The first 2 year measurements from Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) lidar and CloudSat radar were analyzed to study the distribution and phase partition of midlevel liquid-layer topped stratiform clouds (MLTSC, top higher than 2.5 km above the Earth's surface and top temperature warmer than −40°C) globally. The global mean MLTSC occurrence was ∼7.8% and the global mean MLTSC percentage fraction related to all midlevel clouds was ∼33.6%. Strong seasonal and day-night variations of MLTSC occurrence were observed over different latitude regions. In the polar regions, the maximum occurrence was in summer, while the minimum occurred in winter, with small day-night differences. In the tropics, a high MLTSC occurrence band shifted southward from June–July–August to December–January–February with significantly more MLTSC during the nighttime. The global mean MLTSC top height and temperature were ∼4.5 km above the surface and −13.6°C. Overall, 61.8% of MLTSCs were mixed phase and 12.4% were supercooled liquid (contains only liquid phase or with ice below the detection limit). The fraction of mixed-phase MLTSC increased as the cloud top temperature decreased, with a sharp increase between −10 and −15°C and a noticeable latitude difference. This temperature dependence indicated that ice nucleation is active at −10°C in these clouds. The global mean ice water path (IWP) of mixed-phase MLTSCs, estimated based on an empirical temperature–radar reflectivity–ice water content relationship, was ∼13.4 g/m2, and the IWP increased as cloud top temperature decreased. To improve MLTSC parameterizations in global climate models, further studies are needed to better understand the latitude dependence of MLTSC distributions and microphysical properties and how aerosol and water phase cloud properties affecting ice generation in MLTSCs.
 Midlevel clouds have cloud base above 2 km from the Earth's surface and below 7 km or less depending on latitude [World Meteorological Organization, 1988]. Previous studies show that midlevel clouds cover approximately 22% of the Earth's surface [Warren et al., 1986, 1988; Rossow and Schiffer, 1999] and play an important role in the Earth's radiation budget [Sun and Shine, 1995; Hogan et al., 2003]. However, because these clouds seldom produce strong precipitation and severe weather [Gedzelman, 1988], the study of midlevel clouds is often neglected, and their macrophysical, microphysical, and radiative properties still are documented and understood poorly. As a result, these clouds are called “forgotten clouds” in atmospheric cloud studies [Vonder Haar et al., 1997]. In addition, the representations of midlevel clouds as well as other clouds in global climate models (GCMs) are poor and are regarded as the largest uncertainty in predicating future climate change [Intergovernmental Panel on Climate Change, 2007]. Model simulation and observation comparison studies [Zhang et al., 2005; Illingworth et al., 2007] show that almost all GCMs underpredict midlevel cloud occurrence. However, satellite midlevel cloud observations with passive sensors could be affected by high clouds [Hahn et al., 2001]. New active observations from space have demonstrated potentials to better detect midlevel clouds globally [Mace et al., 2009].
 One of the main challenges for better understanding and simulating midlevel clouds is that they often are presented in a mixed-phase state when the cloud top extends above the freezing level. Recently, mixed-phase, midlevel stratiform clouds (i.e., altocumulus, altostratus, “altostratocumulus”) have generated a great deal of interest among atmospheric scientists because their less complex dynamics provide a test system for cloud microphysical theories and represent a “simple” scenario for studying ice generation characteristics and growth in mixed-phase clouds [Field, 1999; Fleishauer et al., 2002; Heymsfield et al., 2006]. Previous observations showed that midlevel stratiform mixed-phase clouds have a thin liquid water-dominated mixed-phase layer at the top and deep ice virga falling below it in both midlatitude and the polar regions [Hobbs et al., 2001; Hogan et al., 2003; Wang et al., 2004; Shupe et al., 2006; Carey et al., 2008]. At temperatures between 0°C and −40°C, liquid water droplets and ice crystals may coexist and persist for long periods of time [Rauber and Tokay, 1991; Shupe et al., 2008]. These complicated three-phase transformations in mixed-phase clouds still are not well understood, and their parameterizations in GCMs are particularly difficult [Gregory and Morris, 1996; Klein et al., 2009]. In addition, the liquid water-dominated layer at the top of midlevel stratiform clouds, although always thin, has a much larger radiative impact than ice clouds of the same water content, especially for solar radiation [Sassen and Khvorostyanov, 2007]. Furthermore, the supercooled liquid water droplets in midlevel clouds may present an icing hazard for small aircraft [Cober et al., 2001]. A global-scale analysis of midlevel stratiform clouds and their phase partition could advance our understanding of those questions and improve the parameterizations of mixed-phase clouds in GCMs.
