Seasonal and Diurnal Variations of Raindrop Size Distribution in ...

2 downloads 0 Views 833KB Size Report
Journal of the Meteorological Society of Japan, Vol. ... variations of raindrop size distribution (DSD) at Gadanki (GD), Singapore ...... use a power law, Z ¼ aRb.
Journal of the Meteorological Society of Japan, Vol. 84A, pp. 195--209, 2006

195

Seasonal and Diurnal Variations of Raindrop Size Distribution in Asian Monsoon Region

Toshiaki KOZU Faculty of Science and Engineering, Shimane University, Matsue, Japan

K. Krishna REDDY, Shuichi MORI Institute of Observational Research for Global Change (IORGC), Japan Agency for Marine-Earth Science and Technology (JAMSTEC), Yokosuka, Japan

Merhala THURAI Colorado State University, CO, U.S.A.

J. Teong ONG Nanyang Technological University, Singapore

D. Narayana RAO National Atmospheric Research Laboratory, India

and Toyoshi SHIMOMAI Faculty of Science and Engineering, Shimane University, Matsue, Japan (Manuscript received 31 October 2005, in final form 15 February 2006)

Abstract Diurnal and seasonal variations of raindrop size distribution (DSD) at Gadanki (GD), Singapore (SG) and Kototabang (KT) are studied to elucidate characteristics of DSD in the Asian monsoon region. It is found that DSDs are affected by diurnal convective cycles and seasonal variations in precipitation characteristics. GD has the most significant seasonal variation in DSD. Clear difference in rainfall characteristics between the Southwest and Northeast monsoon seasons is considered to be the main cause of such clear seasonal variation. KT has the most significant diurnal variation of DSD, which is probably caused by the fact that KT is greatly affected by ocean-land contrast and mountain effects to generate local convection in the afternoon. SG has less diurnal and seasonal variations compared with the other two locations, which is related to the fact that SG is affected both by land and oceanic rainfall. Z-R relations apCorresponding author: Toshiaki Kozu, Faculty of Science and Engineering, Shimane University, Matsue, 690-8504, Japan. E-mail: [email protected] ( 2006, Meteorological Society of Japan

196

Journal of the Meteorological Society of Japan

Vol. 84A

plicable to radar rainfall measurement in these areas are derived. It is shown that the use of the Marshall-Palmer Z-R relation (Z ¼ 200R 1:6 ) gives bias errors of about 1.5 dB or less in rain rate estimation except for the northeast monsoon season in GD, for 12@18 local time during pre-southwest monsoon season in GD, and for 06@12 local time during some monsoon seasons in KT.

1.

Introduction

The information of raindrop size distribution (DSD) is essential for obtaining relations between Integral Rainfall Parameters (IRPs) such as rain rate ðRÞ and the radar reflectivity factor ðZÞ, which is important for quantitative radar rainfall measurement. DSDs are often modeled with the gamma or lognormal functions, and the properties of DSD are described as the characterization of DSD model parameters. IRPs themselves and IRP relations can also be regarded as DSD information. Since it is generally difficult to directly measure DSD or IRP relations, efforts have been made to relate ‘‘macro-scale’’ environmental or rainfall properties with DSD or IRP relations and to clarify their diurnal and seasonal dependences. We define ‘‘macro-scale’’ here as environmental or rainfall properties having gradual space and/or temporal variability. Rainfall type (convective/stratiform) and rain structures (isolated/organized, shallow/deep, etc.) are among the ‘‘macro-scale’’ properties. Diurnal and seasonal variations and their climatological dependences of DSD or IRP relations would also be recognized as relations between ‘‘macroscale’’ parameters vs. DSD. Through this kind of DSD characterization processes, the accuracy in radar rainfall estimation may be improved. A number of researchers have attempted this type of approach since the beginning of radar meteorology (e.g., Stout and Mueller 1968; Battan 1973). In the 1990s revisit to characterizing DSD and/or IRP relations has been made for tropical rainfall in relation to rain estimation algorithms for spaceborne radars. Those studies include DSD variation dependent on wind patterns in Brazil (L’Ecuyer et al. 2003; Tokay et al. 2002), on season in south India (Reddy and Kozu 2003; Rao et al. 2001), and on year and rainfall type in Africa (Nzeukou et al. 2004). Tokay et al. (2001) discussed variability of radar estimates of rain rate due to DSD variation using disdrometer data obtained in tropics for the application to Tropical Rainfall Measuring Mission (TRMM) Precipitation Ra-

dar (PR) algorithms. In spite of such foregoing efforts, global distributions of DSD and Z-R relationships are still unclear. To overcome the difficulty of making direct measurement of DSD world wide, there are two approaches. One is to utilize satellite information itself to estimate DSD parameters as has been attempted in the course of TRMM PR and combined radar/radiometer algorithms (Iguchi et al. 2000; Haddad et al. 1997). However, this type of approach has difficulties to validate the DSD estimates. The other approach is to continue the effort to relate the macro-scale environmental rain properties with DSD as discussed above. We have been conducting rainfall observations at (i) Gadanki (GD), South India, (ii) Singapore (SG), and (iii) Kototabang (KT), West Sumatra, including disdrometer and radar measurements, see Reddy et al. (2002), Thurai et al. (2003) and Fukao (2006) for the GD, SG and KT cases, respectively. These sites, shown in Fig. 1, are in the Asian monsoon climate. In the case of KT and SG, near the equator, seasonal variation of background wind field is generally less clear (Okamoto et al. 2003), which is different from GD. Nevertheless, environmental conditions in Southwest (SW) and Northeast (NE) monsoon seasons are different as can be seen in rainfall and temperature variations (see Fig. 2). Such seasonal variations may be

Fig. 1. Location of Gadanki (GD), Singapore (SG) and Kototabang (KT).

