Climatology of drizzle in marine boundary layer clouds based on 1 year of data from CloudSat and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO)



[1] A survey of the frequency and characteristics of precipitation from low clouds over the oceans based on data from CloudSat and CALIPSO from July 2006 through June 2007 is presented. The low-cloud fraction, drizzle occurrence, and estimated cloud base precipitation rate are examined globally and for eight subtropical and midlatitude stratocumulus (Sc) regions. This analysis is restricted clouds below 4 km. Drizzle detection and characterization is further restricted to clouds with tops above 1 km altitude. The maximum radar reflectivity within an individual CloudSat profile (Zmax) is used to classify the profile as precipitating or nonprecipitating. The distribution of Zmax for all profiles is bimodal with peaks around −23 and −12 dBZ interpreted as originating from populations of nondrizzling and drizzling clouds, respectively. Profiles where Zmax exceeds −18 dBZ are classified as drizzling. Drizzle is detected for 19–34% of cloudy profiles in the subtropical Sc regions and 37–44% of profiles from the midlatitude Sc regions. The cloud base precipitation rate is estimated using the relation: Rcb = 2 · Zmax0.7. The lower quartile/median/upper quartile precipitation rates are 0.25/0.6/2.0 mm d−1 for the subtropical Sc regions and 0.28/0.7/2.3 mm d−1 for midlatitude regions. A consistent nighttime increase in low-cloud fraction and drizzle occurrence is observed for the subtropical Sc regions. For clouds with reff > 17 μm the drizzle occurrence can be treated as a function of LWP alone and exceeds 50% (75%) for a LWP of 50 (110) g m−2. For reff < 17 the drizzle occurrence is strongly dependent on both LWP and reff.

1. Introduction

[2] Marine stratocumulus and other low clouds over the oceans are known to have a significant cooling effect on global climate. The net radiative effect of low clouds on the global energy budget has been estimated at −15 W m−2 and the sensitivity to changes in global low-cloud coverage at −0.63 W m−2 [Hartmann et al., 1992] for each percent increase in low-cloud amount. Parcels within the boundary layer traverse the subtropical Sc regions over the course of a few days before the stratocumulus layer breaks up into a field of trade-wind cumulus [e.g., Albrecht et al., 1995]. Thus, the characteristics of the cloud layer at a given location are the result not only of the initial conditions but also reflect the cumulative effects of an ensemble of boundary layer and cloud-microphysical processes.

[3] That light precipitation, on the order of 1 mm d−1 at cloud base, can impact the microphysical structure, diurnal variability, and evolution of the stratocumulus-topped boundary layer has been recognized for more than 2 decades [e.g., Nicholls, 1984, 1987]. Effects of drizzle include: removal of liquid water from the cloud layer leading to reductions in entrainment and inhibition of mixing between the cloud and subcloud layers [e.g., Stevens et al., 1998], scavenging of Cloud Condensation Nuclei (CCN) [Ackerman et al., 1993; Wood, 2006], cooling and moistening of the subcloud layer as a result of evaporation which acts to further inhibit mixing between the cloud and subcloud layers and which, under certain conditions can lead to decoupling of the cloud and subcloud layers [Nicholls, 1984, 1987] or to the development of a poorly mixed boundary layer where the cloud and subcloud layers are linked by spatially and temporally intermittent convective elements [Stevens et al., 1998]. Recently it has been hypothesized that drizzle plays a key role in driving the rapid transition from unbroken cloud and closed-cellular organization to an open-cellular structure and broken cloud [Petters, 2004; Sharon et al., 2006; Stevens et al., 2005].

[4] Despite the longstanding recognition of the myriad effects of drizzle, and in large part as a result of limited capabilities for detecting and measuring light precipitation, drizzle has received relatively little attention compared to processes such as entrainment. Observations from several recent field studies including: the Atlantic Stratocumulus Transition Experiment (ASTEX) [Albrecht et al., 1995], the East Pacific Investigation of Climate stratocumulus study (EPIC) [Bretherton et al., 2004], and the Second Dynamics and Chemistry of Marine Stratus (DYCOMS-II) [Stevens et al., 2003] and Coastal Stratus field campaigns [Vali et al., 1998] have shown that drizzle is a nearly ubiquitous feature of marine and near-coastal Sc with observed cloud base precipitation rates frequently reaching 1 mm d−1 [Comstock et al., 2004; van Zanten et al., 2005] (hereafter abbreviated CWYB04 and VZ05 respectively). These observations have helped drive a resurgence of interest in drizzle and its impacts on the stratocumulus-topped boundary layer. However, observations from individual field campaigns are inherently ill-suited to assessing the frequency and strength of drizzle or its impacts on low-cloud properties on a global scale, a limitation exacerbated by the concentration of field studies in a few regions, particularly within a few hundred kilometers of the California coast.

[5] In contrast to quantities such as Low Cloud Fraction (LCF), effective radius (reff), and Liquid Water Path (LWP), which can be obtained on a near daily basis from spaceborne sensors such as MODIS, little is known about the frequency, extent, or strength of drizzle on a global scale. Indirect methods for detecting drizzle based on differences in the vertical profiles of LWP and reff between drizzling and nondrizzling Sc have been described [Masunaga et al., 2002a, 2002b; Shao and Liu, 2004] but have not been extensively validated using direct observations of drizzle.

[6] Fox and Illingworth [1997] used in situ cloud droplet spectra data to examine the possibility of detecting stratus/stratocumulus using spaceborne cloud radars and concluded that a system with a sensitivity of −30 dBZ would be able to detect >80% of marine stratocumulus with liquid water contents greater than 0.025 g m−3. This ability has recently been realized with the launch, in mid-2006, of the CloudSat and CALIPSO satellites, part of NASA's A-Train constellation [Stephens et al., 2002].

[7] In this paper, the LCF, Drizzle Occurrence (DO), and the estimated cloud base precipitation rate (Rcb) are examined using CALIPSO-derived cloud top heights, radar reflectivity from the Cloud Profiling Radar (CPR), MODIS retrievals of LWP and reff, and additional data on the atmospheric structure from ECMWF.

[8] A cursory description of the CloudSat and CALIPSO data sets, and the conditions used to calculate LCF, determine DO, and estimate Rcb is given in section 2. The cloud and drizzle characteristics of the individual Sc regions are discussed in section 3. Seasonal variability and day-night differences in LCF, drizzle occurrence and characteristics are examined in sections 4 and 5 respectively. A cursory examination of the dependence of drizzle on cloud properties is presented in section 6. Finally, some concluding remarks and prospective avenues for future work are presented in section 7. A list of acronyms and symbols used in this paper is given in Table 1.

