WANG Lei,LI Chengcai,YAO Zhigang,ZHAO Zengliang,HAN Zhigang,and WEI QiangDepartment of Atmospheric and Oceanic Sciences,School of Physics,Peking University,Beijing 0087

2Beijing Institute of Applied Meteorology,Beijing 100029

1.Introduction

Clouds play an important role in regulating global radiation(Ramanathan et al.,1989).They impact not only on short-term and local weather processes,but also on atmospheric circulation and global climate change.Cloud microphysical properties are important for studying the radiation budget and dynamical processes.In numerical simulations,the distribution of cloud liquid water content(LWC)is an important aspect of the parameterization of cloud microphysics(Morrison and Gettelman,2008).In terms of data collection,the CloudSat satellite has an advantage because it provides vertically-resolved information on clouds with a 94-GHz cloud prof i ling radar(CPR)(Stephens et al.,2002).The resultant 2B-CWC-RO products contain cloud water content prof i les and are produced at the CloudSat data processing center(DPC).

Austin(2007)proposed a CloudSat cloud water content(CWC)retrieval process for the latest standard 2B-CWC-RO product,in which cloud water retrievals are performed separately for the liquid and ice phases(Austin et al.,2009),and the two results are combined by a suitable cloud liquid water/ice water distribution scheme for mixed clouds.The retrievals assume a lognormal size distribution of cloud droplets and obtain an optimal solution by minimizing the cost function based on optimal estimation theory(Austin and Stephens,2001).The algorithms select a priori data based on different cloud types and locations.Several experiments and studies have been conducted to assess the Cloud-Sat cloud product.From November 2006 to March 2007,the Canadian CloudSat-CALIPSO Validation Project(C3VP)was performed to evaluate the CloudSat retrieval algorithm(Barker et al.,2008).Noh et al.(2011)compared aircraft observations with CloudSat retrieval results for winter mixed clouds.Brunke et al.(2010)compared the liquid water path(LWP)retrieved over the ocean from CloudSat,the Cloud-Aerosol Lidar with Orthogonal Polarization(CALIOP)instrument,and the Moderate Resolution Imaging Spectroradiometer(MODIS).Li et al.(2008)compared four sets of satellite LWP results,and found that CloudSat LWPs are signif i cantly higher than all the other observations in boundarylayer stratocumulus regions.

One important problem for CloudSat CWC retrievals is caused by a priori data.The cloud microphysical property retrievals are constrained by a priori data selected from historical aircraft observations during the iteration procedure in the retrievals.Austin(2007)obtained a priori data from a database of cloud microphysical parameters summarized by Miles et al.(2000).From aircraft data reported in the literature,Miles et al.(2000)created a database of stratus cloud droplet(diameter less than 50μm)size distribution parameters as a priori data for liquid phase clouds.Austin(2007)used the same a priori data as Miles et al.(2000)for selection criteria in different regions.The analysis(Austin et al.,2009;Feng et al.,2009)demonstrated that a priori data have signif i cant impacts on CloudSat CWC retrievals.However,aircraft observations(Miles et al.,2000)revealed that cloud microphysical parameters are signif i cantly different in different regions.Therefore,the selected a priori data may cause signif i cant errors in CloudSat CWC retrievals,meaning it is necessary to use regional a priori parameters from aircraft data to constrain the retrievals in the iterations and improve cloud parameter retrievals from CloudSat observations.

Although Miles et al.(2000)summarized many aircraft data in different regions,they did not include cloud microphysical characteristics in China.Recently,several researchers have analyzed cloud microphysical characteristics in China based on Particle Measuring Systems(PMS)data(Yan and Chen,1990;Wang et al.,2005;Zhao et al.,2010;Guo et al.,2013).Many studies have statistically analyzed CloudSat cloud products(Wang et al.,2011;Yang and Wang,2012;Peng et al.,2013),and many statistical results for cloud droplet size distributions over the Beijing region have been published with PMS data.However,no studies have investigated the parameters of lognormal size distribution for this region.Additionally,the PMS data obtained in China have not been applied to validate and improve active microwave satellite retrievals.Although much research has been performedonthevalidationoftheCloudSatCWCretrievalproduct,little work has been conducted on assessing the product in China.In particular,there is a gap in China between in-situ and satellite remote sensing data.

The aim of the present reported study was to improve the CloudSat CWC product with a priori data from aircraft observations in the Beijing area,and to compare the modif i ed CloudSat CWP retrievals with in situ data and MODIS CWP.First,we constructed a database of cloud droplet lognormal size distribution derived from in situ data from 12 f l ights over Beijingduring2008–09.Second,wesettheparametersofthe database as a priori data for the CloudSat CWC retrieval algorithm and compared the different retrieval results with different a priori data.Third,we selected different cloud particle lognormal size distribution parameters as a priori data for CloudSat CWC retrieval over northern China during April–October in 2008 and 2009.And f i nally,we compared the modif i ed CloudSat CWP and corresponding MODIS CWP results.

