Precipitation and cloud statistics in the deep tropical convective regime

Authors


Abstract

[1] Precipitation and cloud statistics in the deep tropical convective regime is investigated through the analysis of grid-scale data from a two-dimensional, cloud-resolving model simulation. The model is forced by large-scale vertical velocity, zonal wind, horizontal advection, and sea surface temperature observed and derived from the Tropical Ocean Global Atmosphere Coupled Ocean Atmosphere Response Experiment. The analysis is conducted by categorizing the grid-scale data into eight rainfall types based on precipitation processes: Water vapor convergence, local vapor change, and hydrometeor change/convergence. Among the eight rainfall types, the rainfall with local atmospheric drying, water vapor divergence, and hydrometeor loss/convergence has the largest contribution (30.8%) to the total rainfall because of large rainfall coverage (35.3%). The hydrometeor loss is mainly caused by water clouds through precipitation and the evaporation of rain. For the three other rainfall types with water vapor divergence, each rainfall type contributes to the total rainfall by less than 5%. Of the total rainfall, 61% is attributed to the four rainfall types with water vapor convergence. Although the rainfall with local atmospheric drying, water vapor convergence, and hydrometeor loss/convergence shows the largest surface rain rate (27.8 mm h−1), it only accounts for a small part (10%) of the total rainfall due to its small rainfall coverage (1.2%). For the three other rainfall types with water vapor convergence, each rainfall type contributes to the total rainfall by 14–19%. The grid-scale precipitation statistics are significantly different from the model domain mean precipitation statistics found by Shen et al. (2010), suggesting a spatial-scale dependence of precipitation statistics.

1. Introduction

[2] Precipitation is an important process in regulating atmospheric moisture, heat budgets, and local hydrological cycles in the deep tropical convective regime. It is greatly affected by local atmospheric thermal processes such as infrared cooling, which produces the nocturnal peak through a decreasing saturation mixing ratio [e.g., Kraus, 1963; Tao et al., 1996; Sui et al., 1997, 1998], solar heating, which generates the afternoon maximum through the warm ocean (a related local destabilization process) [e.g., Sui et al., 1997, 1998], and large-scale dynamic processes such as large-scale vertical velocity, which force the surface rainfall [e.g., Sui et al., 1994; Xu and Randall, 1996; Grabowski et al., 1996; Li et al., 1999; Tao et al., 1999]. These thermodynamic processes appear to lead to a wide range of temporal and spatial scale variability of precipitation from the diurnal cycle to the climate timescale.

[3] As an important step toward improving the understanding of tropical precipitation processes, Gao et al. [2005] derived a vapor-related surface rainfall equation by combining mass-integrated water vapor and cloud budgets. In the diagnostic precipitation budget, the surface rain rate is associated with water vapor processes such as local atmospheric drying/moistening, water vapor convergence/divergence, surface evaporation, and cloud processes, which includes the increase/decrease of local hydrometeor concentration and hydrometeor convergence/divergence. The surface rainfall equation has been widely applied to the analysis of cloud-resolving model simulations throughout the select period of the Tropical Ocean Global Atmosphere Coupled Ocean Atmosphere Response Experiment (TOGA COARE) [Li et al., 1999] in order to study surface rainfall processes [Cui and Li, 2006; Gao and Li, 2008b, 2010], precipitation efficiency [Li et al., 2002a; Sui et al., 2005, 2007], diurnal variations of surface rainfall [Gao et al., 2009], cloud clusters and the merging thereof [Ping et al., 2008], as well as the effects of ice microphysics on rainfall [Gao et al., 2006].

[4] Recently, Shen et al. [2010] examined the roles of large-scale forcing, thermodynamics, and cloud microphysics in tropical precipitation processes by categorizing model domain mean data from the COARE simulation to the eight rainfall types based on surface rainfall processes proposed by Gao et al. [2005]. The analysis of the model domain mean surface rainfall budgets shows that surface rainfall is associated with water vapor convergence, local atmospheric drying, and hydrometeor loss through vapor condensation and depositions. The analysis of grid-scale data is different from the analysis of model domain mean data since the model domain mean data contain information regarding different rainfall types and nonraining regions. Li et al. [2002b] showed that when large-scale upward motions are imposed on the model during COARE, rainfall can be attributed to convective regions where prevailing upward motions throughout the troposphere produce water vapor convergence and stratiform regions where downward motions in the mid and lower troposphere generate water vapor divergence. The collection of cloud water by rain is a dominant microphysical process resulting in rainfall over convective regions, whereas the collection of cloud water by rain and the melting of precipitation ice hydrometeors into rain are the major processes responsible for rainfall over stratiform regions. The transport of hydrometeor concentration from convective regions to raining stratiform regions is an important source of stratiform rainfall [e.g., Cui and Li, 2006]. Wang et al. [2007] revealed that local atmospheric moistening is associated with water vapor convergence during the formation and mature phases of monsoon precipitation system over the South China Sea whereas the local atmospheric drying is related to water vapor divergence during the dissipating phase. Local atmospheric drying is caused by the increase in the net condensation resulting from the decrease in saturation specific humidity induced by nocturnal radiative cooling. Therefore, the model domain contains different types of rainfall, whose statistics can be studied through the rainfall-type categorization of grid-scale data by surface rainfall budget. The questions to be discussed in this study are what precipitation and cloud statistics are in grid-scale calculations and how they differ from the model domain mean calculations by Shen et al. [2010].

[5] In this study, the grid-scale data from the two-dimensional (2D), cloud-resolving model simulation during the selected 17 day period of TOGA COARE are used to document and analyze precipitation and cloud statistics with the different rainfall types categorized by surface rainfall processes. The contribution of each rainfall type to the total rainfall is examined by analyzing the fractional rainfall coverage and the mean surface rain rate. The associated rainfall processes are further investigated to determine the contributions of local vapor and hydrometeor change and of water vapor convergence to the surface rainfall. The cloud budgets are analyzed for water and ice clouds to elucidate how cloud microphysical processes impact local hydrometeor change. The model and experiment are briefly described in section 2. The results are presented in section 3. The conclusions are discussed in section 4.

2. Model and Experiment

[6] The cloud-resolving model simulation data analyzed in this study are from Gao and Li [2008b], in which the 2D version of the cloud-resolving model [Sui et al., 1994, 1998] originally developed by Soong and Ogura [1980], Soong and Tao [1980], and Tao and Simpson [1993], and modified by Li et al. [1999], was used to conduct the model simulation. Li et al. [1999] replaced constant cloud single scattering albedo and asymmetry factors with those varied with clouds and environmental thermodynamic conditions. The time scale in depositional growth of snow from cloud ice increased from 50 μm [Hsie et al., 1980] to 100 μm [Krueger et al., 1995] based on aircraft observations. The model includes prognostic equations for potential temperature and specific humidity, perturbation horizontal and vertical momentum, and cloud hydrometeors (cloud water, raindrop, cloud ice, snow, and graupel). The model uses single-moment cloud microphysical parameterization schemes [Lin et al., 1983; Rutledge and Hobbs, 1983, 1984; Tao et al., 1989; Krueger et al., 1995] (Table 1), and solar and thermal infrared radiation parameterization schemes [Chou et al., 1991, 1998; Chou and Suarez, 1994]. The model uses cyclic lateral boundaries and has a horizontal domain of 768 km with 33 vertical levels. Its horizontal and temporal resolutions are 1.5 km and 12 s, respectively. Detailed model descriptions can be found in Gao and Li [2008a].

Table 1. List of Microphysical Processes and Parameterization Schemes From Rutledge and Hobbs [1983] (RH83), Rutledge and Hobbs [1984] (RH84), Lin et al. [1983] (LFO), Tao et al. [1989] (TSM), and Krueger et al. [1995] (KFLC)
NotationDescriptionScheme
PMLTGGrowth of vapor by evaporation of liquid from graupel surfaceRH84
PMLTSGrowth of vapor by evaporation of melting snowRH83
PREVPGrowth of vapor by evaporation of raindropsRH83
PIMLTGrowth of cloud water by melting of cloud iceRH83
PCNDGrowth of cloud water by condensation of supersaturated vaporTSM
PGMLTGrowth of raindrops by melting of graupelRH84
PSMLTGrowth of raindrops by melting of snowRH83
PRACIGrowth of raindrops by the accretion of cloud iceRH84
PRACWGrowth of raindrops by the collection of cloud waterRH83
PRACSGrowth of raindrops by the accretion of snowRH84
PRAUTGrowth of raindrops by the autoconversion of cloud waterLFO
PIDWGrowth of cloud ice by the deposition of cloud waterKFLC
PIACRGrowth of cloud ice by the accretion of rainRH84
PIHOMGrowth of cloud ice by the homogeneous freezing of cloud water 
PDEPGrowth of cloud ice by the deposition of supersaturated vaporTSM
PSAUTGrowth of snow by the conversion of cloud iceRH83
PSACIGrowth of snow by the collection of cloud iceRH83
PSACWGrowth of snow by the accretion of cloud waterRH83
PSFWGrowth of snow by the deposition of cloud waterKFLC
PSFIDepositional growth of snow from cloud iceKFLC
PSACRGrowth of snow by the accretion of raindropsLFO
PSDEPGrowth of snow by the deposition of vaporRH83
PGACIGrowth of graupel by the collection of cloud iceRH84
PGACRGrowth of graupel by the accretion of raindropsRH84
PGACSGrowth of graupel by the accretion of snowRH84
PGACWGrowth of graupel by the accretion of cloud waterRH84
PWACSGrowth of graupel by the riming of snowRH84
PGDEPGrowth of graupel by the deposition of vaporRH84
PGFRGrowth of graupel by the freezing of raindropsLFO

