You are on page 1of 23
1 Manca 1995 xU Partiti g Mass, Heat, and Moisture Budgets of Explicitly Ensembles into Convective and Stratiform Components KUAN-Man Xu Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado (Manuscript received 22 November 1993, i final form 11 Suly 1994) ABSTRACT ‘Simulated data from the UCLA Cumulus Ensemble Model (CEM) ate analyzed to partition mass, beat, and moisture budgets of cumulus ensembles into convective and stratiform components. A'method based primarily ‘on the horizontal distribution of maximum cloud draft strength below the melting level in & CEM grid column has been developed for this analysis. The stratiform region includes both precipitating and nonpreciptaing anvils ‘The convective and stratiform components of mass, heat, and moisture budgets are distinedy differen, in qualitative agreement with previous observational and modeling studies. Inthe heat and moisture budgets, the Siference is due mainly to that in the phase change processes, In general, condensation/ deposition dominate ‘evaporation/sublimation in the convective region. Al of these processes are more or less equally important in the stratiform region. Freezing occurs only im the convective fegion. Sublimation from saow/graupel in the stfatiform region is much more important than inthe convective region, Radiative effects in the statiform ‘component of the heat udget are as important as effets of phase changes, while radiative effects inthe con- ulated Cumulus 581 ‘The importance of the converge ‘convergences inthe 65 of eddy fluxes, especially in the moisture budget, is confirmed. The tiform component are found tobe parametrizable only ifthe venical motions and the Properties of both mesoscale updrafte and mesoscale downdrafts are known. The horizontal iahomogencty ‘within mesoscale updrats/dowr fluxes in the statiform component, (ss of secondary importance for parameterizing the convergences of eddy 1. Introduction Cumulus convection can substantially influence the behavior of large-scale circulations by transporting heat, moisture, momentum, and energy, as well as by releasing latent heat (e.g., Riehl and Maikus 1958 ), but it cannot be resolved with the conventional grid size (~200-1000 km) of large-scale models. The goal of cumulus parameterization is to formulate the collective effects of subgrid-scale cloud processes, rather than the effects of individual clouds, in terms of the prognostic variables of resolvable scale. A quantitative under- standing of the basic physical processes and macro- scopic behavior of cumulus ensembles is essential in order to address the cumulus parameterization problem. Despite the development of many different cumulus parameterizations (¢.g.. Manabe et al. 1965; Kuo 1965, 1974; Arakawa and Schubert 1974; Emanuel 1991; Donner 1993), questions related to the parameteriza- bility of cumulus convection have rarely been ad- dressed. Corresponding author address: Dr. Kyan-Man Xu, Department of Aumospheric Science, Colorado State University, Fert Collins, CO 50823, © 1995 American Meteorological Society Recently, Xu et al. (1992), Xu and Arakawa (1992), and Xu (1994) have attempted to assess the limit of parameterizability of cumulus convection with simulated data from a cloud-resolving model, oF a so- called cumulus ensemble model (CEM). Xu et al. (1992) examined the modulation of cumulus activity by time-varying large-scale processes. They found that in all simulations cumulus activity is rather strongly ‘modulated by large-scale processes but also undergoes, high-frequency fluctuations. The degree of modulation is somewhat dependent upon the large-scale conditions, such as the large-scale advective effects and the basic wind shear. The modulation is usually stronger for stronger large-scale advective effects. When the basic ‘wind shear is strong, there are some phase delays in the modulation due to the existence of mesoscale convec- tive organization, Xu and Arakawa (1992) investigated the sensitivity of the Arakawa~Schubert (1974) cumulus parameter- ization to the horizontal resolution of large-scale mod- cls. Their main finding is that the Arakawa-Schubert (1974) cumulus parameterization is basically valid in spite of the existence of mesoscale organization in cu- mulus convection. In particular, the assumption of cloud work function (CWF) quasi equilibrium, that is, fan approximate balance between the generation of 952 CWF by large-scale processes and its destruction due to cumulus convection itself, is more accurate for in- puts averaged over shorter distances that can partially resolve mesoscale processes. Here the CWF is defined as the rate of generation of cloud-scale kinetic energy due to work done by the buoyancy force per unit cloud- base mass flux (Arakawa and Schubert 1974). The es planation for this puzzling result is that the large-scale forcing, which is defined as the change of CWF when large-scale processes act alone, increases more than does the net change of the CWF as the averaging dis- tance decreases. For large averaging distances, which do not resolve mesoscale processes, the net change of the CWF is relatively important. Xu (1994) examined the dependency of three types of closure assumptions used in cumulus parameteriza- tions, as classified by Arakawa and Chen (1987), on the averaging distance. Two of them are not signifi- cantly dependent on the averaging distance. This im- plies that a proper combination of closure assumptions ‘can be chosen so that cumuulus convection can be par- ameterizable with the influence of mesoscale processes. However, itis not possible to single out mesoscale pro- cesses for causing the dependency of some closure as- sumptions on the averaging distance. ‘This study extends the work of Xu et al. (1992), Xu and Arakawa (1992), and Xu (1994) by examining the relative role of the mesoscale stratiform component in the mass, heat, and moisture budgets of cumulus en- sembles, and attempts to further address the following question: Is it possible to parameterize cumulus clouds despite the influence of mesoscale processes? Another motivation, other than those mentioned above, for ad~ dressing this question is that mesoscale convective structures have much larger spatial scales and much longer life cycles than individual clouds (e.g., Houze and Betts 1981; Houze and Hobbs 1982). The scale separation between convective and large-scale pro- ‘cesses, which is one of the most basic assumptions in ‘cumulus parameterization, thus becomes ambiguous. ‘What is the mesoscale stratiform region? The Glos- sary of Meteorology (Huschke 1959) defined strati- form as “a description of clouds of extensive horizon- tal development, as contrasted to the vertically devel- ‘oped cumuliform types.” Houze (1977) stated that “precipitation falling from the trailing anvil cloud was stratiform.’” Adler and Negri (1988) defined stratiform regions as “anvils of mature and decaying convective systems."” Recently, Houze (1993) defined stratiform and convective precipitation in terms of their vertical velocity scales. Stratiform precipitation is defined as a precipitation process in which the vertical air motion is, small compared to the fall velocity of ice crystals and snow. In this study, mesoscale stratiform regions are referred to as precipitating and nonprecipitating anvils, ‘of mesoscale convective systems. Mesoscale processes are those occurring in the mesoscale stratiform regions. JOURNAL OF THE ATMOSPHERIC SCIENCES. Vor. 52, No. $ Observations have shown that the cumulus heating and cumulus drying rates in stratiform regions are pos- itive in the upper troposphere and negative in the lower troposphere, which are considerably different from those in convective regions (e.g., Houze 1977, 1982; Leary and Houze 1980; Johnson and Young 1983; Johnson 1984; Cheng and Yanai 1989; Gallus. and Johnson 1991; Lin and Johnson 1994). ‘The distinct profiles of cumulus heating and cumulus drying rates, in stratiform regions are related to the profile of vertical ‘motion. The vertical motion within stratiform regions, is upward in the upper troposphere and downward in the lower troposphere, with a magnitude of a few tenths of a meter per second [ see Houze (1989) for a review ] They are also related to the convergences of the eddy heat and moisture fluxes in stratiform regions. ‘The importance of eddy transports in the stratiform component of heat and moisture budgets is far less cer- tain, For