2.Data

2.1.CloudSat

CloudSat Standard Data Products are distributed by the CloudSat DPC operated by the Cooperative Institute for Research in the Atmosphere(CIRA),a research institute at Colorado State University.In this study,2B-GEOPROF,2BCLDCLASS,and the auxiliary data ECMWF-AUX product distributed by the DPC were required as inputs for the proposed CloudSat cloud water retrieval algorithm.2BGEOPROF provides the ref l ectivity of CPR;2B-CLDCLASS provides information on cloud detection and cloud type;and ECMWF-AUX provides temperature prof i les.The 2CPRECIP-COLUMN product provides a more robust representation of precipitation and was used to remove the precipitation when comparing the CloudSat-and MODIS-derived CWP results,as reported in section 5.2.

2.2.The in situ probe data

The data used to analyze cloud microphysics properties were obtained from two airborne instruments:the Forward Scattering Spectrometer Probe(FSSP-100),and the Two Dimensional-Grey Cloud Optical Array Probe(OAP-2D-GA2).Additionally,instruments for measuring temperature,dew point,altitude,airspeed,and GPS were mounted on the aircraft.

FSSP-100 measures the particles that pass through a laser beamafterascatteringintensityatvariousscalestodetermine the particle distribution of the range(Knollenberg,1976)(the measurement range is set at 2–32 μm).It is assumed that the FSSP-100 measurements are for liquid droplets.OAP-2DGA2 contains a photo diode array of 64 elements and zoom optics that enable image resolutions as small as 30μm along the optical array,which can measure the scale of 30–1800-μm particle images.An automatic particle classif i cation of the particle images was applied to the OAP-2D-GA2 data,which classif i ed the two-dimensional grayscale particle images of the probe into eight categories:tiny,linear,aggregate,graupel,spherical,hexagonal,irregular,and dendrite.The classif i cation process and results are presented in Wang et al.(2014).By assuming spherical and tiny particles to be in the liquid phase,we were able to obtain the liquid phase particle size distribution using FSSP-100 and OAP-2D-GA2 data.

2.3.MODIS data

MODISis carriedon theAqua spacecraft,whichhasbeen a part of the A-train satellite constellation since May 2004.MODIS acquires data in 36 spectral bands in the visible and near-infrared spectral range,with a maximum spatial resolution of 250 m and a scan width of 2330 km.In this study,MODIS cloud LWP from the MYD06L2 products released by the Level 2 Atmosphere Archive and Distribution System(LAADS)were used to validate the CWP from CloudSat.In the MYD06 product,the cloud LWP is indirectly estimated from cloud optical thickness and cloud effective radius at a resolution of 1 km(Platnick et al.,2003).

Seethala and Horv′ath(2010)compared one year’s Advanced Microwave Scanning Radiometer for EOS(AMSRE)and MODIS cloud LWP data for warm marine clouds and found that the correlation coeff i cient was up to 0.95 and the root-mean-square error was 15 g m?2in extensive marine stratocumulus clouds after removing the overall MODIS high bias in overcast domains.AMSR-E overestimated MODIS by 45%on average when all scenes were considered.de La Torre Ju′arez et al.(2009)presented a comparison of LWP from the AMSR-E and MODIS instruments.Their results indicated that MODIS CWP retrievals resemble AMSR-E LWP data,with the differences between the two datasets quantif i ed as a function of cloud top height,cloud fraction,cloud top temperature,LWP,cloud effective radius,and cloud optical thickness.Dong et al.(2008)compared microwave radiometer and MODIS LWP in stratiform clouds and found that the mean and standard deviation of the difference was 0.6±49.9 g m?2.These studies demonstrate that MODIS LWPs correspond well with passive microwave measurements.

3.CloudSat CWC retrieval algorithm

Based on the algorithm proposed by Austin(2007),we established a CloudSat CWC retrieval program.The Cloud-Sat 2B-GEOPROF and 2B-CLDCLASS products,and the auxiliary data ECMWF-AUX product,distributed by the DPC,are the inputs for the CWC program.

Austin and Stephens(2001)developed the retrieval of stratus cloud microphysical parameters based on optimal estimation theory(Rodgers,1976).For the retrieval approach,see Eq.(1),where y represents the vector quantities(radar refl ectivity),x is the state vector(geometric mean radius,number density,and distribution parameter),and F is the forward model.εyrepresents the measurement error.

The retrieval algorithm obtains the optimal solution by minimizing the cost function Φ,which is a weighted sum of the measurement vector-forward model difference and the state vector a priori difference[Eq.(2)].?x is the state vector of the retrieved prof i le,and xais a priori data specif i ed based on the real atmosphere parameters.Sarepresents the uncertainty of a priori prof i le and the covariance between various prof i le elements.Syis the forward model covariance matrix and observation.