[7] The 2D cloud-resolving model was integrated by Gao and Li [2008b] from 0400 Local Standard Time (LST) 22 December 1992 to 0400 LST 08 January 1993 (total of 17 days). The model is forced by large-scale vertical velocity, zonal wind, and horizontal advections derived using six hourly TOGA COARE observations within the Intensive Flux Array (IFA) region (M. Zhang, personal communication, 1999) and hourly sea surface temperatures at the Improved Meteorological (IMET) surface mooring buoy (1.75°S, 156°E; see Figure 1) [Weller and Anderson, 1996].

Figure 1.

Time-height distribution of (a) vertical velocity (cm s−1), (b) zonal wind (m s−1), and (c) time series of sea surface temperature (SST, °C) during a selected period of TOGA COARE [from Gao and Li, 2008b]. Upward motion in Figure 1a and westerly wind in Figure 1b are shaded.

[8] Based on Gao et al. [2005] and Cui and Li [2006], the surface rain rate (PS) can be written as

equation image

where

equation image
equation image
equation image
equation image

Here, qv is specific humidity; u and w are the zonal and vertical components of wind, respectively; equation image is height dependent mean air density; q5 = q2 + q3, q2 = qc + qr, q3 = qi + qs + qg, and qc, qr, qi, qs, qg are the mixing ratios of cloud water, raindrops, cloud ice, snow, and graupel, respectively; overbar denotes a model domain mean; prime is a perturbation from model domain mean; [()]( = equation image ()dz) is mass integration, and zt and zb are heights of the top and bottom of the model atmosphere, respectively; and superscript o is an imposed COARE-observed value.

[9] Hydrometeor change/convergence (QCM) are from water (QCMW) and ice (QCMI) clouds. Following Sui and Li [2005], the cloud budget can be written by

equation image
equation image
equation image

where

equation image
equation image
equation image

Here, (2b) and (2c) are water and ice cloud budgets, respectively. Ice water path (IWP) is the sum of mass integrations of mixing ratios of cloud water and raindrops, and liquid water path (LWP) is the sum of mass integrations of mixing ratios of cloud ice, snow, and graupel. (2f) is the conversion between the ice and water hydrometeors, which is mainly determined by the accretion of cloud water by precipitation ice hydrometeors (PSACW + PGACW) and the melting of precipitation hydrometeors to rain (PSMLT + PGMLT) in the deep tropical convective regime [Li et al., 2002b]. Cloud microphysical processes and their parameterization schemes in (2) can be found in Table 1.

[10] In (1a), QWVT, QWVF, QWVE, and QCM are local water vapor change, water vapor convergence, surface evaporation, and hydrometeor change/convergence, respectively. The QWVT, QWVF, and QCM can be positive or negative, whereas surface evaporation (QWVE) is positive. Following Shen et al. [2010], T and t represent local atmospheric drying (QWVT > 0) and moistening (QWVT < 0), respectively. F and f denote water vapor convergence (QWVF > 0) and divergence (QWVF < 0), respectively. M and m represent hydrometeor loss/convergence (QCM > 0) and hydrometeor gain/divergence (QCM < 0), respectively. Unlike Shen et al. [2010], where calculations of model domain mean simulation data were made, this study will calculate and analyze the eight rainfall types (TFM, TFm, tFM, tFm, TfM, Tfm, tfM, tfm; see Table 2) using grid-scale data.

Table 2. Summary of Rainfall Types
TypeaDescription
  • a

    T and t represent local atmospheric drying and moistening, respectively; F and f represent water vapor convergence and divergence, respectively; M and m represent hydrometeor loss/convergence and gain/divergence, respectively.

