AMSU-B Observations of Mixed-Phase Clouds over Land

AMSU-B Observations of Mixed-Phase Clouds over Land

M. N. DEETER AND J. VIVEKANANDAN

Research Applications Program, National Center for Atmospheric Research,* Boulder, Colorado

(Manuscript received 29 July 2003, in final form 13 July 2004)

ABSTRACT

Measurements from passive microwave satellite instruments such as the Advanced Microwave SoundingUnit B (AMSU-B) are sensitive to both liquid and ice cloud particles. Radiative transfer modeling isexploited to simulate the response of the AMSU-B instrument to mixed-phase clouds over land. Theplane-parallel radiative transfer model employed for the study accounts for scattering and absorption fromcloud ice as well as absorption and emission from trace gases and cloud liquid. The radiative effects ofmixed-phase clouds on AMSU-B window channels (i.e., 89 and 150 GHz) and water vapor line channels(i.e., 183 � 1, 3, and 7 GHz) are studied. Sensitivities to noncloud parameters, including surface tempera-ture, surface emissivity, and atmospheric temperature and water vapor profiles, are also quantified. Mod-eling results indicate that both cloud phases generally have significant radiative effects and that the 150- and183 � 7-GHz channels are typically the most sensitive channels to integrated cloud properties (i.e., liquidwater path and ice water path). However, results also indicate that AMSU-B measurements alone areprobably insufficient for retrieving all mixed-phase cloud properties of interest. These results are supportedby comparisons of AMSU-B observations of a mixed-phase cloud over the Atmospheric Radiation Mea-surement (ARM) Program’s Southern Great Plains (SGP) site with corresponding calculated clear-skyvalues.

1. Introduction

Although mixed-phase clouds are relatively common(Riley 1998), remote sensing techniques specifically tai-lored to retrieve their physical properties [e.g., ice wa-ter content (IWC) and liquid water content (LWC)] aremuch less common than techniques developed forsingle-phase clouds. One remote sensing technique forstudying mixed-phase clouds exploits multiple ground-based instruments, which are able to separately moni-tor the individual phases (e.g., Vivekanandan et al.1997). Specifically, ground-based passive microwavemeasurements of the liquid water path (LWP; exploit-ing measurements at 20 and 31 GHz) are typically as-sumed to be insensitive to nonprecipitating ice particles(Westwater 1993), whereas radar reflectivity is heavilydominated by properties of the ice water path (IWP;Vivekanandan et al. 1997). In contrast, the radiativeeffects of both phases should be considered in studies ofnonprecipitating mixed-phase clouds based on passivemeasurements in the visible and infrared spectral re-

gions (Baum et al. 2000), and for microwaves with fre-quencies greater than about 90 GHz (Deeter and Evans2000; Greenwald et al. 1999). Thus, when remote sens-ing techniques developed for single-phase clouds areapplied to mixed-phase clouds, the presence of the ne-glected phase can generally pose a serious source ofretrieval error. Here, we report forward-modelingsimulations, which demonstrate that both phases mayproduce significant radiative effects at operational mil-limeter-wave frequencies between 89 and 183 GHz.These simulations have direct relevance to the applica-tion of single-phase millimeter-wave retrieval algo-rithms on mixed-phase clouds. In addition, we present acase study to demonstrate empirically that mixed-phaseclouds over land produce measurable radiative effectsat operational millimeter-wave frequencies.

Satellite remote sensing offers features (global cov-erage, regular sampling, etc.) for observational atmo-spheric studies that are not readily achieved using ei-ther in situ sampling or ground-based remote sensingtechniques. However, interpretation of remote sensingobservations of clouds requires a detailed understand-ing of both cloud properties and radiative transfer. Pas-sive satellite sensors have been used to classify andquantify cloud properties for decades. These systemseither operate in the shortwave (visible and near-infrared), thermal infrared, or microwave spectral re-gions. Of the three, microwave systems are the most

* The National Center for Atmospheric Research is sponsoredby the National Science Foundation.

Corresponding author address: M. N. Deeter, National Centerfor Atmospheric Research, P.O. Box 3000, Boulder, CO 80307.E-mail: [email protected]

72 J O U R N A L O F A P P L I E D M E T E O R O L O G Y VOLUME 44

© 2005 American Meteorological Society

JAM2187

capable of probing thick clouds. Millimeter-wave sen-sors (microwave systems operating at frequencies be-tween about 30 and 300 GHz) respond to radiation atwavelengths considerably larger than typical cloud par-ticles, whereas shortwave and thermal infrared sensorsrespond to wavelengths much smaller than typicalcloud particles. Mie theory predicts that the extinctionefficiency Q of typical cloud particles exposed to short-wave and thermal infrared radiation should be on theorder of unity, but should be much smaller than unityfor cloud particles exposed to microwave radiation(Goody and Yung 1995). Thus, for the same cloud, op-tical depths are generally much larger for shortwaveand thermal infrared sensors than for microwave sys-tems. In practical terms, this distinction implies thatshortwave and thermal infrared techniques are usuallymost useful for deducing cloud-top properties (e.g.,cloud-top temperature), whereas microwave tech-niques often provide a more direct measure of verti-cally integrated cloud properties (e.g., liquid waterpath). In this paper, we attempt to quantify the sensi-tivity of satellite-based millimeter-wave sensors tomixed-phase cloud properties, including cloud verticalstructure, liquid water path, ice water path, and ice par-ticle size. Results of this study should be useful forguiding the development of future cloud classificationand retrieval algorithms for studying mixed-phaseclouds using satellite-based millimeter-wave radiom-eters. Results presented here also indicate that the ap-plication to mixed-phase clouds, of millimeter-wave-based retrieval algorithms developed specifically forsingle-phase clouds, is generally not appropriate.

In the next section, we review current satellite-basedmillimeter-wave sensors that could potentially be ex-ploited for characterizing mixed-phase clouds. In sec-tion 3, we describe the radiative transfer models andcloud models used to simulate millimeter-wave mea-surements of mixed-phase clouds. Simulation resultsare then presented that show the sensitivity of the ob-servations both to clear-sky parameters (e.g., surfacetemperature and emissivity) and to mixed-phase cloudparameters. Last, simultaneous Advanced MicrowaveSounding Unit B (AMSU-B) observations and ground-based remote sensing observations of a system of non-precipitating clouds at the Atmospheric RadiationMeasurements (ARM) Program’s Southern GreatPlains (SGP) site in March 2002 are used to demon-strate the effects of a particular mixed-phase cloud onthe AMSU-B channels, and to test some qualitativepredictions from modeling.

2. Cloud studies using millimeter-wave radiometry

Beginning with the Special Sensor Microwave Tem-perature Profiler 2 (SSM/T-2) satellite instrument onthe Defense Meteorological Satellite Program (DMSP)satellite platforms, a standard configuration of five mil-limeter-wave channels has emerged. Similar satellite-

based instruments include AMSU-B on the current Na-tional Oceanic and Atmospheric Administration(NOAA)-K/-L/-M satellites and the Humidity Sensorfor Brazil (HSB) (Jet Propulsion Laboratory 2000) onthe Earth Observing System (EOS) Aqua platform (on-line at http://aqua.nasa.gov). With the exception ofHSB (which lacks only the 89-GHz channel), all ofthese instruments are equipped with two “windowchannels” near 89 and 150 GHz, and three “water va-por line channels” centered on the water vapor absorp-tion line at 183.3 GHz. The primary intended functionof these instruments is to enable retrievals of watervapor profiles even in cloudy conditions (which is notfeasible using infrared sensors, such as HIRS).

a. Single-phase clouds versus mixed-phase cloudsBoth observational and modeling studies have previ-

ously considered the sensitivity of these instruments’observations to cloud liquid water, cloud ice, and vari-ous forms of precipitation (Burns et al. 1997; Isaacs andDeblonde 1987; Muller et al. 1994). These studies dem-onstrated clearly that, although millimeter-wave radia-tion penetrates both liquid and ice clouds more readilythan do visible or infrared radiation, the radiative sen-sitivity of millimeter waves to clouds is not generallynegligible. None of these studies, however, specificallyconsidered millimeter-wave observations of mixed-phase clouds. Because radiation at millimeter wave-lengths is sensitive to both liquid and ice particles, in-terpretation of observations of mixed-phase clouds issubstantially more complicated than for single-phaseclouds. For this reason, perhaps, retrieval algorithmsfor extracting cloud properties from millimeter-wavesensor data have so far assumed single-phase cloudsconsisting of either pure liquid (Rosenkranz 2001) orpure ice (Zhao and Weng 2002). Neither of these algo-rithms were developed to treat mixed-phase clouds,and, therefore, should not strictly be applied in caseswhere mixed-phase conditions might occur.

On the other hand, the fact that the interaction ofmillimeter-wave radiation with liquid and ice particlesinvolves different physical mechanisms suggests thepossibility of developing retrieval algorithms that inde-pendently retrieve both liquid and ice cloud character-istics (e.g., liquid water path and ice water path). Typi-cally, liquid cloud particles respond to millimeter wavesprimarily through absorption and emission, whereasscattering processes dominate in the radiative responsefor ice particles. At the very least, millimeter-wave ob-servations may provide information on mixed-phaseclouds, which complements information from morestandard instrumentation (e.g., radar and lower-frequency microwave measurements).

b. Window channels versus water vaporline channels

Satellite- and aircraft-based millimeter-wave re-trieval algorithms for retrieving cloud properties have

JANUARY 2005 D E E T E R A N D V I V E K A N A N D A N 73

so far primarily exploited window channels (e.g., 89,150, and 220 GHz) because of the interfering effect ofwater vapor on the water vapor line channels (Deeterand Evans 2000; Liu and Curry 2000; Zhao and Weng2002). Under clear-sky conditions, observed brightnesstemperatures for the water vapor line channels primar-ily depend on the atmospheric temperature and watervapor vertical profiles. Thus, interpretation of observa-tions of cloudy scenes using millimeter-wave windowchannels does not require consideration of the atmo-spheric temperature or humidity to the same degreethat it would for the water vapor line channels. On theother hand, because of the much higher atmospherictransmittance, the use of millimeter-wave windowchannels requires greater consideration of the variableeffects of surface temperature and emissivity. Whenthese surface parameters are known and relatively con-stant (e.g., over the tropical oceans), millimeter-wavewindow channels have a more direct response to cloudsthan the water vapor line channels. (Microwave cloudretrievals over oceanic scenes also benefit from the lowemissivity of the ocean surface, because this results ingreater thermal contrast between the atmosphere andunderlying surface.) Over land, however, surface tem-perature and emissivity are both highly variable, andmust be explicitly considered when studying windowchannel observations. For example, several authors(Greenwald et al. 1997; Jones and Vonder Haar 1990)have demonstrated the importance of surface emissivityvariability to retrievals of liquid water path over landusing the SSM/I 85-GHz channel.

Thus, over land, the water vapor line channels mayprove to be more valuable than (or at least complemen-tary to) the window channels because of their de-creased sensitivity to surface parameters. Moreover, asshown below (in section 3b), the higher frequency ofthe water vapor line channels should make them moresensitive to both liquid and ice cloud particles. Last,because of the much narrower weighting functions ofthe water vapor line channels (relative to the windowchannels), they offer greater potential for resolvingcloud vertical structure. A similar concept for studyingcloud vertical structure was previously considered foran instrument with channels near the 118-GHz oxygenabsorption line (Gasiewski 1993).

3. Modeling strategies

In this section, we describe the radiative transfermodel used to explore the sensitivity of the AMSU-Bchannels to mixed-phase cloud properties. Precipitatingclouds are not treated here. Specifically, we considerwintertime midlevel, mixed-phase clouds over land be-cause of their association with aircraft icing conditions(Cober et al. 2001; Politovich 1989; Riley 1998). Theobjective of this research is to quantify the sensitivity ofthe AMSU-B channels to mixed-phase cloud proper-ties, including cloud vertical structure, liquid water

path, ice water path, and ice particle size. Results of thisstudy should be useful for guiding the development offuture cloud classification and retrieval algorithms forstudying mixed-phase clouds using satellite-based mil-limeter-wave radiometers. More specifically, the resultspresented here should be considered before applyingretrieval algorithms developed for single-phase cloudsto mixed-phase clouds.

For radiance I and optical depth �, the nonpolarizedradiative transfer equation (RTE) for a thermally emit-ting atmosphere may be written

�dI��, ��

d�� I��, �� J��, �� I��, ��

� ��

2 ��1

1

I��, ��P��, ��d�� B��1 � ��, �1�

where � and � are the cosines of the observed andincident zenith angles, J(�, �) is the source function, is the single scattering albedo, P(�, �) is the scatteringphase function, and B(�) is the Planck function (Goodyand Yung 1995). For the case of millimeter-wave radia-tion interacting with liquid and ice cloud particles, boththermal emission and scattering sources are potentiallyimportant and have, therefore, been retained in theRTE. For a polydispersed distribution of spheres ofconcentration (number per unit volume per unit incre-mental diameter) N(D), the optical depth � is related tothe vertical coordinate z through the relation

d�

dz� � � �

0

�

��D

2 �2

Q�D�N�D� dD, �2�

where � is the volume extinction coefficient and Q(D)is the extinction efficiency factor. Here, Q(D) can fur-ther be written as the sum of absorptive and scatteringcomponents, that is,

Q�D� � Qa�D� � Qs�D�. �3�

Generally, Qa and Qs depend on particle size andshape, the radiation wavelength, and the complex re-fractive index m.

a. Microphysical properties

For the mixed-phase cloud radiative transfer calcula-tions reported in this work, liquid cloud particles areassumed to be spherical. We also assume a solid spheremodel for ice particles, despite the highly variable bulkdensity and habits of ice particles actually observed incirrus and mixed-phase clouds. Evans et al. (Evans andStephens 1995; Evans and Vivekanandan 1990; Evanset al. 1998) demonstrated that the radiometric sensitiv-ity at millimeter-wave frequencies for cirrus particlesvaries substantially (up to a factor of 2 or 3) with par-ticle shape. However, the sensitivity (e.g., extinction)for solid ice spheres was often found to be intermediatewith respect to the sensitivity for more realistic habits


(columns, rosettes, etc.). Thus, for microwave radiativetransfer, the solid ice sphere is a reasonable nominal iceparticle–shaped model. Effects due to the nonsphericityand varying bulk density of actual ice particles will beexplored in future research.

Cloud particle size distributions must be specified inorder to calculate the basic cloud radiative properties,that is, the single scattering properties (extinction,single scattering albedo, and scattering phase function).For both liquid and ice water particles, we employ thestandard gamma distribution to model the particle sizevariability. By convention, the gamma distribution iswritten.

N�D� � aD� exp�� 3.67�D

D0�, �4�

where N is the concentration (number of particles perunit volume per unit incremental diameter) of sphericalparticles of diameter D, a is a scaling factor related tothe liquid or ice water content (i.e., liquid or ice massper unit volume), � defines the width of the distribu-tion, and D0 is the median diameter of the distribution(Evans et al. 1998). The three parameters in the distri-bution independently control the integrated mass den-sity (liquid or ice water content), characteristic particlesize, and distribution width.

Ice particle size in mixed-phase clouds often exhibitsmuch greater variability than liquid particle size. Nev-ertheless, we consider the dependence of single scatter-ing properties on particle size for both phases. For liq-uid cloud particles, we consider median diameters be-tween 10 and 100 �m. In nonprecipitating clouds,observed liquid cloud droplets are typically less than 40�m in diameter (Liou 1992). For ice particles, we con-sider median diameters between 10 and 500 �m. Par-ticles with maximum dimensions of hundreds of mi-crons have been observed in the lower regions of non-precipitating mixed-phase clouds (e.g., Fleishauer et al.2002).

b. Single scattering properties

Scattering properties are often expressed in terms ofthe size parameter x defined as

x �D, �5�

where � is the radiation wavelength. The Rayleigh scat-tering regime is strictly valid when x � 1 for all particlesin a given distribution. Although this limit is not per-fectly obeyed for all of the particle sizes and radiativewavelengths considered in this work, Rayleigh scatter-ing may be considered a good first-order model forunderstanding the roles of millimeter-wave absorptionand scattering in mixed-phase clouds. (Actual modelingresults presented in section 4 are based on Mie scatter-ing theory rather than Rayleigh scattering theory.) Forspherical particles of the complex refractive index m,

the absorption and scattering efficiency factors in theRayleigh limit are

Qa�x� � 4x Im�m2 � 1

m2 � 2� �6�

and

Qs�x� �8x4

3 �m2 � 1

m2 � 2�2

. �7�

By resolving � into its absorptive and scattering com-ponents �a and �s, we find from Eqs. (2), (6), and (7)that

d�a

dz�

8�

Im�m2 � 1

m2 � 2��

0

�

��D

2 �3

N�D� dD �8�

and

d�s

dz�

83 ��2��

�4�m2 � 1

m2 � 2�2�

0

�

��D

2 �6

N�D� dD. �9�

These equations indicate that �a is proportional to thethird moment of N(D), whereas �s is proportional to thesixth moment. The third moment of N(D) is propor-tional to the total volume occupied by the cloud par-ticles of a given distribution and, therefore, representsliquid or ice mass (assuming that particle mass densityis fixed). In contrast, the scattering optical depth �s isprimarily determined by the largest particles in the par-ticle size distribution (through the sixth moment), andincreases rapidly with increasing particle size.

The use of the solid-sphere model for both liquid andice cloud particles allows the use of Mie theory forcalculating single scattering properties, including ex-tinction, single scattering albedo (defined as the ratio ofthe scattering extinction coefficient and the total extinc-tion coefficient), and asymmetry parameter (i.e., thefirst moment of the scattering phase function) (Liou1992). For homogeneous spherical scatterers, Mietheory is exact (and, therefore, applicable to size pa-rameters well outside of the Rayleigh regime). Calcu-lated values of extinction, the single scattering albedo,and the asymmetry parameter are plotted against themedian diameter D0 for gamma distributions of liquidwater and ice spheres in Figs. 1 and 2. In all singlescattering calculations, the median particle size was var-ied, whereas the liquid water content (or ice water con-tent for solid spheres) was kept fixed at 1 g m�3. Thephase- and frequency-dependent complex refractive in-dex of water is taken from Ray (1972). In both figures,the curves labeled “183.3 GHz” represent the singlescattering properties for all three AMSU-B water vaporline channels. Because all three of these channels arecentered on the same frequency, and because of therelatively weak variability of the single scattering prop-erties over the frequency range from 175 to 191 GHz


(the bounding frequency limits for the three channels),the single scattering properties may be assumed to bethe same for all three channels.

As indicated by Fig. 1, the single scattering albedo forliquid spheres is much less than one throughout therange of expected cloud particle size. Scattering fromliquid cloud particles may, therefore, be assumed to beinsignificant for the AMSU-B channels. Previous workindicates that scattering in liquid clouds may be justifi-ably neglected at AMSU-B frequencies when the moderadius is less than about 20 �m (Muller et al. 1994). Asdescribed previously, the prevalence of absorption overscattering in the Rayleigh limit (i.e., for particles muchsmaller than the radiation wavelength) theoreticallyproduces extinction that is proportional to the thirdmoment of the particle size distribution, which repre-sents liquid water content (liquid mass per unit vol-ume). Because liquid water content was kept fixed dur-ing single scattering calculations, the extinction shouldexhibit negligible dependence on the liquid sphere me-dian particle size. This prediction is confirmed in thetop panel of Fig. 1, up to the typical maximum diameterof liquid cloud particles (�40 �m). The monotonic in-crease of extinction with millimeter-wave frequencyevident in the figure is also consistent with the absorp-

tive Rayleigh limit (size parameter x increases linearlywith frequency).

In the millimeter-wave spectral range, Im(m) (i.e.,the absorptive component) for liquid water exceedsIm(m) for ice by more than two orders of magnitude(Ray 1972). Therefore, for particles of similar size, thesingle scattering albedo for ice particles should be muchlarger than for liquid water particles. This predictedbehavior is confirmed by comparing the middle panelsof Figs. 1 and 2. For ice spheres, the prevalence ofscattering introduces a strong particle-size dependenceto the extinction, unlike the case for liquid waterspheres. In the Rayleigh limit, scattering-dominated ex-tinction increases as the sixth moment of the particlesize distribution. The relatively weak extinction for iceparticles smaller than about 100–200 �m indicates thatsuch particles are effectively insignificant at millimeter-wave frequencies. For example, the extinction producedby liquid phase particles (with equivalent liquid watercontent) is more than an order of magnitude greaterthan for ice particles with D0 values less than 200 �m.

c. Radiative transfer modelingFor mixed-phase cloud radiative transfer simulations,

we employed a modified two-stream scattering model.

FIG. 1. Single scattering properties—(top) extinction, (middle)single scattering albedo, and (bottom) asymmetry parameter—ofliquid water spheres for three AMSU-B microwave bands as afunction of median diameter. Extinction is calculated for fixedliquid water content of 1 g m�3.

FIG. 2. As in Fig. 1, single scattering properties of ice spheres forthe three AMSU-B microwave bands as a function of mediandiameter. Extinction is calculated for fixed ice water content of1 g m�3.


The plane-parallel radiative transfer model RTSPEC(available online at http://nit.colorado.edu) accountsfor absorption and scattering processes using a hybridtechnique that integrates the single scattering solutionwith the Eddington solution (Deeter and Evans 1998).While nearly as computationally efficient as the Ed-dington solution, the hybrid solution retains more in-formation regarding the scattering phase function. (TheEddington solution relies on just a single moment of thescattering phase function.) This is particularly impor-tant under conditions of weak scattering (i.e., small tomoderate scattering optical depths), which is often thecase in microwave atmospheric remote sensing.

Before executing RTSPEC, the user must computelayer absorption optical depths over the spectral rangeof interest for the atmosphere being modeled. For thatpurpose, we employed the MONORTM radiativetransfer code from Atmospheric and EnvironmentalResearch, Inc. (available online at http://rtweb.aer.com). Layer optical depths were calculated on a gridwith 1-km vertical resolution and at a spectral resolu-tion of 0.002 cm�1 (approximately 60 MHz). Absorp-tion effects due to both gases and liquid water particles(representing the liquid component of mixed-phaseclouds) are explicitly included in the layer optical depthcalculations. Based on the conclusion above that scat-tering is insignificant for liquid cloud particles atAMSU-B frequencies, we neglect it completely. Thisassumption is equivalent to the approximation that thesingle scattering albedo for liquid-phase cloud particlesis exactly zero. In this limit, the radiative properties ofthe liquid component of mixed-phase clouds depend onthe liquid water content (liquid mass per unit volume),but not on liquid particle size. For the same reason, theassumed size distribution plays no role in the radiativetransfer calculations.

d. Cloud models

The most simple mixed-phase cloud model is one inwhich both liquid and ice particles are assumed to behomogeneously distributed within a single layer. Morecomplex two- and three-layer models have also beenconsidered in which the top layer is populated exclu-sively by ice particles and the bottom layer by liquiddroplets (Zhang and Vivekanandan 1999). Previous mi-crowave modeling results for ground-based remotesensing studies indicate that substantial brightness tem-perature differences (on the order of 10 K) may resultfrom the choice of the mixed-phase cloud model(Zhang and Vivekanandan 1999). We considered threeplane-parallel models to represent the geometrical(macrophysical) characteristics of mixed-phase clouds.The uniform mixing (UM) model is a single-layermodel characterized by cloud-base and cloud-topheights of 3 and 6 km, respectively. Within thoseboundaries, both LWC and IWC are assumed to behomogeneous. The glaciated cloud-top (GT) model ischaracterized by a homogeneous liquid cloud layer be-

tween 3 and 5 km, and a homogeneous ice cloud layerbetween 5 and 6 km. Finally, the glaciated cloud-base(GB) model is characterized by a liquid cloud layerbetween 4 and 6 km, and an ice cloud layer between 3and 4 km. All three cloud models exhibit the samecloud-top and cloud-base heights, although with sub-stantially different distributions of ice and liquid mass.

Inclusion of the GB model is consistent with the ob-servation of large ice particles (and a peak in IWC)near the bottoms of some mixed-phase clouds (Fleis-hauer et al. 2002). Because the extinction associatedwith ice particles increases rapidly with particle size, thevertical distribution of the largest particles will have amuch stronger effect on radiative transfer modeling re-sults than will the distribution of smaller ice particles.Although the actual macrophysical characteristics ofmixed-phase clouds are extremely variable, the threemodels considered here should provide a constructivebasis to simulate the effects of ice and liquid verticaldistribution on AMSU-B observations.

For each cloud model, the IWP (or vertically inte-grated IWC) within the ice cloud layer and the LWPwithin the liquid cloud layer were independently var-ied. Radiative transfer calculations were carried out foreach cloud model over a grid of IWP and LWP values,and for two ice particle D0 values: 200 and 400 �m.These two D0 values were selected to demonstrate thesignificant role played by ice particle size in determin-ing radiative properties. (For ice particle distributions,a value of 7 was assumed for �. This choice representsa relatively narrow distribution in order to accentuatethe role played by particle size. Sensitivity of the radia-tive transfer model results to � will be addressed infuture studies.)

e. Other modeling parameters

For the “baseline” atmospheric state (i.e., the nomi-nal temperature and water vapor vertical profiles), wechose the McClatchey midlatitude winter profile (Ell-ingson et al. 1991). However, for consistency with thecloud models described above, the water vapor verticalprofile was modified between 3 and 6 km by setting therelative humidity (with respect to ice) within that layerto 100%. Monochromatic clear-sky weighting functions(derivatives of transmittance with respect to height)representing each of the AMSU-B channels for the re-vised McClatchey midlatitude winter profile are shownin Fig. 3. (Weighting functions were calculated on a1-km grid, and then smoothed using cubic spline inter-polation.) The McClatchey midlatitude winter profilewas selected for this study as a reasonable model forrepresenting wintertime continental interior conditions.Surface temperature was taken as the temperature inthe atmospheric profile closest to the surface (272.2 K).Surface emissivity was set to 0.9 for all AMSU-B chan-nels because, over land, typical surface emissivity val-ues at millimeter-wave frequencies range from about0.8 to 1.0 (Felde and Pickle 1995).


4. Simulation results

Prior to simulating the effects of varying cloud pa-rameters on the AMSU-B brightness temperatures,clear-sky simulations were performed to quantify thefirst-order sensitivity of the brightness temperatures torelative humidity, atmospheric and surface tempera-ture, and surface emissivity. Because in real applica-tions these geophysical parameters will generally becharacterized by some uncertainty, each represents apotential source of error in any satellite-based mixed-

phase cloud retrieval algorithm. Four test cases wereconducted in which perturbations were separately ap-plied to 1) the relative humidity (RH) profile, 2) thetemperature profile (Tz), 3) the surface temperature(Ts), and 4) the surface emissivity (�s). Perturbations tothe relative humidity profile (�10%) and temperatureprofile (�2 K) were applied uniformly throughout thetroposphere. Calculated brightness temperatures foreach of the test cases are compared with results for thebaseline state (described in the previous paragraph) inTable 1. As would be expected, perturbations to thesurface parameters are reflected most strongly in thewindow channels, whereas perturbations in the meteo-rological profiles mainly affect the water vapor chan-nels. Because the weighting function for the 183 �7-GHz channel is substantially higher at the surfacecompared to the other water vapor channels (as seen inFig. 3), it exhibits relatively stronger sensitivity to sur-face parameter perturbations and weaker sensitivity tometeorological profile perturbations.

Radiative transfer modeling results for the UM, GT,and GB cloud models are presented as brightness tem-perature contour plots in Figs. 4, 5, and 6. For eachcloud model, LWP and IWP were independently variedbetween 0 and 500 g m�2 for a specific value of D0

(either 200 or 400 �m). All simulations were made forthe nadir view. First-order radiative sensitivities to bothLWP and IWP, that is, �T� /�LWP and �TB/�IWP arealso presented in Tables 2 and 3. These derivatives in-dividually represent the radiative sensitivity to onephase of the cloud in the absence of the other phase(i.e., �TB/�LWP is evaluated at zero IWP whereas �TB/�IWP is evaluated at zero LWP).

a. General observations

As seen in Figs. 4, 5, and 6, brightness temperaturesfor all of the water vapor channels tend to decreasemonotonically both with increasing LWP and IWP. Forthese channels, increasing LWP tends to increasinglyabsorb upwelling radiance from below and reradiate ata relatively colder brightness temperature. In contrast,increasing IWP generally reduces the brightness tem-perature by reflecting upwelling radiation reaching thecloud bottom downward and reflecting the muchweaker downwelling radiance at the cloud top upward.

FIG. 3. Clear-sky weighting functions (derivatives of transmit-tance with respect to altitude) for the AMSU-B channels and themodified McClatchey midlatitude winter profile. Weighting func-tions were calculated on a 1-km grid and then smoothed by cubicspline interpolation. Results were calculated from monochromaticlayer optical depths calculated by MONORTM for the nadir viewat 89.0, 150.0, 176.3, 180.3, and 182.3 GHz. (Actual radiative trans-fer simulations performed with RTSPEC include spectral integra-tion over the instrumental bandpasses. The monochromaticweighting functions shown here describe the radiance sensitivityat the spectral center of each channel.)

TABLE 1. Comparison of clear-sky brightness temperature calculated for baseline atmospheric state (midlatitude winter) with clear-sky brightness temperatures calculated after applying perturbations to the tropospheric relative humidity, tropospheric temperature,surface temperature, and surface emissivity. Values in parentheses indicate difference between values for perturbed case and baselinestate.

Perturbation TB (89) TB (150 GHz) TB (183 � 7 GHz) TB (183 � 3 GHz) TB (183 � 1 GHz)

None 250.31 255.47 261.25 252.85 243.26�RH � �10% 249.90 (�0.41) 254.48 (�0.99) 261.59 (0.34) 253.93 (1.08) 244.79 (1.53)�Tz � �2.0 K 250.11 (�0.20) 255.02 (�0.45) 259.45 (�1.80) 250.66 (�2.19) 241.09 (�2.17)�Ts � �2.0 K 248.76 (�1.55) 254.18 (�1.29) 260.93 (�0.32) 252.84 (�0.01) 243.26 (0.00)��s � �0.05 240.23 (�10.08) 248.46 (�7.01) 260.79 (�0.46) 252.84 (�0.01) 243.26 (0.00)


For the window channels, however, increasing LWPtends to increase the brightness temperature. Thistrend is closely associated with the assumed surfaceemissivity (Greenwald et al. 1997; Jones and VonderHaar 1990). Although the surface emissivity modulatesboth the thermal emission from the surface and thereflected downwelling radiation (through Kirchoff’slaw), its effect on the thermal emission term dominatesfor the window channels (as confirmed in Table 1).Thus, if the physical temperature of the surface is TS,the effective radiative temperature of the surface is�STS. In radiative terms, the surface may, therefore,appear much colder than the overlying atmosphere.Under these circumstances, increasing the absorptionoptical depth of the atmosphere (e.g., by increasing theLWP) will tend to “mask” the effectively colder sur-face, thereby increasing the brightness temperature ob-served at the top of the atmosphere. (This effect is evenmore dramatic over surfaces of open water, for whichthe emissivity is much smaller than it is for land.) Incontrast, increasing IWP monotonically reduces thewindow channels’ brightness temperatures by the samemechanism (scattering) described for the water vaporline channels.

b. Implications for retrieval algorithm development

The results of the simulations shown in Figs. 4, 5, and6 and in Tables 2 and 3 have several implications for thedevelopment of millimeter-wave mixed-phase cloudclassification and retrieval algorithms. First, compari-son of the results for the GB and GT cloud models inTable 3 (where IWP � 0) reveals that vertical displace-ment of the 2-km-thick liquid cloud by 1 km has sig-nificant effects for both the window channels and thewater vapor channels. For the window channels, sensi-tivity to LWP is greatest for the GT model, where theliquid is nearest the surface. For the water vapor chan-nels, in contrast, the GB model exhibits the greatestLWP sensitivity. Thus, the window channels seem mostsensitive to liquid cloud particles near the surface,whereas the 183 � 7-GHz channel is more sensitive toliquid in the midtroposphere. Generally, these resultsdemonstrate that the validity of assumptions regardingthe liquid phase vertical distribution will have a strongeffect on the accuracy of millimeter-wave-based re-trievals of LWP. Overall, the 183 � 7-GHz channelexhibits the highest LWP sensitivity �TB/�LWP.

As indicated in Table 2 and in Figs. 5 and 6, the IWPsensitivity �TB/�IWP depends strongly on both the iceparticle size and the vertical location of the ice layer.The strong increase in IWP sensitivity with increasing

FIG. 4. AMSU-B brightness temperature contours for the UMmixed-phase cloud model with ice particle median diameters of200 and 400 �m. Consecutive contours are separated in incre-ments of 2 K.

FIG. 5. As in Fig. 4 but for the GT mixed-phase cloud model.


particle size is a direct result of the nonlinear increaseof extinction with particle size, as discussed in section3b and shown in Fig. 2. Compared to the correspondingvalues of �TB/�IWP for D0 � 200 �m, the values for D0

� 400 �m are approximately 8 times as large for the89-GHz channel and 4 times as large for the water va-por channels. Thus, the water vapor channels exhibitweaker particle size dependence than the window chan-nels. For the cloud models considered, the 183 � 7- and150-GHz channels exhibit the greatest IWP sensitivity.Sensitivity to IWP increases with the altitude of the icelayer for all channels, because absorption and emission

processes from both water vapor and liquid tend tomask the underlying atmosphere. As the ice layer ismoved higher, there is less absorption to diminish thescattering effect of the ice layer. This effect is muchmore pronounced for the water vapor channels than forthe window channels, as would be expected. For ex-ample, for the same D0, the IWP sensitivity �TB/�IWPfor the 183 � 7-GHz channel more than doubles as theice-layer base moves from 3 (in the GB model) to 5 (inthe GT model) km.

In view of the large number of mixed-phase cloudparameters to which AMSU-B brightness temperaturesare apparently sensitive, AMSU-B data alone are prob-ably insufficient as the basis of a comprehensive mixed-phase cloud retrieval algorithm. The results above dem-onstrate that AMSU-B brightness temperatures aresensitive to cloud geometry (cloud boundaries for eachphase), LWP, IWP, and ice particle size. Additionalinformation, either from 1) complementary satellite orground-based remote sensing instruments, 2) a prioridata, or 3) modeling results would clearly be needed toretrieve basic mixed-phase cloud properties like LWPand IWP. Complementary remote sensing measure-ments might include satellite retrievals of cloud-topheight and/or radar retrievals of cloud boundaries. Apriori data might include statistics of in situ ice particlespectra measured under specific meteorological condi-tions.

5. Mixed-phase cloud case study

As discussed in section 2b, identifying and quantify-ing the effects of clouds in millimeter-wave satellite ob-servations is much more challenging over land thanover open water. Here, we present a case study to dem-onstrate that nonprecipitating mixed-phase clouds overland produce measurable effects on satellite millimeter-wave measurements. Comparisons of observed bright-ness temperatures with corresponding calculated clear-sky values reveal the effects of the clouds. These effectsare quantified in terms of brightness temperature de-pression, that is,

�TB � T Bcalc � T B

obs, �10�

where T obsB and T calc

B are the observed and correspond-ing calculated clear-sky brightness temperatures, re-spectively.

FIG. 6. As in Fig. 4 but for the GB mixed-phase cloud model.

TABLE 2. First-order radiative sensitivities [K (kg m�2)�1] ofAMSU-B channels to IWP (�TB/�IWP) in three cloud models fortwo values of D0. Values are calculated for zero LWP.

Cloudmodel

D0(�m)

AMSU-B channel frequency (GHz)

89 150 183 � 1 183 � 3 183 � 7

UM 200 �0.8 �6.9 �0.7 �4.0 �10.5400 �6.9 �38.3 �2.5 �15.3 �42.6

GB 200 �0.7 �6.2 �0.0 �1.1 �6.7400 �6.5 �35.3 �0.1 �4.2 �28.7

GT 200 �0.9 �7.7 �1.8 �8.1 �15.0400 �7.2 �42.6 �6.8 �35.1 �65.7

TABLE 3. First-order radiative sensitivities [K (kg m�2)�1] ofAMSU-B channels to LWP (�TB/�LWP) in three cloud models.Values are calculated for zero IWP.

Cloudmodel


89 150 183 � 1 183 � 3 183 � 7

UM 18.3 12.2 �2.5 �10.9 �22.8GB 15.5 7.4 �3.5 �14.5 �28.9GT 21.2 17.7 �0.8 �6.2 �15.5


a. Ancillary data

The U.S. