] In the stratiform precipitation cases on 8 and 16 December 2008, W and VZ+airwere measured by the EAR. The EAR uses a circular antenna array which consists of 560 three-element Yagi antennas and has an approximate diameter of 110 m. Each of 560 Yagi antennas is driven by a solid-state transmitter-receiver module with 180 W peak output power, and the total peak output power is 100 kW. For the thorough system description of the EAR, seeFukao et al. [2003]. Table 1lists the observational parameters of the EAR. The EAR can steer its radar beam to off-vertical directions in order to measure horizontal wind velocity. However, during the period described in the study, the radar beam was fixed to the vertical incidence in order to intensively measureW and VZ+air. Because time for data sampling (i.e., time for transmitting and receiving signals) was 131 s and time for transferring observed data to the hard disk drive was 9 s, the time interval of each record was 140 s. Data sampling was not done during the data transfer to the hard disk drive. After storing successive 8 records (18 min 40 s), the W measurement was interrupted by another observation mode with a time length of 122 s. The vertical resolution was 150 m, which corresponds to the transmitted pulse width of 1 μs. During the experiment, 5 carrier frequencies were used in order to improve the range resolution by a range imaging technique [Mega et al., 2010]. However, in order to simplify the data analysis, 5 Doppler spectra each of which was measured by cycling carrier frequencies were averaged into one Doppler spectrum. The averaging corresponds to the incoherent integration. The vertical resolution of 150 m was sufficient to interpret the measurement results presented in the study.

Table 1. Observational Parameters of the Equatorial Atmosphere Radara
ItemSpecification
  • a

    IPP, interpulse period; Ncoh, number of coherent integrations (time domain averaging); NFFT, number of fast Fourier transform points.

Transmitted pulse1 μs × 16 (16 bit optimum code of Spano and Ghebrebrhan [1996])
Beam direction (azimuth, zenith)(0°, 0°)
IPP (μs)400
Ncoh64
NFFT1024
Sampling interval (μs)1.0 (150 m in range resolution)

[10] Using the averaged Doppler spectrum, values of W and VZ+airwere calculated from the first-order moment of clear air echo and hydrometeor echo, respectively. Clear air echo and hydrometeor echo were separated both automatically and manually in order to avoid misestimation ofW and VZ+air. Fine velocity resolution of Doppler spectra (0.024 m s−1) was useful to separate clear air and hydrometeor echoes. In order to estimate the measurement error of W, equation (13) of Yamamoto et al. [1988] was used. In the equation, error of W is determined by the radar wavelength, observation time, and number of incoherent integrations. During the observation period, they were 6.38 m, 131 s, and 5 times, respectively. Because the moment method was used, the value of coefficient k was 0.38. Using the values, the error in W was estimated to be 0.02 m s−1 even for a large spectral width of 0.5 m s−1. Therefore the measurement accuracy of W was sufficient for discussing W which can be ∼0.1 m s−1 or less in the stratiform precipitation. Using equation (6.26) of Doviak and Zrnić [1993], measurement error of VZ+air was also estimated. The effect of incoherent integration was also taken into account. Radar wavelength, NFFT, sampling interval (0.128 s), and number of incoherent integration determine the error estimation. Measurement error of VZ+air for ice phase hydrometeors (i.e., above the altitude of 0°C) was estimated to be less than 0.04 m s−1for signal-to-noise ratio (SNR) greater than 5 dB and spectral width less than 0.3 m s−1. Because Doppler velocity range used for calculating VZ+air was limited in order to maximize SNR, SNR greater than 5 dB was attained in most of VZ+air presented in the study. Measurement error of VZ+air for raindrops was difficult to quantify because assumptions necessary for quantifying the estimation errors did not hold true due to large variability in VZ+air. However, because VZ variation with altitude was more smooth than VZ+air (later shown in Figure 2), it is concluded that VZ+air was measured with the accuracy comparable to W. Because a means to calibrate received power was not available, radar reflectivity factor is not shown.

Figure 2.

Altitude profiles of Doppler spectrum (colored), W (thick black curve on left), VZ+air (thin black curve), and VZ (thick black curve on right) measured by the Equatorial Atmosphere Radar (EAR) from 20:22:51 to 20:25:02 local standard time (LST) on 16 December 2008. Positive values of Doppler velocity indicate that a scatterer is moving toward the ground.

[11] W and VZ+air measured by the EAR were used to retrieve VZ. Figure 2 is an altitude profile of Doppler spectrum. W, VZ+air, and VZ measured by the EAR from 20:22:51 to 20:25:02 local standard time (hereafter LST; LST = UTC +7) on 16 December 2008. Clear air echoes and hydrometeor echoes were detected separately due to larger fall velocity of hydrometeors than W [e.g., Renggono et al., 2006; Wakasugi et al., 1986]. It is stressed that simultaneous detection of clear air and hydrometeor echoes is difficult for instruments other than 50 MHz wind-profiling radars.VZ was computed by subtracting W from VZ+air (i.e., VZ = VZ+air − W). Though VZ+air, a sum of VZ and W, showed vertical fluctuations due to vertical W changes exceeding ±0.5 m s−1 or greater, the fluctuations were removed in VZ.