TFMWater vapor convergence, local atmospheric drying, and hydrometeor loss/convergence
TFmWater vapor convergence, local atmospheric drying, and hydrometeor gain/divergence
tFMWater vapor convergence, local atmospheric moistening, and hydrometeor loss/convergence
tFmWater vapor convergence, local atmospheric moistening, and hydrometeor gain/divergence
TfMWater vapor divergence, local atmospheric drying, and hydrometeor loss/convergence
TfmWater vapor divergence, local atmospheric drying, and hydrometeor gain/divergence
tfMWater vapor divergence, local atmospheric moistening, and hydrometeor loss/convergence
tfmWater vapor divergence, local atmospheric moistening, and hydrometeor gain/divergence

3. Results

[11] The total rainfall is largely attributable to the rainfall associated with local atmospheric drying and hydrometeor loss and water vapor divergence in TfM (30.8%); partly because this rainfall type covers 35.3% of the total rainfall area (Table 3). The mean surface rain rate of TfM is 2.833 mm h−1, which is related to the local atmospheric drying rate and the hydrometeor loss/convergence, while water vapor divergence prevails. The water vapor divergence is associated with downward motions below 7 km, which reaches a maximum of −0.2 m s−1 around 2 km (Figure 2); the downward motions produce downward water vapor mass flux with a maximum around 1.3 km (Figure 3) and weak downward hydrometeor mass flux below 3 km (Figure 4). The perturbation specific humidity of TfM is positive with a maximum of 0.5 g kg−1 around 4 km, whereas it is negative below 1 km (Figure 5). The hydrometeor mixing ratio of TfM has a maximum of 0.4 g kg−1 around 5 km (Figure 6). The hydrometeor loss/convergence of TfM is from the loss/convergence of water (76.1%) and ice (23.9%) hydrometeors (Table 4). The water hydrometeor loss/convergence is associated with the precipitation and evaporation of rain that overcome the conversion from the ice hydrometeor to the water hydrometeor, which leads to the ice hydrometeor loss/convergence. The IWP is 24.2% smaller than the LWP.

Figure 2.

Vertical profiles of mean vertical velocity (m s−1) in TFM (dark solid), TFm (dark dash), tFM (dark long dash), tFm (dark dotted dash), TfM (light solid), Tfm (light dash), tfM (light long dash), and tfm (light dotted dash).

Figure 3.

Vertical profiles of mean water vapor mass flux (10−2 kg m−2 s−1) in TFM (dark solid), TFm (dark dash), tFM (dark long dash), tFm (dark dotted dash), TfM (light solid), Tfm (light dash), tfM (light long dash), and tfm (light dotted dash).

Figure 4.

As in Figure 3 except for mean hydrometeor mass flux.

Figure 5.

Vertical profiles of mean perturbation specific humidity (g kg−1) in TFM (dark solid), TFm (dark dash), tFM (dark long dash), tFm (dark dotted dash), TfM (light solid), Tfm (light dash), tfM (light long dash), and tfm (light dotted dash).

Figure 6.

As in Figure 5 except for mean hydrometeor mixing ratio.

Table 3. Fractional Rainfall Coverage (FRC), Percentage of Rain Amount Over Total Rainfall Amount (PRA), and Mean Values of PS, QWVT, QWVF, QWVE, and QCM by Rainfall Type
 TFMTFmtFMtFmTfMTfmtfMtfm
FRC (%)1.1576.90624.74415.35135.34510.645.8350.022
PRA (%)9.95117.66719.21714.17430.8485.0123.0370
PS (mm h−1)27.8148.2732.5112.9862.8221.5231.6830.005
QWVT (mm h−1)7.5428.142−10.279−12.13012.95512.789−1.733−0.160
QWVF (mm h−1)9.40715.3499.33323.912−15.067−8.336−2.102−0.106
QWVE (mm h−1)0.3850.3570.2490.3020.270.2810.2550.36
QCM (mm h−1)10.479−15.5753.208−9.0994.665−3.2115.264−0.089
Table 4. Mean Values of IWP, LWP, QCM, QCMI, QCMW, PS, −PCND, −(PDEP + PSDEP + PGDEP), PREVP, PMLTS + PMLTG, and C(LWP, IWP) by Rainfall Type
 TFMTFmtFMtFmTfMTfmtfMtfm
IWP (mm)1.4421.7110.660.930.7320.9590.8160.182
LWP (mm)5.5623.8940.8732.0110.9661.2030.7530.178
QCM (mm h−1)10.479−15.5753.208−9.0994.665−3.2115.264−0.089
QCMI (mm h−1)−0.523−5.4360.684−1.3111.115−0.9241.4150.127
QCMW (mm h−1)11.002−10.142.524−7.7883.55−2.2873.849−0.217
PS (mm h−1)27.8148.2732.5112.9862.8221.5231.6830.005
PCND (mm h−1)−19.518−23.257−0.284−12.0010.286−4.7731.694−0.339
−(PDEP + PSDEP + PGDEP) (mm h−1)−1.236−2.726−0.354−1.172−0.364−1.432−0.23−0.144
PREVP (mm h−1)3.3692.0821.2941.0561.8381.4122.010.386
(PMLTS + PMLTG) (mm h−1)0.050.0530.0420.0320.0820.0590.1080.002
C(IWP, LWP) (mm h−1)0.663−2.7630.996−0.1711.3970.4491.5370.269