example, Cheng and Yanai (1989) argued that the convergence of eddy fluxes can be neglected in stratiform regions compared to the net condensation rate. Johnson and Young (1983) and Gallus and John- son (1991) provided some support for this argument on the basis of diagnoses of the heat and moisture budg- ts of the stratiform regions of mesoscale convective systems in the Tropics and midiatitudes. Houze (1982) derived the convergence of eddy heat flux in stratiform regions based on data from radar analyses. It is smaller than the derived net condensation rate. Note that radar analyses cover the precipitating anvil region only, while diagnostic analyses include both the precipitating and nonprecipitating anvil regions. Tao et al. (1993) found that the convergences of eddy moisture fluxes are very important in both convective and stratiform regions of simulated tropical and midlatitude squall systems. Nevertheless, the extent to which the eddy {transports in the stratiform region can be parameterized is not fully known. Tt has been well recognized that the latent heat re- lease is very important in the heat and moisture budgets of cumulus ensembles from diagnostic and modeling studies (e.g., Houze 1982; Johnson and Young 1983; Tao et al. 1993). The distinction of the vertical distr bution of different types of latent heating (i.e., conden- sation/evaporation, freezing/melting, deposition/sub- imation) between stratiform and convective regions is, still not fully known (Houze 1982; Chong and Hauser 1990; Tao et al. 1993). Another objective of this study is, therefore, to distinguish further the relative impor- tance of the different microphysical processes in con- veetive and stratiform components of the heat and moisture budgets of cumulus ensembles, as well as the radiative effects in the heat budget. In this study a method for partitioning simulated cu- ulus ensembles into convective, stratiform, and clear regions is developed. The new method uses the hori- zontal distribution of maximum cloud draft strength in CEM grid column rather than that of the surface pre- 1 Manes 1995 cipitation rate (P) as the primary criterion (e.g., Tao and Simpson 1989; Tao et al. 1993; Sui et al. 1994). ‘A cutoff value of (P) is also used in the conventional (P)-based method to designate convective areas re- Eardless of the horizontal distribution of (P). Histori- cally, partitioning methods were based on observations of surface precipitation rate at a single station (c.g. Houze 1973; Balsley et al. 1988), radar reflectivity (Churchill and Houze 1984), or satelite data (Adler and Negri 1988). The data used in these studies had very low temporal resolutions. CEM data can resolve the temporal evolution and spatial structures of con- vective cells and stratiform anvils much better than the observational data. Thus, (P) alone may not adequately characterize convective and stratiform regions in sim- ulated high-resolution data because the vertical struc- tures of these regions are not considered. The present method is a preliminary attempt to characterize con- veetive and stratiform regions based on kinematic char- acteristics (Houze 1993) of mesoscale convective sys- tems. Further improvement will be needed in the future, {tis emphasized that the results in the present study are better compared with those from diagnostic studies than those from radar analyses, due to the different def- initions of stratiform regions. Section 2 briefly describes the numerical simula- tions. The partitioning method is presented in section 3. Sections 4 and 5 examine the convective and stratiform components of mass, heat, and moisture budgets. Conclusions and a discussion are provided in section 6. 2, Numerical simulation of cumulus ensembles ‘The numerical cumulus ensemble model (CEM) de- veloped at the University of California, Los Angeles, by Krueger (1988) is used in this study. Recent mod- ifications of the model were described by Xu and Krue- ger (1991) and Krueger et al. (1995). The CEM used in this study is based on the two-dimensional (x-z) anelastic system of equations with the Coriolis accel- eration and a three-phase microphysics parameteriza- tion (Lin et al. 1983; Lord et al. 1984; Krueger et al. 1995). The model includes a very sophisticated for- mulation of turbulence based on a third-moment clo- sure (Krueger 1988). An interactive radiative transfer parameterization has been included in the model (Xu and Randall 1995a) Only one simulation, 104, with interactive longwave radiative transfer, will be analyzed in detail for this study. It has been analyzed to study the cloud—radia- tion. interaction mechanisms in Xu and Randall (1995b). In this simulation, the horizontally uniform, large-scale adveetive cooling and moistening rates, as in simulations Q02 and Q03 described in Xu ct al. (1992), are prescribed. They are assumed to vary with time according to F(t) = [1 = cos(2nt/T)1/2, aw xU 553 where T is the period of the time variation, which is chosen to be 27 hours. Another prescribed variable in the model is the y component of the pressure gradient thus, a time-independent x component of the geo- strophic wind (a nonsheared profile of 5 m s~') is pre= scribed. Ageostrophic wind shears can be produced uring the integration, The simulated cloud systems in 104 are somewhat organized, as in Q03 [Fig. 2b of Xu et al. (1992)]. The underlying surface is the ocean with, a fixed sea surface temperature (299.9 K). ‘The horizontal domain size used in 104 is 512 km. ‘The horizontal grid size is 2 km. 104 was run for 15 days of physical time, with a time step of 10s. See Xu and Randall (1995b) for the remaining aspects of 104. ‘The long integration was performed so that the statis- tical behavior of cumulus ensembles can be examined. 3. A partitioning method In the analysis of CEM results, partitioning assump- tions have been used to distinguish convective and stratiform regions. As mentioned in the introduction, ‘Tao and Simpson (1989) proposed a method, based on the horizontal distribution of (P) within the domain (hereafter referred to as the T-S method). The T-S ‘method is briefly described as follows. The convective region includes a core and two adjacent grid columns. In the core region (P) must be either (i) twice as great as the averaged values over the four adjacent grid col- uumns or (ii) more than 20 mm h-! (25 mm h-' in Tao et al, (1993)]. The remaining grid columns with non- zero (P) are treated as the siratiform region. Nonpre- cipitating anvils were separately treated. This method has been revised in several recent studies by Tao and his colleagues (Tao et al. 1993; Sui et al. 1994). For example, additional criteria such as the cloud water mixing ratio below the melting level (>0.1 g ke~) and maximum updraft speed (>5 m s~') are used to identify nonprecipitating convective cells. ‘A new partitioning method is proposed here, which is based largely on observations of the kinematic struc- tures of mesoscale convective systems (Houze 1977, 1993; Zipser 1977). Convective regions are character” ized by strong vertical motion, that is, convective cells, while stratiform regions are characterized by a rela- tively uniform weak ascent/descent and weak precip- itation over a large area. Mesoscale updraft regions (above the melting level) occasionally contain some dissipating convective cells. The vertical velocity within these cells is still much greater than that of the surrounding area. Thus, the draft strength above the ‘melting level may not be a good indicator of stratiform regions (see appendix) “The partitioning method uses the horizontal distri- bution of the maximum cloud draft strength (|Wnse!) below the melting level in a CEM grid column as the primary variable. As in the T-S method, the convective region includes a core and two adjacent grid columns. 