 Currently, midlevel stratiform clouds and their vertical structures can be easily identified from active remote sensing measurements (i.e., lidar, radar). The existing studies of midlevel stratiform clouds are mainly based on in situ measurements, model simulations, ground-based remote sensing or satellite observations. In situ measurements provide direct information on midlevel cloud and aerosol properties and their dynamic environments [Heymsfield et al., 1991; Fleishauer et al., 2002; Carey et al., 2008]. However, in situ measurements usually cover only short time periods and limited regions, making it difficult to accumulate a large database. Cloud resolving models are useful for investigating the radiative properties of midlevel stratiform clouds responding to their microphysical structures and developing better parameterizations for midlevel stratiform clouds in GCMs [Liu and Krueger, 1998; Larson et al., 2001; Sassen and Khvorostyanov, 2007], but models' results are strongly dependent on model microphysical processes parameterizations and require robust observations for validation. Ground-based remote sensing approaches provide long-term, continuous, high-resolution observations of midlevel stratiform clouds [Hogan et al., 2003; Wang et al., 2004]. Shupe et al.  gave a detailed review of ground-based remote observations for mixed-phase clouds, but ground-based remote sensing measurements are limited to small regions over land and are easily blocked by low-level, optically thick clouds. Satellite observations, especially with active sensors, provide a long-term global perspective of midlevel stratiform clouds. Hogan et al.  estimated the global distribution of stratiform supercooled liquid water clouds using Lidar In-Space Technology Experiment (LITE) measurements. Lidar signal, however, is dominated by water phase and might be fully attenuated by the liquid water-dominated mixed-phase layer at the top of clouds, and thus, spaceborne lidar alone has difficulties providing complete cloud vertical structure and cloud phase information. Moreover, their results were limited by the short (53 h) operation period of LITE [Winker et al., 1996]. Fortunately, Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) lidar and CloudSat radar measurements are now available and provide unique data to study midlevel stratiform clouds and their vertical structures [Stephens et al., 2002; Winker et al., 2003]. The different sensitivities of lidar and radar toward water droplets (small size, high concentration) and ice particles (large size, low concentration) allow for the reliable detection and phase determination of midlevel stratiform clouds by combining CloudSat and CALIPSO measurements.
 In this study, collocated CALIPSO and CloudSat measurements were used to investigate global distributions of midlevel liquid-layer topped stratiform clouds (MLTSC). The aim of this work was to study the occurrence and phase partition of MLTSCs from a statistical perspective. The paper is organized as follows: A brief description of the CALIPSO and CloudSat data sets, methods for collocating CALIPSO and CloudSat data, criteria for identifying MLTSCs from the collocated CALIPSO/CloudSat measurements, and algorithms for identifying midlevel stratiform clouds as supercooled liquid and mixed phase are given in section 2. In section 3, the global distributions of MLTSC occurrence and percentage fraction, MLTSC occurrence seasonal and day-night variations, vertical distributions of mixed-phase MLTSC fractions in different latitude bands, and ice water path (IWP) distributions in mixed-phase MLTSCs are presented. Finally, the conclusions and directions for future work are given.
2. Data Analysis
2.1. CALIPSO and CloudSat Measurements
 The NASA A-Train [Stephens et al., 2002] satellite constellation consists of a group of satellites orbiting close together. The satellites of the A-Train constellation are equipped with a variety of Earth observation instruments and provide a unique perspective on global distributions of aerosols, clouds, temperature, relative humidity, and radiative fluxes. More importantly, these measurements are complementary to each other and offer a relatively comprehensive view of the Earth's atmosphere. The CloudSat and CALIPSO satellites were launched together in April 2006 to join in the A-Train constellation and have been acquiring data since June 2006.