July 2006

T. KOZU et al.

197 Table 1. Site parameters and observation periods at GD, SG and KT.

Site information

Period of disdrometer data

Gadanki (GD): 79.10 E, 13.47 N, 165 m ASL

Sep. ’97–Dec. ’97, Jan. ’99–Sept. ’01

Singapore (SG): 103.70 E, 1.38 N, 20 m ASL

Sep. ’94–May ’95, Aug. ’95–Oct. ’95, Dec. ’96–Apr. ’97, Oct. ’97–Nov. ’98

Kototabang (KT): 100.32 E, 0.20 S, 865 m ASL

Aug. ’01–Jul. ’03

DSD (Atlas and Ulbrich 1974; Kozu 1991). In the following sections, we will first outline seasonal and diurnal variations of DSD separately. Next, two-dimensional analysis of seasonaldiurnal variations of DSD is performed to obtain additional information for extracting the effects of local and large-scale rainfall conditions on DSD, taking into consideration the fact that the magnitude and phase of the diurnal cycle itself may have seasonal dependences. 2. Fig. 2. Seasonal variation of monthly rainfall and temperature in GD, SG and KT. Period of observation is 1999– 2000 (GD), average of 1982–2000 (SG rain), average of 1981–2000 (SG temperature) and May 2002–Apr. 2004 (KT). SG data are extracted from Chronological Scientific Tables (NAO, 2002).

overlapped with diurnal variation of precipitation. In this paper, we will present disdrometer data analyses to elucidate diurnal and seasonal variations of DSD, more specifically, results of statistical analyses of DSD parameters at the three different sites in the Asian monsoon region. Throughout this paper, we focus our main interest on characteristics of relatively large drop diameters, which are reflected in the behavior of higher order moments of DSD used to derive DSD parameters. This is related to the fact that IRPs for radar remote sensing and microwave communication applications are mainly proportional to 3 rd@6 th moments of

Observation sites and rainfall characteristics

Table 1 lists the location and the period of disdrometer data at GD, SG and KT. Joss RD69 disdrometers have been used at all of these sites. GD (79.10 E, 13.47 N) is located around 100 km inland of Chennai, India, and has a relatively dry climate. During the mid to late North-East (NE) monsoon season, i.e. January to April, there is little rainfall at GD. Conversely, SG (103.70 E, 1.38 N) is one of the most rainy regions in the world and there is rainfall throughout the year, although some rainfall peak exist in December due to monsoon. KT (100.32 E, 0.20 S) is located in a mountainous region near Padang, West Sumatra, Indonesia, and is affected by both local convective activities and monsoon. Although all sites are in the Asian monsoon climate, seasonal variations in ground temperature and rainfall are much different, as is shown in Fig. 2. Seasonal variations in ground temperature and rainfall are most significant in GD, and less significant in SG and KT. This is reasonable when we consider that GD is lo-

198

Journal of the Meteorological Society of Japan

Vol. 84A

Fig. 3. Seasonal and annual variation of TRMM PR derived (product 3A25) monthly mean storm height for convective rain in south India and around Singapore and Sumatra.

cated at 13.5 degrees of latitude, and SG and KT are close to the equator. This also suggests that GD has more seasonal variation in DSD characteristics than SG and KT. In South India, the convective storm top height extracted from TRMM PR 3A25 product peaks in March to May (7–9 km), and is at its lowest in November to December (5–6 km). Around SG and KT, the convective storm height from the 3A25 shows a similar seasonal variation to that in South India, but the variation is around half (7.5 to 6 km over the whole year), as is shown in Fig. 3. To describe the seasonal variation of DSD, we divide a year into four seasons; preSouthwest (pre-SW), Southwest (SW), preNortheast (pre-NE) and Northeast (NE) monsoon seasons (Reddy and Kozu 2003). These correspond to April–May, June–September, October–November, and December–March, respectively. Note that in SG and GD, zonal wind is generally weaker than 3 m/sec, and seasonal variation of wind is unclear (Okamoto et al. 2003), so the designations ‘‘northeast’’ and ‘‘southwest’’ do not necessarily correspond to actual wind direction. If rainfall is generated by local convective activities and has characteristics of precipitation

Fig. 4. Diurnal variation of rainfall during a 2-hour period for four monsoon seasons in GD, SG and KT.

over land, afternoon convection development would play a major role to form diurnal variation of rainfall. Conversely, large-scale cloud systems are expected to cause continuous or intermittent rainfall, making the diurnal cycle of rainfall less significant. Another fact affecting the diurnal cycle is that the phases of diurnal cycles for rain over ocean and rain over land are different (Takayabu 2002; Mori et al. 2004). Due to the diurnal cycle for oceanic rain having a peak around midnight to early morning, which is different from the afternoon peak over land, if an observation site is located along the coast line, rainfall characteristics would be affected both by oceanic and land rainfall, again causing a less significant diurnal cycle or ‘‘double peak’’ in rainfall. Figure 4 shows diurnal variations of rainfall in GD, SG and KT for