2.3. Lidar

[12] The measurement system and signal processing of the lidar are described. The laser used in this study was radiated to the vertical incidence with a wavelength of 532 nm, output power of about 200 mJ, pulse repetition frequency of 10 Hz, beam diameter of 10 cm, and half laser beam divergence of 0.1 mrad. Abo et al. [2006] describe the detail of the lidar system. In order to measure δin the lower and middle troposphere (below ∼10 km MSL), a telescope receiver for polarization measurement was additionally installed in 2008. Data obtained with the new telescope receiver were used for data analysis. The telescope receiver was aligned with the laser axis and consisted of Makustov-Cassegrain telescope with 11 cm diameter, polarization prism, and two photomultipliers. The receiver field of view (hereafter FOV) was 0.5 mrad. Altitude profiles of the lidar return measured in the same direction as the laser polarization (hereafterP(z)) and that of the lidar return measured perpendicularly to the laser polarization (hereafter P(z)) were recorded after integrating P(z) and P(z) from 600 laser shots (i.e., every 1 min). In order to compensate the nonlinearity of photomultiplier tube, dead time effects on photon counting were corrected by following the supplier's instruction. Effects of dark count were not significant. Note that z denotes altitude. δ(z) is defined as the ratio of P(z) to P(z); that is, inline image. Backscattered intensity (hereafter P(z)) was calculated as the sum of P(z)and P(z) (i.e., P(z) = P(z) + P(z)). Profiles of P(z) and δ(z) were used for data analysis.

3. Results and Discussion

3.1. Stratiform Precipitation on 8 December 2008

[13] Time and altitude variations of stratiform precipitation on 8 December 2008 are described. Figure 3a is a time altitude plot of W from 02:30 to 06:00 LST. The period before 02:30 LST was not shown because lidar signals were attenuated by hydrometeors near the surface and could not penetrate the altitudes above 2.0 km MSL. In the altitude range ∼6.0–9.4 km MSL, weak upward W up to ∼0.2 m s−1 was continuous from 02:30 to 04:45 LST. Nishi et al. [2007] concluded that among the mechanisms that produce upward W in stratiform region (i.e., latent heat release resulting from deposition growth, gravity waves, old cells which have origin in the convective region, and latent heat release resulting from riming), latent heat release resulting from deposition growth is the most plausible cause that produces the moderate and continuous upward motion as observed. After 04:45 LST, upward W in the altitude range ∼6.0–9.4 km was not continuous. Below 6.0 km MSL, downward W was dominant. Radiosondes measured that an altitude of 0°C was ∼4.8–4.9 km MSL (Figure 1). Downward W below the 0°C altitude was common in stratiform precipitation, and can extend to altitudes higher than the 0°C altitude [Houze, 1989]. Surface rainfall was absent except the period of 03:31–03:38 LST, and rainfall amount during the period was 0.019 mm (see Figure 3d). Tokay and Short [1996] classified the rainfall category (very light, light, moderate, heavy, very heavy, and extreme) from surface rainfall intensity. Following their rainfall category, it is concluded that very light rainfall with rainfall intensity of less than 1 mm h−1 was observed in the 8 December case.

Figure 3.

Time altitude plots of (a) W, (b) P, and (c) δ from 02:30 to 06:00 LST on 8 December 2008. (d) Time series of surface rainfall intensity measured by the collocated rain gauge. Surface rainfall intensity was calculated every 1 min. Arrow on the top of Figure 3a is the period used for producing Figure 4. Positive W values indicate that wind velocity is upward.

[14] Figures 3b and 3c are time altitude plots of P and δ, respectively. Throughout the period, significant vertical variations were observed in δ and P in the altitude range where hydrometeor melting is able to occur (i.e., the altitudes just below the 0°C level). Note that the lidar measured volume δ, which include contributions from hydrometeors within the sampling volume. δ was close to zero below 4.5 km MSL, increased with altitude above 4.5 km MSL, and was dominantly greater than 0.10 at 4.70–4.80 km MSL. A local minimum of P (less than 500) was observed at 4.60–4.70 km MSL. The P minimum at the melting altitude occurs because size decrease of melting aggregates causes their smaller backscattering intensity than aggregates which do not melt, and raindrops have greater backscattering intensity than melting aggregates due to surface waves and the front face axial reflection [Sassen et al., 2005]. The vertical decrease of P at melting altitudes is named “lidar dark band” [Sassen and Chen, 1995]. The vertical increase of δ above 4.5 km MSL indicates the presence of nonspherical (i.e., ice phase or mixed phase) hydrometeors. δvalue peaked in the altitude range 4.5–4.9 km MSL, where hydrometeor melting occurred. It is speculated that the complex interfaces produced by water-coated ice crystal branches create the enhancement ofδ at the melting altitudes [Sassen, 1975; Sassen and Chen, 1995]. The features observed in δ and P indicate the presence of melting layer (hereafter ML). After 06:30 LST, the features of ML as observed in P and δ disappeared (not shown). After 04:15 LST, δ above the ML altitude (i.e., 4.9 km) was dominantly less than 0.05 though high nonsphericity can be expected for ice hydrometeors. Because the laser beam was pointed to the vertical incidence, specular reflection from crystal's plane face and/or horizontally oriented plates explains the small δ above the ML [Bissonnette et al., 2001]. Because Bissonnette et al. [2001] also showed that multiple scattering effects for snow are negligible for FOV of 1 mrad, small FOV used in this study (i.e., 0.5 mrad) contributed to minimize the increase of δ by multiple scattering effects.

[15] In order to describe features in W, VZ, P, and δ further quantitatively, they were averaged in time. A period from 03:30 to 04:50 LST, when the lidar continuously detected returns up to 7.0 km MSL and the EAR detected hydrometeor echoes with a data rate greater than 70%, was used for time averaging. Figure 4a is an altitude profile of W. Above 6.0 km MSL, the average W was upward and had amplitudes up to 0.1 m s−1. The disturbance of W observed in its standard deviation was ∼0.05 m s−1. Below 6.0 km MSL, the average W was downward, and its amplitude was less than 0.1 m s−1 in the altitude range 3.0–6.0 km MSL. The disturbance of W in the altitude range 3.0–6.0 km MSL was mostly less than 0.1 m s−1. Figure 4b is an altitude profile of VZ. Values of VZ above 7.6 km MSL and below 3.2 km MSL were not plotted due to insufficient data quality caused by the low data rate of less than 70%. The average VZ was ∼0.8–1.0 m s−1 above 7.0 km MSL, increased with decreasing altitude, and was ∼1.3 m s−1 at 5.1–5.4 km MSL. The average VZ was 1.6 m s−1 at 4.9 km MSL, which was 0.3 m s−1 greater than at 5.1 km. Because 4.9 km MSL is the altitude of 0°C, this increase in VZ indicates that hydrometeor melting commenced around 4.9 km MSL. The average VZ increased with decreasing altitude, and was 3.7 m s−1 at 4.6 km MSL. The average VZ was 3.6–3.9 m s−1 below 4.6 km, and its disturbance was ±0.8–1.0 m s−1. The increase in VZ in the altitude range 4.6–4.9 km MSL indicates the presence of melting of hydrometeors [see Houze, 1993, Figure 6.2], and the thickness of ML was ∼300 m. Figures 4c and 4d are altitude profiles of P and δ, respectively. Values of δ above 7.6 km MSL were not plotted because P was not measured due to lidar signal attenuation. The increase in δ and decrease in P at the melting altitudes (i.e., 4.6–4.9 km MSL) were consistent with the presence of ML, which was confirmed by the vertical increase of VZ with decreasing altitude. The average δ had the maximum value of 0.18 at 4.8 km MSL. The average VZ of raindrops was 3.6–3.9 m s−1 in the altitude range 3.2–4.6 km MSL. Owing to high sphericity of raindrops, the average δ was ∼0.01 below 4.6 km MSL. In section 3.3, the features of W, VZ, P, and δ in the 8 December case are compared with those observed in the 16 December 2008 case.

Figure 4.

Altitude profiles of (a) W, (b) VZ, (c) P, and (d) δ averaged from 03:30 to 04:50 LST on 8 December 2008. Positive W values indicate that wind velocity is upward, and positive VZ values indicate that hydrometeors fall toward the ground. Arrows at the right of each plot show the altitude of 0°C estimated by the radiosonde soundings. The thick curves show the average values, and the dotted curves on the both sides of the averaged values show disturbance determined by the standard deviation. In Figure 4c, some parts of the dotted curve on the left are missing because the standard deviation value was greater than the average value.

3.2. Stratiform Precipitation on 16 December 2008

[16] Time and altitude variations of stratiform precipitation on 16 December 2008 are described. Figure 5d is a time series of surface rainfall intensity from 19:00 to 24:00 LST on 16 December 2008. The period before 19:00 LST was not shown because lidar signals were attenuated by hydrometeors near the surface and could not penetrate the altitudes above 2.0 km MSL. Surface rainfall was continuously recorded until 23:07 LST, and its total rainfall amount from 19:00 to 24:00 LST was 7.62 mm. Figure 5a is a time altitude plot of W. During the period when surface rainfall was mostly greater than 2 mm h−1 (from 20:14 to 21:41 LST), upward W greater than 0.2 m s−1 was dominant above 6.2–7.5 km MSL. After 22:00 LST, the upward W greater than 0.2 m s−1 was absent, and the rainfall rate was less than 1 mm h−1. Following the rainfall category by Tokay and Short [1996], it is concluded that rainfall was moderate (rainfall intensity of 2–5 mm h−1) from 20:14 to 21:41 LST, then changed to very light (rainfall intensity of less than 1 mm h−1) after 21:41 LST. During the period shown in Figure 5, downward W was dominant below 6.0 km MSL as observed in the 8 December case. Figure 5b is a time altitude plot of P. The P value had a vertical minimum (less than 400) at the melting altitude. The P minimum located in the altitude range 4.0–4.4 km MSL from 19:45 to 21:45 LST, then moved in the altitude range 4.4–4.6 km MSL. The minimum of P indicates the presence of lidar dark band. Figure 5c is a time altitude plot of δ. Owing to the hydrometeor melting, δ was greater than 0.10 above 4.2 km MSL from 19:55 to 21:30 LST and at ∼4.4–4.8 km after 21:30 LST.

Figure 5.

Same as Figure 3 except the period is from 19:00 to 24:00 LST on 16 December 2008. Arrow on the top of Figure 5a is the period used for producing Figure 8.

[17] Figure 6 shows a horizontal distribution of Ze measured by the weather radar at 20:46 LST. Elevation angle of the radar beam was 29.5°. Around ∼6.0–7.0 km away from the observatory (i.e., 4.3–4.8 km MSL), radar bright band in which Ze was dominantly greater than 40 dB Ze was observed. Presence of radar bright band confirms the presence of stratiform precipitation [Houze, 1993; Steiner et al., 1995]. Figure 7 shows a time altitude plot of Ze. Because the weather radar did not measure the vertical incidence, the altitude profiles of Ze were computed by averaging Ze values collected by the azimuth beam scan with the elevation angle of 29.5°. An increase of Ze, which indicates the presence of radar bright band, was observed around 4.5 km MSL from 19:00 to 24:00 LST. Ze at the radar bright band was greater than 40 dBZe from 19:40 to 21:50 LST.

Figure 6.

Horizontal distribution of Ze at 20:46 LST on 16 December 2008 measured by the weather radar. Elevation angle of the radar beam was 29.5°.

Figure 7.

Time altitude plot of Ze collected by the weather radar from 19:00 to 24:00 LST on 16 December 2008. Data rate of Ze for every azimuth scan at the elevation angle of 29.5° is shown by contour lines. Ze at each of the range gates is computed by averaging Ze values obtained every azimuth beam scan. Ze data with data rate less than 10% are not shown.

[18] In order to focus on features in W, VZ, P, and δ under the presence of enhanced surface rainfall, they were averaged during the period when the surface rainfall intensity was continuously greater than 2 mm h−1 (i.e., from 20:14 to 21:41 LST). Figure 8a is an altitude profile of W. The average W was upward above 6.2 km MSL, and its amplitude was greater than 0.2 m s−1 above 7.0 km MSL. The average W reached ∼0.5 m s−1 in the altitude range 8.8–9.1 km MSL. The disturbance of W above 6.2 km MSL was mostly greater than 0.2 m s−1. Below 6.2 km MSL, the average W was downward. The downward W was greater than 0.1 m s−1 in the altitude range 3.4–4.3 km MSL. The disturbance of W below 6.2 km MSL was mostly greater than 0.1 m s−1. Figure 8b is an altitude profile of VZ. Values of VZ above 8.7 km MSL were not plotted because hydrometeor echoes were absent above 8.7 km MSL. The average VZ was ∼1.3 m s−1 above 7.0 km MSL, increased gradually with decreasing altitude, and was 1.8 m s−1 at 5.1 km MSL. The average VZ was 2.0 m s−1 at 4.9 km MSL, which was 0.2 m s−1 greater than VZ at 5.1 km MSL. This increase in VZ indicates that hydrometeor melting commenced around 4.9 km MSL. The average VZ in the altitude range 4.0–4.9 km increased rapidly with decreasing altitude, and reached 7.5 m s−1 at 4.0 km MSL. In the altitude rage 2.3–4.0 km MSL, the average VZ was 7.0–7.5 m s−1. The increase in VZ in the altitude range 4.0–4.9 km MSL indicates the presence of ML with a thickness of ∼900 m.

Figure 8.

Same as Figure 4 except that the values are calculated from 20:14 to 21:41 LST on 16 December 2008.

[19] It is noted that accurate measurement of W is necessary for correctly retrieving VZ, especially in presence of significant W disturbances. During the case shown in Figures 4a and 4b (hereafter case A), the average VZ+air above 7.0 km MSL in case A was ∼0.7–0.9 m s−1 because the average VZ was ∼0.8–1.0 m s−1 and the average upward W was ∼0.1 m s−1. During the case shown in Figures 8a and 8b (hereafter case B), the average VZ+air above 7.0 km MSL was ∼0.8–1.0 m s−1 because the average VZ was ∼1.3 m s−1 and the average upward W was ∼0.3–0.5 m s−1. The values of average VZ+air above 7.0 km MSL in cases A and B were nearly the same, though the value of average VZ in the 16 December case was ∼0.3–0.5 m s−1 greater than in case A. This result demonstrates that Wmeasurement by 50 MHz wind-profiling radar is useful for retrieving hydrometeor fall velocity in stratiform precipitation region.

[20] Figure 8c is an altitude profile of P. The average P had the vertical minimum in the altitude range 4.0–4.4 km MSL, which indicates the presence of lidar dark band. The average P decreased sharply with increasing altitude above 4.7 km MSL because of strong lidar signal attenuation by hydrometeors. Figure 8d is an altitude profile of δ. Values of δ above 5.2 km MSL were not plotted because P was not measured sufficiently due to lidar signal attenuation. The average δ above 4.0 km MSL showed an increase and was greater than 0.2 in the altitude range 4.4–4.9 km MSL. The average δ had the maximum value of 0.25 at 4.6 km MSL. The increase in δ and decrease in P at the melting altitudes (i.e., 4.0–4.9 km MSL) were consistent with the presence of ML, which was confirmed by the vertical increase of VZ with decreasing altitude (Figure 8b). Interestingly, the average δ was not close to 0 (i.e., ∼0.08–0.10) in the altitude range 2.0–4.0 km MSL, where raindrops exist. The reason for the increase in δ is discussed in section 3.3.3.

3.3. Interpretation

3.3.1. W and VZ

[21] Differences in W, VZ, P, and δ between cases A and B are described. Upward W above 6.2 km MSL (greater than 0.2 m s−1) in case B was greater than that in case A, which was less than 0.1 m s−1 (Figures 4a and 8a). The disturbances of W above 6.2 km MSL in case B (dominantly greater 0.2 m s−1) was greater than that in case A (∼0.05 m s−1). Latent heat release resulting from deposition growth is the most plausible cause that produced the moderate upward motion in case A [Nishi et al., 2007]. On the other hand, the greater upward W and W disturbance in case B indicates that all the mechanisms that are able to produce upward W in stratiform region (i.e., latent heat release resulting from deposition growth, gravity waves, old cells which had origin in the convective region, and latent heat release resulting from riming) are candidates that produced the observed upward W. It is noted that riming was limited to the altitude below ∼7.6 km MSL, where temperature was greater than 258 K (i.e., −15°C) (see Figure 1) and hence riming can be observed [Woods et al., 2005].

[22] VZ at 7.0 km MSL in case B (1.3 m s−1) was greater than in case A (1.0 m s−1). Because D0 estimated at the bottom of ML in case B was ∼2.8 times greater than in case A (described later), the greater VZ at 7.0 km MSL in case B indicates that hydrometeors with larger size existed in case B, though fall velocity of ice hydrometeors depends on their shape and density [Heymsfield et al., 2002]. Because both vapor condensation occurring in stratiform region and riming in convective cells which have typical updrafts of several m s−1or more significantly contribute to the size growth of ice-phased hydrometeors in the upper part of stratiform region [Houze, 1993], they are candidates that produced larger hydrometeors size in case B. Let us mention again that because the values of VZ+air above 7.0 km MSL in cases A and B were nearly the same (see section 3.2), detailed considerations on hydrometeor fall velocity cannot be made if accurate W measurement was not available.

[23] The increase of VZ with decreasing altitude was observed in the altitude range above 4.9 km MSL, where temperature was below 0°C (Figures 4b and 8b). Above ∼6.0 km MSL, where upward W was observed, both vapor condensation and aggregation are candidates which produced the increase in hydrometeor size. Though it is possible that riming played a role in the size growth of hydrometeors, the contribution of riming was limited to the altitude below 7.6 km MSL, where temperature was greater than 258 K (i.e., −15°C).

[24] The increase of VZ was seen even in the altitude range 4.9–6.0 km MSL, where upward W was not observed. In case A, VZ was ∼1.2 m s−1 around 6.0 km MSL and ∼1.3 m s−1 around 5.1 km MSL (Figure 4b). In case B, VZ was ∼1.4 m s−1 around 6.2 km MSL and ∼1.8 m s−1 around 5.1 km MSL (Figure 8b). Hereafter it is explained that aggregation is the most plausible cause that produced the increase of VZ in the altitude range 4.9–6.0 km MSL. In a 1.0–1.5 km thick layer lying just above the 0°C level, riming is an important process that contributes to size growth of hydrometeors [Houze, 1993]. Riming is closely related to upward W because freezing of liquid drops produces upward W through latent heat release and upward W is necessary for maintaining water saturation [Houze, 1993]. However, because upward W in the altitude range ∼4.9–6.0 km MSL was not observed both in cases A and B (Figures 4a and 8a), it is concluded that riming was not active or absent. Though aggregation is also the process that contributes to size growth of hydrometeors, it does not produce heat. Temperature in the altitude range ∼4.9–6.0 km MSL was 267–273 K (i.e., −6°C–0°C) (Figure 1), and occurrence frequency of aggregation becomes much greater for temperature above ∼−5°C [see Houze, 1993, section 3.2.4; Pruppacher and Klett, 1997]. In situ measurements of hydrometeors using aircrafts have confirmed the hydrometeor size growth by aggregation in the ∼1 km thick layer just above the 0°C level [McFarquhar et al., 2007; Stewart et al., 1984; Willis and Heymsfield, 1989].

[25] The increase of VZ in the altitude range 4.9–6.0 km MSL was greater in case B than in case A (Figures 4b and 8b). In addition to the temperature above ∼−5°C, the presence of planar dendritic crystals or spatial dendrites is necessary for producing aggregation [Rauber, 1987]. The growth of arms of dendritic crystals becomes entangled between −10 and −16°C [Houze, 1993, p. 86], and the altitudes with temperature between −10 and −16°C (i.e., between 263 and 257 K) was 6.7–7.7 km MSL (Figure 1). Because the greater upward W in the altitude range 6.7–7.7 km MSL in case B provided more preferable condition for entangled arm growth of dendritic crystals than in case A, it is concluded that the greater upward W in case B contributed to the greater increase of VZin the altitude range 4.9–6.0 km MSL through enhanced occurrence of aggregation. For further investigation, simultaneous in situ measurement of hydrometeors and 50 MHz wind-profiling radar are necessary. However, it is stressed that accurate measurement ofW provides the opportunity to discuss the microphysical processes which relate to stratiform precipitation in detail. In case A, δ at 7.0 km was 0.015, increased with decreasing altitude, and was 0.06 at 5.0 km MSL (Figure 4d). Though the disturbance observed in δ was significant probably due to returns which produced small δ (i.e., specular reflection from crystal's plane face and/or horizontally oriented plates), this increase in δ indicates the increase of nonsphericity by vapor condensation and aggregation.

[26] By using the model of raindrop fall velocity and assuming DSD, median volume diameter of raindrops (hereafter D0) just below the ML was estimated. VZ was calculated using the relation

display math

where D is the raindrop diameter in m, N(D) is DSD, σ(D) is backscattering cross section of raindrops, and Vt(D) is the terminal fall velocity in m s−1. In the measurement, W and VZ+air are spatially weighted by the antenna beam pattern [Fang and Doviak, 2008]. Because the two-way half-power full width of a radar beam and vertical resolution were 2.4° and 150 m, respectively, the EAR observed ∼260 m × 150 m volume at 7.0 km MSL. However, the horizontal extent of radar beam was small compared to the advection scale. For the sampling time length of 131 s, advection scale of a scatterer is greater than the horizontal extent of radar beam even for the small horizontal wind velocity of 2 m s−1. Inhomogeneity of clear air reflectivity can occur by shear instability, and the reflectivity inhomogeneity can produce misestimation of W by contaminating Doppler velocity measured by the vertical radar beam with horizontal wind velocity [e.g., Yamamoto et al., 2003]. However, the possible misestimation of W is negligible in the study because horizontal wind velocity was small (i.e., mostly less than 10 m s−1). Under the condition that water drop is spherical and D is small compared to the radar wavelength λ (i.e., D ≤ λ/16), σ(D) is expressed by

display math

where Km = (m2 − 1)/(m2 + 2) and m is the complex refractive index of water. Equation (2) is called the Rayleigh approximation. We used the Rayleigh approximation because the condition D ≤ λ/16 is satisfied for wind-profiling radars which are operated below 3 GHz and hence the Rayleigh approximation has been widely used for raindrop measurements by the wind-profiling radars [Renggono et al., 2006; Schafer et al., 2002, Williams, 2002].

[27] The raindrop fall velocity is given by

display math

where Vt(D) is the terminal fall velocity in m s−1, ρ is the air density, and ρ0 is the air density at a pressure of 1013 hPa and temperature of 293.15 K. For details of modeling for raindrop fall velocity, see section 8.2 of Doviak and Zrnić [1993]. ρwas determined by pressure and temperature measured by the radiosonde soundings at the observatory. The Marshall-Palmer distribution is given by

display math

where N0 = 8 × 103 m−3 mm−1, Λ = 4100 × R−0.21 m−1, and R is the rain rate in mm h−1. For details of the Marshall-Palmer distribution, see section 8.1.2 ofDoviak and Zrnić [1993]. Because only Λ (i.e., R) is a variable and the relation between Λ and D0 is given by inline image, D0 is able to be related to VZ by combining equations (1) to (4). The rainfall amount measured by the surface rain gauge from 19:40 to 22:00 LST on 16 December was 6.8 mm, and that estimated at the lowest altitude that the EAR measured (i.e., 2.4 km MSL) were 6.5 mm. The agreement in rainfall amount between the surface rain gauge and the EAR indicates that the Marshall-Palmer distribution was useful for estimatingD0. For VZ of 3.7 m s−1 at 4.5 km MSL (i.e., case A), D0 was 0.4 mm. For VZ of 7.6 m s−1 at 3.9 km MSL (i.e., case B), D0 was 1.1 mm. Though VZ of raindrops had disturbances and the assumption was used for DSD, the estimation result suggests that D0 at the bottom of ML in case B was ∼2.8 times greater than in case A. The greater raindrop size in case B is consistent with the greater VZ above the ML in case B. The uncertainty of D0 was evaluated by assuming the modified gamma distribution (i.e., N(D) = N0Dμ exp(−ΛD) where μ is the shape parameter; see Ulbrich [1983] for details). μ was varied from 0 to 3 because Tokay and Short [1996] showed that μ was generally less than 3 for very light, light, and moderate rains which have rainfall rate less than 5 mm h−1. D0 is given by inline image. Note that the value of N0 did not affect the computation (see equation (1)). For μ of 0–3, D0 for cases A and B varied 0.4–0.5 mm and 1.1–1.4 mm, respectively.

3.3.2. Melting Layer Thickness

[28] As described in sections 3.1 and 3.2, the thickness of ML in case B (900 m) was greater than in case A (300 m). Most of ice-phased hydrometeors melt rapidly as they reach to the 0°C level and only large-sized aggregates determine the thickness of melting layer [Willis and Heymsfield, 1989]. Further, the estimation using VZ at the bottom of ML suggests that D0 at the bottom of ML in case B was ∼2.8 times greater than in case A. Both the thickness of ML and the VZ value at the bottom of ML indicate that hydrometeors with greater size existed in case B. The results support the idea that the size growth of hydrometeors in convective cells by riming and that in stratiform region by condensation above ∼6.0 km MSL and aggregation at 4.9–6.0 km MSL (see section 3.3.1) contributed to the formation of thicker ML in case B. Though δ above the ML was not able to be measured in case B, the peak value of δ in the ML in case B (0.25) was greater than in case A (0.18) and δ in case B was greater than 0.2 over the 400 m altitude range (see Figures 4d and 8d). This result indicates that the nonsphericity of ice hydrometeors was greater in case B, and supports the conclusion that the greater upward W in the altitude range 6.7–7.7 km MSL in case B provided more preferable condition for the entangled arm growth of dendritic crystals than in case A, and contributed to the greater size increase of hydrometeors in the altitude range 4.9–6.0 km MSL through enhanced aggregation.

3.3.3. Volume δ of Raindrops

[29] Volume δ of raindrops (i.e., below 4.0 km MSL) was 0.08–0.10 in case B, though it was close to zero in case A (see Figures 4d and 8d). Using a three-dimensional polarization-dependent ray-tracing algorithm,Roy and Bissonnette [2001] demonstrated that static and dynamical raindrop deformation from the sphere (i.e., steady state flattening and oscillation as a resonant response to eddy shedding) can produce strong increase of δfor single raindrop at off-zenith angles. However, their calculation results also showed that the increase ofδ at the vertical direction (i.e., the direction of the lidar laser beam) does not show significant increase for the raindrop models that are consistent with the measurement results of the scanning lidar [see Roy and Bissonnette, 2001, Figures 16 and 17]. Therefore δ for single raindrop is not likely a cause for the increase of δ in case B. Multiple scattering can generate the increase of volume δunder the presence of large-sized raindrops [Tatarov and Kolev, 2001]. Because scattering by single raindrop is not limited to the backward direction, multiple scattering can increase volume δ even when the lidar laser beam points to the vertical direction. Other lidar measurements showed that the volume δ (i.e., δ including the effects by multiple scattering) for raindrops at the vertical direction can increase ∼0.1 or greater for the FOV as small as 1 mrad [see Roy and Bissonnette, 2001, Figures 6–9]. Because our lidar measurement used the FOV similar to the work of Roy and Bissonnette [2001](i.e., 0.5 mrad), stronger multiple scattering triggered by large-sized raindrops is likely a cause that generated the increase of volumeδ in case B.

[30] Effects of turbulence on raindrop oscillations need to be further investigated [Szakáll et al., 2010]. Therefore it is worth noting that wind disturbance is able to increase δ of single raindrop at the vertical incidence by distorting the orientation of raindrops from the horizontal. Standard deviation of W below the ML altitude was greater than 0.1 m s−1 in case B, while it was less than 0.1 m s−1 in case A. Altitude variation of W in case B was also greater than in case A (Figures 4a and 8a), and hence it is speculated that the greater disturbance of W in case B caused the greater δ in case B. Further theoretical end experimental studies are necessary to prove our speculation.

[31] Another interesting feature observed in δ is a dip around 4.0 km MSL (i.e., the bottom of ML) in case B (see Figure 8d). Most of ice-phased hydrometeors melt rapidly as they reach to the 0°C level and only large-sized aggregates contribute to the ML formation [Willis and Heymsfield, 1989]. The large-sized aggregate disrupts into small-sized particles at the final stage of melting, and the disrupted small-sized particles display round, plate, or lens-like shapes [Fujiyoshi, 1986; Knight, 1979]. The decrease of nonsphericity at the final stage of melting explains the decrease of volume δ at the bottom of ML than that at the altitudes above 4.0 km MSL. We speculate that the stronger multiple scattering effect for raindrops explains the greater volume δ below 4.0 km MSL. By producing stronger scattering to various directions, surface wave of raindrops can cause stronger multiple scattering than melting particles. The presence of lidar dark band, which indicates the stronger raindrop backscattering than melting particles, supports the speculation (see Figures 4c and 8c). Unfortunately there are no theoretical and experimental studies that treated the effects of multiple scattering on melting particles. Further theoretical and experimental studies are required for thorough understanding of δ both for raindrops and melting particles.

4. Summary

[32] Simultaneous measurements of vertical air velocity (W), particle fall velocity, and hydrometeor phase in stratiform precipitation were carried out using the 47 MHz wind-profiling radar referred to as the Equatorial Atmosphere Radar (EAR) and the 532 nm polarization lidar installed at the Equatorial Atmosphere Observatory, West Sumatra, Indonesia (0.2°S, 100.32°E, 865 m MSL). Using the capability of 50 MHz wind-profiling radars to detect clear air echoes and hydrometeor echoes separately, the EAR measuredWand reflectivity-weighted particle fall velocity relative to the ground (VZ+air) simultaneously (Figure 2). Reflectivity-weighted particle fall velocity relative to the ground (VZ) was retrieved by subtracting W from VZ+air (i.e., VZ = VZ+air − W). The lidar measured linear depolarization ratio (δ) for differentiating the phase and sphericity of hydrometeors.

[33] A case of stratiform precipitation observed on 8 (case A) and 16 December 2008 (case B) were compared to describe their differences in W, VZ, and δ (Figures 4 and 8). Raindrops evaporated until they reached to the ground in case A, and surface rainfall intensity was greater than 2 mm h−1 in case B. The altitude of 0°C was ∼4.9 km MSL. Upward W above 6.2 km MSL was greater than 0.2 m s−1 in case B, while upward W above 6.0 km MSL was ∼0.1 m s−1 or less in case A. Latent heat release resulting from deposition growth is the most plausible cause that produced the moderate upward motion in case A. On the other hand, the greater upward W and W disturbance in case B indicate that all the mechanisms that are able to produce upward W in stratiform region (i.e., latent heat release resulting from deposition growth, gravity waves, old cells which had origin in the convective region, and latent heat release resulting from riming) are candidates that produced the observed upward W.