 The CALIPSO satellite payload includes three nadir-viewing instruments: the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP), the Infrared Imaging Radiometer (IIR), and the Wide Field Camera (WFC) [Winker et al., 2003, 2007]. CALIOP lidar measurements provide high vertical resolution profiles of aerosols and clouds (especially optically thin clouds) globally. The vertical and horizontal resolutions of the CALIOP are 30 and 333 m below 8.2 km, and 60 and 1000 m between 8.2 and 20.2 km, respectively. CALIOP is also the first spaceborne polarization lidar, providing linear depolarization measurements at 532 nm [Winker et al., 2003]. The primary measurements (level 1 data) of CALIOP are attenuated atmospheric backscattering coefficients at 532 and 1064 nm and linear lidar depolarization ratio profiles at 532 nm. CALIPSO level 2 science products are also available [Vaughan et al., 2004; Anselmo et al., 2007]. In this work, CALIPSO level 1B data sets were used for the detection of aerosol layers, cloud top boundaries, and the liquid water-dominated layers at the top of MLTSCs.
 CloudSat carried the first spaceborne cloud radar: a 94.05 GHz Cloud Profiling Radar (CPR) with sensitivity of ∼−30 dBZ [Stephens et al., 2002, 2008]. The CPR measurements provide radar reflectivity (Ze) profiles with a vertical resolution of ∼240 m and a footprint of 1.4 km × 1.8 km (cross and along track). By combining radar measurements with CALIOP lidar, MODIS measurements, and other ancillary data, CloudSat provides global cloud distributions and vertical profiles, cloud types, liquid and ice water content profiles, particle effective radius measurements, and precipitation data [Sassen and Wang, 2008; Stephens et al., 2008]. In this work, CloudSat level 2B-GEOPROF and 2B-GEOPROF-lidar products [Mace et al., 2009] were used for the determination of cloud structure and ice particle information in mixed-phase MLTSCs.
 The formation flying of CloudSat and CALIPSO ensures that the CALIOP footprint overlaps with the CPR footprint more than 90% of the time [Stephens et al., 2008]. This makes it possible to collocate CloudSat and CALIPSO measurements by averaging CALIOP profiles within a given CPR footprint. MODIS level 1 radiance, together with cloud retrievals and temperature from the European Centre for Medium-Range Weather Forecasting (ECMWF) forecast model, were also collocated with CloudSat and CALIPSO measurements for analysis. The data analyzed in this work were collected between June 2006 and May 2008.
2.2. Identification of MLTSC Based on Collocated CALIPSO/CloudSat Measurements
 Because of the highly reflective liquid layer at the top of MLTSC, the cloud top boundary is easily identified from lidar measurements by using the threshold method [Winker and Vaughan, 1994]. The criteria for identifying MLTSC from collocated CALIPSO/CloudSat data included the following.
 1. For midlevel clouds, the cloud base height is greater than 2 km above the surface (based on the world meteorological organization definition); however, CALIOP lidar generally cannot penetrate the liquid layer to detect the water layer base below. Considering that MLTSCs are physically thin, lidar observed cloud top height (i.e., higher than 2.5 km above the surface) was used to determine which clouds were midlevel. To include all possible stratiform mixed-phase clouds, the top limit of midlevel cloud is set at −40°C rather than using latitude depending height.
 2. A liquid water layer at the top of MLTSCs is detected if there is a strong peak lidar backscattering near the cloud top (maximum attenuated lidar backscatter coefficient exceeding a threshold of 0.06 sr−1 km−1) and strong attenuation of lidar signal after the peak (small “cloud geometrical depth” observed by lidar, i.e., less than 500 m). Due to strong multiple scattering effect in CALIPSO depolarization measurements and interference of low lidar depolarization ratio produced by horizontally oriented ice crystals [Winker et al., 2007], CALIPSO depolarization ratio measurements are not used for water layer detection.
 3. For clouds to be regarded as stratiform clouds, relatively flat cloud tops are required, for example, with a small standard deviation of cloud top height in midlevel cloud systems (i.e., smaller than 300 m). A midlevel cloud system is defined as a group of midlevel horizontally adjacent cloud profiles with similar vertical coverage; however, in some cases, the cloud tops decreased or increased linearly along the track (for example, the MLTSCs between profiles 120 to 180 in Figure 1). In these cases, the linear trend was subtracted to calculate cloud top standard deviation.