July 2006

T. KOZU et al.

199

four different monsoon seasons. In this figure, each rainfall indicates the yearly averaged rainfall amount during a given season and two-hour period for the period of disdrometer data shown in Table 1. From Fig. 4, it appears that diurnal variation and the afternoon peak of rainfall are most clear in KT, whereas diurnal variation in rainfall is unclear in SG although some double peaks are observed in the diurnal variation except for the SW monsoon season. This result indicates that rainfall in SG is affected by oceanic rainfall. Mori et al. (2004) examined the diurnal cycle of rainfall around Sumatra using TRMM satellite data. Their results from TRMM PR (Fig. 3 in Mori et al. 2004) indicate that rainfall in KT has a clear diurnal cycle involving an afternoon peak, whereas rainfall in SG has more complicated diurnal variation, affected by oceanic rainfall. In GD, diurnal variation of rainfall appears in general. However, the afternoon peak is less significant than in KT, and diurnal variation is not clear in the pre-NE and NE monsoon seasons, which is different from KT. 3.

Outline of DSD and Z-R relations

Before discussing the details of DSD characteristics, we outline the general feature of DSD. Considering that the most direct property relating to DSD for radar remote sensing is the Z-R relation, we examine Z-R relations as well as average DSDs. It is noted that the Z-R relation is used rather for practical viewpoint for radar rainfall measurements and mainly represents properties of intermediate to large drop diameters because R and Z are proportional to higher order DSD moments (Kozu 1991), while the average DSD represents overall DSD characteristics. Figure 5 shows average DSDs at the three sites for specific rain rates of 3 mm/h and 30 mm/h for the four monsoon seasons. For this processing, DSDs having a rain rate of 4.77 dBR (3 mm/h) G 0.5 dB or 14.77 dBR (30 mm/h) G 0.5 dB are extracted and used for the averaging. These rain rates are used as typical values representing ‘‘light’’ and ‘‘heavy’’ rain rates. As shown in these figures, the most significant seasonal variation is found in GD in which DSDs in the NE monsoon season are much narrower than other seasons. This distinct feature

Fig. 5. Averaged DSDs in four monsoon seasons around 3 mm/h and 30 mm/h ranges. In the legend, XX_YYnn, XX, YY and nn represent location, season and rain rate, respectively.

was also found in Reddy and Kozu (2003) and Rao et al. (2001). Conversely, DSDs in KT and SG have much less seasonal variation than in GD. For the 30 mm/h range, DSDs in SG are

200

Journal of the Meteorological Society of Japan

Vol. 84A

Table 2. Comparison of Z-R relations (Z ¼ aR b ) for four seasons at three locations, GD, SG and KT, derived from Joss-disdrometer data. Site

a

b

GD preSW

305.5

1.527

GD SW

264.3

1.468

GD preNE

179.3

1.559

GD NE

72.3

1.697

SG preSW

248.3

1.471

SG SW

285.1

1.385

SG preNE

261.1

1.471

SG NE

239.7

1.453

KT preSW

208.5

1.475

KT SW

199.0

1.508

KT preNE

167.6

1.548

KT NE

225.9

1.453

somewhat narrower in the SW monsoon season, and those in KT are somewhat broader in the pre-NE monsoon season than other seasons. They are not necessarily significant since these seasonal differences appear only in the range of log10 NðDÞ A 1 or less. In fact, there are no significant differences in Z-R relationships among different monsoon seasons in SG and in KT , as is shown in Table 2 in which Z-R relations for the four monsoon seasons at the three sites are summarized. The Z-R relation for the SW monsoon in SG is Z ¼ 285R 1:39 , whereas the averaged Z-R relation for other seasons is Z ¼ 250R 1:47 . The Z-R relation for the pre-NE monsoon in KT is Z ¼ 168R 1:55 , whereas averaged Z-R relation for other seasons is Z ¼ 211R 1:48 . To see the variability of DSD having a diurnal cycle, in Fig. 6, we plot the DSDs averaged over specific 6-hour intervals; 0–6, 6–12, 12–18 and 18–24 local hours. Diurnal variations of DSD are significant at the 30 mm/h range, especially in KT. SG has smaller diurnal variation than other sites, which is probably caused by the effect of the oceanic nature of rainfall in SG. In general, morning (00–12 LT) DSDs for the intense rain rate at the 30 mm/h range are

Fig. 6. Averaged DSDs for 0–6, 6–12, 12–18 and 18–24 LT around 3 mm/h and 30 mm/h ranges.

narrower than afternoon ones. Considering that the local convective cycle over land is expected to cause the peak rainfall activity in the afternoon, this result indicates that the rainfall

July 2006

T. KOZU et al.

associated with the local convection produces broader DSDs. We note here on the number of DSD samples to generate Figs. 5 and 6 and the variability of NðDÞ. The number of DSD samples is generally more than 30 (all cases for the 3 mm/h range and more than 70% of the cases for the 30 mm/h range) and the normalized standard deviation is typically 50% at each disdrometer channel for a given rain rate range (3 or 30 mm/h G 0.5 dB). Therefore, the normalized standard deviation of mean NðDÞ estimates would generally be about 10%. However, in the cases of the 30 mm/h range for the NE monsoon season and for the 0–6 LT period at GD, the number of samples is less than ten, which causes degraded accuracies of mean NðDÞ estimates (normalized standard deviation A 20– 30%). Nevertheless, qualitative discussions should still be valid on the seasonal and diurnal variation involving these cases, considering that the differences between the log10 NðDÞ of 30 mm/h range during the NE monsoon season and others (see Fig. 5(a)) and between the log10 NðDÞ of 30 mm/h range during 00–06 LT and those in the afternoon (see Fig. 6(a)) are about 5 dB or more for the diameter range of 3.5 to 4 mm. From the above discussions, it has become evident that DSD characteristics in GD are different from those in SG and KT in terms of seasonal variations, and that KT and GD have diurnal variations greater than in SG. However, these evidences are only from ‘‘marginal’’ distributions of DSDs as a function of season or local time, and there are some problems in the ‘‘quality’’ of average DSD when the number of DSD samples is small as mentioned above. In the following, we study the seasonal-diurnal variations of DSD in more detail, using DSD parameters instead of DSD itself to make the expression of DSD variation simpler. 4.