354 A core consists of at least one grid column. A core satisfies one of the following three conditions: 1) [wna i8 twice as large as the average over the four adjacent grid columns, 2) |Wmuc| is greater than 3 m s~',or3)(P) exceeds 25 mm h™". Approximately 75% Of the core areas are actually identified by condition 1 When several adjacent grid columns have very large values of | Wie condition 1 might not be satisfied and thus condition 2 is used. When a convective cell is in its dissipating stage, its [Waa i8 likely to be small but (P) may still be large. Thus, condition 3 might be sat- isfied inthis situation. The dissipating cells aloft behind the leading convective line with small (P) may not be identified well by condition 3. These grid columns, where vertical velocities above the melting level are significantly larger than their surrounding area, may have intrinsic characteristics more similar to stratiform regions than to convective regions (see appendix). ‘The stratiform region consists of the remaining grid ‘columns where the total liquid water path (TLWP) c) ceeds 0.2 kg m-?. This criterion is different from that used in the TS method. Obviously, regions with zero {P)! are classified as “stratiform” by the present ‘method. Some of these regions may, however, be oc- cupied by shallow convection or new convective cells beneath the anvils. These clouds may be associated with an upward layer mean vertical motion, a suffi- ciently large cloud water, but little rainwater. Addi- tional criteria, that is, positive layer mean (ground to °C level) vertical velocity, rainwater path below the melting layer less than 0.1 kg m~® and cloud water path below the melting layer greater than 0.4 kg m™?, are used to identify these shallow clouds as a part of the convective region (e.g., at 32-40 km of Fig. La and at 16-18 km of Fig. Ib). See the appendix for the impact of these criteria on the results. The remaining grid col- umns with TLWP less than 0.2 kg m~? are treated as clear regions. For convenience, this will be referred to as the Wmax method. Figures | and 2 may be used to illustrate the Wmax method. Figure 1 shows xz cross sections of vertical velocity (contours) and cloud water plus cloud ice (hatched) and precipitating water mixing ratio (dot- ted), and (P) at 40:20 and 40:50 of simulation 104. Convective regions usually have large (P) except for the shallow cloud regions mentioned above. Stratiform regions, for example, at 0-10 km and 50-70 km of Fig. Ia and at 50-80 km of Fig. 1b, have relatively small (P) or zero (P). The vertical velocities in the "upper troposphere of these regions are not necessarily small, while those in the lower troposphere are small negative values. Some of these regions ((P) > 0.1 * Because ofthe time evolution of anvil clouds, dstnetion between nonprecpitating and precipitating anvils based on instantaneous (P) alone Is highly questionable. Therefore, we do not separately west these two types of anvils inthis study. JOURNAL OF THE ATMOSPHERIC SCIENCES, Vou. 52, No.5 mm h~) would be classified as “stratiform’” using the T-S method igure 2 shows a decaying squall system from one of the sensitivity tests in Xu and Randall (1995a). The convective region is only ~12 km wide, while stratiform anvils cover a much wider region. Nonprecipitating anvils or thin high an- vils, for example, at 70-90 km of Fig. 2a, at 60-95 km of Fig. 2b, are a major part of the stratiform regions. ‘Therefore, the nonprecipitating anvils will not be treated separately in this study. This will create a dif- ficulty for comparison of current results with radar observations. This is, however, reasonable for under- standing the parameterizability problem. 4, Mass budget ‘A cumulus ensemble contains an ensemble of cu- ‘mulus clouds and their environment. The continuity of mass gives the following p= M, + i @ Here pit is the large-scale mass flux, the overbar de- notes the large-scale average, p is the density of the air, Mis the cloud mass flux, and M is the mass flux in the environment. The latter is commonly known as the en- vironmental subsidence M, The cloud mass flux is the sum of mass fluxes of all clouds, that is, saturated or unsaturated drafts with liq- uid water, inside the cumulus ensemble; that is, M.= 2M, GB) where M, is the mass flux of the ith cloud, defined as M, f pwido;, @ where 9; is the fractional area of the ith cloud ata given level. The cloud mass flux can then be written as the sum of updraft mass flux M, and downdraft mass flux Mg; that is, M. = M, + Mg. 6) A cloud updraft is always saturated, while a cloud, downdraft can be saturated or unsaturated. The sum of the cloud water and cloud ice mixing ratios, qc, is used. to identify the cloudy grids (2 km wide). 100% cloudy arid is assumed if q.,, exceeds 1% of the saturation water Vapor mixing ratio g* or a q../001q* fraction of the arid if qo, is less than 0.01q* (Xu and Krueger 1991). If the cloudy grid has a positive (negative) vertical ve- locity, an updraft (downdraft) is assumed, Thus, M, does. not include the upward motion at unsaturated grids. An unsaturated downdraft grid is defined similarly except us- ing the total precipitating water mixing ratio and 0.1 g kg as the criterion: any grid with downward motion but ‘without any liquid water is not treated as a cloud down- draft. That is, M, is the sum of saturated and unsaturated 1 Manen 1995 xU 995 xX ckm) Fic. 1. The 2-2 eros sections of vertical velocity at (a) 40:20 and (b) 40:50 of simultion 104 ‘The Contour interval is Em swith soli lines for w = O'm and dashed lines for w <0 ims"! Hatched ares represent cloud water plus ice mixing ratio exceeding 0.1 g kg-* and dotted fc represent precipitating water mixing ratio exceeding 0.1 g ke’. The surface precipitation rate coresponding tothe cross sections in (a) and (b) 8 shown in (c. Thick lines below (a) snd (b) indicate convective regions. downdrafts. The environment is defined as any grid with- ‘out any liquid water/ice. ‘The partitioning method described in section 3 only categorizes a CEM column as a convective, stratiform, or clear region. The total mass flux at a given level over the convective or stratiform regions of the CEM do- ‘main is the sum of M,, Mz, and M, because liquid wa- ter/ice may not exist at all levels of a given convective ‘or stratiform column; that is, = My + My + My, 4%, 6) where subscript f stands for convective (Cv) or strati- form (St) regions, and M, is the total mass flux of con- vective (stratiform) regions. Thus, (2) can be rewritten p57 = Mo + May + My o where M7, is the mass flux in the clear regions, _ Figure 3 shows the vertical profiles of Ma. and Ms, averaged over all 15 days of simulation 104, using the ‘Wimax and modified TS methods (Tao etal. 1993). The 556 Height (km) ight (km) JOURNAL OF THE ATMOSPHERIC SCIENCES, Vou. 52, No.5 P] $204 Xam Fic. 2. Same as Fig | except for at (3) 8:40 and (b) 9:10 of a sensitivity test ‘in Xu and Randall (19952), representing a decaying squall system. vertical profiles of Mo, and My as obtained by the two methods are generally similar, except for the magnitudes Of the peaks and the locations of zero mass flux level in ‘Mg, relative to the melting level (~4.6 km). Here Mo, shows a single peak in the lower troposphere and is pos- itive throughout the entire troposphere; My is positive in the upper troposphere and negative in the lower tropo- sphere. The latter basically agrees with previous obser- vational (Houze 1977, 1982, 1989; Johnson 1982; John- son and Young 1983; Johnson et al. 1990) and modeling studies (Tao and Simpson 1989; Sui et al. 1994), ‘The difference in Ms, between the two methods is due to the inclusion of nonprecipitating anvils in the Wmax method. These anvil regions have upward mo- tion in the upper troposphere and downward motion beneath the anvils whose bases are typically far above the melting level (see Fig. 2). Thus, the zero mass ffux level is located above the melting level, and the nega- tive mass flux is also greater in the lower troposphere. ‘The difference in Mo, is generally small at any level except for the lower troposphere. Figure 4 shows the vertical profiles of M,, Mz, and M, in convective and stratiform regions of 104. Comparisons between Figs. 4b and 4d confirm the explanation described above. Other differences between the two methods are the following. Stratiform areas are 16.4% from the Wmax 1 Maxx 1995 so. Wax metiod e@ESmetbod Fic. 3, Vertical profiles ofthe total mass flux in convective regions and mesoscale stratiorm regions of simulation 104 from (a) Wimax ‘method and (b) T=S methods, ‘method but only 6.9% from the T-S method due to the inclusion of nonprecipitating anvils by the Wmax ‘method. Convective areas from the two methods are more comparable (6.0% vs 8.7%). A significant por- tion (2.2% out of 8.7%) is classified by the additional criteria used in the modified T-S method. Stratiform precipitation is 38% from the Wmax method and 34% from the T-S method. The aforementioned compari- sons indicate that the two methods are not equivalent in practice, though their differences appear minor. Hereafter, only the results obtained using the Wmax ‘method will be examined Partitioning of the mass budgets of convective and stratiform regions into different components (updraft downdrafts, and environment) has not been studied be- fore. Tao and Simpson (1989) and Sui et al. (1994) only discussed the total mass flux in both regions. Par- titioning of the mass budget of entire cumulus ensem- bles into different components has been discussed in several modeling studies (e.g., Lipps and Hemiler 1986; Tao and Soong 1986; Tao et al. 1987; Krueger 1988). In the convective regions (Fig. 4a), M, dominates ‘Mg. This dominance is far greater than that found in catlier studies for an entire cumulus ensemble (c.