 4. Because MLTSCs generally do not produce precipitation [Gedzelman, 1988], it was required that there would be no strong radar signal extending the entire distance from the cloud top to the surface. Figure 1 shows an example of identified MLTSCs from collocated CALIPSO/CloudSat measurements. Obvious ice appearances were observed among and below the liquid-dominated top layer, as indicated by strong radar reflectivity factor (Ze) measurements, which most likely represent a typical mixed-phase cloud structure in the midtroposphere and polar boundary layer.
2.3. Mixed-Phase Cloud Determinations Based on Collocated CALIPSO/CloudSat Measurements
 Because liquid and ice particles in clouds have different physical properties (e.g., sizes, shapes, densities, terminal velocities, etc.) which greatly impact their radiative properties and precipitation efficiencies, determining the phase of clouds is critical for understanding the impacts of clouds on the atmospheric radiation balance and hydrological cycle. Midlevel water clouds are often supercooled, and ice crystals can be initiated in the layer when cloud top temperatures fall below approximately −6°C [Hobbs and Rangno, 1985]. These crystals then grow larger and fall out of the layer to form common midlevel mixed-phase clouds. Although lidar depolarization measurements are adequate for identifying this type of mixed-phase cloud from ground [Wang and Sassen, 2001], it is challenging to achieve this with CALIPSO lidar due to the strong attenuation of lidar signals from the top liquid-dominated layer, as well as to strong multiple lidar scattering. Fortunately, collocated CloudSat measurements provide an alternative means for determining the presence of ice particles in MLTSCs as illustrated in Figure 1. For cloud droplets and small ice crystals, CloudSat Ze (mm6 m−3) is proportional to the sixth power of particle diameter within the Rayleigh scattering regime. In mixed-phase clouds, since the ice particles are typically much larger than liquid water droplets, the contribution from ice particles dominates radar reflectivity even in the presence of substantial liquid water droplets [Shupe et al., 2006; Sassen and Khvorostyanov, 2007]; therefore, the radar measurements provide useful information for detecting ice particle in midlevel mixed-phase stratiform clouds.
 Temperature-dependent Ze thresholds were developed and applied to CloudSat measurements to identify ice occurrence in MLTSCs. For drizzle-free stratiform clouds, many empirical relationships between Ze and liquid water content (LWC) were developed based on in situ aircraft observations. As suggested by Fox and Illingworth , the Ze-LWC relationship can be expressed as
 The unit of LWC is g/m3. Thus, Ze can be approximately estimated with this relationship for drizzle-free stratiform clouds with a given LWC; therefore, for a given cloud top temperature (CTT) and thickness, the layer possible maximum Ze can be estimated by using cloud layer maximum adiabatic LWC. If CTT is colder than −6°C and the CloudSat CPR radar-observed Ze is larger than the estimated maximum Ze, ice crystals are likely to be present in the liquid water-topped cloud layer; that is, the MLTSC is likely to be mixed phase. The estimated layer maximum Ze decreases as cloud top temperature (CTT) decreases and drops below −28 dBZ at ∼−15°C for a typical water layer thickness of 400 m. Considering CloudSat sensitivities of ∼−29 dBZ, −28 dBZ was used as the threshold when the estimated maximum Ze was smaller than −28 dBZ. This combined temperature-dependent threshold was represented in Figure 2 as the dashed line. Meanwhile, Figure 2 also shows the occurrence of MLTSCs in terms of cloud top temperature (CTT) and observed layer maximum Ze (Ze_max) from collocated CALIPSO/CloudSat measurements. The occurrence was normalized for each temperature bin. These thresholds enabled the detection of 100 μm (diameter) ice particles at a concentration of ∼1/L. Because ice particles can grow quickly under mixed-phase conditions, ice crystals in these clouds typically generate Ze values much higher than these thresholds, as illustrated in Figure 2. Considering 500 m CloudSat radar pulse width, Ze threshold estimated based on maximum LWC is larger than actual measured Ze for the assumed adiabatic clouds. An increase/decrease of the Ze threshold by 2 dBZ cause ∼2% to 10% difference in the determined mixed-phase cloud amount depending on temperature as is shown in Figure 2.
