Statistics of Drop Size Distribution Parameters and Rain Rates for Stratiform and Convective Precipitation during the North Australian Wet Season

Transcription

1 3222 M O N T H L Y W E A T H E R R E V I E W VOLUME 141 Statistics of Drop Size Distribution Parameters an Rain Rates for Stratiform an Convective Precipitation uring the North Australian Wet Season GUILLAUME PENIDE Laboratoire Optique Atmospherique, UMR CNRS 8518, Universite es Sciences et Technologies e Lille, Villeneuve Ascq, France VICKAL V. KUMAR Centre for Australian Weather an Climate Research,* an School of Mathematical Sciences, Monash University, Melbourne, Victoria, Australia ALAIN PROTAT AND PETER T. MAY Centre for Australian Weather an Climate Research,* Melbourne, Victoria, Australia (Manuscript receive 10 September 2012, in final form 18 February 2013) ABSTRACT C-ban polarimetric raar measurements spanning two wet seasons are use to stuy the effects of the large-scale environment on the statistical properties of stratiform an convective rainfall aroun Darwin, Australia. The rainfall physical properties presente herein are the reflectivity fiels, aily rainfall accumulations an raining area, rain rates, an rop size istribution (DSD) parameters (meian volume iameter an normalize intercept parameter). Each of these properties is then analyze accoring to five ifferent atmospheric regimes an further separate into stratiform or convective rain categories following a DSDbase approach. The regimes, objectively ientifie by raiosone thermoynamic an win measurements, represent typical wet-season atmospheric conitions: the active monsoon regime, the break perios, the builup regime, the trae win regime, an a mixture of inactive/break perios. The large-scale context is foun to strongly moulate rainfall an clou microphysical properties. For example, uring the active monsoon regime, the aily rain accumulation is higher than in the other regimes, while this regime is associate with the lowest rain rates. Precipitation in this active monsoon regime is foun to be wiesprea an mainly compose of small particles in high concentration compare to the other regimes. Vertical profiles of reflectivity an DSD parameters suggest that warm rain processes are ominant uring this regime. In contrast, rainfall properties in the rier regimes (trae win/builup regimes) are mostly of continental origin, with rain rates higher than in the moister regimes. In these rier regimes, precipitation is mainly forme of large rainrops in relatively low concentration ue to a larger contribution of the ice microphysical processes on the rainfall formation. 1. Introuction It has been recently shown (Stephens et al. 2010) that the time-integrate accumulations of precipitation prouce by global climate moels closely match * A partnership between the Bureau of Meteorology an the Commonwealth Scientific an Inustrial Research Organisation. Corresponing author aress: Guillaume Penie, Laboratoire Optique Atmospherique, UMR CNRS 8518, Universite Lille1, Villeneuve Ascq CEDEX, France. guillaume.penie@univ-lille1.fr observations when they are globally composite (Stephens et al. 2010). However, this goo match is prouce by global moels through a compensation of errors, with moels proucing precipitation approximately twice as often as that observe an making rainfall far too light (Stephens et al. 2010). Thus, global moels are unable to represent the spatial variability of the precipitation frequency an intensity (Sun et al. 2006). But, if we are to better preict how precipitation patterns might vary in a changing climate, global moels must correctly preict both the frequency of occurrence an the instantaneous intensity of rainfall in the present climate. DOI: /MWR-D Ó 2013 American Meteorological Society