DSD parameters studied

Considering that the most direct DSD parameter for radar remote sensing is the Z-R relation, we use a DSD parameter DZMP (dB) defined by: DZMP ¼ dBZðmeasuredÞ 10 log10 ð200R 1:6 Þ ð1Þ where R is the rain rate. Since R is approxi-

201

mately proportional to the 3.67 th moment of DSD, M3:67 , i.e. R ¼ cR M3:67 (Atlas and Ulbrich 1977), we can express DZMP as follows (Kozu et al. 2005):   M6 DZMP ¼ 10 log10 aM3:67 R 0:6   M6 ¼ C þ 10 log10  6 log10 R M3:67 ¼ C þ 10 log10 DR2:33  0:6dBR

ð2Þ

where a ¼ 200cR , C ¼ 10 log10 a, and dBR ¼ 10 log10 R. The definition of DR2:33 is the D 3:67 weighted mean of D 2:33 . DZMP is expected to have similar characteristics as the median drop diameter D0 or mass-weighted mean diameter Dm 1 M4 /M3 except that the positive correlation between D0 or Dm and R is corrected with the term 0:6dBR. That is, DZMP has the characteristics useful for practical radar applications, and is a measure of ‘‘mean diameter’’ weighted to intermediate to large drop diameter range since it is derived from M6 and M3:67 . The other DSD parameter studied is the shape parameter m which is one of the parameters of the gamma DSD model: NðDÞ ¼ N0 D m eL D

ð3Þ

where D is drop diameter, NðDÞ is DSD, N0 and L are amplitude and scaling parameters of the gamma DSD model. The shape parameter m is derived from the 3 rd , 4 th and 6 th moments of DSD, M3 , M4 , and M6 respectively using the following equation (Kozu and Nakamura 1991): pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 11G  8 þ GðG þ 8Þ m¼ ð4Þ 2ð1  GÞ with G ¼ M43 /ðM32 M6 Þ:

ð5Þ

Note that xth moment of DSD, Mx is defined as ðy Mx ¼ D x NðDÞ dD: ð6Þ 0

The purpose to examine m is to investigate the effects of micro-physical processes on the shape of DSD. 5.

Seasonal-diurnal variation diagram of DSD

To study the details of seasonal-diurnal variations of DSD parameters, statistical process-

202

Journal of the Meteorological Society of Japan

Vol. 84A

ing for each 2-week by 2-local hour box is performed. Figure 7 shows 2D seasonal-diurnal diagrams of DZMP derived from disdrometer data at GD, SG and KT for which the average of DZMP for a 2-week by 2-local hour box for the rain rate ranges from 1 to 3 mm/h and from 10 to 30 mm/h. Although we do not employ radar data to classify rainfall types, 10 to 30 mm/h data would approximately represent characteristics of convective DSDs. Investigation of DSD characteristics with radar classification of rainfall type is a subject of further study. Note that positive and negative values of DZMP indicate broad and narrow DSDs respectively, in comparison with a DSD generating Z ¼ 200R 1:6 . General characteristics of DZMP obtained from Fig. 7 are as follows. (1) In GD, the seasonal variation of DZMP is evident for both the 1@3 mm/h and the 10@30 mm/h ranges. DZMP monotonically decreases from pre-SW to NE monsoon seasons. Although there is clear diurnal variation in rainfall especially during the preSW and SW monsoon seasons (see Fig. 4), Fig. 7. Day (Julian day) and local time plots of DZMP at 1–3 mm/h and 10–30 mm/h ranges. (a) GD 1–3 mm/h, (b) GD 10–30 mm/h, (c) SG 1–3 mm/h, (d) SG 10–30 mm/h, (e) KT 1–3 mm/h, (f ) KT 10–30 mm/h.

July 2006

T. KOZU et al.

203

Fig. 8. Seasonal variation of DZMP at 1–3 mm/h and 10–30 mm/h ranges for four local hour periods, 0–6, 6–12, 12–18 and 18–24 LT.

DZMP does not show such clear diurnal variation at the 1@3 mm/h range. At the 10@30 mm/h range, diurnal variation in DZMP is unknown because these intense rain events are infrequent in the morning. (2) In SG, the seasonal variation of DZMP is not clear for 1@3 mm/h or for 10@30 mm/h range. Diurnal variation is not clear for the 1@3 mm/h range. Some diurnal variations appear in the 10@30 mm/h range, which shows that DZMP in the afternoon is somewhat larger than that for the morning case. (3) In KT, diurnal variation of DZMP is evident for both 1@3 mm/h and 10@30 mm/h ranges. It appears that for the 1@3 mm/h range, DZMP increases slightly from early afternoon to late evening and midnight. Conversely, for the 10@30 mm/h range, early afternoon rains (12@14 LT) have larger DZMP values than other local hours.