In LWC retrievals,a lognormal size distribution of cloud droplets is assumed[Eq.(3)],where NTis the droplet number density,r is the droplet radius,rgis the geometric mean radius,andσlogis the distribution width parameter.For ice phase clouds,the ice particles also follow a lognormal distribution[Eq.(4)].In the equation,NTiis the ice particle number concentration,D is the diameter of an equivalent mass ice sphere,Dgis the geometric mean diameter,andσlogiis the width parameter.The a priori data are specif i ed by the mean and standard deviation of NT,rg,andσlog.

The cloud LWC(2B-LWC-RO)and the ice water content(2B-IWC-RO)are retrieved assuming that the entire prof i le is liquid and ice,respectively.A cloud warmer than 0°C is considered to be pure liquid phase,and the cloud water content is 2B-LWC-RO.When the cloud temperature is colder than?20°C,the phase is considered pure ice,and the cloud water content is 2B-IWC-RO.Between ?20°C and 0°C,the ice and liquid solutions in mixed-phase clouds are scaled linearly with temperature.

4.Application of in situ data

4.1.Determination of a priori data

The a priori data for retrievals were selected from the cloud droplet size distribution of the in situ data.The aircraft experiments were conducted during April–October in 2008 and 2009 over the Beijing area in China.We selected sorties for non-precipitation clouds.The aircraft measurements were in the latitude range of 37°–42°N and the longitude range of 115°–120°E.

For liquid-phase and mixed-phase clouds,the liquid droplet size distributions from FSSP-100 and OAP-2D-GA2 were combined.Based on the method proposed by Miles et al.(2000),we calculated cloud droplet lognormal size distribution parameters for the diameter ranges of 2–500 μm and 2–1500 μm.The lognormal size distribution parameters for cloud particles with a diameter of 2–500 μm and 2–1500 μm were derived from in situ data collected over Beijing.These parameters,and the continental cloud parameters from Miles et al.(2000),are presented in Table 1.D＜500μm listed in Table 1 means that those droplets greater than 500μm in diameter were excluded when estimating the cloud size distribution parameters.D＜50μm and D＜1500μm carry the same meaning as D＜500μm.The NTof continental clouds(Miles et al.,2000)is more than twice that of in situ clouds.rgin 2＜D＜500μm clouds is 3.32μm,which is larger than that in 2＜D＜1500μm clouds.Theσlogof in situ data over Beijing is nearly twice that of continental clouds(Miles et al.,2000).

Three cloud droplet lognormal distribution parameters are shown in Fig.1.The shapes of the lognormal distributions are different due to the different cloud particle size ranges.The highestσlogfrom cloud particles with diameters less than 1500μm was 0.86μm,compared with 0.63μm for cloud particles with diameters less than 500μm.Theσlogofin situ data over Beijing was larger than that of continental cloud particles with diameters less than 50μm.The differences among the three distributions indicate that the lognormalsizedistributionderivedfromtheaircraftdatausedinthis study considered larger particles than those typically present in continental clouds(Miles et al.,2000).

Table 1.The lognormal size distribution of liquid cloud droplets from aircraft data with diameters less than 500μm and 1500μm,and from continental clouds with diameters less than 50μm.

Fig.1.Comparison of the lognormal size distribution plot of liquid cloud droplets from aircraft data with cloud particle diameters less than 500μm and 1500μm and from continental clouds with diameters less than 50μm.

4.2.Retrieval schemes

Three CloudSat CWC retrieval schemes were used to assess the inf l uence of a priori data on CloudSat CWC retrievals.For LWC retrievals,Scheme 1 selected continental cloud parameters(Miles et al.,2000)as the a priori data.Scheme 2 selected the parameters of clouds with particle diameters less than 500μm derived from aircraft data over Beijing as the a priori data.Scheme 3 selected the parameters of clouds with particle diameters of less than 1500μm obtained from aircraft data over Beijing as the a priori data.The lognormal size distribution parameters used in Scheme 1 were the same as the a priori data adopted by Austin(2007).As the a priori data of CloudSat ice water content(IWC)retrieval in the three schemes,the ice particle size distribution parameters for the mean and standard deviation value of log(NTi)were 1.5 and 0.555,respectively;for the mean and standard deviation value of log(D)they were?1.0 and 0.226,respectively;and for the mean and standard deviation value ofσlogithey were 0.45 and 0.117,respectively.In mixed clouds,the ice and liquid solutions were scaled linearly with temperature in all schemes.

4.3.Comparison with aircraft observations

The f l ight over Beijing occurred during the period 0045–0320 UTC 25 September 2006.Figure 2 presents the Multifunctional Transport Satellites(MTSAT)1R(10.3–11.3 μm)brightness temperature at 0300 UTC and 0500 UTC.The black line over the Beijing area denotes the f l ight track,and the black line over the map denotes CloudSat’s ground track.The CloudSat and Aqua overpass occurred at about 0520 UTC over Beijing.Figure 3 presents the MODIS CWP at the Aqua overpass time.The blue lines are the f l ight track and the black lines are CloudSat’s track,also shown in Fig.3.