2 SEPTEMBER 2013 P E N I D E E T A L Precipitation patterns an their associate physical processes remain ifficult to characterize at ifferent scales because of their large spatial an temporal variability. Thus, long-term observations from active an passive remote sensing instruments such as raars, liars, an raiometers on the groun or on boar satellites represent key atasets that can help improve our unerstaning of clou an rainfall properties. The unerlying motivation of this work is to contribute to the improvement of climate moels an satellitebase retrievals such as the Tropical Rainfall Measuring Mission (TRMM; Kummerow et al. 1998) an the future Global Precipitation Measurement (GPM) project by proviing an observational basis to assess their outputs an, in a secon step, to help constrain or buil new parameterizations of microphysics for numerical moels that woul account for the large-scale context [as propose, for instance, as a new framework for parameterization evelopments in Jakob (2010)]. The approach consiere in the present paper, builing on the prior stuies of Bringi et al. (2009) an Thurai et al. (2010), uses long time series of observations from the Darwin (Northern Territory, Australia) ual-polarize C-ban raar to etermine the effect of the large-scale atmospheric environment on the rainfall properties. The unerlying question here is: How variable are the rainfall physical properties as a function of the large-scale context? Moreover, this work aims to complement the stuies of Protat et al. (2011) on the variability of tropical ice clou properties (ice water content, visible extinction, effective raius, total concentration) an Kumar et al. (2013) on convective clou characteristics (life cycle, clou-top height, electrical activity, cell volume). These stuies make use of the same large-scale regime efinitions an the same observational perios [as efine in Pope et al. (2009)] in orer to buil a comprehensive test be for large-scale moel verification of clou an precipitation properties an also to help improve clou resolving moel (CRM) microphysical parameterizations. To achieve this, the ientification of physically base large-scale regimes is of primary importance, as the synoptic environment inherently rives clou an rainfall properties at smaller scales. Various large-scale environment efinitions coul be use for this kin of stuy (Protat et al. 2011), such as the convective moulation over the tropics efine by the Maen Julian oscillation (MJO; Maen an Julian 1972) inex (Wheeler an Henon 2004) or the clou regimes in the tropical western Pacific as efine by Jakob an Tseliouis (2003) within the framework of the International Satellite Clou Climatology Project (ISCCP; Rossow an Schiffer 1991). But, as the aim of this work is to provie small-scale statistics useful for the evaluation of numerical moels an satellite-base retrievals, as well as to inform parameterization evelopment, there is a nee to link the rainfall properties to large-scale atmospheric regimes ientifie with thermoynamical (temperature, ewpoint temperature) an kinematic (horizontal wins) profiles. Inee, these basic variables, which can be easily measure with raiosones an simulate in numerical atmospheric moels, provie a strong founation to buil an compare long-term statistics. Statistics of rop size istribution (DSD) parameters an rain rates retrieve from C-ban polarimetric raar measurements over Darwin uring two consecutive wet seasons (October April 2005/06 an 2006/07) are analyze. DSD parameters were retrieve following the technique escribe in Bringi et al. (2009) an further separate accoring to five large-scale regimes an two ifferent rain types: stratiform an convective. The ataset use herein is larger than in the previous stuies of Bringi et al. (2009) an Thurai et al. (2010), which use only a few ays to stuy the DSD variability as a function of only two regimes (builup an active monsoon). 2. Methoology a. C-PoL raar an DSD retrievals C-Pol (Keenan et al. 1998) is a C-ban (5.5 GHz) ualpolarization raar locate near Darwin. The C-Pol raar performs a volumetric scan every 10 min within a 150-km scan raius, using 15 elevation angles (from 0.58 to 43.18) an a range resolution of 300 m. It transmits an receives linear vertical an horizontal polarizations, which give access to key polarimetric variables such as the horizontal reflectivity Z h, the ifferential reflectivity Z r, an the specific ifferential phase K p, which are the inputs into the DSD retrieval algorithm of Bringi et al. (2009) that is use in the present stuy. In this metho, DSD parameters an rain rates are retrieve on each plan position inicator (PPI) scan using a normalize gamma DSD functional form (Testu et al. 2001) escribe by the meian volume iameter D 0 an the generalize intercept parameter N w. In this stuy, N w is the same as the intercept parameter of an exponential DSD with the same D 0 an liqui water content as the gamma DSD. This algorithm uses a multiparameter approach to take avantage of the complementary information containe in the polarize backscattere signals an has been evelope using 6 months of isrometer ata. First, Z h an Z r are correcte for attenuation using the ZPHI metho [correction of Z h using the ifferential propagation phase F p (Testu et al. 2000; Bringi et al. 2001)] an the Tan et al. (1995) approach, respectively. The thir input variable, K p,is

3 3224 M O N T H L Y W E A T H E R R E V I E W VOLUME 141 FIG. 1. Location of the C-Pol raar (iamon) near Darwin (square) an of the analyze area (ashe circles elimitating the [20 120] km aroun the raar center). erive from the measure ifferential propagation phase F p using the finite impulse response (FIR) range filter (Hubbert an Bringi 1995). Note that as a result of large gate-to-gate fluctuations, Z r was also treate using the FIR range filter. Then, D 0 is retrieve from the ifferential reflectivity using polynomial fits [i.e., D 0 5 f (Z r )], N w is then estimate using a power law from both Z h an D 0 [N w 5 f (Z h, D 0 )], an finally the rain rate is estimate using a function of the form R 5 f (K p ), R 5 f (Z h, Z r ), or R 5 f (Z h ), epening on various threshols an a ecision tree (Bringi et al. 2009). Constant-altitue plan position inicators (CAPPIs) at an altitue of 2.5 km are use so that each analysis is realize at the same altitue. This avois biases from microphysical properties, which vary with height, such as those introuce by the evaporation coalescence breakup of rainrops. Each CAPPI is horizontally efine within a raius of 140 km an a horizontal resolution of 2.5 km. To avoi any issues that might appear uring the interpolation from PPIs to CAPPIs, only the values in the range km are analyze, which correspons to a total coverage area of km 2, as shown in Fig. 1. b. Convective stratiform separation Many stuies have been evote to the classification of rain types into stratiform an convective parts using rain gauges an/or raar ata (Williams et al. 1995; Steiner et al. 1995; Tokay an Short 1996; Biggerstaff an Listemaa 2000; Ulbrich an Atlas 2007; Bringi et al. 2009; Thurai et al. 2010) or visible infrare microwave satellite ata (Aler an Negri 1988; Anagnostou an Kummerow 1997; Hong et al. 1999). Convective an stratiform parts of clou systems show significant ifferences in terms of ynamics an thus in terms of microphysics, so it is obvious that ifferent rain types have to be treate separately in orer to extract consistent conclusions. The approach consiere in the present paper was to use a DSD-base separation technique as escribe in Bringi et al. (2009). They introuce a thir class of precipitation, efine as mixe rain or transition rain forme from ecaying convective cells that have enough microphysical ifferences from purely stratiform an convective rain types to be consiere as a (self-consistent) separate rain class (Williams et al. 1995). They showe that the inherent microphysical ifferences within the three regimes can be easily ientifie in the log 10 (N w ) 2 D 0 space using a simple linear function (separator criterion) given by log 10 (Nw sep ) 521:6D 0 1 6:3, (1) where D 0 is in units of millimeters, N w is in units of per millimeters per meter cube, an the superscript sep