This feature is less evident in the pre-NE and the early part of NE monsoon seasons (November–December). Considering that in the early afternoon, rains due to local convection are dominant, and that in the evening-to-midnight period, organized convective/stratiform rain systems over land to coast lines are dominant (Mori et al. 2004), the results suggest that the local convective rain and organized rain systems have different micro-physical processes. For more quantitative comparison of DZMP , and to show the seasonal-diurnal characteristics of the shape parameter m at GD, SG and KT, DZMP and m values averaged over each monsoon season and each 6-hour local time period are plotted in Figs. 8 and 9 as seasonal dependences of DZMP and m for 0@6, 6@12, 12@18 and 18@24 local hour intervals. A number of

204

Journal of the Meteorological Society of Japan

Vol. 84A

Fig. 9. The same as Fig. 8 but for m.

characteristics can be drawn from these figures, which are summarized in Table 3, showing clear differences in seasonal and diurnal characteristics at the three locations. In addition to Table 3, supplementary notes are given below: (1) In GD, significant seasonal variations are found in m as well as in DZMP . The m value increases from 6 to 7@8 (pre-SW to NE monsoon seasons), while DZMP shifts from around þ2 to 3 (pre-SW to NE monsoon seasons). Since higher order moments are used to derive m (Eqs. 4 and 5), the variation of m should be greatly affected by the behavior of raindrops of relatively large diameters. Therefore the large m values suggest that particle growth in large diameter drops is suppressed, which may be due to the weak convective activity in the NE monsoon season.

(2) In SG, no significant diurnal or seasonal variations are found in DZMP or m except for values in early afternoon (12@18 LT) at the 10@30 mm/h range (DZMP is 1 to 2 dB higher and m is slightly smaller than other local hours). This means that the DSD in SG is affected by the local convective cycle to some extent, which appears in the DSD characteristics at 12@18 LT. (3) In KT, diurnal variation is evident in m as well as in DZMP . The magnitude of diurnal variation in DZMP reaches about 3 to 4 dB in which morning (0@12 LT) DZMP is generally smaller than afternoon ones. At the 1@3 mm/h range, m in 6@12 LT is about 7 in comparison with other local times during which m is 5 to 6. Conversely, at the 10@30 mm/h range, the highest m occurs from midnight to early morning (0@6 LT). (4) From the comparison of Fig. 8 with Fig. 9,

July 2006

T. KOZU et al.

205

Table 3. Summary of seasonal and diurnal variation characteristics of DZMP and m at GD, SG and KT. (a) DZMP (b) m GD DZMP Seasonal 1–3 mm/h

10–30 mm/h

Large

Large

SG DZMP Seasonal 1–3 mm/h 10–30 mm/h

Small

Small

KT DZMP Seasonal 1–3 mm/h

10–30 mm/h

Some

Small

Diurnal

Remarks

No

The smallest at NE monsoon.

Some

The smallest at pre-NE and NE monsoons. Afternoon DZMP is somewhat large.

Diurnal Small

Remarks No significant seasonal dependence.

Afternoon DZMP Moderate is larger. Diurnal

Remarks

Large

Morning DZMP is small. Small diurnal variation in NE monsoon.

Large

Midnight to morning DZMP is small. No diurnal variation in NE monsoon.

we find that high (low) DZMP values correspond to low (high) m values. This means that large m values are given by the reduction of the number density of large raindrops rather than the depression of the number density of small raindrops. (5) DZMP for the 1@3 mm/h range is found to have small diurnal variation in GD and SG, while there is a clear diurnal variation in KT. In contrast, DZMP for the 10@30 mm/h range generally has large values in the afternoon (12–18 hours LT) at all the three sites. 6.

Discussions

From the above described analysis results, several points have become clear.

GD m

Diurnal

Remarks

Some

No clear systematic trend in diurnal variation.

Large

Some

No clear systematic trend in diurnal variation.

SG m

Seasonal

Diurnal

Remarks

1–3 mm/h

No

Small

m is stable around 5 to 6.

10–30 mm/h

No

Small

Afternoon m is slightly small.

KT m

Seasonal

Diurnal

1–3 mm/h

Small

Some

Morning m is clearly large.

10–30 mm/h

Small

Some

Midnight to early morning m is slightly large.

1–3 mm/h

10–30 mm/h

Seasonal

Large

Remarks

The first is the difference between equatorial regions (SG and KT) and the location at 13degrees north latitude (GD). GD should be more directly affected by seasonal change in the nature of rainfall. As discussed in Reddy and Kozu (2003), rainfalls in NE monsoon originate from the Bay of Bengal, while those in SW monsoon are dominated by intense local convection. This results in the seasonal change in rain top height as shown in Fig. 3. The clear difference in rainfall nature between the SW and NE monsoon seasons would directly affect the DSD properties, both in DZMP and m. Conversely, a local convective cycle is still active, even in the NE monsoon in KT and SG. Such local convection would cause broader DSDs during the NE monsoon season in KT and SG than in GD. The second is the difference between mountainous location near coastline (KT) and others (no high mountains around GD and SG). Clear diurnal variation of DSD parameters in KT is