CloudSat passed over the Beijing area approximately 120 min later than the aircraft’s return f l ight,and the distance between the two different areas is about 70–80 km.By comparing the brightness temperatures at 0300 UTC and 0500 UTC over Beijing and its surrounding area(Fig.2),we can see that the entire cloud system was moving eastward.We also simulated the wind f i eld at 4 km during the period 0300–0500 UTC 25 September 2006 using version 3.4 of the Weather Research and Forecasting(WRF)model,driven by National Centers for Environment Prediction(NCEP)Global Final Analysis(FNL)data.Figure 4 presents the 4-km wind f i eld at 0400 UTC 25 September.In the area of aircraft observation and CloudSat overpass,the wind direction was 270°–280°,and the wind speed was about 30–40 km h?1.Since the horizontal distribution pattern of clouds did not change signif icantly,we considered that the clouds measured by CloudSat over Beijing were similar to those detected by the aircraft observations.

Using 2B-GEOPROF,2B-CLDCLAS,and ECMWFAUX from the CloudSat products at this overpass as inputs,we applied the three schemes in the area from 40.43°N to 40.46°N(Fig.3,red points).The LWC and IWC derived from aircraft data were averaged in 240-m intervals,which was consistent with CloudSat products.

Figure 5a presents the LWC prof i les of the CloudSat retrievals with the three schemes and the in situ data.The cloud LWCs were primarily located at an altitude of 2400–5100 m,and the values ranged from 0.05 to 0.015 g m?3.The highest LWC of about 0.15 g m?3was observed at 4000 m.The maximum LWC in Scheme 1 was 0.45 g m?3.The LWC values inSchemes1and2were0.1–0.15gm?3and0.05–0.1gm?3,respectively.In the three schemes,Scheme 3 had the lowest value.The LWC of Scheme 3 was closest to the value of the in situ data.

Fig.2.Brightness temperature of MTSAT IR1(10.3–11.3 μm)at 0300 UTC(left)and 0500 UTC(right)on 25 September 2006.

Fig.3.TheMODISCWPoverBeijingat0520UTC25September 2006.

Fig.4.The simulated 4-km wind f i eld at 0400 UTC 25 September 2006.

Figure 5b presents the IWC prof i les derived from the CloudSat observations with Scheme 1 and in situ measurements.The IWC from the aircraft was less than 0.001 g m?3,and the maximum was at 4200 m.CloudSat’s IWCs were less than 0.0018 g m?3,and the maximum occurred at the same height as in the in situ data.

Fig.5.Comparison of the LWC prof i les from the three CloudSat schemes and aircraft observations(left)and of IWC prof i les from CloudSat Scheme 1 and aircraft observations(right).(Scheme 1—continental clouds with diameters less than 50μm;Scheme 2—with diameters less than 500μm;Scheme 3—with diameters less than 1500μm).

In this case,aircraft-detected clouds contained more particles with diameters larger than 50μm.In Fig.5,the curves of Scheme 3 and in situ show the best agreement.A possible explanation for this is that the size distribution of the a prior data in Scheme 3 was closer to the real situation than in Scheme 1,so the cloud water content derived from Scheme 3,including information on larger particles,was in better agreement with aircraft observations.

5.Comparison with MODIS CWP

Because aircraft observations have been carried out in quite a limited number of regions and periods,other data must be used to assess the results of the schemes applied to CloudSat CWC retrievals.Because the CWP can be obtained from the CloudSat CWC prof i le and the cloud LWP can be obtained from MODIS observations simultaneously,the performance of the CloudSat CWC retrievals can be evaluated by comparison with the MODIS CWP.

In the comparison with CloudSat CWP data,we generated a match-up dataset of MODIS CWP.The collocation method proposed by Partain(2007)was used.The closest matched MODIS CWP footprint in a 3 km(along-track)by 5 km(across-track)box of MODIS data centered on every CloudSat position was used.The mean value of the 15 MODIS CWPs was selected as the matching value of the CloudSat CWP.

5.1.Case analysis

A comparison of the CloudSat CWP retrievals based on the three schemes with the MODIS CWP for 25 September 2006,which were described in section 4,is presented.As shown in Fig.3,the MODIS CWP values were primarily distributed in the range 200–1000 g m?2.We performed Cloud-Sat CWC retrievals based on the three schemes in the latitude regions of 39.6°–40.7°N.Figure 6 presents the CloudSat CWPs,MODIS CWP,and CloudSat ref l ectivity factor in the selected regions,in which the red points represent the rain clouds obtained from the CloudSat 2C-PRECIP-COLUMN products.Table 2 lists the means and standard deviations of the CloudSat CWP from each scheme and the MODIS CWP,and also presents the differences between the two satellites’observations.