¢., Lipps and Hemler 1986; Tao et al. 1987). Another fea- ture in Fig. 4a is that Mf, is small due to exclusion of cumulus-induced subsidence around cumulus clouds, which occurs in clear regions. In the stratiform regions (Fig. 4b), the magnitudes of M, and M, are comparable, but their distributions ate different. The M, maximum is located in the upper troposphere, which’ agrees with observations (¢.g., Houze 1977; Gamache and Houze 1982, 1985). How- ever, M, in the lower troposphere is not negligible, parts of which are probably caused by shallow con- vection below anvils with high bases (¢.g., ~55 km in Fig. 2b). In the lower troposphere, M,is very large due largely to unsaturated downdrafts. This is also consistent with classical observations (e.g., Houze xu 957 1977, 1989). Another feature in Fig. 4b is that large values of M, exist above the melting level. This has recently been observed in midiatitude and tropical squall systems (Biggerstaff and Houze 1993; Sun et al. 1993; Keenan and Rutledge 1993). A significant portion of M, between the melting level and 10 km is related to the anvils with high bases (e-g., Figs. 1a and 2a,b). Other portions of M, in the upper troposphere are associated with saturated downdrafts (e.g., ~68 km in Fig. 1b and ~65 km in Fig. 2a), which could be driven by the pressure gradient force (e.g., Johnson and Hamilton 1988; Biggerstaff and Houze 1993) This type of saturated downdrafts was also simulated by Lipps and Hemler (1986) but without ice phase microphysics. _ _In summary, the vertical distributions of Hic, and Mg, are distinctly different, in qualitative agreement with previous observational and modeling studies; that is, Mc, is positive throughout the troposphere with a single peak in the lower troposphere, while Ms is pos- itive in the upper troposphere and negative in the lower troposphere. Further partitioning of convective and stratiform regions into updrafts, downdrafis, and en- vironments reveals that (i) M, in the convective region is much more dominant than that in the entire cumulus ensemble, and (ii) in the stratiform region, M, in the lower troposphere and M, in the upper troposphere can be important in the mass budget. 5. Heat and moisture budgets In diagnostic studies, the heat and moisture budgets are diagnosed from the large-scale variables (¢.g., Yanai et al. 1973; Johnson and Young 1983; Cheng 1989). The convective and stratiform components of the heat and moisture budgets cannot be easily sepa- rated. These components can be directly calculated from cloud-scale variables (c.g., deviations from the large scale and the latent heat release associated with the phase changes) as simulated by cloud-resolving, models. The apparent heat source (Q,) and apparent moisture sink (Q3) for a cumulus ensemble (Yanai et al, 1973) can be written as where c,d, e, f,m, and s are the rates of condensation, deposition, evaporation, freezing, melting, and subli- mation, respectively; e. and e, are the rates of evapo- ration of cloud water and evaporation of rainwater, re- 558 (2) Wmax~ Cony JOURNAL OF THE ATMOSPHERIC SCIENCES, Vou. 52, No.5 (b) Wmax -- Strat 1s: Height km) (OTS ~ Conv (@TS Strat 2 2 in r 1 0 77 4 ds ® Br 7 ge « s “ 3 x x x 1 r © ° 44 sm 64202 4 6 aw on 6 “oxy ‘ors, Fio, 4, Same as Fig, 3 except for updraft mass flux (Mu), downdraft mass fax (Md), ‘and environmental subsidence (Me). Mass flux associated with saturated dovwndrafs is also shown (Std_s) spectively; s; and s, are the rates of sublimation from ice to vapor and sublimation from graupel/snow to va- Por, respectively; Qe is the radiative heating rate of the ‘cumulus ensemble; cp is the specific heat at constant pressure; L,, Land L, are the latent heats of vaporiza- tion, fusion, and sublimation, respectively; and the overbar represents the average over the cumulus en- semble, that is, the entire domain. ‘The first terms on the right side of (8) and (9) are the convergences of eddy heat (,) and moisture fluxes (Ez), respectively. (The horizontal convergences are usually negligible.) These convergences also include the contributions from turbulent processes. Equations (8) and (9) state that Q; ~ Qz and Q; include the ef fects of the eddy flux convergences and the conversion rates associated with phase changes. Simply speaking, both Q, — Qp and Q, include an ‘eddy’ component and a “phase change’ component. Equations (8) and (9) give a straightforward way to calculate Q; and Q; for convective (1. and Qz.) and stratiform regions (Qin and Qn), as well as the indi- vidual terms in the budgets, by adding the values at each CEM grid point. No assumption, other than the partitioning method, is needed to calculate the budgets. Al variables are first averaged over 1 h in time and over the entire domain in space from data sampled every 5 min. Then, they are averaged/composited in time with respect to the phase of the imposed large- scale advective process (Xu et al. 1992). In the follow- ing, Qic, Qacr Qin and Qo» are first compared with those from’ previous observational and modeling, studies. 4. Qy and Qs Figures 5 and 6 show time—height cross sections of Ques Ores Orns ANd Qan for 104. Here Qj- and Qn tep- 1 Maxcw 1995 Qe Height (km), xU 559 . 3s 68 $8 #2 Time (H) Fic. 5, Time sequence of the ensemble mean of) the apparent heat source and (b) the apparent ‘moisture sink in the convective regions of 104, The abselssa isthe phase of the imposed large- scale advective processes. The conto resent the right-hand side of (8). The Qi. maxima are located at 3 km, about 2 km higher than Q>. maxima (Fig. 5). This indicates the existence of eddy transports due to convective-scale motions (Yanai et al. 1973). This distribution of Q,. and Q,, is basically simitar to that from a diagnostic analysis (Johnson 1984) and a modeling study (Tao et al. 1993) of cloud systems in the tropical western Pacific except for slightly lower levels of Q,. and Q2. peaks. This difference is due to different large-scale conditions. ‘The most prominent feature in Fig. 6 is that both Qrm and Orn ate positive in the upper troposphere and negative in the lower troposphere. This is qualitatively consistent with previous observations (¢.g., Houze 1982, 1989; Johnson and Young 1983; Gallus and Johnson [991 and simulations (‘Tao et al. 1993). The location of zero Qj, and Qan levels in the middle tro- posphere deduced from radar analyses (Houze 1989) is slightly lower than that from diagnostic analyses because nonprecipitating anvils are included in the di- agnostic analyses, as well as in this study. The results shown in Fig. 6 are similar to those of a tropical Pa- cific case simulated by Tao et al. (1993). Figure 6 interval is 2K d"! Contours over Id Ke are hatched, also shows that Qs, maxima are located between 8 and ‘9-km, which is approximately 1.5 km higher than Qoy maxima. The Qin and Q:q minima are not located at the same level either. These results indicate that the eddy transports due to mesoscale motions and turbu- Ience are not negligible, as shown later. In addition, the positive Qiq and Q2q, below 1 km may be related to turbulence-scale motion. Jn essence, convective and stratiform components of the heat and moisture budgets are similar to those found from previous observational studies. (Houze 1982; Johnson and Young 1983; Gallus and Johnson 1991; Chong and Hauser 1990); that is, there are no major differences from the observed results as far as Qe, Ores Qin, and Qsq are concerned. This adds credence to the partitioning method proposed in this study. b. Eddy and phase change components ‘igure 7 shows time—height cross sections of the total latent heat release, that is, the phase change com- ponent in the heat budget, in