206

Journal of the Meteorological Society of Japan

probably caused by the fact that KT is greatly affected by ocean-land contrast and mountain effects, which generates local convection in the afternoon (Mori et al. 2004). However, such clear diurnal variation becomes smaller in the NE monsoon season than other seasons, which is seen both in DZMP and in m. This is probably due to the fact that during the NE monsoon season, rainfall from large cloud systems is predominant, causing smaller diurnal variation not only in rainfall but in DSD parameters. A similar result has been found in intraseasonal variation of DSD (Kozu et al. 2005). The third is the difference between semioceanic location (SG) and others. In SG, both DZMP and m show small seasonal and diurnal variations in comparison with the other locations, although the afternoon peak in DZMP is still evident, indicating the existence of a local convection cycle to some extent. A mixture of land and oceanic rainfall characteristics, as seen in SG, may be found in other places having geographical conditions similar to SG. The degree of diurnal cycle may be determined by the degree of ocean/land mixture. The small seasonal variation in DSD would be related to the relatively constant rainfall over the year at SG, although there are some peaks in rainfall due to NE monsoon in November to December. It is reported from the disdrometer observation in the tropical West Pacific (Tokay and Short 1996) that there is a distinct difference in DSD between convective and stratiform rain types; the former clearly has narrower DSDs and the latter has broader DSDs. Conversely, Yuter and Houze (1997) who analyzed PMS probe data over the tropical Western Pacific reported that there was no clear distinction in DSDs between convective and stratiform rainfall types. Rao et al. (2001) examined rain-type classified Z-R relations at GD. Their result is Z ¼ 178R 1:51 (convective) and Z ¼ 251R 1:48 (stratiform). In GD and SG, a boundary layer radar (Reddy et al. 2002) and a vertical looking S-band rain radar (Thurai et al. 2003) have been operated, respectively. Using Doppler and Z-factor profiles, we have made an approximate convective and stratiform classification. It is found that in both GD and SG, the ratio of the number of stratiform rain cases to that of convective rain (S/C ratio) has its maximum in the middle of the NE monsoon, and its minimum in

Vol. 84A

the pre-SW monsoon. A similar trend has also been observed in Indonesia (Renggono et al. 2001). This information helps to detect seasonal variation of rainfall types and underlining environmental conditions. It appears that the S/C ratio has a clear correlation with DSD in GD, but not so clear in both SG and KT. In the former case, stratiform rain in the NE monsoon season is associated with oceanic cloud systems, causing relatively shallow, light rain. In contrast, in the case of SG and KT, a diurnal convective cycle is still evident, which would be reflected to broader DSD than in the NE monsoon season in GD. The difference between the tropical West Pacific study (Tokay and Short 1996) and the present study is probably due to the different characteristics involving oceanic and land convective clouds. 7.

Z-R relations

A usual way to express the relation between Z and R for radar rainfall measurement is to use a power law, Z ¼ aR b . As a summary of the above discussion on 2-D variations of DSD parameters, power-law fitted Z-R relations from principal component analysis of dBR and dBZ are shown in Fig. 10. A more detailed list of Z-R relations is given in Table 4. In Fig. 10,

Fig. 10. Z-R relations expressed as the dB difference from MP Z-R relation (dBZ ¼ 23 þ 1:6dBR) ðDZMP Þ for four monsoon seasons in GD, SG and KT. Representative Z-R relations during the specific 12@18 LT period in GD and 06@12 LT period in KT are also plotted.

July 2006

T. KOZU et al.

Table 4. Coefficient ðaÞ and exponent ðbÞ in Z ¼ aR b relation for four local time periods and four monsoon seasons in GD, SG, and KT, derived from Jossdisdrometer data. (a) Gadanki LT

00–06 a

b

06–12 a

b

12–18 a

b

18–24 a

b

pre-SW

293.3 1.553

246.5 1.503

362.0 1.516

334.0 1.498

SW

268.4 1.441

305.9 1.499

260.6 1.495

249.3 1.483

pre-NE

176.0 1.580

164.6 1.521

188.2 1.560

187.9 1.538

NE

54.78 1.783

110.2 1.782

77.32 1.655

66.21 1.773

06–12

12–18

18–24

(b) Singapore LT

00–06 a

b

a

b

a

b

a

b

pre-SW

191.0 1.444

268.6 1.424

260.4 1.507

289.8 1.376

SW

269.9 1.421

259.5 1.378

313.3 1.407

312.3 1.362

pre-NE

246.7 1.418

270.6 1.424

266.8 1.526

259.8 1.430

NE

244.8 1.416

235.3 1.372

239.3 1.475

242.7 1.434

06–12

12–18

18–24

(c) Kototabang LT

00–06 a

b

a

b

a

b

a

b

pre-SW

222.0 1.452

138.2 1.513

199.8 1.480

227.3 1.460

SW

239.9 1.375

112.5 1.585

185.2 1.547

235.7 1.446

pre-NE

175.1 1.613

110.9 1.506

133.5 1.593

204.0 1.539

NE

233.9 1.498

173.4 1.512

186.4 1.491

272.8 1.401

DZMP ¼ 0 dB represents that the Z-R relation is the same as the MP Z-R relation (200R 1:6 ), and positive (negative) DZMP represents that Z for a given R is greater (smaller) than the value given by 200R 1:6 . In addition to the Z-R relation for each season at each site, the Z-R relation for 06@12 LT during the pre-NE monsoon season in KT and that for 12@18 LT during the pre-SW monsoon season in GD are also plotted, which are significantly different from the MP Z-R relation. To keep the readability of Fig. 10, only Z-R relations for the limited cases are plotted from many Z-R relations for different monsoon seasons and local times because most of the other Z-R relations are relatively