The CloudSat CWPs based on Schemes 1,2 and 3 followed the same trends.The CWPs from Scheme 1 were the largest,and its mean value was about three times as large as MODIS CWP.The CWPs based on Scheme 2 were approximately 1.5 times as large as the MODIS CWP.The minimumCWPsoccurredbasedonScheme3;themeanvaluewas 273.8 g m?2,which was close to the MODIS CWP value of 267.3 g m?2.A few CWPs based on Scheme 3 were smaller than the corresponding MODIS CWP when it was less than 250 g m?2.The smallest difference between the CWP from Scheme 3 and MODIS CWP was 105.8 g m?2.It should be notedthatmanycloudsattheoverpasstimefeaturedinFig.3,except those in the latitudinal area from 40.43°N to 40.46°N,were f l agged as precipitating clouds.In fact,it was not reasonable to attempt to retrieve CWP in such cases.CWP values derived from Schemes 2 or 3 seemed close to MODIS CWP,and so it is not possible to indicate whether or not the schemes are more suitable for CloudSat CWP retrievals.To evaluate the CloudSat schemes under non-precipitating conditions accurately,the 2C-PRECIP-COLUMN products were used to remove precipitating clouds before the comparison of CloudSat CWP and MODIS CWP was conducted,the results of which are reported next,in section 5.2.

5.2.Results during April–October in 2008–09

To assess the CloudSat CWC retrievals,we applied Schemes 1 and 2 to the CloudSat overpass of northern China and surrounding areas during April–October in 2008 and 2009.Then,the CloudSat CWPs were compared with the MODIS CWP.In Fig.7,the box designates the selected comparison regions.

Fig.6.MODIS CWP and CloudSat CWPs from Schemes 1,2,and 3,and CloudSat ref l ectivity factor along the CloudSat track on 25 September 2006.The red points indicate rain clouds derived from 2C-PRECIP-COLUMN.The three CloudSat LWC retrieval schemes are the same as in Fig.5.

Table 2.Mean and standard deviation values of the CloudSat CWPs from the three schemes and MODIS CWP.

Fig.7.Illustration of the selected region for the comparison of the CloudSat CWP and the MODIS CWP.

In the data collocation,the following cases were removed:(1)those in which any cloud type among cirrus,nimbostratus,and deep convective clouds occurred at any level in the vertical prof i le;(2)those in which the cloud cover from the 15 MODIS pixels that surrounded the Cloud-Sat pixel was less than 75%;(3)those in which precipitating clouds obtained from 2B-CLD and 2C-PRECIP-COLUMN occurred(apart from if the quality of the data was good and no precipitation was detected);and(4)those in which multilayered clouds occurred.

We used the CloudSat cloud classif i cation product 2BCLD,which provides the cloud type,precipitation f l ag,and other information on each layer.It classif i es clouds into cirrus(Ci),altostratus(As),altocumulus(Ac),stratus(St),stratocumulus(Sc),cumulonimbus(Cu),nimbostratus(Ns),and deep convective clouds(DC).

There were 3431 collocated pairs for comparison.After performing the CloudSat CWC retrievals with different a priori data,we obtained the CWPs from Scheme 1(CWPs1),Scheme 2(CWPs2),and the LWP from MODIS(LWPm).According to the value of CWPs1,the data were divided into eight groups:(1)CWPs1less than 100 g m?2;(2)CWPs1within 100–200 g m?2;(3)CWPs1within 200–300 g m?2;(4)CWPs1within 300–400 g m?2;(5)CWPs1less than 400 g m?2(the sum of groups 1–4);(6)CWPs1within 400–800 g m?2;(7)CWPs1within 800–1200 g m?2;and(8)CWPs1greater than 1200 g m?2.There were 1230,701,262,195,2388,636,227,and180collocatedpairsavailableintheeight groups,respectively.Figure 8 presents the frequency of cloud type from 2B-CLD in groups 5–8.As and Ac were the primary cloud types for the comparison,and the results show a maximum frequency of As in the four groups.Sc had its maximum frequency at 400＜CWPs1＜1200 g m?2.Cu had a greater frequency in groups CWPs1larger than 800 g m?2.

Fig.8.Cloud type occurrence frequency in the four data groups:(a)CWPs1＜ 400 g m?2;(b)400＜CWPs1＜800 g m?2;(c)800＜CWPs1＜1200 g m?2;(d)CWPs1＞1200 g m?2.

The means and standard deviations of the differences among the CWPs from Scheme 1,Scheme 2,and the MODIS CWP in the eight groups are presented in Table 3.Also listed are the correlation coeff i cients between the CWPs and the MODIS CWP.The means of CWP–LWP from Scheme 2 are less than from Scheme 1 when CWPs1was greater than 200 g m?2,and the standard deviations of the difference from Scheme 2 were less than those from Scheme 1 when CWPs1was greater than 400 g m?2.The correlation coeff i cients betweentheCWPfromScheme2andMODISCWPweremuch better than Scheme 1 in most groups except CWPs1＞1200 g m?2.In the group with CWPs1＜100 g m?2and 100＜CWPs1＜200 g m?2,the mean values of the difference from CWPs2and LWPmwere larger than CWPs1.The mean values of the difference of CloudSat CWP and MODIS CWP from Schemes 1 and 2 were comparable at CWPs1＜400 g m?2,and the correlation coeff i cient increased from 44.6%to 53.6%.In the group 400＜ CWPs1＜800 g m?2,CWPLWP from Scheme 1 was almost 1.5 times that from Scheme 2,and the correlation coeff i cients increased from 12.5%to 53.0%.