convective and stratiform regions. The time evolution and the vertical distribution 560 Fic. 6. Same as Fig 5 except fort ‘Contours over 4 Kd ate Of the total latent heat release are almost identical to those of Q,. (Fig. Sa) and Qi, (Fig. 6a) except for the negative Values near the surface in the convective regions, which are due to evaporative cooling associ- ated with downdrafis. Figures 8a and 8b show that the convergences of eddy heat fluxes (hereafter E,. and E), for convective and stratiform regions, respectively) are small in both regions, compared to the phase change components shown in Fig. 7. Neither E,. nor E,, exceeds 1.5 Kd“! at any level above 1 km. The E\< and Ey, maxima are 2.1 and 1.7 K d-', respectively, and appear near the surface. Because the magnitude of the phase-change component in the stratiform region is very small at some levels, for example, above 10 km (Fig. 7b), Ei. may not be negligible. The small values of Ey. and Ey, shown in Figs. 8a and 8b are related to the slow vari- ation of potential temperature perturbations with height. Figures 9a and 9b are similar to Figs. 8a and 8b except for the convergences of eddy moisture fluxes (hereafter E;- and £>,). A prominent feature in Figs. ‘9a and 9b is that the magnitudes of E- and E>, are large and are comparable to those of the phase change com- JOURNAL OF THE ATMOSPHERIC SCIENCES, Vor. 52, No. 5 rr) stratiform regions. The contour interval is 1K 2 ched. Contours less han 4 Ka" are dated ‘ponent in the moisture budget (not shown). This result is similar to that simulated by Tao et al. (1993). Note from (8) and (9) that the phase change component in the moisture budget is very similar to that in the heat ‘budget (Fig. 7), except for minor differences above 4 km, The large Values of E,, and E,, are related to the approximately exponential variation of water vapor mixing ratio with pressure. There are some noticeable differences between E,. and E,, (Figs. 9a and 9b). Here E. shows a very large drying below 2 km and a slightly smaller moistening above 2 km at all times; E,, shows a smaller drying below 1 km and an additional drying layer between 4.5 and 6.5 km. The latter is related t0 larger moistening, effects due to sublimation/melting and stronger down- draft speeds at the bottom of the layer than at the top of the layer, which creates a moisture divergence in this layer, that is, drying. ‘The above comparison between the eddy and phase change components thus indicates that (i) Ea. and Ea, play a very important role in the moisture budget of both regions, agreeing with Tao et al. (1993), and (ii) E,_ and Ey, are much less important but not totally neg- ligible, especially in the stratiform region. 1 Manet 1995 () Convective xu 561 () Stratiform Height (km) ‘Time (H) FFG. 7, (@) Same as Fig. Sa and (b) same as Fig. 6a except forthe phase change ‘component of the ©. Parameterizability of eddy components in stratiform regions ‘The results discussed above suggest that it is nec- essary to parameterize E;, and E:,. Thus, one may raise the following questions: Is there a straightforward way to parameterize them? Is the horizontal inhomogeneity of the stratiform regions negligible? Cheng and Yanai (1989) derived the following for- ulas for the convergences of eddy heat and moisture fluxes, denoted by E,,,and E,, respectively, by assum- ing that the stratiform regions are horizontally homo- geneous Em 13 [My ~ onbB)(Sq~ 1, (10) & ap Lyd a p Ean = — 2 [My ~ Onl Gm— DI, aap ‘elt Gm — D1, 1D ‘where subscript m stands for the stratiform region, the tilde represents the clear region, M, is identical to Ma, in section 4, M = piv, ois the fractional area of the stratiform region, sis the dry static energy, and is the ater vapor mixing ratio, 5 budget Equations (10) and (11) allow for the coexistence of mesoscale updrafis and downdrafts but do not dis- tinguish between the mean properties of updrafts and downdrafts. If this distinction is considered, Bi and E,,, can be written as follows, denoted, respectively, by Ejae and Bou: 1a , =p Man — Onl (Sou = 8) + (Mya ~ Onl )(Sma ~ 8], (12) he a , © ap Mew — nlf (dna — + Mg ~ nllf)\Qna ~ BI. 13) Here subscripts mu and md stand for mesoscale up- drafts and downdrafts, respectively. Houze (1982) used a formula similar to (12) to estimate Ey, except that 1) he did not consider M because |M| < |My and 0, <€ 1 and 2) he did not take into account that updrafis (downdrafts) exist below (above) the melting level due to the lack of good data. 562 (@) Convective Ly PPI Lilia 6 Zi JOURNAL OF THE ATMOSPHERIC SCIENCES, Vou. 52, No. 5 sho LY 5 Ok Fic, 8, Same as Fig. 7 except forthe convergences of eddy heat ures a,b). The parameterized ‘convergences inthe srtiform regions ac shown i () and (d). The contowe interval is 0.5 Kd! ‘Contours over D Kd arc hatched. Figures 8c and 8d show time—height cross sections of Ein and Ea, using (10) and (12), respectively. Fig- ures 9c and 9d show those of Ean and Eau using (11) and (13), respectively. All variables in (10)~ (13)can bbe calculated from simulated data once the updraft/ downdraft grid boxes in the stratiform regions are iden- -d with the criteria described in section 4. A com- parison between Figs. 8b and 8c shows that E, is not well estimated by (10). The distinction of mesoscale updraft and downdraft properties (Fig. 84) greatly im- proves the results. Since (10) - (13) do not account for horizontal inhomogeneity, the aforementioned results thus suggest that horizontal inhomogeneity within mesoscale updrafts/downdrafts may be important at certain levels, for example, 6-9 km. On the other hand, E>y gives a reasonably good estimate of Fs,, especially below 7 km (comparing Figs. 9c and 9b). The magnitudes of E>, are under- estimated at most levels. ‘The distinction of updraft and downdraft properties greatly improves the esti- mate of E>, (comparing Figs. 9d and 9b), especially above 7 km and below 3 km. The magnitude of Frag is generally more comparable to that of E>, (Fig. 9b) than that of Ez» (Fig. 9c). Therefore, Eis reason- ably well parameterized if the mean properties of both mesoscale updrafts and downdrafts are consid- ered. The horizontal inhomogeneity within meso- scale updrafts/downdrafts is clearly of secondary im- portance for parameterizing E>, The properties in the clear regions are usually ap- proximated by those of the large scale in any cumulus parameterization without mesoscale effects (e.g., Ar- akawa and Schubert 1974). The estimate of Ei, and Esp, with F and J, respectively, replacing ¥ and, is quantitatively similar to that shown in Figs. 8c and 9c, except that the magnitudes are underestimated by a fac- tor of 2 (not shown). Similar estimates for Ey. and Eng ate not greatly affected by this approximation be- cause of the large difference in the properties of up- drafts/downdrafts from those in the environment. Therefore, # and q in (12) and (13) can be replaced by Sand q, respectively. In summary, E,, and E,, are parameterizable only if the vertical motions and the mean properties of both mesoscale updrafts and downdrafts are known. The horizontal inhomogeneity within mesoscale updrafts 1 Maku 1995 xu 563 c. L7 Li ZEEE. Height (km) . 8 0 Lz 24 Fic. 8. (Continued) and downdrafts of stratiform regions may be of sec- ‘ondary importance for the parameterization. 