Table 1. List of Acronyms and Symbols
LWPliquid water pathg m−2
CCNcloud condensation nuclei 
reffeffective radiusμm
Ndcloud droplet number concentrationcm−3
Rcbestimated precipitation rate at cloud basemm d−1
Zmaxmaximum reflectivity within a CPR profiledBZ
LCFlow cloud fraction 
DOdrizzle occurrence 
LTSlower tropospheric stability°C
Z-RReflectivity-rain rate relationship 
CALIPSOCloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation 
CPRCloud Profiling Radar 
ECMWFEuropean Center for Medium-range Weather Forecasting 
MODISModerate Resolution Imaging Spectroradiometer 

2. Data and Methods

[9] Data from CloudSat and CALIPSO from July 2006 through June 2007 is used in conjunction with ancillary data from ECMWF and Aqua-MODIS. The CALIPSO L2 1-km resolution vertical feature mask [Winker et al., 2003] is used to compute the LCF and to calculate average cloud top heights for CALIPSO. CPR reflectivities are used to detect drizzle and provide a rough estimate of precipitation rate at cloud base. These data are combined onto a 2.5° × 2.5° latitude-longitude grid for mapping and global analyses and are aggregated by region for the eight subtropical and midlatitude Sc regions.

[10] Each 2.5° × 2.5° grid box is sampled by an average of 5–8 CloudSat orbits per month resulting in an average of ∼1000 CloudSat profiles per month (day + night) for each grid box. The average number of orbits (profiles) per month for each of the eight Sc regions is listed in Table 2 and ranges from 53 orbits (105 profiles) for the Canarian region to 640 orbits (106 profiles) for the Circumpolar region.

Table 2. Sc Regionsa
RegionLatitudeLongitudeArea (×106 km2)Orbits (Profiles) per Month (D+N)
  • a

    Region numbers correspond to labels in Figure 3.

1Californian10°N30°N150°W110°W9.176 (1.4 × 105)
2Namibian30°S0°S2°5W15°E14.179 (2.0 × 105)
3Canarian10°N30°N45°W20°W5.753 (0.94 × 105)
4Australian40°S15°N60°E115°E14.8111 (2.5 × 105)
5Peruvian30°S0°S120°W70°W17.697 (2.5 × 105)
6North Pacific30°N55°N180°W130°W11.2106 (2.3 × 105)
7North Atlantic30°N60°N45°W10°W9.090 (1.9 × 105)
8Circumpolar55°S40°S180°W180°E44.4640 (9.6 × 105)

[11] The dependence of DO and Rcb on LWP and reff is examined using visible/near-infrared retrievals from Aqua-MODIS [King et al., 1997] which are available for the daytime portions of each orbit.

[12] The difference between the ECMWF 700 mb and 1000 mb potential temperatures is used as a measure of the stability of the lower troposphere: LTS = θ700−θ1000. The stability of the lower troposphere has been shown to be the primary factor underlying changes in low-cloud amount [e.g., Klein and Hartmann, 1993; Wood and Bretherton, 2006; Wood and Hartmann, 2006] and has been used in parameterizations to predict the low-cloud amount in global circulation models [Rasch and Kristjansson, 1998; Slingo, 1987].

2.1. Low-Cloud Detection and Drizzle Identification

[13] In this study we consider low clouds to be those having a cloud top <4 km. Separate cloud layers above that altitude may also be present. The LCF is defined as the fraction of profiles where a cloud top below 4 km is identified in either the CALIPSO feature mask or in the CloudSat 2B-GeoProf Cloud-Mask. The maximum reflectivity within a single CPR profile (<4 km altitude), denoted Zmax, is used to classify profiles as drizzling/nondrizzling and to estimate the cloud base precipitation rate.

[14] No attempt was made to stratify profiles by cloud type in this study. Thus, the results presented here are an aggregate all low-cloud types. Warren et al. [1988] (hereafter W88), analyzed ship based observations of cloud type and coverage included in the Comprehensive Ocean-Atmosphere Data Set (COADS) [Woodruff et al., 1987] to map the frequency of occurrence and amount as a function of cloud type over the oceans. The W88 climatology divided clouds into six types, including three low-cloud types (defined as having a base below 2500 m): Stratus (St), Cumulus (Cu), and Cumulonimbus (Cb). At latitudes above 30° St dominated all other cloud types, while at lower latitudes, the zonally averaged cloud amount due to Cu and St was comparable, with the amount of St slightly exceeding that of Cu outside the ITCZ (W88). The distribution of Cu average cloud amount is fairly uniform zonally with maximum values occurring in the central ocean basins, whereas St are concentrated toward the eastern reaches of the Atlantic and Pacific oceans. On the basis of the W88 climatology, we expect that the average amount of stratiform cloud (St or Sc) is at least twice that of Cu in the subtropical Sc regions. In this analysis we use the term stratocumulus (Sc) as a catchall for all low-cloud types. While the St or Sc classification is representative of the predominant cloud type in the regions examined, other low-cloud types may effect the distribution of precipitation rates. Seasonal variations and day-night differences in the relative frequency of occurrence of different cloud types may underlie some of the observed variability in DO, LCF, and precipitation statistics.

[15] The CPR Level 2B-GeoProf Cloud-Mask (revision R04) is used to eliminate CPR bins with weak, spurious, or surface contaminated signals. The GeoProf Cloud-Mask algorithm incorporates signal-to-noise ratio and spatial continuity to characterize the strength and validity of the signal in each CPR bin on a 10-level scale including flag values for missing and surface contaminated data [Marchand et al., 2008] (hereafter MMAS08). In this analysis, use of the CPR data is restricted to bins with a Cloud-Mask value ≥30 (see MMAS08 Table 1 for a listing of the CloudMask values and their meanings), a criteria designed to have a <2.5% incidence of false detection and which has been shown to have a <5% incidence of false detection as compared to CALIPSO (MMAS08).

[16] The minimum height above the ocean surface for which the CPR reflectivity can be considered good depends on the characteristics of the CPR (pulse length, shape of the leading edge of the transmitted pulse) and on the strength of the meteorological signal relative to the surface return. As of revision R04, an estimate of surface contamination is subtracted from the signal in the 4 lowest bins above the ocean surface. The cloud mask is set to a value of 5 for bins where the corrected signal is insufficiently strong relative to the uncertainty in the estimated surface return (MMAS08). These bins are eliminated by the Cloud Mask ≥ 30 threshold. Prior to (following) this correction, rain and heavy drizzle can be detected at altitudes of ∼720 m (∼480 m) and above, while moderate drizzle can be detected at altitudes of ∼860 m (∼480 m) and above (MMAS08; CloudSat 2B GEOPROF Quality Statement: May 2007 (Version R04)). Since inclusion of altitudes where only strong (or moderate) drizzle could be detected would introduce a bias the detection and characterization of drizzle, use of the CPR reflectivities is restricted to profiles with CALIPSO-derived cloud top heights ≥1 km.