Figure 8 and Table 3 shows that the correlation coeff icients between the CloudSat CWP and the MODIS CWP were signif i cantly improved for As,Ac,and Sc when using Scheme 2.It is well known that the MODIS instrument is generally sensitive to small cloud particles and the radar of CloudSat is more sensitive to larger size droplets.As a result,as shown in Table 3,for clouds with smaller CWP,the MODIS CWPs were larger than the CloudSat CWPs.Conversely,the CWPs derived from CloudSat were larger than from MODIS for clouds with high value CWPs.If a priori data including information on larger particles is applied,it will make the CloudSat CWP even smaller compared with the MODIS CWP.

Table 3.Statistical results of the comparisons of the CloudSat CWPs from the three schemes and the MODIS CWP in the data group with CWPs1＜100 g m?2,100＜CWPs1＜200 g m?2,200＜CWPs1＜300 g m?2,300＜CWPs1＜400 g m?2,CWPs1＜400 g m?2,400＜CWPs1＜800 g m?2,800＜CWPs1＜1200 g m?2,and CWPs1＞1200 g m?2.

Fig.9.Comparison of the CloudSat CWPs from the three schemes and the MODIS CWP in four data groups:(a)CWPs1＜100 g m?2;(b)100＜CWPs1＜200 g m?2;(c)200＜CWPs1＜300 g m?2;(d)300＜CWPs1＜400 g m?2.The three CloudSat LWC retrieval schemes are the same as in Fig.5.

Comparisons between the CloudSat CWPs1/CWPs2and the MODIS CWP of the eight groups are given in Figs.9 and 10.The groups including(1)CWPs1＜100 g m?2,(2)100＜CWPs1＜200 g m?2,(3)200＜ CWPs1＜300 g m?2,and(4)300＜ CWPs1＜400 g m?2are shown in Fig.9;while those including(5)CWPs1＜400 g m?2,(6)400＜CWPs1＜800 g m?2,(7)800＜ CWPs1＜1200 g m?2,(8)CWPs1＞1200 g m?2are presented in Fig.10.The dotted line(Figs.9 and 10)indicates the ideal situation for the CloudSat CWP and the MODIS CWP comparison.Scheme 2 slightly reduced the CWP compared with Scheme 1 at CWPs1＜200 g m?2and almost all CWPs2values were less than LWPm.Where CWPs1＜ 400 g m?2(Fig.9),there were several CWPs1valuessignif i cantlygreaterthanLWPm,whereasmost CWPs2values were smaller than LWPm.Most CWPs1and CWPs2values were greater than LWPmat CWP＞400 g m?2(Fig.10).After large cloud particles with diameters greater than50μmasaprioridataweretakenintoaccountinScheme 2,a much better agreement was achieved between the Cloud-Sat CWP and the MODIS CWP.

6.Discussion and conclusion

Several studies have been published on the prevalence of liquid phase droplets with diameters greater than 50μm in mixed clouds below 0°C.Cober et al.(2001)analyzed microphysical characteristics of supercooled large drops greater than 50μm in diameter in icing conditions and found that the largest median volume diameters(MVDs)observed were approximately 1000μm in extreme icing conditions.Vidaurre and Hallett(2009)investigated 81 h of aircraft data for mixed clouds,and the results indicated that if supercooled droplets with diameters greater than 50μm and frozen drops(diameters less than 500μm)occur in a cloud,precipitation is facilitated.It appears that large particles exist in mixed clouds.Incloudwatercontentretrievals,becauselargerdrops have a disproportionate effect on ref l ectivity(one sixth),the LWC results are sensitive to the particle size(Comstock et al.,2004).Although the concentration of large particles is relatively small,neglecting them in a priori data will lead to an overestimation of the LWC in CloudSat cloud water retrievals.However,the a priori data in CloudSat cloud water retrievals(Austin,2007)were adopted from a database of stratus cloud droplets with diameters less than 50μm(Miles et al.,2000).

Fig.10.Comparison of the CloudSat CWPs from the three schemes and the MODIS CWP in four data groups:(a)CWPs1＜400 g m?2;(b)400＜ CWPs1＜800 g m?2;(c)800＜ CWPs1＜1200 g m?2;(d)CWPs1＞1200 g m?2.The three CloudSat LWC retrieval schemes are the same as Fig.5.