[12] Rainfalls with water vapor convergence make significant contributions to total rainfall. Of the total rainfall, 10%, 17.7%, 19.2%, and 14.2% are from the rainfall associated with local atmospheric drying and hydrometeor loss/convergence in TFM, the local atmospheric drying and hydrometeor gain/divergence in TFm, the local atmospheric moistening and hydrometeor loss/convergence in tFM, and the local atmospheric moistening and hydrometeor gain/divergence in tFm, respectively (Table 3). The large contributions of rainfall with local atmospheric moistening may be associated with the large fractional coverage of rainfall (24.7% in tFM and 15.4% in tFm). Although they cover small area (1.2% in TFM and 6.9% in TFm), rainfalls with local atmospheric drying have large magnitudes (27.814 mm h−1 in TFM and 8.273 mm h−1 in TFm).

[13] In TFM, water vapor convergence, local atmospheric drying and hydrometeor loss/convergence have similar contributions (∼30%) to the rain rate. Although the weak downward motions occur near the surface, strong upward motions with their maximum of about 1 m s−1 around 3.5 km produce the strong upward water vapor and hydrometeor mass fluxes (Figures 24). Maximum perturbation specific humidity is 1.8 g kg−1 at 2 km (Figure 5), whereas maximum hydrometeor mixing ratio is about 1.2 g kg−1 below 5.5 km (Figure 6). The analysis of the cloud budget in TFM reveals that the hydrometeor loss/convergence is only from water clouds (Table 4). The water hydrometeor loss/convergence corresponds to the surface rainfall and the evaporation of rain, which is largely offset by vapor condensation. The IWP is about 25.9% of LWP, which indicates the dominance of water clouds.

[14] In TFm, the water vapor convergence is nearly balanced out by the hydrometeor gain/divergence, whereas rainfall is related to the local atmospheric drying (Table 3). The water vapor convergence rate is larger in TFm than in TFM because the mid and upper tropospheric upward motions are much stronger in TFM than in TFm and upward motions occur in TFm whereas downward motions appear in TFM near the surface (Figure 2). Upward motions in TFm generate the upward water vapor mass flux throughout the troposphere, though the maximum upward water vapor mass flux is smaller in TFm than in TFM (Figure 3). The upward hydrometeor mass flux is weaker in TFm than in TFM (Figure 4). Perturbation specific humidity and hydrometeor mixing ratio in the mid and lower troposphere are generally smaller in TFm than in TFM (Figures 5 and 6). The cloud budget in TFm shows that the large hydrometeor gain/divergence is from both water (65.1%) and ice (34.9%) clouds (Table 4). The water hydrometeor gain/divergence is primarily associated with vapor condensation, whereas the ice hydrometeor gain/divergence is related to vapor depositions and the conversion from the water hydrometeor to the ice hydrometeor. The IWP is about 43.9% of LWP.

[15] In tFM, water vapor convergence is largely used to moisten the local atmosphere (Table 3). As a result, the rainfall corresponds to the hydrometeor loss/convergence. The upward motions are rather weak, but they extend from the surface to the tropopause (Figure 2) and yield weak upward water vapor and hydrometeor mass fluxes throughout the troposphere (Figures 3 and 4). Perturbation specific humidity reaches the maximum of 0.5 g kg−1 between 2 and 5 km (Figure 5) and hydrometeor mixing ratio shows a maximum of 0.3 g kg−1 around 5 km (Figure 6). The hydrometeor loss/convergence is primarily from water clouds (Table 4), in which the rain rate and evaporation rate of rain are higher than the conversion rate from the ice clouds to the water clouds. The IWP is 24.4% smaller than LWP.

[16] In tFm, the water vapor convergence is largely used to moisten the local atmosphere and to increase hydrometeor concentration and/or to advect hydrometeors out, which consequently produces a small surface rain rate (Table 3). The upward motions occur through the troposphere while they reach their maximum around 2 km (Figure 2). In particular, the vertically averaged upward motions are about 0.2 m s−1, the largest among the eight rainfall types. The strong lower tropospheric upward motions lead to the strong upward water vapor and hydrometeor mass fluxes near the surface (Figures 3 and 4). Maximum perturbation specific humidity is 1 g kg−1 at 2 km (Figure 5) and maximum hydrometeor mixing ratio is about 0.5 g kg−1 around 2–5 km (Figure 6). The large hydrometeor gain/divergence is mainly from water clouds due to the vapor condensation (Table 4). The IWP is about half of the LWP.