4 SEPTEMBER 2013 P E N I D E E T A L FIG. 2. Two-year mean profiles of raiosone measurements of (a) horizontal wins an (b) relative humiity, for the five wet-season regimes. stans for the separator. From this separator criterion, a simple inex i efine as the ifference between the retrieve log 10 (Nw CPOL ) an Eq. (1) is calculate: i 5 log 10 (N CPOL w ) 2 log 10 (N sep ). (2) As a result, large positive values of i refer to convective regions, large negative values of i represent stratiform regions, an low-magnitue values of i, both positive an negative, inicate transition regions. Bringi et al. (2009) have also compare their classification metho with the nonpolarimetric Steiner et al. (1995) approach. Bringi et al. showe by using probability ensity function (PDF) comparisons of the two techniques that the DSD-base approach allows a more constraine efinition of the convective regions without any small values of rain rates, D 0, an N w (smaller than 10 mm h 21, 0.7 mm, an 3.2, respectively). They conclue that this approach seeme more realistic in terms of the efinition of convection, an that stratiform regions were as accurately ientifie as with the texturebase metho with the aition of a smoother transition between the two regimes via the introuction of the mixe rain class. As the mixe or uncertain category represents only 3% of the ataset, only purely stratiform an purely convective precipitation categories w have been use in the following. In aition, two-sample Kolmogorov Smirnov tests have been realize on every pair of PDFs presente herein in orer to test the inepenency an the statistical significance of the results. The test gives 0 if the two PDFs being teste are totally inepenent an 1 if the two samples come from the same istribution. We foun that the maximum value of all the tests that we realize is about 10 26, which inicates that all the PDFs are clearly inepenent an that all the comparisons between the PDFs are statistically significant. c. Thermoynamic an kinematic large-scale regimes Pope et al. (2009) performe a cluster analysis on win an thermoynamical profiles from aily raiosones (2300 UTC) launche at Darwin corresponing to 49 wet seasons between 1957/58 an 2005/06. They efine five atmospheric regimes corresponing to ifferent horizontal win, temperature, an ewpoint temperature profiles. The mean profiles of horizontal win an relative humiity corresponing to the five atmospheric regimes within our two-wet season ataset are shown in Figs. 2a an 2b, respectively. Aitionally, Table 1 summarizes the number of points an the occurrence corresponing to each of the five large-scale regimes, which are then separate into stratiform an convective

5 3226 M O N T H L Y W E A T H E R R E V I E W VOLUME 141 TABLE 1. Number of ays an occurrences of the five large-scale regimes, an the corresponing stratiform an convective number of points an occurrence use to buil the statistics. Regimes Dry east Deep west East Shallow west Moist east No. of ays (%) 37 (10.3) 64 (17.8) 25 (6.9) 59 (16.4) 175 (48.6) No. of points (%) Stratiform (88.7) (91.9) (89.2) (92.6) (91.4) Convective (8.1) (5.0) (7.6) (4.8) (5.5) precipitation. The following main features of the five regimes are ientifie: Dry easterly regime (DE) This regime is characterize by rier than normal southeasterly wins from the surface up to the mitroposphere (7 8 km) an weak northwesterlies above 8 km. As relative humiity ecreases ramatically with height compare to the other regimes, the DE regime is consiere to be the riest of the five regimes an can be viewe as a trae win regime. This regime occurs on about 10% of the ays in our ataset (Table 1). Deep westerly regime (DW) This regime is compose of northwesterly wins below 7 8 km associate with high humiity an southeasterly wins above 8 km. This regime is ientifie with the typical active monsoon regime at Darwin (Drosowsky 1996) an represents about 18% of the ays. Easterly regime (E) This regime represents a transition environment (builup) between the trae wins (DE regime) an the active monsoon (DW regime) (cf. Keenan an Carbone 1992). It has almost the same horizontal win profile (with lower magnitues) as the DE regime but is characterize by a moister profile. This regime represents about 7% of the ataset. Shallow westerly regime (SW) This regime is associate with a weaker monsoon circulation than occurs uring the DW regime ue to an eastwar propagation of the MJO (Pope et al. 2009). It is compose of shallow southwesterly wins below 2 km an weak southeasterly wins above with an intermeiate moisture profile. This regime can be seen as a mixe inactive an break monsoon regime with a relative occurrence of about 16% uring the wet season. Moist easterly regime (ME) This regime is characterize by weak easterlies throughout the entire atmosphere associate with higher humiity profiles than the other easterly regimes. The ME regime correspons to break monsoon conitions an occurs on about half of the ays uring the north Australian wet season. To illustrate the variability an length of each regime, the temporal evolution of the regime classification is shown in Fig. 3. One can see that the active monsoon regime (DW) an the break perios (ME) are the only two regimes that, once establishe, ten to last several FIG. 3. Temporal evolution of the five large-scale regimes uring the two wet seasons (2005/06 an 2006/07).