To evaluate the inf l uence of the cloud droplet size distribution of particles on the CWC retrievals from CloudSat,we characterized the cloud lognormal size distribution derived from 12 fl ights over Beijing China in 2008–09 and presented the parameters of stratiform clouds with D＜500μm(drizzle)and D＜1500μm(rain).Then,the parameters from the aircraft data were applied to CloudSat LWC retrievals.The CloudSat CWPs were compared with the MODIS CWP.For clouds with small CWPs,the MODIS CWPs were larger than CloudSat CWPs.The CWPs derived from CloudSat were larger than from MODIS for those clouds with high CWP values.The results also indicated that the lognormal parameters for large particles in the retrieval algorithm could signi ficantly reduce the overestimation of CloudSat CWP compared with MODIS CWP.In the cloud lognormal size distribution including larger sized droplets,it will be more accurate for large particles,but smaller particles in the distribution may deviate from the actual circumstances.Kahn et al.(2008)suggested that CloudSat can penetrate clouds but is not suffi ciently sensitive to smaller hydrometeors.Further analysis in the present study showed that if only small particles in the distribution were considered,there was little in fl uence on LWC retrievals by changing the cloud number concentration of a priori data.Because of the absence of large particles,LWC was generally overestimated in comparison with theMODISCWP.Therefore,theaccuratedescriptionoflarge particles is important in CloudSat LWC retrievals.

When additional information is available,it can be used to constrain the a priori data to improve the CloudSat LWC retrievals.Austin et al.(2009)proposed a new retrieval that includes temperature information to assist in determining the correct region of state space.Matrosov et al.(2004)evaluated the performance of radar-ref l ectivity-based relations for retrievals of marine stratiform cloud LWC.In mixed clouds,the a priori data could be combined with other information to reduce the uncertainties of LWC retrievals.When the radar ref l ectivity is small,the a priori data should contain fewer large particles.In contrast,more information on large droplets should be considered in a priori data with high radar ref l ectivity.

Acknowledgements.This work is supported by China public science and technology research funds projects of meteorology(Grant No.GYHY201406015),the Chinese Academy of Sciences(Grant No.XDA05040000),the National High-Tech R&D Program of China(Grant No.SQ2010AA1221583001),National Science Foundation program(Grant Nos.41375024,40775002,41175020,and 41375008),and the basic research program(Grant No.2010CB950802).We would like to acknowledge the NASA CloudSat project for making CloudSat data available to the scientif i c community.We are also grateful to NASA/GSFC for the use of their MODIS Level 2 cloud products.

Austin,R.T.,2007:Level 2B radar-only cloud water content(2BCWC-RO)process description document.CloudSat project report,24 pp.[Available online at http://www.cloudsat.cira.colostate.edu/ICD/2B-CWC-RO/2B-CWC-ROPD 5.1.pdf]

Austin,R.T.,and G.L.Stephens,2001:Retrieval of stratus cloud microphysical parameters using millimeter-wave radar and visible optical depth in preparation for CloudSat:1.Algorithm formulation.J.Geophysi.Res.,106(D22),2 8233–2 8242.

Austin,R.T.,A.J.Heymsf i eld,and G.L.Stephens,2009:RetrievaloficecloudmicrophysicalparametersusingtheCloud-Sat millimeter-wave radar and temperature.J.Geophys.Res.,114,D00A23,doi:10.1029/2008jd010049.

Barker,H.W.,A.V.Korolev,D.R.Hudak,J.W.Strapp,K.B.Strawbridge,and M.Wolde,2008:A comparison between CloudSat and aircraft data for a multilayer,mixed phase cloud system during the Canadian CloudSat-CALIPSO Validation Project.J.Geophys.Res.,113,doi:10.1029/2008JD009971.

Brunke,M.A.,S.P.de Szoeke,P.Zuidema,and X.Zeng,2010:A comparisons of ship and satellite measurements of cloud properties with global climate model simulations in the southeast Pacif i c stratus deck.Atmos.Chem.Phys.,10,6527–6536.

Cober,S.G.,G.A.Isaac,and J.W.Strapp,2001:Characterizations of aircraft icing environments that include supercooled large drops.J.Appl.Meteor.,40,1984–2002.

Comstock,K.K.,R.Wood,S.E.Yuter,andC.S.Bretherton,2004:Ref l ectivity and rain rate in and below drizzling stratocumulus.Quart.J.Roy.Meteor.Soc.,130,2891–2918.

de La Torre Ju′arez,M.,B.Kahn,and E.Fetzer,2009:Cloud-type dependencies of MODIS and AMSR-E liquid water path differences.Atmospheric Chemistry And Physics Discussions,9,3367–3399.

Dong,X.Q.,P.Minnis,B.Xi,S.Sun-Mack,and Y.Chen,2008:Comparison of CERES-MODIS stratus cloud properties with ground-based measurements at the DOE ARM Southern Great Plains site.J.Geophys.Res.,113,doi:10.1029/2007JD008438.

Feng,W.W.,Z.G.Yao,Z.G.Han,and Z.L.Zhao,2009:Simulation analysis on microphysical parameters retrieval of liquid water clouds with satellite-based millimeter-wave radar.Journal of PLA University of Science and Technology:Natural Science Edition,10,95–102.(in Chinese)

Guo,X.L.,D.H.Fu,and Z.X.Hu,2013:Progress in cloud physics,precipitation,and weather modif i cation during 2008–2012.Chinese J.Atmos.Sci.,37,351–363.(in Chinese)

Kahn,B.H.,and Coauthors,2008:Cloud type comparisons of AIRS,CloudSat,and CALIPSO cloud height and amount.Atmospheric Chemistry and Physics,8,1231–1248.