3.1. Global Distributions of MLTSCs
 The global distribution of MLTSC occurrence is presented in Figure 3 (top). The MLTSC occurrence, as calculated from collocated CAPLIPSO/CloudSat measurements, is defined as the ratio of the MLTSC profile amount over all collocated CALIPSO/CloudSat profile amounts in a 2.5 × 2.5° box. The 2 year global mean of MLTSC occurrence was found to be roughly 7.8%. A large land-ocean contrast in occurrence was observed: 6.1% (land) versus 8.6% (ocean). High MLTSC occurrence regions were found in the tropics and in polar regions. In the tropics, MLTSCs were mainly distributed over land, e.g., northern South America and southern Africa. In north midlatitude regions, high MLTSC occurrence was located over eastern China. This MLTSC distribution pattern in the tropics and midlatitudes was similar to the altocumulus cloud distribution observed from the first year CloudSat measurements [Sassen and Wang, 2008]. A high occurrence of MLTSCs was also observed over the southern ocean surrounding the Antarctic. Overall, the Southern Hemisphere (7.9%) had almost the same occurrence of MLTSCs as the Northern Hemisphere (7.7%). It should be noted that in some cases, such as thick anvil clouds, the lidar signal will be attenuated significantly and will fail to detect MLTSC; therefore, our result might be biased low for regions with high occurrence of optically thick anvil associated with deep convective clouds. Area coverage of these clouds, however, is relatively small and mainly in the tropics [Sassen and Wang, 2008].
 To show the relative occurrence of MLTSC, the MLTSC percentage fraction related to all midlevel clouds was also calculated based on collocated CloudSat 2B-GEOPROF-Lidar product [Mace et al., 2009] in a 2.5 × 2.5° box. Figure 3 (bottom) shows the global distribution of MLTSC percentage fraction. The global mean MLTSC percentage fraction was 33.6%. A large land-ocean contrast in MLTSC percentage fraction was observed: 28.3% (land) versus 35.9% (ocean). Compared with MLTSC occurrence distribution, MLTSC percentage fraction distribution shows smaller spatial contrast. Although MLTSC only count for ∼1/3 of the total midlevel clouds globally, its overall contribution to midlevel clouds is larger than this. After the top water layer dissipates at their late lifecycle, the remaining ice-phase clouds can stay in the air for a longer time in many situations. Ice nuclei activated in MLTSC and released after ice evaporation could be activated downwind more easily to favor ice cloud formation [Vali, 1996]. A high-percentage fraction near the west coast of North America and a lower percentage fraction on land show this possible impact. MLTSCs have a higher contribution to midlevel clouds in the Southern Hemisphere (37.3%) than in the Northern Hemisphere (30.5%) though MLTSCs have similar mean occurrences over both hemispheres. This trend also was observed by Hogan et al.'s  study. The Southern Hemisphere also showed smaller spatial variations in percentage fraction.
Figure 4 shows seasonal variations in global MLTSC occurrence distributions. These seasonal variations were noticeably different over different regions. In the polar regions, occurrence was highest in summer (JJA for the Arctic and DJF for the Antarctic) and lowest in winter (DJF for the Arctic and JJA for the Antarctic). This seasonal variation may be related to the water vapor supply from lower latitudes and to warmer summer temperatures. Over the Arctic, there was also a large seasonal variation in geographic distribution. During summer, these clouds were found evenly over land and ocean while in winter they mainly located over oceans and coastal regions and extended over land downwind of oceans. In north midlatitude regions, more MLTSCs were observed during winter than summer although magnitudes of seasonal variations varied by region. In the tropics, a band of MLTSC shifting southward from JJA to DJF was observed. This seasonal shift followed the seasonal change of deep convective clouds, which are primary sources of water vapor supply for the tropical middle troposphere. However, exact mechanism on how deep convection supply water vapor for tropical MLTSC formation is needed to be better understood. Detrainment from deep convection and cumulus congestus was regarded as an important source of midlevel water vapor, but Yasunaga et al.  argued that melting within the stratiform cloud rather than the detrainment of deep convection is the source of water vapor for midlevel thin clouds in the tropics.