[17] Although the rainfall in TfM is the largest contributor to the total rainfall, the other rainfall types associated with water vapor divergence play minor roles in the total rainfall (Table 3). The rainfall in Tfm and tfM contribute 5.0 and 3.0% of the total rainfall, respectively; partially due to smaller coverage and weak rainfall intensity. The rainfall in tfm is virtually zero, but the samples are very small as indicated by small fractional rainfall coverage.

[18] In Tfm, the small rain rate corresponds to the local atmospheric drying while water vapor divergence prevails and hydrometeor gain/divergence occurs (Table 3). The water vapor divergence is associated with downward motions near the surface although the upward motions appear above 3 km (Figure 2). Water vapor mass flux is upward above 2.5 km (Figure 3), whereas hydrometeor mass flux is generally upward (Figure 4). Perturbation specific humidity is positive above 1 km (Figure 5). Maximum hydrometeor mixing ratio reaches a maximum of 0.5 g kg−1 around 5.5 km (Figure 6). Hydrometeor gain/divergence is from both water (71.2%) and ice (28.8%) clouds (Table 3). The water hydrometeor gain/divergence results from vapor condensation, which is much larger than the surface rainfall. The ice hydrometeor gain/divergence corresponds to vapor depositions. The IWP is 20.2% smaller than LWP.

[19] In tfM, hydrometeor loss/convergence is responsible for the surface rainfall while it overcomes the water vapor divergence to moisten the local atmosphere (Table 3). The water vapor divergence is associated with downward water vapor mass flux (Figure 3) forced by the tropospheric downward motions (Figure 2). The weak downward hydrometeor mass flux occurs below 6 km (Figure 4). Perturbation specific humidity is positive above 1.5 km (Figure 5). The vertical profile of hydrometeor mixing ratio in tfM is generally similar to that in TfM (Figure 6). The hydrometeor loss/convergence is from both water (73.1%) and ice (26.9%) clouds (Table 4). The water hydrometeor loss/convergence results from the surface rainfall and evaporations of cloud water and rain (PCND < 0 when the evaporation of cloud water occurs). The IWP is 8.4% larger than the LWP.

[20] To study precipitation statistics in the evolution of precipitation systems, we choose a 2 day rainfall case from 2000 LST 23 December to 1800 LST 25 December 1992. The evolution of the rainfall event is analyzed over four phases: The onset phase (2000–2200 LST 23 December), the development phase (2300 LST 23 December to 1400 LST 24 December), the mature phase (1500 LST 24 December to 0900 LST 25 December), and the decay phase (1000–1800 LST 25 December). Model domain mean surface rain rates are 0.157 mm h−1 for the onset phase, 0.358 mm h−1 for the development phase, 0.677 mm h−1 for the mature phase, and 0.189 mm h−1 for the decay phase. During the onset phase, rainfall is largely attributable to TfM, tFm, and TFm. TfM makes a significant contribution to rainfall even though rainfall with water convergence is slightly larger than rainfall with water vapor divergence (Table 6). The large contributions to rainfall made by tFm and TFm signify a large growth of clouds. During the development phase, rainfall is created byTFM, TfM, and TFm. During this phase, TfM's contribution to rainfall is significantly lower. The large contributions to rainfall from TFM and TfM are mainly caused by local atmospheric drying induced by the decrease in saturation mixing ratio, which is brought on by nocturnal radiative cooling. During the mature phase, TfM, tFM, tFm, and TFm account for the total rainfall. The large contribution by tFM indicates that the rainfall is associated with water vapor convergence and hydrometeor loss/convergence. During the decay phase, TfM and tFM make the largest rainfall contributions. The rainfall contribution by TfM is increased during this final phase.