Knollenberg,R.G.,1976:Three new instruments for cloud physics measurements: The 2-D spectrometer,the forward scattering spectrometer probe,and the active scattering aerosol spectrometer.Intel.Conf.on Cloud Physics,American Meteorological Society,July 1976,554–561.

Li,J.L.F.,and Coauthors,2008:Comparisons of satellites liquid water estimates to ECMWF and GMAO analyses,20th century IPCC AR4 climate simulations,and GCM simulations.Geophys.Res.Lett.,35,doi:10.1029/2008GL035427.

Matrosov,S.Y.,T.Uttal,and D.A.Hazen,2004:Evaluation of radar ref l ectivity-based estimates of water content in stratiform marine clouds.J.Appl.Meteor.,43,405–419.

Miles,N.L.,J.Verlinde,and E.E.Clothiaux,2000:Cloud droplet size distributions in low-level stratiform clouds.J.Atmos.Sci.,57,295–311.

Morrison,H.,and A.Gettelman,2008:A new two-moment bulk stratiform cloud microphysics scheme in the Community Atmosphere Model,version 3(CAM3).Part I:Description and numerical tests.J.Climate,21,3642–3659.

Noh,Y.J.,C.J.Seaman,T.H.Vonder Haar,D.R.Hudak,and P.Rodriguez,2011:Comparisons and analyses of aircraft and satellite observations for wintertime mixed-phase clouds.J.Geophys.Res.,116,doi:10.1029/2010JD015420.

Partain,P.,2007:Cloudsat MODIS-AUX auxiliary data process description and interface control document.Cooperative Institute for Research in the Atmosphere,Colorado State University,23 pp.

Peng,J.,H.Zhang,and X.Y.Shen,2013:Analysis of vertical structure of clouds in East Asia with cloudSat data.Chinese J.Atmos.Sci.,37,91–100.(in Chinese)

Platnick,S.,M.D.King,S.A.Ackerman,W.P.Menzel,B.A.Baum,J.C.Riedi,and R.A.Frey,2003:The MODIS cloud products:algorithms and examples from terra.IEEE Transactions on Geoscience and Remote Sensing,41,459–473.

Ramanathan,V.,R.D.Cess,E.F.Harrison,P.Minnis,B.R.Barkstrom,E.Ahmad,and D.Hartmann,1989:Cloud-radiative forcing and climate:Results from the Earth Radiation Budget Experiment.Science,243,57–63.

Rodgers,C.D.,1976:Retrieval of atmospheric temperature and composition from remote measurements of thermal radiation.Rev.Geophys.,14,609–624.

Seethala,C.,and′A.Horv′ath,2010:Global assessment of AMSRE and MODIS cloud liquid water path retrievals in warm oceanic clouds.J.Geophys.Res.,115,doi:10.1029/2009JD 012662.

Stephens,G.L.,and Coauthors,2002:The CloudSat mission and the A-Train:A new dimension of space-based observations of clouds and precipitation.Bull.Amer.Meteor.Soc.,83,1771–1790.

Vidaurre,G.,and J.Hallett,2009:Ice and water content of stratiform mixed-phase cloud.Quar.J.Roy.Meteor.Soc.,135,1292–1306.

Wang,L.,C.C.Li,Z.L.Zhao,Z.G.Yao,Z.G.Han,and Q.Wei,2014:The Application of the 2D habit classif i cation in cloud microphysics analysis.Chinese J.Atmos.Sci.,38(2),201–202.(in Chinese)

Wang,S.H.,Z.G.Han,Z.G.Yao,Z.L.Zhao,and X.Jie,2011:Analysis on Cloud vertical structure over China and its neighborhood based on cloudSat data.Plateau Meteorology,30,38–52.(in Chinese)

Wang,Y.F.,H.C.Lei,Y.X.Wu,W.A.Xiao,and X.Q.Zhang,2005:Size distributions of the water drops in the warm layer of stratiform clouds in Yanan.Journal of Nanjing Insitute of Meteorology,28,787–793.(in Chinese)

Yan,C.F.,and W.K.Chen,1990:The stratus cloud droplet number/size distributions and spectral parameters calculation.J.Appl.Meteor.,1,352–359.(in Chinese)

Yang,D.S.,and P.C.Wang,2012:Tempo-spatial distribution characteristics of cloud particle size over China during Summer.Climatic and Environmental Research,17,433–443.(in Chinese)

Zhao,Z.L.,J.T.Mao,Q.Wei,Y.J.Ying,L.Wang,Z.G.Han,and C.C.Li,2010:A study of vertical structure of spring stratiformcloudsinNorthwestChina.MeteorologicalMonthly,36,71–77.(in Chinese)