Figure 5 presents the daytime (1300 to 1400 local time) and nighttime (0200 to 0400 local time) distributions of MLTSC occurrence in different seasons. Although the global nighttime mean (8.0%) was only slightly larger than the daytime mean (7.6%), day and night differences for some regions were much larger. The largest day-night variation occurred over the tropics, especially over land. During nighttime, MLTSCs not only occurred more often for any given high-occurrence region, but also extended over wider areas. Although deep convections peaked in the afternoon over tropical land due to solar heating, the moisture supplied by these convections was available to form MLTSCs in the night, when longwave cooling took place. For most midlatitude regions, up to 5% more MLTSCs were observed during nighttime. Meanwhile, small day and night variations were observed in the polar regions.
3.2. Vertical Distributions of MLTSCs
 CALIPSO lidar and CloudSat radar measurements provided vertical distributions of MLTSCs. The vertical probability distribution functions (PDF) of MLTSCs in terms of cloud top height (above the surface) and top temperature for different latitude bands are shown in Figures 6a and 6b. The global mean MLTSC top height was 4.5 km above the surface with a standard deviation of 1.6 km. The MLTSC fraction decreased with increasing cloud top height except in the tropics which show a peak between 5 and 6 km. Many studies observed the “midlevel peak” at around 5 km of cloud distributions in the tropics [Johnson et al., 1999; Yasunaga et al., 2006]. The global mean MLTSC top temperature was −13.6°C with a standard deviation of 12.4°C. The MLTSC top temperature distributions were generally in a “bell shape” for each latitude band with the temperature of the peak fraction decreasing as the latitude increased. The fraction of MLTSCs with cloud top temperature below 0°C increased from 69.7% in the tropics to 99.6% in the polar regions. For the same latitude band over the Northern and Southern Hemispheres, although MLTSC cloud top height distributions were similar, there were large differences in cloud top temperature distributions with more cold clouds in the south midlatitude and Antarctic.
Figures 6c and 6d show the MLTSC percentage fractions as a function of cloud top height and top temperature for different latitude bands. The peak heights of percentage fraction increased from about 3 km in polar regions to ∼5.5 km in the tropics. In Figure 6d, two peaks at about −22°C and −12°C were observed for all latitude bands, but there was up to a 20% difference in magnitude among different latitude bands. MLTSC percentage fractions over the southern midlatitude and polar regions were higher than those over the northern midlatitude and polar region with peak differences over 20% between −20 and −30°C. The mean MLTSC percentage fraction decreased from 41.8% at 2.5 km above the surface to 11.6% at 8 km and from 39.8% at 0°C to 1.7% at −35°C. Although the vertical trend of MLTSC percentage fraction in terms of CTT from Hogan et al.'s  results was similar to results presented here, results based on longer-term CloudSat and CALIPSO showed higher-percentage fractions of midlevel clouds with stratiform supercooled water.
 An important feature of MLTSCs is that ice particles are often generated in these clouds when the cloud top temperature is colder than −6°C [Hobbs and Rangno, 1985]. The fraction of MLTSCs with cloud top temperature lower than −6°C was 74.2%. Assuming that all observed CloudSat maximum Ze values larger than the thresholds resulted from ice particles rather than supercooled drizzle drops, 61.8% of MLTSCs were mixed phase, and 12.4% of MLTSCs were supercooled liquid (pure liquid or with ice below the detection limit). Figure 7 shows the occurrence mixed-phase MLTSCs as a function of cloud top temperature. There were steady increases of the mixed-phase MLTSC fraction when cloud top temperatures decreased to ∼−15°C. Below that temperature, small decreases in mixed-phase MLTSC fraction were observed until −20°C followed by a constant increase of mixed-phase MLTSC fraction. When cloud top temperature was colder than −30°C, almost all MLTSCs were mixed-phase. The mixed-phase MLTSC fraction increased from 43.7% at CTT of −6°C to 95.2% at −30°C, which was consistent with the results of Korolev et al.'s  study. Moreover, supercooled liquid MLTSCs persisted at temperatures as low as −30°C, especially in the tropics (about 10.1%). At the same cloud top temperature, the tropics had smaller fractions of mixed-phase MLTSCs than either midlatitude or polar regions.