[17] Distributions of cloud top height for the eight Sc regions are shown in Figure 1. Cloud top heights generally lie between 0.4 and 1.8 km altitude. Thus, the exact value of the cloud top height cutoff is expected to have a minimal impact. However, the cloud top height ≥1 km restriction for the detection and characterization of drizzle excludes a significant fraction of low clouds, particularly in eastern reaches of the subtropical Sc regions. In the subtropical and midlatitude Sc regions the fraction of cloudy profiles with cloud tops <1 km ranges between 14% and 24% of nighttime cloudy profiles and between 19% and 36% of daytime cloudy profiles.

Figure 1.

Distribution of CALIPSO-derived cloud top heights for the eight Sc regions. Horizontal dashed line indicates the 1-km cloud top height cutoff for analysis of the CloudSat CPR data.

[18] The probability distribution function (PDF) of Zmax, shown in Figure 2, is bimodal with peaks around −23 and −12 dBZ, which we interpret as the result of distinct populations of nondrizzling and drizzling Sc (the peak around −23 dBZ may be an artifact of the minimum detectible signal of the CPR and the CloudMask threshold used). A threshold value of Zmax > −18 dBZ, corresponding to the approximate location of the minimum separating the nondrizzling and drizzling peaks, was used as the indicator of drizzle. We define the DO as the number of profiles where Zmax exceeded −18 dBZ divided by the number of profiles with CALIPSO-derived cloud top heights between 1 and 4 km. Because of this normalization, the DO can be used directly to compare the probability that a cloudy profile is drizzling between regions, day and night, and seasons. Absolute quantities, such as the areal extent of drizzle, can be computed by scaling the DO by the LCF, subject to an assumption regarding the DO for clouds with tops <1 km.

Figure 2.

PDF of Zmax for points with cloud top height ≥1 km. Vertical dashed line marks the −18 dBZ threshold used to indicate the presence of drizzle. PDF is computed using the 1-year CloudSat data set for each stratocumulus region.

[19] The selection of a reflectivity threshold for identifying precipitation is of interest for since the presence of a drizzle mode in the droplet size distribution, which tends to dominate radar reflectivity (the sixth moment of the drop-size distribution for droplets in the Rayleigh regime) but contributes minimally to lower moments of the droplet size distribution, breaks the linkage between reflectivity and LWC [e.g., Fox and Illingworth, 1997; Matrosov et al., 2004]. Wang and Geerts [2003] analyzed in situ droplet size spectra and near-coincident radar data to determine thresholds ranging from ∼−20 dBZ to −15 dBZ depending on the cloud droplet number concentration (Nd) and height within the cloud layer. The separation between the contribution to reflectivity from cloud droplets and the contribution from drizzle was greatest in the lower parts of the cloud layer [Wang and Geerts, 2003]. The persistence of the minimum in the PDF of Zmax, evident in Figure 2, appears to be the result of two complementary factors. First, radar reflectivity at a given level in cloud is higher when drizzle is present: Frisch et al. [1995] and Baedi et al. [2002] reported jumps in radar reflectivity of +16 and +10 dBZ respectively between nondrizzling and drizzling Sc. Second, in warm, nonprecipitating clouds reflectivity increases rapidly with height above cloud base [Baedi et al., 2002], whereas in precipitating clouds reflectivity increases from cloud top downward to cloud base and then decreases below cloud base owing to evaporation. Thus, for clouds of a given thickness, averaging over the 500 m pulse length (greater than the thickness of most nondrizzling cloud layers) results in the CPR-measured reflectivity being reduced more (relative to the maximum reflectivity at any level in cloud) for nondrizzling cloud layers than for drizzling ones.

2.2. Precipitation Rate Estimation

[20] An estimate of the precipitation rate at cloud base (Rcb) is obtained by applying a Reflectivity-Rain rate (Z-R) relationship based on an average of the cloud base parameters reported by VZ05 for DYCOMS-II and similar to the values reported by CWYB04,

equation image

where Zmax has units of mm6 m−3 and the Rcb is expressed in units of mm d−1. The −18 dBZ threshold corresponds to a cloud base precipitation rate of ∼0.1 mm d−1, while a radar reflectivity of 0 dBZ (typically the threshold for “heavy drizzle”) corresponds to a precipitation rate of 2 mm d−1.

[21] This Z-R relationship does not take into account the characteristics of the CPR, particularly the 500 m pulse length, which will result in a reduction in Zmax compared to the reflectivity at cloud base [Baedi et al., 2002]. Thus, Rcb obtained from (1) will be systematically lower than would be obtained using the radar reflectivity at cloud base. Additional factors that may impact the Z-R relationship include cloud type, the size of the CPR footprint, and attenuation. Quantitative application of the Rcb values presented here, for instance in the development of parameterizations for use in numerical models, is discouraged as more accurate, better characterized estimates of Rcb are expected to be forthcoming. The estimate of Rcb presented here is intended for qualitative applications such as assessing the importance of drizzle on global and regional scales.

[22] Some perspective on the significance of the Rcb values presented can be gained by noting that a precipitation rate of 1 mm d−1 (1 kg m−2 d−1) corresponds to a latent heat flux of slightly less than 30 w m−2. This precipitation rate, which is equivalent to a mass flux of 1 kg m−2 d−1, is also sufficient to deplete typical cloud water contents (50–200 g m−2) on a timescale of a few hours and to reduce the boundary layer CCN concentration by ∼100 cm−3 d−1 [Wood, 2006].

[23] The precipitation rate at the ocean surface is generally less than half that at cloud base (CWYB04, VZ05) with drizzle frequently evaporating completely before reaching the surface. The relation between cloud base reflectivity and the surface precipitation rate is highly sensitive to both cloud base height and the size distribution of drizzle drops at cloud base (CWYB04). No estimate of the precipitation rate at the ocean surface is made in this analysis.

[24] The distribution of Zmax spans nearly four decades from the −18 dBZ threshold to +20 dBZ and is positively skewed such that strong, but infrequent drizzle events contribute disproportionately to the average precipitation rate. VZ05 and CWYB04 reported similar skewness in radar and radar-lidar derived precipitation rates in marine Sc. Since the motivation for studying drizzle is its effects on the characteristics and evolution of the cloud layer, rather than its contribution to annual precipitation, we use the median and upper and lower quartiles (abbreviated M50%, Q25% and Q75% respectively and collectively referred to as quartiles) rather the mean to examine seasonal variations and regional differences in the distributions of Zmax and Rcb.

3. Stratocumulus Regions

[25] The eight subtropical and midlatitude Sc regions examined in this study are listed in Table 2 and shown alongside the LCF Figure 3. These regions correspond to those analyzed by Klein and Hartmann [1993] but are defined somewhat differently in terms of geographical area. With the exceptions of the (small) Canarian region and the (large) Circumpolar region, these regions range in size from 9 × 106 to 17 × 106 km2. Four of these regions (the Californian, Canarian, Namibian, and Peruvian) are classic subtropical marine Sc layers that occur where strong trade inversions limit mixing between the BL and free atmosphere. Three midlatitude Sc regions are also examined: the North Atlantic, North Pacific, and Circumpolar. The distinction between the subtropical and midlatitude Sc regions and the geographic boundaries of the individual regions is somewhat arbitrary: the Australian region, generally categorized as subtropical, appears to share characteristics of both the midlatitude and subtropical regions, while parts of the North Atlantic and North Pacific Sc regions might be better categorized as subtropical. the Arctic stratus/stratocumulus are not considered here, in part because the prevalence of mixed-phase clouds complicates comparison of radar reflectivities with those from warm clouds in the subtropical and midlatitude Sc regions.

Figure 3.

Single-year averages for (left) day and (right) nighttime. (a) LCF. (b) Average cloud top height. (c) Fraction of low clouds with tops <1 km. Sc regions are marked by white boxes and numbered in Figure 3a (left). Subtropical regions: 1, Californian; 2, Namibian; 3, Canarian; 4, Australian; 5, Peruvian. Midlatitude regions: 6, North Pacific; 7, North Atlantic; 8, Circumpolar.

[26] The 1-year average of LCF, cloud top height and DO for the Sc regions are listed in Table 3 along with the quartiles of Zmax/Rcb. The subtropical and midlatitude Sc regions have comparable LCF, but are distinguished by larger day-night differences in LCF and cloud top height in the subtropical regions and the significantly larger DO observed for the midlatitude Sc regions as can be seen in Figure 3.

Table 3. Single-Year Average Values for the Sc Regionsa
RegionLCFCloud Top Height (km)Cloud Top < 1 kmDOZmax (dBZ) 25%/50%/75%Rcb (mm d−1) 25%/50%/75%LWP (g m−2)reff (μm)
  • a

    Values are averages of day and nighttime observations with the exception of LWP and reff, which are only available for daytime and are computed for cases where low clouds are present.

North Pacific64%1.623%40%−12.2/−6.7/0.60.29/0.68/2.313116.5
North Atlantic60%1.722%37%−12.0/−5.7/2.20.29/0.80/2.814416.4

[27] The Californian, Namibian, and Peruvian Sc regions each consist of a well-defined, contiguous core where the 1-year average daytime (nighttime) LCF exceeds 70% (85%). LCF in the Australian region increases from northwest to southeast with no clear distinction between this region and the adjacent Circumpolar region. Maximum LCF in the Canarian region lie in a band roughly parallel to the NE–SW orientation of the African coastline, but displaced ∼1000 km to the west.

[28] Cloud top heights for the Californian, Namibian, Peruvian, and Canarian Sc regions have a well-defined gradient with average cloud top heights increasing with distance away from the coastline (Figure 3). The gradient in cloud top heights is adjacent to the Australian west coast is minimal and poorly defined. In the Californian region, the minimum 1-year average cloud top heights (about 0.6 km) are found adjacent to the coastline in the region of the California Bight (32°N−35°N, 115°W–125°W). Average cloud top heights (Figure 3) increase gradually away from the coast up to an average of ∼1.5 km at 140°W. The Namibian region follows a similar pattern with 1-year average cloud top heights increasing from 0.5 to 0.6 km around the Namibia-Angola border (17°S, 12°E) to 1.25 km at the prime meridian. The Peruvian stratocumulus region is generally deeper with 1-year average cloud top heights increasing from 0.8 to 0.9 km along the coastline to 1.25 km ∼ 1000 km west of the coastline and increasing more gradually farther west. The disproportionate impact of the 1 km cloud top height cutoff for detection of drizzle on the near-coastal portions of the Californian, Namibian, and Peruvian Sc regions is shown in Figure 3c. As a result of the persistent spatial patterns evident in Figure 3, the LCF, DO, and other values computed for the Californian, Namibian, and Peruvian Sc regions are expected to depend somewhat on the exact geographical definitions of these regions. In contrast, values computed for the Australian and midlatitude Sc regions should be only minimally affected by the choice of boundaries.

[29] The three midlatitude Sc regions lack the well-defined geographical patterns in LCF and cloud top height that characterize the subtropical regions. The DO, shown in Figure 4, is higher in the midlatitude than in the subtropical regions. This difference appears to be largely attributable to differences in LWP between the midlatitude and subtropical regions, which are on the order of 100–150 g m−2 compared to <100 g m−2 in the subtropical regions. The DO increases significantly at night in the subtropical regions. This increase is concentrated in the core regions of the Californian, Namibian, and Peruvian regions. Since the LCF and DO both increase at night, there is a net increase of 35–45% in the areal extent of detectible drizzle for the subtropical regions.

Figure 4.

(a) (left) DO for the daytime observations and (right) DOs for the nighttime observations. DOs are not shown for grid points where >50% of cloud top heights are below 1 km. (b) MODIS retrievals of LWP. (c) MODIS retrieved reff. LWP and reff are average values for daytime profiles where low clouds are present. Individual Sc regions are outlined in white and are identified in Figure 3 and listed in Table 2.

[30] Drizzle rates are fairly consistent for the Subtropical regions with Rcb (in mm d−1) of 0.25 ± 5% (Q25%), 0.6 ± 10% (M50%), 2 ± 15% (Q75%) with the Q25%/M50%/Q75% values moving together for different regions (i.e., regions having the highest M50% values also have the highest Q25% and Q75%). Drizzle is stronger in the midlatitude regions with Rcb (in mm d−1) of 0.29 ± 5% (Q25%), 0.75 ± 10% (M50%), 2.5 ± 10% (Q75%). Comparison of the of the Rcb quartiles listed in Table 3 with previously reported precipitation rates and distributions is complicated by differences in methodology and presentation. Since the values reported here are computed on a profile-by-profile basis, while in situ measurements of drizzle rate are typically reported as flight- or leg-averaged values (including drizzling and nondrizzling segments), we expect the values reported to be higher than leg- or flight-averaged values. Nonetheless, median Rcb values in Table 3 for the subtropical Sc regions (∼0.6 mm d−1) lie within the range of mean Rcb values (0.38–1.29 mm d−1 including drizzling and nondrizzling regions) reported by VZ05 for DYCOMS-II flights where drizzle was present. The Q75% Rcb values (2.3–2.5 mm d−1) for midlatitude Sc are consistent with the ∼2 mm d−1 conditionally averaged precipitation rate (for segments where the drizzle rate exceeded 0.5 mm d−1) values reported by Wood [2005a]. Both VZ05 (Figure 11b) and CWYB04 (Figure 9) reported a strong positive skewness in the precipitation rate such most of the precipitation originated from a small fraction of the total area: VZ05 showed that precipitation rate 70 m above the ocean surface was skewed such that >70% of the total precipitation originated from <25% of the drizzling area, while CWYB04 found that 50% (75%) of the cloud base precipitation originated from 5% (22%) of the area where the precipitation rate exceeded 9 (2.4) mm d−1. These observations are broadly consistent with the skewness evident in factor of ∼10 between the Q25% and Q75% Rcb values listed in Table 3. The VZ05 and CWYB04 results and, to a lesser degree, the quartiles of Zmax/Rcb presented in Table 3 emphasize the importance of heavily drizzling cores embedded within more weakly drizzling Sc and point out the need to consider the effects of the horizontal averaging over the CloudSat footprint on the drizzle statistics.

4. Seasonal Variability

[31] Monthly and seasonally averaged values for the eight Sc regions are used to examine seasonal variability in the occurrence and characteristics of drizzle. Seasons are referred to by initialisms of the constituent months: March–April–May (MAM), June–July–August (JJA), September–October–November (SON), and December–January–February (DJF). The monthly averaged LCF, DO and the distribution of Zmax are shown in Figure 5. The peak LCF for the Californian, North Atlantic, and North Pacific occur in JJA. The Circumpolar region also has a maximum LCF during these months, however seasonal variability in LCF is minimal for this region (∼10% of the mean). Maximum LCF for the Namibian and Peruvian regions occurred in SON. The seasons and values of peak LCF are in general agreement with those reported by Klein and Hartmann [1993], although the timing of the maximum LCF for the Canarian and Australian regions differs by a few months. Monthly averaged values of cloud top height, cloud top temperature, and LTS are shown in Figures 5d5f. Correlations between increasing LTS and increasing LCF have been widely reported [e.g., Klein and Hartmann, 1993; Wood and Bretherton, 2006]. A strong correlation between LTS and LCF for the subtropical Sc regions (other than the Australian) is readily apparent in Figures 5a and 5d. A weaker correlation between LTS and LCF is evident for the subtropical regions.

Figure 5.

Monthly average values for the eight Sc regions. Daytime (nighttime) values are marked by a solid (dashed) line. Horizontal dashed line marks the 1-year combined day and night average value for each region. (a) LCF. (b) DO. (c) Box-and-whiskers plot of the monthly distribution of Zmax for the Sc regions. Corresponding Rcb values are shown on the right axis. The bottoms and tops of the boxes mark the Q25% and Q75% values. The horizontal mark in the box marks the median. Whiskers are plotted at the 5% and 95% percentiles. Monthly mean values of Zmax are marked by solid circles. (d) Monthly average LTS. (e) Cloud top height. (f) Cloud top temperature.

[32] Seasonal variability is also apparent in DO, particularly for the midlatitude Sc regions (Figure 5b). For these regions the DO is clearly anticorrelated with LTS, while in the subtropical regions the there is no clear correlation between LTS and DO, nor does the max/min DO consistently occur in the same season as for LWP (Table 4). For the North Atlantic and North Pacific regions DO is at a minimum in JJA at the same time that the LCF is at its peak (the maximum DO for the circumpolar region also occurs in JJA). Interestingly, the amplitude of the seasonal cycle in DO for the midlatitude Sc regions, particularly the North Pacific, is larger for the daytime than for the nighttime observations.

Table 4. Maximum/Minimum Values of Low Cloud and DOs, Q25%, Median, and Q75% of Zmax and Rcb, Average LWP, reff When Low Clouds are Present, and the Seasons During Which the Maximum/Minimum Values Occurreda
RegionLCFDOZmax (dBZ) (25%/50%/75%)Rcb (mm d−1) (25%/50%/75%)LWP (g m−2)reff (μm)
  • a

    For each region, top row shows maximum and bottom row shows minimum. Note that the maximum/minimum seasons for the quartiles and median of Rcb are the same as for Zmax.

Californian73% (JJA) 66% (SON)33% (JJA) 25%(MAM)−13.0/−7.31/0.6 (SON) −13.3/−8.3/−1.3(MAM)0.25/0.6/2.2 0.24/0.52/1.698 (SON) 79 (MAM)16.3 (SON) 15.5 (JJA)
Namibian77% (SON) 60% (MAM)33% (SON) 29% (JJA)−12.9/−7.5/0.2 (MAM) −13.3/−8.7/−2.4 (SON)0.25/0.6/2.1 0.23/0.49/1.489 (JJA) 78 (DJF)16.2 (MAM) 13.9 (SON)
Canarian58% (MAM) 43% (SON)20% (DJF) 18% (SON)−12.6/−6.2/2.1 (SON) −13.5/−8.4/−0.6 (JJA)0.26/0.74/2.8 0.23/0.51/1.898 (DJF) 66 (JJA)18.4 (SON) 14.9 (JJA)
Australian67% (SON) 61% (DJF)37% (JJA) 27% (DJF)−12.4/−6.4/1.8 (JJA) −12.8/−7.3/0.3 (SON)0.27/0.71/2.6 0.25/0.61/2.1128 (JJA) 90 (DJF)18.6 (JJA) 16.0 (DJF)
Peruvian74% (SON) 64% (MAM)37% (JJA) 29% (DJF)−12.7/−6.9/0.8 (MAM) −12.9/−7.9/−1.2 (SON)0.26/0.66/2.3 0.25/0.55/1.6104 (JJA) 84 (SON)17.0 (JJA) 15.2 (DJF)
North Pacific72% (JJA) 55% (DJF)44% (DJF) 36% (JJA)−11.4/−4.9/2.1 (DJF) −12.9/−8.5/−2.3 (JJA)0.32/0.91/2.8 0.25/0.5/1.4156 (DJF) 110 (JJA)18.7 (DJF) 15.3(MAM)
North Atlantic63% (JJA) 57% (SON)44% (DJF) 31% (JJA)−11.3/−4.4/3.8 (DJF) −12.7/−7.6/−0.7 (JJA)0.32/0.99/3.7 0.26/0.59/1.8173 (DJF) 114 (JJA)18.0 (SON) 14.9 (JJA)
Circumpolar67% (JJA) 64% (DJF)50% (JJA) 36% (DJF)−11.6/−5.6/2.0 (JJA) −12.5/−6.9/0.5 (DJF)0.31/0.81/2.7 0.26/0.65/2.1179 (JJA) 122 (DJF)19.2 (JJA) 14.8 (DJF)

[33] Monthly variations in the distribution of Zmax are shown in Figure 5c. The distributions of Zmax for the different regions are generally similar (a characteristic also evident in Figure 2). However, much greater seasonal variability is evident for the North Atlantic and North Pacific regions with seasonal max-min differences of 1.4 dBZ (Q25%), 3.7 dBZ (M50%), and 4.5 dBZ (Q75%) and 0.07 mm d−1 (Q25%), 0.4 mm d−1 (M50%), and 1.8 mm d−1 (Q75%), while the seasonal max-min differences for the Circumpolar and Australian regions are smaller they follow a similar pattern marked by a reduction in the strength of drizzle across all quartiles and that is clearly correlated with the seasonal cycle in DO such that the strength of drizzle also increases with increasing DO. In contrast, a much weaker seasonal variability in the distribution of Zmax is observed for the subtropical regions and is not clearly correlated with either DO, LCF, or LTS.