6 SEPTEMBER 2013 P E N I D E E T A L FIG. 4. Bar graph of aily total raar omain rainfall (left bars labele R ) an average raining area (right bars labele A ) within 120 km of the raar using all ata an separately for the five large-scale atmospheric regimes. Amounts that are convective in nature are shown in a arker shae an the remaining lighter shaes represent the stratiform contributions. The stratiform contributions have also been expresse as the percentage of the total. ays. Except uring the first two months (October November) of the 2006/07 wet season, the DE regime is very sparse. The E regime is infrequently represente (only about 7%) an mainly appears between two break perios. Finally, the SW regime has an intermeiate pattern of behavior since it tens to last for a few ays once establishe an it almost always appears after the DW regime, which is consistent with the hypothesis that the SW an DW regimes are connecte with the eastwar propagation of the MJO (Pope et al. 2009). 3. Results In this section we present an overview of the stratiform an convective rainfall properties as a function of the five Darwin wet season regimes. First, the general patterns of the stratiform an convective precipitation are iscusse. Next, the vertical structure of the reflectivity fiels an DSD parameters will be analyze. Finally, the variability of the DSD parameters at 2.5-km height will be presente. a. Overview of stratiform convective precipitation uring the five Darwin wet-season regimes Figures 4 an 5 summarize the general patterns of stratiform an convective precipitation corresponing to the five large-scale regimes. Figure 4 shows the aily average rainfall an raining area retrieve from C-Pol measurements, an Fig. 5 illustrates the rainfall spatial variability. It is clear from Fig. 4 that the active monsoon regime (DW) correspons to the largest precipitation area an the highest aily rainfall accumulation. The DE an E are the riest regimes. The SW an ME regimes are quite similar in terms of rainfall accumulation an raining area with magnitues closer to the DW regime than to the DE an E regimes. A main ifference between the regimes is the convective contribution to the total rainfall, which is only 49% within the DW regime (lowest percentage) while it reaches 65% an 69% within the E an DE regimes, respectively. In aition, the convective contribution to the total raining area varies from 8% (DW, SW, an ME) up to 13% for the DE regime. More stratiform rain is prouce uring the DW regime as compare to the other regimes (51% versus 48% for ME) ue to longer-live anvil clous generate by eep convection (May an Ballinger 2007; Protat et al. 2011; Kumar et al. 2013). The active monsoon regime shows a wiesprea spatial variability over the whole raar scan (Fig. 5) with clear evience of oceanic an coastal influences (e.g., Keenan an Carbone 1992). The oceanic convection is usually embee within a clean maritime environment associate with high concentrations of coarse- an fine-moe sulfate aerosols (Allen et al. 2008), while the coastal influence is mainly ue to the lan- an sea-breeze convergence aroun Darwin. Within all the easterly regimes, large aily rainfall accumulations are locate on the western part of the Tiwi Islans, off the coast of northern Australia (Fig. 5). This particular feature of the easterly regimes can be attribute to the strong interactions between sea breezes an col pools that often evolve in mesoscale convective systems (MCSs; a.k.a. Hectors) when gust fronts amplify the convection along the zonal breeze fronts (Carbone et al. 2000). During the ry easterly regime, convective precipitation is mainly present over the Tiwi Islans with only a weak coastal influence while the E regime exhibits a much more wiesprea area of coverage with a noticeable coastal influence. The ME regime also exhibits seconary maxima over the inlan sector an over the Australian western coast. Finally, precipitation within the SW regime has no real preferential geographical location (Fig. 5). During this regime, the accumulate rainfall an the convective contribution to rainfall are slightly larger than average (54% versus 52%; Fig. 4). b. Vertical structure of stratiform convective precipitation uring the five Darwin wet-season regimes To further analyze the regime epenence of rainfall parameters in terms of microphysical processes, vertical information is neee. Figure 6 shows the heightepenent probability ensity functions of reflectivities [hereafter HPDF, as efine in Protat et al. (2010)], an Fig. 7 shows the mean profiles of D 0 an N w together with, at each CAPPI level, a bar representing the interquartile range (i.e., the range within which 50% of

7 3228 M O N T H L Y W E A T H E R R E V I E W VOLUME 141 FIG. 5. Daily rainfall accumulations uring the five ifferent large-scale atmospheric regimes an separately for (left) stratiform an (right) convective precipitation. The total area is shown in pixels of 2.5 km km.

8 SEPTEMBER 2013 P E N I D E E T A L FIG. 6. HPDFs of (a) stratiform reflectivity an (b) convective reflectivity (% BZ 21 km 21 ); together with the HPDFs of reflectivity anomalies within the five regimes (e.g., regime total, %BZ 21 km 21 ). (left) The stratiform precipitation an (right) convective rainfall. (top) White lines in the HPDFs of reflectivity anomalies represent the moe values corresponing to the total HPDFs. From top to bottom, regimes are organize as follow: DE, E, DW, SW, an ME.

9 3230 M O N T H L Y W E A T H E R R E V I E W VOLUME 141 FIG. 7. Mean vertical profiles of (a),(b) D 0 an (c),() log 10 (N w ) for the five large-scale regimes an separate into (a),(c) stratiform an (b),() convective precipitation. The shae area represents the whole ataset. At each CAPPI level (i.e., every 500 m), bars representing the interquartile range for each regime are also represente. The interquartile range represents the ifference between the thir an the first quartiles (i.e., the range within which are locate 50% of the population aroun the meian value). the population aroun the meian value is locate). Two kins of HPDFs are presente in Fig. 6; the top two HPDFs are built consiering all the stratiform an convective points (Figs. 6a an 6b, respectively), an the remaining panels represent the HPDFs of reflectivity anomalies (i.e., regime 2 total) corresponing to the five regimes an separate in stratiform (left-han sie in Fig. 6) an convective precipitation (right-han sie). Figures 6a an 6b exhibit very common features relative to the stratiform an convective precipitation (Yuter an Houze 1995). The HPDF of stratiform precipitation (Fig. 6a) shows a slow increase in the reflectivities with ecreasing height from 14 to 5 km (the approximate altitue of the 08C isotherm), followe by a slight ecrease in liqui phase towar the surface. The increase in the reflectivities from 14 to 5 km can be explaine by the growth processes of ice particles, such as the aggregation an iffusional growth processes, which are preponerant in stratiform clous (e.g., Houghton 1968; Yuter an Houze 1995; Houze 1997; McFarquhar et al. 2007; Penie et al. 2010; Protat et al. 2011). The ecreasing reflectivities from the melting level towar the surface can