[34] The values of VZ+air above 7.0 km MSL in cases A and B were nearly the same, though the value of VZ in the 16 December case was ∼0.3–0.5 m s−1 greater than in case A (see section 3.2). The result demonstrates that detailed considerations on hydrometeor fall velocity cannot be made when accurate W measurement is not available, and also indicates that Wmeasurement by 50 MHz wind-profiling radar is useful for retrieving hydrometeor fall velocity.

[35] The increase of VZ with decreasing altitude was observed in the altitude range above the 0°C altitude (Figures 4b and 8b). Above ∼6.0 km MSL, where upward W was observed, both vapor condensation and aggregation are candidates that produced the increase in hydrometeor size. The increase of VZ was also observed in the altitude range ∼4.9–6.0 km MSL (i.e., temperature between −6 and 0°C), where upward W was not observed. Because the absence of upward W indicates that latent heat release was not active or absent, it is concluded that aggregation played a major role in the increase of hydrometeor size in the altitude range. Because the greater upward W in the altitude range 6.7–7.7 km MSL in case B provided more preferable condition for the entangled arm growth of dendritic crystals at temperatures between −10 and −16°C, it is concluded that the greater upward W in case B contributed to the greater increase of VZ in the altitude range 4.9–6.0 km MSL than in case A through enhanced occurrence of aggregation. The greater peak value of δ in the ML (0.25), greater VZ (1.8 m s−1 at 5.1 km MSL), and greater ML thickness (900 m) in case B indicate that the nonsphericity of ice hydrometeors was greater in case B, and hence supports the conclusion.

[36] Using VZand assuming the Marshall-Palmer distribution, median volume diameter of raindrops (hereafterD0) at the bottom of ML was estimated. D0 was 0.4 mm in case A and 1.1 mm in case B. The greater raindrop size in case B is consistent with the greater VZ above the ML in case B (Figures 4b and 8b). The thickness of ML determined by the altitude profiles of VZ was 300 m in case A and 900 m in case B. The thicker ML in case B is also consistent with the greater D0 at the bottom of ML.

[37] In the ML, the lidar measured the increase in δ caused by melting aggregates and decrease in lidar backscattered intensity (P) referred to as lidar dark band (Figures 4c, 4d, 8c, and 8d). The altitude and thickness of the increase in δ and those of lidar dark band were consistent with the thickness of ML determined by VZ (Figures 4b and 8b).

[38] δ of raindrops was 0.08–0.10 in case B, while it was close to zero (∼0.01) in case A (see Figures 4d and 8d). Stronger multiple scattering triggered by large-sized raindrops is likely a cause that generated the increase of volumeδ in case B. In case B, a dip of δ was measured at the bottom of ML (around 4.0 km MSL) (see Figure 8d). The decrease of nonsphericity at the final stage of melting and stronger multiple scattering for raindrops than melting hydrometeors are candidates that produced the dip.

[39] In order to carry out further studies, limitations in both observed precipitation cases and measurement means exist. A calibration means to measure radar reflectivity factor (Ze) is necessary to interpret precipitation nature in more detail. Recently Yamamoto et al. [2011] demonstrated that a combination of surface rainfall measurement and 24 GHz small weather radar named micro rain radar (MRR) [Meteorologische Messtechnik GmbH, 2005] is useful for Zecalibration. Volume-scanning weather radar is necessary for observing three-dimensional structure of precipitation system [e.g.,Lang and Rutledge, 2008; Steiner et al., 1995]. Because future observation using the EAR, lidar, MRR, and volume-scanning radar is necessary for further understanding of interactions betweenW and hydrometeor characteristics, efforts to realize the further observation are being made. DSD retrieval using Doppler spectra measured by the EAR would be useful to relate δ of raindrops to DSD. Further, accurate measurement of Ze and VZby 50 MHz wind-profiling radars would be useful to assess the relation between ice hydrometeor mass and fall velocity, because their wavelength of ∼6 m is able to neglect the non-Rayleigh scattering of hydrometeors which occurs atX band (8–12 GHz) or higher frequencies. Rajopadhyaya et al. [1994]suggested that Doppler spectra measured by 50 MHz wind-profiling radars are useful for retrieving drop size distribution of solid hydrometeors. Collocated lidar measurement ofδ would be useful to obtain information on the shape of solid hydrometeors. Simultaneous in situ measurement of hydrometeor shape and size using aircraft is also highly desirable.

Acknowledgments

[40] The Equatorial Atmosphere Radar (EAR) belongs to Research Institute for Sustainable Humanosphere (RISH), Kyoto University, and is operated by RISH and Indonesian National Institute of Aeronautics and Space (LAPAN). The EAR and lidar observations were operated as a part of the observation campaign named Cloud Observation Campaign by Lidar and the Equatorial Atmosphere Radar (CLEAR), which is supported by the Hydrometeorological Array for ISV-Monsoon Automonitoring (HARIMAU) project of the Japan EOS Promotion Program (JEPP) funded by the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan; Grant-in-Aid for Scientific Research (B) (19403008) funded by Japan Society for the Promotion of Science (JSPS); and Grant-in-Aid for Young Scientists (B) (19740293) funded by MEXT. The work was also financially supported by Grants-in-Aid for Scientific Research (B) (22403009 and 23340142) funded by JSPS and the research grant for Exploratory Research on Sustainable Humanosphere Science from RISH.