[34] The seasonal variability in DO for the subtropical regions does not follow a consistent pattern: for the Australian and Peruvian regions the DO peaks in phase or possibly slightly ahead of the peak in LCF, while the peak DO for the Californian region lags the peak LCF by ∼2 months, and for the Nambian region there is a peak in the nighttime DO in SON, in phase with the maximum LCF, but the daytime DO peaks in JJA. Given the variety of patterns encountered, analysis of multiple years of data will be needed to differentiate between seasonal and interannual variations.

5. Day-Night Differences

[35] Day-night differences are examined using the regionally averaged or gridded global data from the daytime and nighttime portions of each orbit. Numerous observations have shown a pronounced diurnal cycle in LCF and LWP over the oceans [Bretherton et al., 2004; Ciesielski et al., 2001; Fairall et al., 1990; Rozendaal et al., 1995; Wood et al., 2002; Zuidema and Hartmann, 1995; Zuidema et al., 2005] with peak LCF and LWP occurring at night. The diurnal cycle in LCF and LWP has generally been attributed to shortwave absorption within the cloud layer [Blaskovic et al., 1991; Turton and Nicholls, 1987] an effect which may be amplified for deeper boundary layers that become partly or completely decoupled during daytime [Turton and Nicholls, 1987]. The timing of peak LCF and LWP varies by region but tends to peak between 2 and 4 h with respect to local sun time [Rozendaal et al., 1995; Wood et al., 2002]. The sun-synchronous, polar orbit used by the A-train satellites [Stephens et al., 2002] and the resulting, early morning/early afternoon overpass times of the A-train satellites are displaced only ∼2 h from the maximum/minimum times in the diurnal cycle of LWP [Wood et al., 2002]. Thus, the day-night differences in LCF examined here are expected to resolve most of the amplitude of the diurnal cycle and have the additional advantage of not aliasing seasonal variability onto the diurnal cycle and visa versa [Salby and Callaghan, 1997; Wood et al., 2002].

[36] Comstock et al. [2005, Figure 4] reported a maximum (minimum) in DO at 7 (15) hours local time during EPIC with the minimum occurring simultaneously with the peak decoupling. Given the importance of local effects in the Peruvian Sc region [Garreaud and Munoz, 2004], the broader applicability of these results is unclear. If the timing reported by Comstock et al. [2005] holds for other regions, then day-night differences in DO obtained from CloudSat will vastly underestimate the magnitude of the diurnal cycle. Alternatively, given the dependence of drizzle on LWP [Pawlowska and Brenguier, 2003; van Zanten et al., 2005; Wood, 2005b] it may also be reasonable to assume that peak drizzle production occurs roughly in phase with peak LWP.

[37] Global day-night differences in LCF and DO are shown in Figure 6. These differences are small in the midlatitude regions and much larger in the subtropical regions (particularly the Californian, Namibian, and Peruvian). The day-night difference in DO appears to be restricted to the core region of the Californian, Namibian and Peruvian regions. In contrast, the difference in LCF is more consistent, although a region of enhanced day-night LCF difference displaced offshore from the core regions can be seen in Figure 6a. Garreaud and Munoz [2004] show that the timing of the overnight minimum in divergence varies across the Peruvian Sc region. Such a consistent propagating feature (the “upsidence wave”) would instead appear as a geographical pattern in day-night differences with maximum day-night differences where the sampling times of the A-train occurred in phase with the disturbance.

Figure 6.

(a) Day-night difference in LCF. (b) Day-night difference in DO. The differences are computed on a 2.5° × 2.5° grid and have been smoothed using a boxcar average over a nine-grid-point neighborhood (7.5° × 7.5°). Negative values indicate larger LCF or DO at night.

[38] Regionally averaged day-night differences in LCF and DO are listed in Table 5 and can also be seen in Figures 5a and 5b. All the regions show an increase in nighttime LCF and DO. The day-night difference in LCF is larger for the subtropical than for the midlatitude Sc regions. Nighttime LCF for the subtropical regions are higher than the daytime LCF by 0.1–0.15, while for the midlatitude regions the increase in nighttime LCF is ∼0.05. There is some suggestion, evident in Figure 5, that for the midlatitude Sc regions the day-night differences are seasonally dependent with the largest (smallest) day-night differences in cloud fraction and DO in JJA (DJF) for the North Atlantic and North Pacific regions and in DFJ (JJA) for the circumpolar region.

Table 5. Day−Night Differences in Cloud Fraction, Cloud Top Height, and DO
 LCF, %Cloud Top HeightCloud Top < 1 km, %DO, %
North Pacific6462671.6−20+20232521403841
North Atlantic6056631.7−10+10222420373742

[39] The subtropical Sc regions show a consistent increase in cloud top height at night, generally between by 100 and 200 m, although average values such as these can be misleading owing to day-night differences in the prevalence of >1 km and <1 km cloud modes evident in Figure 7. Large day-night differences are observed for the Canarian region, but these are the result of a day-night switch in the predominant cloud mode, rather than an increase in the height of clouds within one of the modes of the distribution.

Figure 7.

Distribution of Cloud top heights for the subtropical Sc regions. Solid line with symbols is combined day + night distribution. Dotted line is night only. Dash-dotted line is day only. Note that the day-only and night-only values are normalized by one half the total number of day and night points to facilitate direct comparison.

[40] While it is not possible to unambiguously separate increases in LCF (left-right shifts of the distribution) from increases in cloud top height (vertical shifts in the distribution) it appears that a upward shift of 100–200 m depending on region, coupled with a small increase in LCF, would account for most of the day-night differences in the distribution above 1 km. Below 1 km the day-night differences are more consistent with a significant daytime increase in LCF for clouds below 1 km coupled with a nighttime increase in cloud top height for clouds below 1 km. In contrast to the fairly consistent behavior of the >1 km mode (with the exception of the Canarian region), the nighttime increase in cloud top height for the <1 km mode varies significantly between regions.

[41] Bretherton et al. [2004] reported a similar (∼200 m) nighttime increase in BL depth in the Peruvian Sc region based on soundings taken every 3 h during the EPIC field campaign, contrary to expectations that the diurnal cycle in LWP was primarily attributable to changes in cloud base height resulting from the diurnal cycle of solar radiation coupled in some regions with a diurnal cycle in decoupling [Turton and Nicholls, 1987]. Garreaud and Munoz [2004] modeled the diurnal circulation above and adjacent to the South American continent and found a significant diurnal cycle in subsidence over the Peruvian Sc region consistent with the observed nighttime increase in cloud top height reported by Bretherton et al. [2004]. Dai and Deser [1999] examined surface wind observations and found that the maximum divergence over land regions occurred around dawn, while that over the adjacent ocean had a maximum in the evening. The persistent nighttime increase in cloud top height for the subtropical Sc regions (Figure 7) suggests that diurnal cycles in subsidence may underlie the diurnal cycles in LCF and LWP to a greater degree than generally recognized, a conclusion also hinted at by Wood et al. [2002] and Rozendaal et al. [1995] for respect to the Peruvian and Namibian Sc regions.