4. Phase change processes ‘The phase change component in the heat budget, that is, the total latent heat release, will be further decom- posed into individual terms associated with various phase changes, as shown in (8). Figures 10 and 11 show time-height cross sections of the latent heat re- Tease associated with the following processes: con- densation, Lz, evaporation of cloud water, — Lz, evaporation of rainwater, ~ L,, net freezing/melting, LAF ~ m), deposition, Ld, sublimation from ice t0 vapor, —L,5;, and sublimation from snow/graupel to vapor, Lp, in the convective and stratiform regions of simulation 104, respectively. In the convective regions, condensation is the dom- inant process, especially in the lower and middle tro- posphere (Fig. 10a). Deposition in the upper tropo- sphere significantly contributes to the total latent heat release there (Fig. 10e). Evaporation of cloud water is much less than condensation and is mainly concen- trated in the lower troposphere (Fig. 10b). Evaporation of rainwater is somewhat important only below 3 km, especially in the subcloud layer (Fig. 10c). Freezing and melting occur only around the melting level, with approximately equal magnitudes of freezing above the melting level and melting below the melting level (Fig. 10d). ‘The least important process is sublimation, =L,(s; + 5,) (Fig. 10F), due to the lack of downdrafts in the middle and upper troposphere (Fig. 4a) ‘The time evolution of each latent heat release term in the convective regions is very simple; that is, the maximum occurs at 13 h (Fig. 10). The secondary ‘maxima are dictated by those in 1,¢, which are related to maxima of M, in convective regions (not shown) In the stratiform regions, all phase change processes except for freezing contribute, more or less equally, t0 the total latent heat release, as far as the magnitudes are concemed (Fig. 11). Condensation occurs through- ‘out the entire troposphere (Fig. 1a), with a much smaller magnitude than its counterpart in the convec- tive region (Fig. 10a), especially in the lower tropo- sphere, Condensation in the lower troposphere of the stratiform region should not occur if there are no up- drafts there. Updrafts below the melting level exist in the stratiform region, however (Fig. 4b). Deposition in the upper troposphere is much larger than that in the convective region (comparing Figs. He and 10e) be- 564 (@) Convective €E ppm LE JOURNAL OF THE ATMOSPHERIC SCIENCES. Vor. $2, No.5 yg Fic. 9. Same as Fig. 8 except forthe convergences of eddy moisture Nuxes. "The contour interval cause of larger updraft mass fluxes in the stratiform region (Fig. 4b). Evaporation of rainwater (Fig. 11c) and cloud water (Fig. 11b) become much larger in the stratiform region; their magnitudes are as large as those of condensation. Melting dominates freezing because of the presence of extensive stratiform precipitation and, weak updrafts (Fig. 11d). This result is consistent with observations (e.g., Leary and Houze 1979). Sublimation is almost as important as deposition (Figs. L1e—g). This finding has not been quantitatively identified by observations. Sublimation from ice to va- por occurs in the upper troposphere with its maximum at 9 km (Fig. 11f). Sublimation from snow/ graupel to vapor is large in the middle troposphere (Fig. 11g). ‘The latter is not surprising because snow/graupel is abundant in the layer 3-km thick above the melting level. In summary, Fig. 11 has shown that in the strat- iform region the phase changes from liquid/ice to va- por are at least as important as the corresponding trans- formations from vapor to liquid/ice, and that freezing is negligible compared to melting. ‘The time evolution of the latent heat release terms in the stratiform region (Fig. 11) differs from that in the convective region (Fig. 10), due to slow develop- "DK a" for (a) and 1 Kd for (0), (0) and (a), ‘ment of stratiform anvils from cumulus towers. The ‘maxima occur between 15 and 18 h in the stratiform region, instead of at 13 h. ‘A comparison of these latent heat release terms be- ‘ween convective and stratiform regions shows that the phase changes from liquid/ice to vapor (evaporation and sublimation) are much more important in strati- form regions than in convective regions. This is to be expected because of the existence of a large area of ‘mesoscale precipitation in the stratiform region and ‘weaker mesoscale updratts e. Radiative effects The effects of radiation on cumulus ensembles can bbe examined by comparing the radiative heating rate Qe with the individual terms on the right-hand side of (8). Figures 12a~c show time—height cross sec- tions of Qy averaged over convective regions, strat- iform regions, and clear regions, respectively. The infrared radiative effects are ‘basically cooling throughout the entire troposphere in all three regions. The lack of warming at cloud bases in Figs. 12a and 12b is due to the coexistence of clouds with varying 1 Manes 1998 Height (km) Height (km), °o s 6 8 @ xu LEP 565 Z sk Time (H) Fic. 9. Continued) cloud bases and the presence of grid points without liquid water/ice. The small magnitude of Qy in Figs. 12a and 12b is related to this cancellation effect of varying cloud bases and cloud tops and the approx- imately 50% of grid points without liquid water/ice in each region. In the convective region, there are two distinct cooling maxima (~3 K d~'). One is produced at 13 km by deep convection, and the other at about 4 km by shallow convection, Just below the cooling max- ima between 8 and 16 h, there are cooling minima (=1K d"), which are associated with those grid columns where nonprecipitating shallow clouds un- derlie thin anvil clouds. Figure 12a shows that Qx is more important than £,, (Fig. 8a), but is much smaller than the individual latent heat release terms in (8). In the stratiform region, the cooling maxima (~4 K d"') are located at about 12 km, while the cooling minima (~0.2 K d~') are located at about 8 km. This vertical distribution promotes the development of up- pet-tropospheric convection through continuous desta- bilization (Xu and Randall 199Sb). In addition, be- ‘cause of the small magnitudes of the individual terms (Figs. 9 and 11) on the right-hand side of (8), the radiative effects in the stratiform region are expected to be very important. ‘The importance of solar radiation in the convective and stratiform components of the heat budget { from simulation 105 described in Xu and Randall ( 1995b); not shown] is similar to that of infrared radiation. For example, the solar heating rate is generally smaller in the convective region than in the stratiform region by a factor of 2 in the upper troposphere. The solar heating rate in the upper troposphere of the stratiform region is greater than that due to net dep- osition/ sublimation. 6. Conclusions and discussion Data simulated with a cumulus ensemble model (CEM) have been analyzed to further examine the parameterizability of cumulus convection. Specitfi- cally, the relative role of the mesoscale stratiform component in the mass, heat, and moisture budgets of cumulus ensembles has been investigated. A kin- ematically based partitioning method has been de- veloped to separate simulated cumulus ensembles into convective, stratiform, and clear regions. Strat- iform regions include both precipitating and nonpre- cipitating anvils. 566 JOURNAL OF THE ATMOSPHERIC SCIENCES Vou. 52, No.5 @Ly€ 1s ep ° od os a0 & Height (km), Height (km) Fic. 10. Same as Fig. S excep forthe individual phase-change terms, that is, (a) condensation, () evaporation within clouds, (€) evaporation of rainwater, (net freezing/melting, (e) deposition, nd (f) sublimation, in the convective regions. The contour interval ix 2 K d~ for (a) and 0.3 Kd * forthe remaining panels. Contours over 14 K d~* are hatched in (@), ‘The main conclusion of this study is that the effects of mesoscale stratiform processes may have to be in- cluded in any parameterization of cumulus convection for use in large-scale models if cumulus convection is organized on the mesoscale. This conclusion can, how- ever, be substantially weakened by several factors re lated to partitioning of convective and stratiform regions. Specific results supporting this conclusion are summarized as follows. ‘The convective and stratiform components of mass, heat, and moisture budgets are distinctly different, in qualitative agreement with previous observational and modeling studies (e.g., Houze 1977, 1982, 1989; John- son and Young 1983; Johnson et al. 1990; Tao et al 1 Marcu 1995 @Lyq Em) xu 567 Height (km), OL, (545) os 6 8 Time (H) ic. 10. (Continued) 1993). In the mass budget, the total mass flux in con- vective regions is positive throughout the troposphere with a single peak in the lower troposphere. The total mass flux in stratiform regions is positive in the upper ‘troposphere and negative in the lower troposphere. The inclusion of nonprecipitating anvils in stratiform regions results in a zero mass-flux level above the melt- ing level. A unique aspect of this analysis of mass budget is that the relative importance of updrafts, sat- turated/unsaturated downdrafts, and environment can be identified at all levels of convective and stratiform regions. Detailed analyses of the heat and moisture budgets reveal that the various phase change processes in the ‘two regions are very different. In general, the phase changes from vapor to liquid/ice (i.e., condensation and deposition) dominate those from liquid/ice to vapor (ie., evaporation and sublimation), while freezing and melting are only