10 SEPTEMBER 2013 P E N I D E E T A L be interprete as a combination of the evaporation an breakup processes, which ten to ecrease the particle iameters an thus the reflectivities. Moreover, Figs. 7a an 7c clearly show that, when consiering all the ata (shae area), there is a quasi-constant D 0 value from 4 to 2 km associate with increasing values of N w, which can be interprete as a competition between coalescence an breakup leaing to an increase in the number of both small an large particles that coul explain the constant meian volume iameter (hyrometeors such as aggregates, once melte, can lea to large an unstable rops). In aition, the slight increase in LWC (not shown) from 4 to 2 km can be explaine by the growth of the particles in a humi environment. Inee, at such altitues aitional conensational growth an/or accretion of clou water may contribute to the increase in mass. These processes are sensitive to the local ynamics an thermoynamics as well as to the DSD shape, given that small roplets evaporate first an that breakup is mainly relate to the larger rainrops. In contrast, the HPDF of the convective precipitation (Fig. 6b) exhibits a ifferent pattern of vertical behavior. There is a steeper increase in the reflectivity moe from 14 km own to 4-km altitue (compare to the stratiform HPDF; Fig. 6a), followe by a quasi-constant moe of reflectivity towar the surface (aroun 42 BZ). The faster increase of the convective reflectivity moe with ecreasing height is ue to the ifference between the growth processes involve. Convective clous have larger uprafts than o stratiform clous, resulting in ifferent ice crystal microphysics notably via the activation of the riming process. Usually, large rime hyrometeors such as hail or graupel are create exclusively within the convective parts of tropical thunerstorms an mainly above the melting layer where there is supercoole water available (e.g., Lulam 1950; Houghton 1968; Musil 1970; Browning et al. 1976; Houze 2004). Reflectivities associate with these types of hyrometeors increase rapily with ecreasing height, which explains the ifferent slope foun between the stratiform an convective regions at these levels. Once create, these large an ense hyrometeors fall rapily an thus take time to melt after they fall through the 08C isotherm so that they are not associate with a well-efine melting layer (bright ban). Thus, the reflectivity maximum is observe between 2 an 4 km (i.e., below the 08C level) ue to the presence of both melting hail (an/or graupel) an large rainrops (coming from melte ice crystals) that are not as affecte by the evaporation process as stratiform precipitation, but where the collisional processes (breakup an coalescence) ominate (Hu an Srivastava 1995). The increase of the HPDF with close to the groun is ue to the cumulative effect of both large rainrops breakup an coalescence processes, which have opposite consequences on the DSD parameters (i.e., the coalescence process increases D 0 an ecreases N w, whereas breakup ecreases D 0 an increases N w ). Therefore, a wie range of possible reflectivity values can be observe accoring to the ifferent relative efficiencies of these two processes within the 2-yr ataset. Nevertheless, the vertical profiles of convective D 0 an N w (Figs. 7b an 7) show that, within all the regimes, there is a slow ecrease of D 0 from 4 to 3 km associate with a large increase in N w, which is the signature of the breakup process. But, as D 0 ecreases faster between 2 km an the surface, the breakup process between 3 an 4 km must be in competition with the coalescence process that has the opposite effect on the DSD parameters, as explaine previously. Then, from 3 km to the surface there is a steeper ecrease of D 0 associate with a slightly increasing N w for DW, ME, an SW, while it is almost constant for DE an E. This may be interprete as being a competition between an enhance breakup process (an/or the activation of the evaporation process in aition to the breakup one, but this is less probable in convective precipitation) that is compensate, at least in terms of N w, by the coalescence process (an/or, to a lesser extent, by the evaporation process). The HPDFs of reflectivity anomalies show that there are some funamental ifferences between the five regimes. For example, uring the DE an E regimes, convection is rare but very intense (Figs. 6 an 6f) an correspons to larger than average reflectivities in the lower levels (between 40 an 50 BZ). These large reflectivity values are correlate with the larger relative occurrence of high-level ice clous with reflectivities in the BZ range that are mainly generate by eep convection (Protat et al. 2011; Kumar et al. 2013). These large reflectivities in the DE an E regimes are also associate with the largest mean meian volume iameters (Fig. 7b) an with the largest rain rates (see the tails of the black an re PDFs of rain rates in Fig. 8b), with averages of 33.4 an 32.1 mm h 21, respectively. These two regimes also generate very ifferent signatures in the stratiform part (compare to the other regimes) in response to ifferent convective parts. This inclues a much wier istribution of the reflectivities (Figs. 6c an 6e) in the ice-only region (positive anomalies on both sies of a negative anomaly inicate a broaer istribution at these levels) as well as aitional structures with, notably, a strong bimoality (two maxima at a given altitue) above 8 km that is not present within the other regimes. These large ifferences, compare to the other regimes, can be explaine by the larger variability of the convective cell intensities within