207

close to the MP Z-R relation. It is found that variations in Z-R relation are at most G1.5 dB from the MP Z-R relation except for the NE monsoon in GD, for 12@18 LT during the preSW monsoon in GD, and several Z-R relations for 06@12 LT in KT. Since 06@12 LT is generally the ‘‘bottom’’ of the diurnal cycle of rainfall (see Fig. 4), the number of samples to generate Z-R relation is relatively small especially in KT and GD. Moreover rainfall during 06@12 LT may be a ‘‘leakage’’ of 00@06 or 12@18 LT rainfall, having different DSD characteristics. It has been found that the Z-R relation during the 06@12 LT period is somewhat unstable compared with other local time periods. Such leakage may be a cause of the instability of the Z-R relation during 06@12 LT. Finally we consider appropriate Z-R relations applicable to radar rainfall measurements in these Asian monsoon areas. From Fig. 10 and Table 4, we can conclude that the use of the MP Z-R relation for GD, SG and KT gives bias errors of about 1.5 dB or less in rain rate estimation except for the two cases: (1) the NE monsoon season and 12–18 LT during the preSW monsoon season in GD where seasonal variation of DSD is significant, and (2) 06@12 LT during the pre-SW, SW and pre-NE monsoon seasons in KT where diurnal variation of DSD is significant. If we use a ‘‘best-fit’’ single Z-R relation which may be an average of Z-R relations at all sites and local hours, the bias error will be reduced to about 1 dB. However, care has to be taken concerning the following two points. One is the possible large seasonal variation of DSD in the area where the origin of rainfall greatly changes depending on the season, such as at GD. Second is the possible large diurnal variation of DSD in the areas where diurnal variation of rainfall is significant. In the case of KT, morning DSD is generally narrower than at other local hours. If we can develop DSD or Z-R relation models for such season and local hours in addition to the single model for other locations, seasons and local times, overall estimation accuracy will be improved. This requires further study. Nevertheless, it is encouraging that we can use a single Z-R relation with a reasonable bias error for the three locations having different climatic characteristics within the Asian monsoon region.

208

8.

Journal of the Meteorological Society of Japan

Concluding remarks

Diurnal and seasonal variations of DSD in Gadanki (GD), Singapore (SG) and Kototabang (KT) have been studied to elucidate characteristics of DSD and Z-R relations in the Asian monsoon region. It is found that DSDs are affected by both diurnal convective cycle and seasonal variation of rainfall characteristics. GD has the most significant seasonal variation in DSD. The seasonal variation is evident both in DZMP and m, for light and heavy rain rate ranges. The clear difference in rainfall characteristics between SW and NE monsoon seasons would be the main cause of such clear seasonal variation. KT has the most significant diurnal variation of DSD. The diurnal variation is evident in both DZMP and m, for light and heavy rain rate ranges. Changes in the DSD parameters observed from early afternoon to midnight suggest that local convective rain and organized rain systems have different micro-physical processes. The clear diurnal variation in KT is probably caused by the fact that KT is greatly affected by ocean-land contrast and mountain effects to generate local convection in the afternoon. However, this feature is less evident in the early part of NE monsoon season (December). SG has less diurnal and seasonal variations of DSD compared with the other two locations, although there appear some diurnal variations in the heavy rain rate range in which early afternoon DZMP is somewhat larger than in the morning. These characteristics are related to the fact that SG is affected both by land and oceanic rainfall. It was shown that the use of the MP Z-R relation (Z ¼ 200R 1:6 ) for the three disdrometer sites gives bias errors of about 1.5 dB or less in rain rate estimation except for the two cases; (1) the NE monsoon season and 12–18 LT during the pre-SW monsoon season in GD, and (2) 06@12 LT during the pre-SW, SW and pre-NE monsoon seasons in KT. From the characteristics of DSD and Z-R relations at these sites, it was suggested that the Z-R relation may show significant seasonal variation in the area where the origin of rainfall and surrounding environment depend heavily on the season. A similar caution would need to be taken in the area

Vol. 84A

where there is a significant diurnal variation of rainfall. In this paper, rainfall type classification was not seriously examined since we thought that DSD characteristics at 1@3 mm/h and 10@30 mm/h ranges roughly represent characteristics of stratiform and convective DSDs. More detailed DSD study with rainfall type classification using radar data would be useful to study the relation between microphysical processes and DSD. In addition, the vertical structure of DSD needs to be studied further in order to elucidate the details of DSD evolution during raindrop falling. Acknowledgments We thank Profs. Y. Fujiyoshi and Y.N. Takayabu for valuable discussions, and Profs. S. Fukao and M. Yamamoto for managing the CPEA project. Disdrometer observation at Gadanki has been conducted by National Atmospheric Research Laboratory, India, and National Institute of Information and Communications Technology, Japan. This work is supported by Grant-in-Aid for Scientific Research on Priority Area-764 of the Ministry of Education, Culture, Sports, Science, and Technology of Japan (Grant No. 13136206), and by CREST of JST (Japan Science and Technology Corporation). References Atlas, D. and C.W. Ulbrich, 1974: The physical basis for attenuation-rainfall relationships and the measurement of rainfall parameters by combined attenuation and radar methods. Jour. Res. Atmos., 8, 275–298. ——— and ———, 1977: Path- and area-integrated rainfall measurement by microwave attenuation in the 1–3 cm band. Jour. Appl. Meteorol., 16, 1322–1331. Battan, L.J., 1973: Radar Observation of the Atmosphere, Univ. of Chicago Press, 324pp. Fukao, S., 2006: Coupling Processes in the Equatorial Atmosphere (CPEA): A project overview. J. Meteor. Soc. Japan, this issue. Haddad, Z.S., E.A. Smith, C.D. Kummerow, T. Iguchi, M.R. Farrar, S.L. Durden, M. Alves, and W.S. Olson, 1997: The TRMM ‘Day-1’ radar/ radiometer combined rain-profiling algorithm. J. Meteor. Soc. Japan, 75, (4), 799–809. Iguchi, T., T. Kozu, R. Meneghini, J. Awaka, and K. Okamoto, 2000: Rain profiling algorithm for the TRMM precipitation radar. J. Applied Me-