[21] To study precipitation statistics over convective and stratiform regions, the convective stratiform rainfall partitioning method developed by Tao et al. [1993] and modified by Sui et al. [1994] is used to identify convective and stratiform rainfall regions. In this separation scheme, when the surface rain rate at the model grid point is twice as large as the average taken over the surrounding four grid points (i.e., two neighbors on each side in the 2D framework), the model grid point and the grid point on either side are considered convective. In addition, any grid point with a rain rate of 20 mm h−1 or more is designated as convective, regardless of the above criteria. Additional information on maximum updraft above 600 hPa and cloud water is also used as criteria to identify convective rainfall points. All nonconvective rainfall points are characterized as stratiform. Over convective rainfall regions, rainfall with water vapor convergence (68.29%) contributes more to total rainfall than rainfall with water vapor divergence (31.71%) contributes. Over stratiform rainfall regions, rainfall with water vapor divergence (52.27%) makes a larger contribution to rainfall than rainfall with water vapor convergence (47.73%) makes.

4. Conclusions and Discussions

[22] The rainfall and associated cloud statistics in the deep tropical convective regime are investigated by analyzing the grid-scale data from a 2D cloud-resolving model simulation during TOGA COARE. The analysis is carried out by categorizing eight rainfall types based on different rainfall processes. There are four major results. 1. Among the eight rainfall types, the largest contributor to total rainfall is rainfall with local atmospheric drying, hydrometeor loss/convergence and water vapor divergence, which occupies about 35% of rainfall areas. The estimate of hydrometeor loss/convergence is crucial in the determination of the surface rain rate since the local atmospheric drying rate is lower than the water vapor divergence rate. The hydrometeor loss/convergence is primarily from the water clouds. The water hydrometeor is reduced by the surface rainfall and the evaporation of rain, the rates of which are larger than the vapor condensation rate. 2. Although the three other rainfall types with water vapor divergence cover about 16% of rainfall areas, they only contribute to approximately 8% of the total rainfall. 3. Four rainfall types contribute to the total rainfall by about 61%. The largest mean surface rain rate (27.8 mm h−1) is from rainfall with water vapor convergence, local atmospheric drying, and water hydrometeor loss/convergence, in which the vapor condensation is not enough to offset the loss of water hydrometeor from the surface rainfall and the evaporation of rain. Due to the small rainfall coverage (1.2%), this rainfall type only accounts for about 10% of the total rainfall. 4. Three other rainfall types with water vapor convergence contribute to the total rainfall by 14–19%. The water vapor convergence is nearly balanced by the hydrometeor gain/divergence of both water and ice clouds through vapor condensation and deposition, and the rainfall corresponds primarily to local atmospheric drying. The water vapor convergence is almost used to moisten the local atmosphere, and the rainfall is associated with the hydrometeor loss/convergence of water clouds due to surface rainfall and the evaporation of rain. The strong water vapor convergence is used to moisten the local atmosphere and to increase surface rainfall and rain hydrometeor through vapor condensation.

[23] The four rainfall types with water vapor convergence account for about 61% of the total rainfall in the calculation of grid-scale data in this study, which is significantly smaller than that (86.6%) calculated from the model domain mean simulation data by Shen et al. [2010] (Table 5). The large contributions of the rainfall with water vapor divergence are excluded by simple model domain mean calculations for mean rainfall types. In the analysis of grid-scale data conducted in this survey, the rainfall with local atmospheric drying, water vapor divergence and hydrometeor loss/convergence (TfM) contributes 30.8% the total rainfall, which is about five times that (6.2%) found through the analysis of model domain mean simulation data by Shen et al. [2010]. This is because TfM is the largest rainfall contributor for each mean rainfall types. Thus, the contribution of each rainfall type to the total rainfall in the analysis of grid-scale data could be significantly different from that determined by the analysis of model domain mean data. Therefore, the comparison in precipitation statistics between grid-scale and domain mean calculations suggest a spatial-scale dependence of precipitation statistics (Table 6).

Table 5. Percentage of Rain Amount (PRA) Over Total Rainfall Amounta
 Mean Rainfall Type
TFMTFmtFMtFmTfMTfmtfM
  • a

    Calculated using model domain mean simulation data for seven mean rainfall types and by rainfall types calculated using grid-scale simulation data for each mean rainfall type.

 Domain Mean Simulation Data
 26.57234.78916.6278.6316.2074.3012.387
 Grid-Scale Simulation Data
TFM12.2649.56711.1177.7677.2826.7214.376
TFm15.67521.71614.37517.99015.89519.3387.330
tFM19.01619.39022.06316.60019.18115.17018.153
tFm12.17416.16811.29417.16612.66720.10810.091
TfM33.98325.40333.58931.16537.19427.62844.101
Tfm3.3215.2924.5516.1594.6428.16410.502
tfM3.5672.4653.0123.1533.1382.8715.446
Table 6. Percentage of Rain Amount Over Total Rainfall Amounta
PhaseTFMTFmtFMtFmTfMTfmtfM
  • a

    Calculated using grid-scale data during the onset phase (2000–2200 LST 23 December), the development phase (2300 LST 23 December to 1400 LST 24 December), the mature phase (1500 LST 24 December to 0900 LST 25 December), and the decay phase (1000–1800 LST 25 December) in a rainfall case from 2000 LST 23 December to 1800 LST 25 December 1992.