11 3232 M O N T H L Y W E A T H E R R E V I E W VOLUME 141 FIG. 8. PDFs of (a) stratiform an (b) convective rain rates separate accoring to the five large-scale regimes. Means an stanar eviations (in brackets) for each PDF are also represente. Bins of 1 mm h 21 have been use. the E an DE regimes (Kumar et al. 2013), yieling a wier range of stratiform clou properties. Nevertheless, an important part of the istributions in the upper levels (above 8 km) is associate with large reflectivities, so ice particles are larger than in the other regimes, yieling a weaker vertical reflectivity graient (variation of the reflectivity towar the 08C level). At lower levels (i.e., below the 08C level), one can see clear signatures of large positive anomalies that rapily ecrease towar the groun. This is the result of a more efficient evaporation process ue to the rier environment in the E an DE regimes. Inee, Fig. 7a shows that the E an DE regimes are the only two regimes that exhibit an increase in D 0 from 2 km towar the surface associate with a ecrease in N w (Fig. 7c), which is the signature of the evaporation process at those levels (Li an Srivastava 2001; Kumjian an Ryzhkov 2010). Moreover, the positive anomaly moe within the DE regime (Fig. 6c) is slightly shifte towar smaller reflectivity values compare to the E regime (Fig. 6e), which inicates an even more important role of the evaporation within the DE regime that is also clearly visible in the vertical profiles (Fig. 7) as D 0 increases not only from 2 km towar the groun but from 4 km in altitue to the groun. Convective precipitation within the DW regime (active monsoon; Fig. 6h) is characterize by a large shift (5 10 BZ) of the main reflectivity moe towar smaller values through the whole troposphere (similar to the stratiform precipitation) with a larger occurrence of lowreflectivity/high-level clous compare to the climatology. This feature emonstrates that uring the active monsoon regime, convection is less intense. Inee, 25% of the DW convective points have reflectivities below 40 BZ (at 2.5 km), which is a higher fraction than in the other regimes: DE 5 8%, E 5 13%, SW 5 14%, an ME 5 20%. Moreover, the vertical profile of the mean D 0 (Fig. 7b) shows smaller magnitues within the DW regime compare to the others from the 08C level towar the surface. These results may be explaine by the presence of more cells embee within large areas of stratiform clous typical of monsoon conitions. Some of these cells may be in their mature/ecaying phase, but consiering the large ataset use herein it seems more probable that these cells are overall associate with weaker uprafts (May an Ballinger 2007). Moreover, this is consistent with the vertical reflectivity graient in the DW regime, which is larger than in the other regimes, especially between 4 an 10 km in altitue. This implies that ice microphysical processes (mainly aggregation an riming) play a less important role in the convective precipitation pattern, as it tens to act on a smaller epth (only between 4 an 8 km roughly) compare to the E an DE regimes. Thus, convective rain rates within this regime are naturally smaller than in the other regimes (Fig. 8b) with a mean value of 25.3 mm h 21 (compare to 33.4 mm h 21 within the DE regime). During the SW regime, convective parts are characterize by larger reflectivities than average in liqui phase (from the HPDF anomaly; Fig. 6j) that correspon

12 SEPTEMBER 2013 P E N I D E E T A L FIG. 9. PDFs of the DSD parameters: D 0 an log 10 (N w ) for the five large-scale regimes an separate accoring to the precipitation type. The (a),(b) stratiform precipitation an (c),() convective rainfall. Means an stanar eviations (in brackets) for each PDF are also represente. Bins of 0.1 mm (D 0 ) an 0.2 [log 10 (N w )] have been use. to larger rainrops (Fig. 7b). This is consistent with the fast increase in the reflectivity moe with ecreasing height in the ice phase, which inicates the presence of large ice crystals as explaine earlier. As shown in Kumar et al. (2013), the SW regime is the most efficient regime at proucing eep convective, electrically active, an long lasting (MCS like) storms. Both statistics for the 5-BZ clou-top heights an the electrical activity are clearly highest in the SW regime, followe by the easterly regimes (DE/E an ME) an, finally, the DW regime. These convective parts are associate with increase frequencies of occurrence of large rainfall (mean value of 31.7 mm h 21 ; Fig. 8b). Interestingly, these larger frequencies of occurrence of liqui phase reflectivities are associate with a wier istribution of ice-phase reflectivities (i.e., a larger ifference between the two extrema of positive anomalies). But, compare to the other regimes that exhibit the same bimoality in the istribution of the convective reflectivity anomalies (i.e., E an DE), this variability is not translate in the resulting stratiform ice parts of the storms uring the SW regime (no more bimoality). This might inicate that the transfer of ice particles from the SW convective part to the stratiform part is ifferent than in the other regimes. Presumably, as the moe associate with the large reflectivities has isappeare, smaller ice particles are transferre an/or generate in the SW stratiform part. Finally, the ME regime (break) shows a clear an simple signature both in the convective an stratiform parts (Figs. 6k an 6l). The vertical reflectivity graient is quite similar to that of the DW regime but associate with larger reflectivities, which correspon to bigger hyrometeors than in the DW regime (Fig. 7b). This is consistent with the fact that ME an DW are the two regimes that generate most of the total precipitation an exhibit ientical convective stratiform contributions to the raining area (Fig. 4), but with more rain prouce by the convective part in the ME regime (52% versus 49%). Thus, the convective rain rate is slightly larger than in the DW regime with a mean value of 28.6 mm h 21 (Fig. 8b). c. PDFs of DSD parameters for the five large-scale regimes As shown in the previous section, each large-scale regime is associate with ifferent preominant cell types: eep shallow convection, stratiform convective rainfall contribution, life cycle, geographical location, an all of these features are epenent on both the thermoynamical environment an the vertical velocity fiel (May an Ballinger 2007). Moreover, as these thermoynamic an vertical velocity fiels rive the clou microphysics, it is worthwhile to analyze the statistical ifferences that may exist on the DSD parameters. PDFs of DSD parameters [D 0 an log 10 (N w )] for the five wet-season regimes an separate accoring to the rain type are presente in Fig. 9. The first noteworthy result is that whatever the parameter an precipitation