important around the melting level in the convective region. All of these 568 JOURNAL OF THE ATMOSPHERIC SCIENCES @Ly Vou. $2, No. 5 . s 6 9s ww SBN Time (H) Fc, 11. Same as Fig. 10 except fr the stratiform regions. Sublimation from iee to vapor (f) and sublimation from graupeUsnow to vapor (g) a separately shown. The contour interval is 05 Kd" forall panels, Contours over 4 Kd" ae hatched. Contours less than 4 Kd" are doted, phase changes are more or less equally important in the stratiform region, except for the negligible effects of freezing. In particular, sublimation from graupel/ snow to vapor is important in the stratiform region ‘but is not important in the convective region. Radi- ative effects in the stratiform component of the heat budget are as important as effects of phase changes, while radiative effects in the convective component are far less important. Solar and infrared radiation have more or less equally important impacts on the heat budget of cumulus ensembles. ‘The convergences of the eddy heat and moisture fluxes are particularly important in the stratiform component of the heat and moisture budgets because the individual phase-change terms are relatively small, This finding indicates that a modification of the assumption used by Cheng and Yanai (1989) in their diagnostic study of the mesoscale effects on the @) yy -m) E o-—__—»—___4 jeseeee 0 ——eneeaaeenaEE Sere Height (km) ° . . ° “ime eo io. 11. (Continued) 570 (#) Convective region Height (km), E ss 68 SO JOURNAL OF THE ATMOSPHERIC SCIENCES Vor. $2, No.5 ts 18S Time (H) Fic. 12. Same as Fig. 5 except forthe radiative heating rates in (a) the convective region, (b) the stratiform region, and (c) the environment. The contour interval is 0.5 K d-!. Contours less than 25 K dae dotted. heat and moisture budgets is needed. The most no- ticeable feature in the eddy moisture fluxes of strat- iform regions is the middle-level (4.5—7 km) drying. ‘The convergence of the eddy moisture fluxes is much more important than that of the eddy heat fluxes in both regions, agreeing with previous studies (e.g., Tao et al. 1993). ‘The convergences of the eddy heat and moisture fluxes in the stratiform component are found to be par- ameterizable only if the vertical motions and mean properties of both mesoscale updrafts and downdrafts are known. The mean properties of the stratiform re- gion cannot give a good estimate of the convergence of eddy fiuxes at all levels, especially for that of the eddy moisture flux. This conclusion may be related to the inclusion of nonprecipitating anvils in the stratiform region, which increases the horizontal inhomogeneity. On the other hand, the horizontal inhomogeneity within 1 Manes 1995 mesoscale updrafts/downdrafts of stratiform regions ‘may be of secondary importance for parameterizing the convergences of eddy fluxes in the stratiform region. ‘The partitioning method proposed in this study only distinguishes a CEM column into convective or strati- form region. Convective drafts can tilt with height and their diameters may increase with height. If these two factors are taken into account, a significant portion of strong drafts above the melting level may be classified as convective. Thus, the relative importance of the stratiform component in the mass, heat, and moisture budgets may be significantly diminished. A trajectory analysis as adopted by Lin (1994) may be a useful tool to more objectively partition convective and stratiform regions and to distinguish precipitating from nonpre- cipitating anvils. Nevertheless, the results presented in this study may provide useful guidance for developing cumulus pa- ameterizations with the effects of mesoscale processes (eg., Donner 1993), in particular, the formulation of| eddy and phase-change components in the heat and ‘moisture budgets. ‘Another approach for parameterizing the mesoscale effects is to consider a parameterization of strong drafts with their bases in the middle troposphere of the strat- iform region, which are the major contributor to the eddy fluxes. The latent heat release associated with the remaining stratiform regions can be dealt with a cloud microphysics parameterization (e.g., Fowler et al 1995). This approach is currently being explored at the Colorado State University. Because of the limitation of the partitioning method, this study may not give a firm answer to the question proposed in the introduction: Is it possible to Parameterice cumulus clouds despite the influence of mesoscale processes? Xu and Arakawa (1992) con- cluded that the Arakawa~Schubert cumulus parame- terization is basically valid in spite of the existence of mesoscale organization in cumulus convection. The Arakawa—Schubert cumulus parameterization does not include the effect of mesoscale processes. Thi study tends to suggest that some mesoscale effects should be included in a parameterization, although the exact importance of mesoscale effects needs to be identified further. The difference between these two studies may be attributed to the fact that some meso- scale effects are likely treated as those of deep con- ection in the parameterization because cumulus ef- fects are forced to approximately balance the effects of large-scale processes. Further study is needed to clarify this complicated issue. Acknowledgments, The author would like to thank Professors David A. Randall and Richard H. Johnson for their valuable comments and discussion on an ear- lier version of the paper, Ms. Tammy M. Weckwerth for improving the manuscript, and three anonymous re- viewers for their constructive comments. This study xu sm was supported by the U.S. Department of Energy’s At- mospheric Radiation Measurement program under Grant DE-FG02-92ER61363 to the Colorado State University (Randall, PI) APPENDIX Sensitivity Tests of the Partitioning Method This appendix provides some results on the sensitiv- ity tests of the partitioning method. In particular, the importance of the dissipating cells aloft behind the leading convective line and the nonprecipitating shal- ow clouds is examined. ‘An additional criterion, Wow. above the melting level ‘greater than 3m s~!, is used to identify the importance Of dissipative cells aloft to the budgets in the stratiform regions. Figure Al shows a comparison of the vertical profiles averaged over all 15 days and the stratiform regions of simulation 104, as obtained from the standard Wmax method and the Wmax method with the addi- tional criterion. These dissipative cells are indeed very important to the mass flux profile (Fig. Ala), the con- oMy Ey vt etal tent et ese Ta? a Fic. Al. Vertical profiles of () the mass fux,(b) the convergence of eddy moisture fhxes, and (¢) the total latent heat release of the tratiform regions of 104, using the standard Wmax method (solid Tine) and the Winax method with an additional riteron for dssipa- tive eels alot (dashed lin). 572 Me re toe "Tsetse Wiyeteep) van? Tina fa tenet ease uot fons Fic. A2. Same as Fig. Al except for the vertical profiles in the convective regions for testing the criteria for nonprecipitating shal- Tow clouds. Solid lines are forthe standard Wmax method and dashed ime forthe Wrnax method without the eteia for nonprecipiating hallow clouds vergence of the eddy moisture fluxes (Fig. Alb), and the total latent heating rate (Fig. Alc) only in the upper troposphere. The relative importance of the eddy com- ponent to the phase-change component is the same (Figs. Alb,c). Since (i) the impact of these dissipative cells occurs only in the upper troposphere and (ii) the surface precipitation rate is only slightly affected, these dissipative cells may more likely be a part of the strat- iform regions. Other dissipative cells with large surface precipitation (>25 mm h~") have been considered as a part of the convective regions by the Wmax method ‘The criteria for nonprecipitating shallow clouds are also evaluated. Figure A2 shows a comparison of the results for the impact of these clouds on the profiles in the convective regions. The mass flux (Fig. A2a) in- creases slightly (~10%). These clouds cause more moistening above 1.5 km and more drying below 1.5 km (Fig. A2b). These clouds also cause slightly more latent heat release below the melting level (Fig. A2c) due to the net condensation (Fig. A2d). These are typ- ical characteristics of nonprecipitating shallow clouds. ‘Therefore, these criteria in the Wmax method are very well justified JOURNAL