Onset0.00020.6784.43825.87046.8361.3480.831
Development27.35723.44711.7047.75325.0632.1482.529
Mature7.32214.05325.17415.91129.7403.2454.556
Decay3.41610.68929.1019.67538.7276.2302.162

[24] The spatial-scale dependence of precipitation statistics implies spatial-scale dependence of precipitation efficiency (PE), which can be demonstrated through the comparison of rainfall source between the accumulation from grid-scale data over the model domain for PE2 and the direct calculation from model domain mean data for PE1 in Figure 7. The precipitation efficiency is defined based on surface rainfall budget (1) by Sui et al. [2007]. The PE2 is much smaller than PE1 because the model domain mean rainfall source is much smaller than the grid cumulative rainfall source over the entire model domain. This is due to the fact that the rainfall source is largely offset by the sink in the model domain mean calculations.

Figure 7.

Precipitation efficiency is calculated by accumulating rainfall sources from each model grid over the model domain in PE2, whereas it is calculated using model domain mean data in PE1. Precipitation efficiency is defined based on surface rainfall budget (1) by Sui et al. [2007].

[25] One of the applications of the analysis of precipitation statistics is the evaluation of convective stratiform rainfall partitioning schemes. These classic separation schemes, like the scheme used in this study, largely incorporate the intensities of radar reflectivity or rainfall. Table 7 shows that convective rainfall regions include a large amount of rainfall with water vapor divergence whereas stratiform rainfall regions include a large amount of rainfall with water vapor convergence. This indicates that a convective stratiform rainfall separation scheme using rainfall intensity may fail to adequately distinguish between convective and stratiform rainfall.

Table 7. Percentage of Rain Amount Over Total Rainfall Amounta
 TFMTFmtFMtFmTfMTfmtfM
  • a

    Calculated using grid-scale data over convective and raining stratiform regions (convective and raining stratiform regions identified using the convective stratiform rainfall separation scheme originally developed by Tao et al. [1993], modified by Sui et al. [1994], and used in the cloud-resolving model in this study).

Convective13.8222.7115.1116.6525.285.011.41
Stratiform2.808.3826.939.6241.225.016.04

[26] As described by X. Cui and X. Li (A cloud-resolving modeling study of short-term surface rainfall processes, submitted to Meteorology and Atmospheric Physics, 2010), the model domain mean surface rainfall lags the mean water vapor convergence associated with imposed large-scale vertical velocity by about 3 hours. Calculations of lag correlation between the rainfall and the mean water vapor convergence show that rainfall with local atmospheric drying, water vapor convergence, and hydrometeor loss/convergence lags the mean water vapor convergence by 4 hours, whereas rainfall with local atmospheric drying, water vapor convergence, and hydrometeor gain/divergence lags the mean water vapor convergence by 2 hours (Table 8). The other types of rainfall lag the mean water vapor convergence by 3 hours. This suggests that the maximum surface rainfall created by rainfall with local atmospheric drying, water vapor convergence, and hydrometeor loss/convergence occurs an hour later than the maximum mean surface rainfall does and the maximum cloud growth associated with rainfall with local atmospheric drying, water vapor convergence, and hydrometeor gain/divergence occurs an hour before the maximum mean surface rainfall.

Table 8. Maximum Lag Correlation Coefficients and Lag Hours Between Rainfall Type and Model Domain Mean Water Vapor Convergence
 Lag Correlation CoefficientLag Hours
TFM0.444
TFm0.602
tFM0.673
tFm0.643
TfM0.663
Tfm0.413
tfM0.463

[27] Caution should be exercised since the analysis is carried out using 2D model simulation data. Further examination of three-dimensional model simulation data is necessary to demonstrate generalizations from the results of the two-dimensional analysis.

Acknowledgments

[28] The authors thank W. -K. Tao at NASA/GSFC for his cloud resolving model, M. Zhang at SUNY, Stony Brook for his TOGA COARE forcing data, and three anonymous reviewers for their constructive comments. This study is supported by the National Key Basic Research and Development Project of China under Grant 2009CB421503, the National Natural Science Foundation of China under Grant 40775033, the Chinese Special Scientific Research Project for Public Interest under Grants 41075039 and GYHY200806009, and the Qinglan Project of Jiangsu Province of China under Grant 2009.

Ancillary

Advertisement