[42] Inspection of the PDFs of cloud top height shown in Figure 7 reveals an unexpected bimodality with peak densities at ∼1.3 km altitude and 0.7–0.9 km. At night the LCF associated with the ∼1.3 km peak increases accompanied by an upward shift (by some 200 m) in this part of the distribution. In contrast, the LCF associated with the 0.7–0.9 km mode actually increases during the day. The anticorrelated day-night differences in LCF for the high and low modes accounts for the peculiarities in the combined day+night PDFs of cloud top height shown in Figure 7. The day-night difference in cloud top height for the midlatitude and Australian regions is smaller, with average nighttime increases in cloud top height of <50 m, but broadly similar to the differences for the subtropical regions in that the nighttime increase in LCF is concentrated above the ∼1.3 km peak while there is a slight daytime increase in LCF below 1 km. However, this difference may also be due to a small nighttime increase in cloud top height for clouds below 1 km since no separate sub-1-km mode is apparent. Day-night differences in cloud top height and cloud top temperature for the subtropical regions are small, but anticorrelated generally consistent with overnight deepening of the boundary layer and the increased prominence of the <1 km mode during the day.

[43] The quartiles of Zmax and Rcb for the day and nighttime observations are listed in Table 6. In the subtropical regions changes in the distribution of Zmax and Rcb are concentrated toward lower end of the distribution. The Q25% of Zmax is ∼0.6 dBZ higher at night, while M50% is higher by ∼0.7 dBZ and Q75% is essentially unchanged. For the midlatitude regions, Q25% is ∼0.3 dBZ higher at night, while M50% is ∼0.2 dBZ higher at night and Q75% is actually higher during the day (by ∼0.2 dBZ). Thus, day-night differences in drizzle consist mainly of slight increases in drizzle strength for the weaker drizzle events

Table 6. Quartiles and Median of Zmax and Rcb for Day Versus Nighttime Observations
RegionZmax (dBZ) 25%, 50%, 75%Rcb (mm d−1) 25%, 50%, 75%
CalifornianDay−13.5, −8.3, −0.40.23, 0.52, 1.9
Night−12.7, −7.5, −0.40.26, 0.60, 1.9
NamibianDay−13.6, −8.7, −1.20.22, 0.49, 1.6
Night−12.8, −7.7, −0.90.25, 0.56, 1.7
CanarianDay−13.2, −7.7, 0.30.24, 0.58, 2.1
Night−12.9, −7.3, 0.30.25, 0.62, 2.1
AustralianDay−12.9, −7.3, 0.90.25, 0.62, 2.3
Night−12.5, −6.7, 1.10.27, 0.68, 2.4
PeruvianDay−13.3, −8.2, −0.70.24, 0.53, 1.8
Night−12.5, −7.1, −0.20.27, 0.63, 1.9
North PacificDay−12.5, −6.8, 0.80.27, 0.67, 2.3
Night−12.1, −6.6, 0.40.29, 0.69, 2.1
North AtlanticDay−12.1, −5.8, 2.30.29, 0.79, 2.9
Night−11.9, −5.7, 2.10.29, 0.80, 2.8
CircumpolarDay−12.3, −6.4, 1.30.27, 0.72, 2.5
Night−12.0, −6.1, 1.20.29, 0.75, 2.4

6. Dependence of Drizzle on Cloud Properties

[44] The combined LWP and reff from Aqua-MODIS are used with Rcb and Zmax from the CPR to determine under what conditions drizzle is likely and how its strength varies as a function of LWP and reff. To address these questions profiles with low clouds, where MODIS retrievals of LWP and reff were available, and where the cloud top height was ≥1 km, were binned by LWP and reff using bin widths of 10 g m−2 and 1 μm respectively for examination of the DO and 20 g m−2 and 2 μm for examination of M50% values of Zmax and Rcb. LWP bins extended from 0 to 400 g m−2, while reff bins ranged from 5 to 30 μm. An additional restriction, that cloud top temperature be ≥273 K, was applied to profiles used for examination of the dependence of Zmax and Rcb on cloud properties. Misleading results that might arise for LWP, reff pairs with relatively few observations were avoided by truncating the results where the number of points for each bin was less than 2.5% of the maximum number of points in any bin. This truncation had the effect of eliminating bins below a line running from 6 μm reff and an LWP of 10 g m−2 to an reff of 15 μm and an LWP of 360 g m−2. Less than 5% of the total profiles were eliminated by this truncation. Broken clouds, which are recognized to result in systematic overestimates of reff [e.g., Marshak et al., 2006], were not eliminated from this analysis and may significantly impact the results, especially for larger reff.

[45] Contours of the DO as a function of LWP and reff are shown in Figure 8. The DO for the subtropical and midlatitude Sc follows a similar pattern, but the DO increases more rapidly as a function of LWP for the subtropical Sc regions. For reff larger than ∼17 μm the DO depends primarily on LWP, while for reff < 17 μm the DO depends on both LWP and reff. The DO, shown in Figure 9 as a function of LWP for 19 ≤ reff ≤ 22 μm, increases rapidly for LWP below ∼100 g m−2 reaching a typical value of 0.75 at a LWP of 100 g m−2. Above 100 g m−2 the DO increases more gradually before reaching a plateau of ∼0.9 for LWP above ∼250 g m−2. Zuidema et al. [2005, Figure 9] reported a similar increase in DO as a function of LWP and strong correlation between DO and LWP. However, the DO shown in Figure 9 is ∼50% higher than reported by Zuidema et al. [2005] for LWP < 250 g m−2 despite both analyses using similar reflectivity thresholds (−17 versus −18 dBZ). Plotting DO versus LWP for all reff (not shown) reduces this discrepancy slightly, but the underlying cause remains unclear. Spatial averaging over the CloudSat footprint is one possible candidate (owing to the positively skewed distribution of reflectivity CloudSat exceeds the drizzle threshold owing to the inclusion of a few strong fall streaks within the sample volume), but an in depth analysis of the spatial statistics will be needed to determine what fraction of the discrepancy in DO could be explained by differences in spatial averaging.

Figure 8.

(a) Contours of the DO as a function of LWP and reff for all subtropical Sc regions. (b) Same as Figure 8a, but for the midlatitude Sc regions.

Figure 9.

DO as a function of LWP for 19 ≤ reff ≤ 22 μm.

[46] The drizzle rate is generally recognized to be to be strongly dependent on LWP and less-strongly dependent on Nd [Pawlowska and Brenguier, 2003; Comstock et al., 2004; van Zanten et al., 2005] although the forms of the parameterization of precipitation rate in terms of cloud properties vary and the resulting precipitation rates vary even more dramatically [Wood, 2005b]. Assuming an adiabatic liquid water profile LWPh2, so the Pawlowska and Brenguier [2003] and VZ05 relationships can be expressed as RcbLWP2Nd−1 and RcbLWP1.5Nd−1 respectively, comparable to RcbLWP1.75Nd−1.75 reported by CWYB04. Given the acknowledged limitations of the analysis presented here (use of an off-the-shelf Z-R relationship and inclusion of broken clouds and accompanying biases n reff), our intent is to investigate the possibility that a single, simple parameterization can adequately represent drizzle from all marine stratocumulus [Wood, 2005b] or whether a set of parameterizations for different regions and conditions is needed rather than the actual determination of such a parameterization. Median Zmax and Rcb values for the combined subtropical and midlatitude Sc regions are shown in Figure 10. The consistent and relatively simple dependence of the median of Zmax and Rcb on LWC and reff suggests that a simple parameterization of drizzle in terms of these quantities should be possible, but that different parameterizations will be needed for the subtropical and midlatitude Sc regions. As with the DO, the median values of Zmax and Rcb have a similar form, but different gradients in terms of LWP/reff. The difference between Q25% and Q75% is fairly consistent at about 8 dBZ. Thus, any parameterization of Rcb in terms of cloud properties will be subject to an uncertainty in excess a factor of 2 (although some of this uncertainty results CPR specific factors, such as the 500 m vertical averaging). The median values of Zmax and Rcb are more strongly dependent on reff than the DO. Some curvature in the contours of Zmax and Rcb is apparent, but less than in the contours of DO.

Figure 10.

(a) Median Zmax as a function of LWP and reff for all subtropical Sc regions. (b) Same as Figure 10a but for the midlatitude Sc regions. (c) Median Rcb for all subtropical Sc regions. (d) Same as Figure 10c but for midlatitude Sc regions.

7. Conclusions

[47] Data from the CloudSat and CALIPSO satellites has been examined to determine LCF and DO in eight subtropical and midlatitude Sc regions. The distribution of Zmax is bimodal with peaks around −23 and −12 dBZ. While the existence of distinct distributions corresponding to nondrizzling and drizzling clouds is not unexpected given the results of Baedi et al. [2002] and Frisch et al. [1995], it is nonetheless reassuring in that it facilitates the selection of a single threshold for identifying drizzle while reducing the impact of the exact threshold used. The CPR reflectivity exceeds the ad hoc threshold of −18 dBZ for ∼30% (40%) of subtropical (midlatitude) Sc with tops ≥ 1 km. As such, this analysis reaffirms the prevalence of drizzle in marine Sc reported by Vali et al. [1998], VZ05, and Bretherton et al. [2004] among others. While the DO values reported here suggest that drizzle is a frequent, rather than ubiquitous, feature of marine Sc, it is important to note that, especially in strongly precipitating Sc, drizzle occupies only a fraction of the total area (VZ05). CWYB04 reported that only 41% of the area at cloud base exceeded the −12 dBZ detection threshold of the scanning C-band radar. Thus, the fraction of Sc affected by drizzle is larger than indicated by the DO since drizzling and nondrizzling patches frequently exist side-by-side and are linked by circulations within the boundary layer [Comstock et al., 2007; Kropfli and Orr, 1993; Leon, 2006].

[48] A number of differences between the subtropical and midlatitude Sc regions are evident. Both the LCF and the DO are higher in the midlatitude Sc regions. Drizzle also tends to be stronger in the midlatitude regions as measured by the quartile values (−12/-6/+1.5 dBZ average for the midlatitude regions versus −13/−7.5/0 dBZ average for all subtropical Sc regions). Not surprisingly, the North Pacific and North Atlantic Sc regions also have a larger and more consistent seasonal cycle in LCF, DO, drizzle strength, and underlying cloud properties. Seasonal variations in the distribution of drizzle rates are concentrated toward the high end of the distribution of Zmax with a ∼4.5 dB difference between the max/min seasonal Q75% values for the North Pacific and North Atlantic regions. These increases in the strongest drizzle events occur in phase with the maximum in DO.

[49] In contrast, seasonal variations are relatively minor for the subtropical Sc regions, but day-night differences are more pronounced and more consistent for the subtropical regions. The LCF and DO both increase at night resulting in a net increase of ∼40% the areal extent of detectible drizzle at night. Day-night differences in the distribution of Zmax are concentrated toward the lower end of the distribution with a nighttime increase in the median Rcb of about 0.1 mm d−1 in the subtropical Sc regions.

[50] The importance of the subtropical Sc regions from the perspective of global climate is widely accepted. However, the midlatitude Sc regions have received substantially less attention despite having a comparable impact on global climate [Klein and Hartmann, 1993]. Our analysis reaffirms the prevalence of low clouds in the midlatitude regions and that drizzle is likely to be particularly important in determining cloud characteristics in the midlatitude Sc regions.

[51] The work presented here is preliminary and there are many promising avenues for future research, in addition to addressing specific limitations the analysis presented here. The most pressing of these needs is for the development and validation of a CloudSat specific Z-R relationship including an estimate of the uncertainties in Rcb. Given the inherent difficulty in collecting in situ observations within the single vertical slice sampled by CloudSat and CALIPSO the most promising approach would apparently be to simulate the effects of the CloudSat and CALIPSO sampling on existing data sets such as those from the DYCOMS-II and EPIC field campaigns.

[52] The effect of surface-contamination of the CPR reflectivity for altitudes below 1 km and the resulting implications for the analysis of low clouds should also be revisited. Profiles from the near-coastal portions of the subtropical Sc regions are disproportionately excluded owing to the decrease in cloud top heights toward the adjacent continents. Wood and Hartmann [2006, Figure 16] showed that the preferred types of MCC structure shifted from closed cells toward open-cellular or cellular-but-disorganized structures as the boundary layer depth increased above 1 km. Additionally, shallower boundary layers (less than ∼700–750 m deep) are less prone to decoupling [Bretherton and Wyant, 1997]. Thus, the results presented here may not be representative of shallower Sc layers, including the extensively studied region adjacent to the California coast.

[53] Stratification of the CloudSat and CALIPSO data set by cloud type is likely to shed light not only on the differences in DO and drizzle characteristics between cloud types, but also allow the contributions to the day-night differences and seasonal variations in DO, LCF, and drizzle characteristics due to variations in the relative frequency of occurrence of different cloud types to be separated from variations within a single cloud type. Finally, stratification by type of MCC organization for clouds identified as marine Sc may prove enlightening, particularly with respect to relationships between precipitation, cloud properties, aerosol spectra, and mesoscale organization.


[54] This work was supported by grant JPL48251 from NASA-JPL. We are indebted to the CloudSat and CALIPSO science teams and project managers whose efforts made this work possible. We would specifically like to thank the CloudSat PI Graeme Stevens and Deputy-PI Deb Vane, and CALIPSO PI Dave Winkler. The efforts of Simone Tanelli in characterizing and correcting for surface contamination in the CPR reflectivity are of particular importance to this study. Comments and questions from Rob Wood and two anonymous reviewers helped improve this manuscript.