13 3234 M O N T H L Y W E A T H E R R E V I E W VOLUME 141 type consiere, the with (stanar eviation, marke in Fig. 9) of the PDF is smaller uring the DW regime (active monsoon), where ice microphysical processes are expecte to have a smaller impact on precipitation. The main ifferences between the regimes appear when comparing the moist regimes (DW an ME) with the ry ones (DE an E); the shallow westerly regime exhibits an intermeiate pattern of behavior. This observation is well marke in the convective PDFs (Figs. 9c an 9) while it is less obvious in the stratiform ones (Figs. 9a an 9b). Even though the ifferences are small in Figs. 9a an 9b, a noticeable fraction of the stratiform precipitation uring the E an DE regimes (consiering the number of points use to buil the PDFs; Table 1) have higher values of D 0 (above 1.2 mm) an lower values of log 10 (N w ) (below 3.5; e.g., 3160 mm 21 m 23 ) compare to the other regimes. These ifferences might be attribute to the larger influence of ice processes on the prouction of rainfall (Thurai et al. 2010) in aition to the evaporation of the smallest roplets uring their fall, which tens to ecrease the number of small hyrometeors, thereby increasing the meian volume iameter. Contrary to the ry regimes, higher concentrations of small rainrops are observe uring the active monsoon (DW) an break perios (ME), both in stratiform an convective precipitation, which strengthens (from a statistical point of view) the results of previous stuies (May an Ballinger 2007; Bringi et al. 2009; Thurai et al. 2010; May et al. 2011; Munchak et al. 2012). Differences among regimes in the convective PDFs are larger, inicating that the large-scale environment clearly affects the DSD characteristics. Mean meian volume iameters for convective rain vary from 1.40 mm uring the DW regime to 1.71 mm uring the DE regime, which correspons to a variation of about 18% in terms of mean D 0 while there is a factor of 2.8 between the means of the corresponing N w. Even though the behavior of D 0 an N w is generally opposite (as shown in Figs. 7c an 7), it is interesting to combine those two variables to explore the variation of the associate LWC. Figure 10 represents the PDF of the LWC calculate following Testu et al. (2001) an assuming a constant shape parameter of m 5 1. It is interesting to note in Fig. 10 that, whatever the regime consiere, the LWC of the convective rainfall is very similar ue to the compensation between the size an the concentration. However, the mean LWC corresponing to the active monsoon an the break regime are slightly greater than for the DE an E regimes, which can be explaine by the fact that LWC is mainly carrie by the smaller hyrometeors an that both ice processes an evaporation play a key role within the E an DE regimes. This also FIG. 10. PDFs of LWC (g m 23 ) calculate following Testu et al. (2001) an assuming a shape parameter m 5 1 for the convective precipitation within the five large-scale regimes together with the means an stanar eviations (in brackets). Bins of 0.05 g m 23 have been use. explains the ifferences in term of skewness in the istributions. The DE an E regimes have a small positive skewness (0.11, an 0.07, respectively) whereas DW, SW, an ME have a well-marke positive skewness (0.34, 0.23, an 0.29, respectively), which represents the signature of the larger amount of small roplets. May et al. (2011) investigate the impact of aerosols on DSD parameters an notice that high positive values of skewness in the N w istributions were correlate to enhance aerosol loaings. They also showe that in low aerosol concentration regimes larger supercoole roplets are avecte within the convective towers, which is consistent with the presence of larger ice crystals an rainrops within the E an DE regimes. To further investigate these ifferences an make sure they are not simply as a result of ifferences in reflectivity PDFs among regimes, conitional PDFs were also built in reflectivity bins. Figure 11 shows conitional PDFs of rain rates an DSD parameters relative to convective precipitation an separate accoring to both large-scale regimes an reflectivity ranges (bins of 5 BZ were use). First, one can see that below 40 BZ, PDFs of rain rates are very similar within the five regimes with only a few ifferences in terms of relative occurrence. PDFs of DSD parameters show more ifferences an notably a clear seconary moe within the DE an E regimes, where the other regimes exhibit a slightly ecreasing (D 0 ) or increasing (N w ) slope of the PDF. Size sorting within eveloping cells an/or evaporation (which can be efficient since Z h, 40 BZ) coul be goo caniates to explain this seconary moe, as these processes prouce an increase in D 0 an a ecrease in N w (Li an Srivastava 2001; Kumjian an Ryzhkov 2012). This seconary peak in the DE an E regimes coul also

14 SEPTEMBER 2013 P E N I D E E T A L FIG. 11. Conitional PDFs of (top) rain rates, (mile) D 0, an (bottom) log 10 (N w ) for convective precipitation accoring to various reflectivity ranges (5-BZ bins between 35 an 50 BZ) an separate accoring to the five large-scale regimes. Means an stanar eviations (in brackets) also appear in each panel. Note that, in the first row, the relative occurrences of the number of points use, within each regime, to buil the corresponing PDFs are also represente in square brackets. be ue to the lack of representativeness in these regimes. For example, less than 8% of the DE convective points were use to buil the PDF in the BZ range compare to almost 21% of the DW convective points, an the secon peak represents 27% of these points (i.e., a total of about 280 points scattere within the 2-yr scans). The main statistical ifferences appear in the an BZ ranges because they represent together about 75% 80% of all the convective points. One can see that PDFs of rain rates are very ifferent between the an BZ ranges, as means an the stanar eviations increase by about 50% whereas there is only a slight shift (same shape) in the corresponing PDFs of D 0 an N w towar larger values without any increase in the stanar eviations. Inee, a small increase in the mean iameter of 10% leas to an increase of about 30% in terms of the mean rainrop volume, which, associate with a 17% increase in the number concentration, explain, in part, the 50% variation of the rain rate. 4. Conclusions In this paper, a 2-yr ataset collecte using a C-ban ual-polarization raar (C-Pol) locate near Darwin, Australia, is use to extract statistics of stratiform an convective precipitation within five large-scale regimes. The statistical properties presente in the present paper are the total aily rainfall accumulations an raining area, the spatial variability of the aily rainfall accumulations, the vertical variability of the reflectivity fiels an DSD parameters retrieve following Bringi et al. (2009), an the probability ensity functions of rain rates an DSD parameters at a constant altitue of 2.5 km. Each of these properties is then examine as a function of five large-scale atmospheric regimes an separately for stratiform an convective rainfall. The main results can be summarize as follows: First, as this paper is base on preliminary stuies by Bringi et al. (2009) an Thurai et al. (2010) but using a much larger ataset an a ifferent approach concerning the efinition of the regimes, their main conclusions on the active monsoon regime (DW) an the builup one (DE) neee to be checke/strengthene (from a statistical point of view) an extene using vertical variability analysis. Precipitation is mostly of convective origin within the DE an E regimes while the convective stratiform precipitation ratio is aroun unity for the other regimes. SW an ME have almost the same total rainfall