OF THE ATMOSPHERIC SCIENCES, Vou. 52, No.5 REFERENCES ‘Adler, RF, and A.J. Negri, 1988: A satellite infared technique to estimate tropical convective and sratiform rainfall. J Appl Me- teor, 27, 30-81 ‘Arakaw, A, and W. H. Schubert, 1974: Interaction of a cumulus ‘oud ensemble with the large-scale environment. Part. J- At mos. Set, 31, 674-701. Vand J-M. Chen, 1987: Closure assumptions in the cumulus ation problem. Short and Medium Range Numeri ian, Collection of Papers at the WMO/TUGG NWP Symposium, Tokyo, Japan, 107-131 Balsley, B-B., W. L. Ecklund, D. A. Carer, A.C. Riddle, and K. S. Gage, 1988: Average vertical motion in the tropical atmosphere ‘observed by a radar wind profiler on Pohnpei (7°N, 157E). J. ‘Amos. Set, 45, 396-405, Biggerstaff,M. 1, and R. A. Houze Jr, 1993: Kinematics and micro- ‘Physics of the transition zone of the 10=11 June 1985 squall Tine. J. Atmos. Sei, $0, 3091~3110. ‘Cheng, M-D., 1989: Effects of downdraft and mesoscale convective ‘organization on the heat and moisture badges of tropical cloud clusters, Par I: A diagnostic cumulus ensemble model. Jt mos. Sc, 46, 1517-1538. land M, Yanai, 1989: Effects of downdrats and mesoscale con- ‘eetve organization on the eat and moisture budgets of topical loud clusters. Part IT: ects of mesoscale convective orga ization. J. Amos. Sci, 46, 1566~1588. (Chong, M., and D. Hauser, 1990: A tropical squall line observed ‘uring the COPT 81 experiment im West Africa. Par Ill: Heat and moisture budgets. Mon. Wea. Rev:. 18, 1696-1706, (Churchill, D.D,, and R. A. Houze It, 1984: Development and struc ture of winter monsoon cloud clasters on 10 December 1978. J Atmos. Sei. Al, 933-960. Donner, L. J., 1993: A cumulus parameterization including mass fluxes, vertical momentum dynamics, and mesoscale effects. J Aamos. Sci. 0, 889-906, Emanuel, K. A. 1991: A scheme for representing cumulus convec- tio in large-scale models. J. Arnos. Sct, 48, 2313-2335. Fowler, L. D.,D. A. Randall, and S.A. Ruledge, 1995: Liguid and {ee cloud microphysics in the CSU general circulation mod. Part 1: Model description and simulated microphysical pro- cestes. J. Climate, submited. Gals, W.A, Je, and R. H Johnson, 1991: Heat and moisture budge ‘fan intense milteude squall ine. J. mos. Set, 48, 122-146 GGamacke, J.P and RA. Houz, Jc 1982: Mesoscale ar motions a sociated witha tropical squall line. Mon, Wea. Rev, 110, 118~135. and ——, 1985: Further analysis of dhe composite wind and ‘thermodynamic structure ofthe 12 September GATE squall line. ‘Mon. Wea. Rew, 113, 1241-1259. Houze,R. A. Je, 1973: A climatological study of vertical ransporss by cumulus-sale convection. J. Annas. Sei, 30, 1112-1123. 1977 Structure and dynamics ofa wopicalsqual-line system. ‘Mon. Wea. Rev, 10, 1540-1567. — 1982: Cloud clusters and large-scale vertical motions in the Twopies. J Meteor Soe. Japan, 60, 396-810, —, 1989: Observed structure of mesoscale convective systems and implications for large-scale heating. Quart. J. Roy. Meteor. So. M18, 425-462. 1993: Cloud Dynamics. Neademic Press, $70 pp. and AK Bets, 1981: Convection in GATE. Rev. Geophys ‘Space Phyr, 19, $41~576, —, and P. V: Hobos, 1982: Organization and structure of precip- iaaing cloud systems, Advances in Geophysics, Vol. 24, Aca- demic Press, 225-315. Huschke, RE, Ed, 1959: The Glossary of Meteorology: Amer. Me tor. Soc. 638 pp. Johnson, RH. 1982: Vertical motion in neat-equatorial winter mon- ‘S000 convection. J. Meteor. Soc. Japan. 6D, 632688. 1984: Pattioning tropical heat and moisture budgets into cv ‘nulus and mesoseale components: Implications for curmalus pa- ‘ameterizaion. Mon. Wea, Rev, 112, 1590-1601 1 Mane 1995 and G. S, Young, 1983: Heat and moisture budgets of topical ‘mesoscale anvil clouds. J. Atmos. Sel, 40, 2138-2147 and PJ. Hamilton, 1988: The relationship of surface features, To the precipitation and airflow structure of an intense midt- tude squall line. Mon. Wea. Rev., 16, 14441872, —— and W. A. Gallus Jr, and M.’D. Vescio, 1990: Near-ropo- “phere vertical motion within the tailing stratiform region of @ ridlatitude squall line. J. Atmos. Sei, 47, 2200~2210. Keenan, T.D., and 5. A. Rutledge, 1993: Mewoseale characteristics ‘of monsdonal convection and associated stratiform precipita: tion. Mon. Wea. Re., I, 352-374. Krueger, S. K., 1988: Numerical simulation of topical cumulus ‘clouds and ther interaction with the subcloud layer. J. Atmos. Sei, 48, 221-2250. 0) Fu, K-N. Liou, and H.-N. Chin, 1995: Improvement of fan ice-phase microphysics parameterization for use in au- retical simulation of tropical convection. J. Appl. Meteor, submited Kuo, H. [-, 1965; On formation and intensification of tropical cy: ‘clones through latent heat release by cumulus convection, J ‘Armos. Se, 22, 40-63. 1974: Further studies of the parameterization of the influence ‘of cumulus convection on large-scale flow. J. Atmos. St, 31, 1232-1240. Leary, CA, and R. A. Houze Ir, 1979: Melting and evaporation of hhydrometeors in precipitation from the anvil clouds of deep tropical convection. J. Atmos. Sc, 36, 65-679 — and — 1980: The contribution of mesoscale motions to the ‘mass and heat fluxes of an intense ropial convective system, J. Aumos. Sci, 37, 1-796 Lin, C,, 1994: Development ofan improved cloud model for use in cumulus parameterization. Ph.D. dissertation, University of Cal- fornia, Los Angeles, 252 pp. X., and R. H. Jobson, 1994: Heat and moisture budgets and circulation characteristics of atonal squall line. J. Armes. Sct, 51, 1661-1681 YL, RD. Fatly, and H. D. Orville, 1983: Bulk parameter- ination of the snow field in a cloud model. J. Climate Appl Meteor. 22, 1065~1002. Lipps, FB. and R.S. Hemler, 1986: Numerical simulation of deep ‘wopical convection associated with large-scale convergence. J ‘Armes. Sci, 43, 1796-1816 Lord, $"1, H-E. Willoughby, and J. M. Piotrowicz, 1984: Role of a parameterized ice-phase microphysics in an axisymmetric, non hydrologic wopieal cyclone model. J. Atmos. Sci, 41, 2836- 2848, Lin, Lin, xu 373, Manabe, S., J. Smagorinsky, and R. F. Strickler, 1965: Simulated climatology of a general circulation model with a hydrologic cycle. Mon. Wea. Rev, 93, 769-798. Riehl, Hand J 8. Malkus, 1958: On the heat balance of the equ ‘orial wough zone. Geophysic, 6, 503-538. Sui, C-H,, K-M. Lau, W.-K Tao, and J. Simpson, 1994: The tropical water and energy cycle in a cumulus ensemble model. Part I Equilibrium climate. J Atmos. Sc, $1, 111728, ‘Sun, J1'S. Braun, ML. Biggerstalf, R. ©. Fovell,and R. A. Houze 5, 1993: Warm upper-level downdraft associated with a squall, ine. Mon. Wea. Kev., 121, 2919-2927 ‘Two, W. K, and S."T- Seong, 1986: A study ofthe response of deep twopical clouds to mesoscale processes: Three-dimensional nu terial experiments. J. Atmos. Se, 43, 2683-2676, and J. Simpson, 1989: Modeling study of a tropical squall-type ‘eonvectve line, J. Atmos. Sc, 46, 177202. ind ST. Soong, 1987: Statistical properties of cloud ‘ensemble: A numerical seudy, J. Atmos. Sei, 4, 3175~3187. CH. Sui, B. Fee, S. Lang, j Scala, M-D. Chou, and K. Pek ‘xing, 1993: Heating, moistre and water budgets of wopicl and midlatude sual lines: Comparcons and senitvity to longwave radon. J. Amos. Si, $0, 673-650 Xu, KM, 1996: A statistical analysis ofthe dependency of closure assumptions in cumulus parameterization on the horizontal res: ‘olution, J. Atmos. Sc, 81, 3674-3691, sand Krueger, 199i: Evaluation of lone ions ‘ing a cumulus ensemble model. Mon. Wea Rey, 118, 42~367 and A Arakawa, 1992: Semi-prognostic tests ofthe Arakawa~ ‘Schubert cumulus parameterization using simulated data J. At mos. Sl, #9, 2421-2436. Vand D. A. Randall, 1995a: Impact of interactive radiative trans Teron the macroscopie behavior of cumulus ensembles. Part: Radiation parameterization and sensitivity tests. J. Aamos. Se, 52, 800-817. 1 and ——, 1995b: lenpact of interactive radiative wanser on the ‘macroscopic behavior of cumulus ensembles. Part I: Mechanisms {or cloud-radiation interactions. J. Atmos. Se, 52, 818-837 A. Arakawa, and 8. K. Krucger, 1992: The macroscopic be- ihavior of cumulus ensembles simulated by a cumulus ensemble model. J. Atmos. Se, 48, 2402-2420, Yanai, M, 8. Esbensen, and J-H. Ch, 1973: Determination of bulk ‘roperties of tropical cloud clusters from large-scale heat and moisture budgets J. Atmos. Sci, 30, 611-627 Zipset, EJ, 1977: Mesoscale and convectve-scale downdraft as distinc components of squal-ine structure. Mon. Wea. Rev. 105, 1568-1589

You might also like