3.3. IWP Distributions for Mixed-Phase MLTSCs
 Ice microphysical properties have significant impact on the radiative transfer of mixed-phase stratiform clouds [Sassen and Khvorostyanov, 2007]. CloudSat radar measurements provide a good opportunity to retrieve ice properties in mixed-phase MLTSC globally. So far, there are several IWC retrieval methods using radar and ancillary measurements as discussed by Heymsfield et al. . Considering the performance of each method and the available ancillary measurements for CloudSat radar, we used the radar and temperature method [Hogan et al., 2006] to estimate the IWC in mixed-phase MLTSCs,
The units of IWC, Ze and T are g/m3, dBZ, and °C separately. The IWP was then obtained by integrating IWC from the radar-detected mixed-phase MLTSC cloud base to top. When the ice virga bases below the freezing level, ice particles begin to melt or evaporate, and drizzle might be formed, which will bias the IWP calculation. As a result, we only calculated the cases with ice virga base temperatures colder than 0°C. Although equation (2) performed well [Heymsfield et al., 2008], retrieving IWC from radar reflectivity and temperature for mixed-phase clouds may have larger uncertainty than that for ice clouds because in situ data used for empirical relationship development are from ice clouds. In the future, ground-based lidar and radar measurements [Wang et al., 2004] and in situ measurements in MLTSCs can be used to further improve IWC-Ze, IWC-T-Ze relationships for MLTSC study.
 The global IWP distribution of mixed-phase MLTSCs is presented in Figure 8. Overall, the global mean IWP of mixed-phase stratiform clouds was 13.4 g/m2 with a standard deviation of 19.5 g/m2. Compared to the mean cirrus IWP of 12.2 g/m2 observed at the Southern Great Plain [Wang and Sassen, 2002], there were radiatively significant ice particles in the mixed-phase MLTSCs although the water phase may still play a dominant role radiatively. Large IWPs of mixed-phase MLTSCs were distributed over midlatitude and polar regions. The Northern Hemisphere and Southern Hemisphere mean IWP of mixed-phase MLTSCs were 14.2 g/m2 and 12.6 g/m2, respectively. Globally, the mean IWP of mixed-phase MLTSCs over land (13.2 g/m2) was almost the same as that of over the ocean (13.5 g/m2). Unlike MLTSC occurrence, the global daytime mean (14.0 g/m2) is slightly higher than the nighttime mean (12.7 g/m2) probably because the daytime mean MLTSC CTT (−19.9°C) was slightly colder than that of the nighttime (−19.7°C).
Figure 9 shows the mean IWP as a function of cloud top temperature over six latitude bands, indicating strong temperature dependence. Unlike the temperature dependence of IWP in cirrus which decreases with the decreasing of cloud top temperature [Wang and Sassen, 2002], the IWP in mixed-phase MLTSCs generally increased with the decreasing of cloud top temperature. This reflects different ice formation and growth mechanisms in these two types of clouds. Although the IWC decrease with decreasing cloud top temperatures in mixed-phase clouds [Korolev et al., 2003], the total cloud depth (liquid-dominated layer and ice virga) observed from collocated CALIPSO/CloudSat lidar and radar measurements increased with decreasing cloud top temperatures (data not shown). For different latitude regions, the Arctic and north midlatitude region had larger IWP values than other regions at the same cloud top temperature. There was a local peak IWP value of roughly −15°C for all latitude regions, which is consistent with the high growth rate of ice crystals around −15°C under mixed-phase conditions.
 In this study, we defined the midlevel liquid-layer topped stratiform cloud (MLTSC) as liquid-layer topped stratiform cloud with a cloud top higher than 2.5 km above the surface and a cloud top temperature warmer than −40°C. The first 2 year CALIPSO lidar and CloudSat radar measurements were analyzed to study global distributions of MLTSCs and their phase partition. The formation flight of the CloudSat and CALIPSO satellites provided global collocated lidar and radar measurements for these cloud studies. Collocated CALIPSO lidar backscattering, CloudSat radar reflectivity and ECMWF-AUX temperature profiles were used to identify MLTSCs, study their vertical distributions, and determine the presence of ice in MLTSCs based on temperature-dependent Ze thresholds.
 The main findings of our analysis are as follows.
 1. The global mean MLTSC occurrence was ∼7.8%, with ∼8.6% over ocean and 6.1% over land. This high occurrence, together with the optically thick liquid-dominated layer at the top, indicates that MLTSCs play a significant role in the Earth's energy budget. There were significant geographical variations on MLTSC distribution with high occurrences over northern South America, southern Africa, eastern China, and the polar regions. There were strong seasonal variations of MLTSC occurrence over different latitude regions as well. In the polar regions, the maximum occurrence resulted in summer and the minimum in winter. In the tropics, a band of high MLTSC occurrence shifted southward from JJA to DJF.
 2. The global mean MLTSC percentage fraction related to all midlevel clouds was 33.6%, with 28.3% over land and 35.9% over ocean. The MLTSC percentage fraction distribution showed a smaller special contrast. The Southern Hemisphere had a significant higher mean MLTSC fraction than that of the Northern Hemisphere. MLTSCs can impact midlevel ice cloud formation directly by evolving into ice clouds after top water layer dissipated and indirectly by supplying more efficient ice nuclei for downstream cloud formation.
 3. Although global nighttime MLTSC occurrence (8.0%) was close to the daytime occurrence (7.6%), there were much larger regional day and night differences. The largest day-night contrast of MLTSC occurrence was over tropical land with more MLTSCs forming and extended area of high occurrence at night. There were small day-night variations over polar regions.
 4. The global mean MLTSC top height and temperature were ∼4.5 km above the surface and −13.6°C, respectively. Although cloud top height distributions between the same north and south latitude bands were similar (hemispheric symmetry), there were more cold MLTSC clouds in south midlatitude and Antarctic regions than in north midlatitude and Arctic regions. The peak heights of percentage fraction varied greatly with different latitude bands with up to a 20% difference of peak percentage fraction in magnitude. Although the vertical trend of MLTSC percentage fraction in terms of CTT was similar to the results of Hogan et al. , longer-term CloudSat and CALIPSO measurements showed higher stratiform supercooled water percentage fractions in midlevel clouds.
 5. Overall, 61.8% of MLTSCs were mixed phase, and 12.4% were supercooled liquid (pure liquid or with ice below the detection limit). The fraction of mixed-phase MLTSCs increased as the cloud top temperature decreased with sharp increases between −10° and −15°C. The temperature dependency of phase partition show noticeable latitude difference. The supercooled liquid MLTSCs persisted at temperatures as low as −30°C, especially in the tropics.
 6. The global mean IWP of mixed-phase MLTSCs was 13.4 g/m2. For all latitude bands, the IWP generally increased with decreasing cloud top temperature. The Arctic and north midlatitude regions had larger IWPs than those over other regions for the same cloud top temperature. A better IWC-Ze relationship or new multiple sensor approach for mixed-phase clouds is needed to improve IWC and IWP retrievals in mixed-phase clouds from space.
 The high occurrence of mixed-phase MLTSCs and the relatively large IWP indicate the importance of understanding ice generation in MLTSCs in order to accurately model their radiative impact and improve the MLTSC parameterizations in GCMs. The temperature dependence of mixed-phase MLTSC occurrence indicates that ice nucleation is active at −10°C in these clouds with noticeable latitude dependencies. Further studies are needed to understand the latitude dependence of MLTSC distributions and microphysical properties and how aerosol and water phase cloud properties affect ice generation in MLTSCs. Although satellite data provide a global view of these clouds, they are not adequate to characterize the diurnal cycles and lifecycle of MLTSCs. Ground-based and in situ observations indicated that ice generation in these clouds was strongly dependent on lifecycle stage; therefore, it is important to combine long-term CloudSat and CALIPSO measurements with in situ and ground-based remote sensing measurements to better understand the formation and evolution of MLTSCs.
 This research was funded by NASA grant NNX07AQ83G. Many thanks go to CALIPSO PI Dave Winkler, CloudSat PI Graeme Stephens, and the CALIPSO/CloudSat data group. We are also grateful to the anonymous reviewers whose comments led to improvements in this paper.