July 2006

T. KOZU et al.

teorology (TRMM special issue), 39, (12) Pt. 1, 2038–2052. Kozu, T., 1991: Estimation of raindrop size distribution from spaceborne radar measurement, Dr. E dissertation, submitted to Kyoto University, Kyoto, Japan, 196pp. ——— and K. Nakamura, 1991: Rainfall parameter estimation from dual-radar measurements combining reflectivity profile and pathintegrated attenuation. Jour. Atmos. Ocean. Tech., 8, 260–270. ———, T. Shimomai, Z. Akramin, Marzuki, Y. Shibagaki, H. Hashiguchi, 2005: Intraseasonal variation of raindrop size distribution at Kototabang, West Sumatra, Indonesia. Geophys. Res. Letters, 32, L07803, doi:10.1029/ 2004GL022340. L’Ecuyer, T.S., C. Kummerow, and H. Matsunaga, 2003: Relationships between the microphysics of precipitating cloud systems and their radar reflectivity patterns. 31st Conf. on Radar Meteorol., P3A.1, 399–402, Seattle. Mori, S., J. Hamada, I.T. Yudi, M.D. Yamanaka, N. Okamoto, F. Murata, N. Sakurai, H. Hashiguchi, and T. Sribimawati, 2004: Diurnal landsea rainfall peak migration over Sumatra Island, Indonesian maritime continent observed by TRMM satellite and intensive rawinsonde soundings. Mon. Wea. Rev., 132, 2021–2039. NAO, 2002: Chronological Scientific Tables, National Astronomical Observatory, Japan. Nzeukou, A., H. Sauvageot, A.D. Ochou, and C.M.F. Kebe, 2004: Raindrop size distribution and radar parameters at Cape Verde. J. Appl. Meteorol. 43, 90–105. Okamoto, N., M.D. Yamanaka, S. Ogino, H. Hashiguchi, N. Nishi, T. Sribimawati, and A. Numaguchi, 2003: Seasonal variations of tropospheric wind over Indonesia: Comparison between collected operational rawinsonde data and NCEP reanalysis for 1992–99. J. Meteor. Soc. Japan, 81, 829–850. Rao, T.N., D.N. Rao, and K. Mohan, 2001: Classification of tropical precipitating systems and associated Z-R relationships. Jour. Geophys. Res., 106, No. D16, 17,699–17,711. Reddy, K.K., T. Kozu, Y. Ohno, K. Nakamura, A. Higuchi, K.M.C. Reddy, V.K. Anandan, P. Srinibasulu, A.R. Jain, P.B. Rao, R.R. Rao, G. Vis-

209

wanathan, and D.N. Rao, 2002: Planetary boundary layer and precipitation studies using lower atmospheric wind profiler over tropical India. Radio Science, 37, (4), 14-1–14-17. ——— and ———, 2003: Measurements of raindrop size distribution over Gadanki during southwest and north-east monsoon. Indian J. of Radio & Space Physics, 32, 286–295. Renggono, F., H. Hashiguchi, S. Fukao, M.D. Yamanaka, S.-Y. Ogino, N. Okamoto, F. Murata, B.P. Sitorus, M. Kudsy, M. Kartasasmita, and G. Ibrahim, 2001: Precipitating clouds observed by 1.3 GHz boundary layer radars in equatorial Indonesia. Annales Geophysicae, 19, 889–897. Stout, G.E. and E.A. Mueller, 1968: Survey of relationships between rainfall rate and radar reflectivity in the measurement of precipitation. J. Appl. Meteorol., 7, 465–474. Takayabu, Y.N., 2002: Spectral representation of rain profiles and diurnal variations observed with TRMM PR over the equatorial area. Geophys. Res. Lett., 29, 1584, doi:10.1029/ 2001GL014113. Thurai, M., T. Iguchi, T. Kozu, J.D. Eastment, C.L. Wilson, and J.T. Ong, 2003: Radar Observations in Singapore and their implications for the TRMM precipitation radar retrieval algorithms. Radio Science, 38, (5), 1086, doi:1029/ 2002RS002855. Tokay, A. and D.A. Short, 1996: Evidence from tropical raindrop spectra of the origin of rain from stratiform and convective clouds. Jour. Appl. Meteorol., 35, 355–371. ———, R. Meneghini, J. Kwiatkowski, E. Amitai, T. Kozu, T. Iguchi, C. Williams, M. Kulie, and C. Wilson, 2001: On the role of dropsize distribution in the TRMM rain profiling algorithm. 30th Int’l Conf. on Radar Meteorol., 6.6, 345– 347, Munich, Germany, July. ———, A. Kruger, W.F. Krajewski, P.A. Kucera, and A.J.P. Filho, 2002: Measurements of drop size distribution in the southwestern Amazon basin. J. Geophys. Res., 107, (D20), 8052, doi:10.1029/2001JD000355. Yuter, S.E. and R.A. Houze Jr., 1997: Measurements of raindrop size distributions over the Pacific warm pool and implications for Z-R relations. J. Appl. Meteorol., 36, 847–867.