15 3236 M O N T H L Y W E A T H E R R E V I E W VOLUME 141 accumulation an average raining area. The DW regime prouces the most aily precipitation. However, since the ME regime occurrence composes 50% of the wet season, it is the main contributor to the annual rainfall in this region. Spatial maps of total rainfall accumulation show that stratiform rainfall an convective rainfall are strongly correlate in space. All of the easterly regimes (DE, E, an ME) have a maximum over the Tiwi Islans an have a noticeable coastal influence. Westerly regimes show a more scattere precipitation pattern over the whole omain. The DW regime has a strong coastal influence, mainly over the western part of the raar omain, whereas uring the SW regime precipitation is wiesprea with a weak lan influence. The ifferences in terms of vertical reflectivity graient an DSD parameter profiles highlight clear ifferences between the regimes such as the larger impact of the evaporation process uring the E an DE regimes, or the fact that the active monsoon regime is preominantly associate with warm rain processes whereas the other regimes are more sensitive to ice microphysical processes. PDFs of stratiform rain rates show a nearly ientical pattern of behavior whatever the regime consiere. PDFs of convective precipitation show that all the regimes have a peak of occurrence at 18 mm h 21 an then iffer accoring to the shape of the ecreasing slopes. The moist regimes (DW an ME) are associate with the lowest convective rain rates an stanar eviations compare to the ry regimes (DE an E). PDFs of DSD parameters exhibit significant ifferences accoring to the environmental conitions. The active monsoon regime (DW) exhibits the narrower istributions of both D 0 an N w for both rain types. The DW regime has the highest concentrations of small roplets, followe by the ME (break), the SW, an finally the riest regimes (E an DE). Conitional PDFs have also shown that the shape of the PDF of the DSD parameters was very similar between the an BZ ranges, which represent more than 75% of the convective points. Base on the present stuy an the previous ones of Protat et al. (2011) an Kumar et al. (2013), it is clear that clou an precipitation statistical properties are strongly linke to the environmental conitions (i.e., large-scale forcing). Such complementary results on ice clou (Protat et al. 2011) an convective cell properties (Kumar et al. 2013), as well as on precipitation characteristics (this paper), provie a comprehensive picture of the statistical properties of clou systems as a whole in this area. Moreover, they offer a nearly complete observational basis to assess moel parameterizations an satellite retrievals in light of this regime framework. Further evelopments in moel parameterizations an retrieval algorithms can also be guie by comparing the statistics as a function of large-scale atmospheric regimes. As a future work, an with the aim of having an even more complete (an unpreceente) picture of the statistical properties of the tropical troposphere in this region, aerosol physical an chemical properties shoul also be stuie as a function of the same large-scale atmospheric regimes. Acknowlegments. Part of this work has been supporte by the U.S. Department of Energy Atmospheric Raiation Measurement Program. The authors woul like to acknowlege the contributions of Bra Atkinson, Dennis Klau, an Michael Whimpey in supporting the Darwin CPOL raar operations an ata management, as well as Susan Rennie, Justin Peter, an the three anonymous reviewers for proviing helpful comments on the manuscript. REFERENCES Aler, R. F., an A. J. Negri, 1988: A satellite infrare technique to estimate tropical convective an stratiform rainfall. J. Appl. Meteor., 27, Allen, G., an Coauthors, 2008: Aerosol an trace-gas measurements in the Darwin area uring the wet season. J. Geophys. Res., 113, D06306, oi: /2007jd Anagnostou, E. N., an C. D. Kummerov, 1997: Stratiform an convective classification of rainfall using SSM/I 85-GHz brightness temperature observations. J. Atmos. Oceanic Technol., 14, Biggerstaff, M. I., an S. A. Listemaa, 2000: An improve scheme for convective/stratiform echo classification using raar reflectivity. J. Appl. Meteor., 39, Bringi, V. N., T. D. Keenan, an V. Chanrasekar, 2001: Correcting C-ban raar reflectivity an ifferential reflectivity ata for rain attenuation: A self-consistent metho with constraints. IEEE Trans. Geosci. Remote Sens., 39, , C. R. Williams, M. Thurai, an P. T. May, 2009: Using ualpolarize raar an ual-frequency profiler for DSD characterization: A case stuy from Darwin, Australia. J. Atmos. Oceanic Technol., 26, Browning, K. A., an Coauthors, 1976: Structure of an evolving hailstorm. Part V: Synthesis an implications for hail growth an hail suppression. Mon. Wea. Rev., 104, Carbone, R. E., J. W. Wilson, T. D. Keenan, an J. M. Hacker, 2000: Tropical islan convection in the absence of significant topography. Part I: Life cycle of iurnally force convection. Mon. Wea. Rev., 128, Drosowsky, W., 1996: Variability of the Australian summer monsoon at Darwin: J. Climate, 9, Hong, Y., C. D. Kummerov, an W. S. Olson, 1999: Separation of convective an stratiform precipitation using microwave brightness temperature. J. Appl. Meteor., 38,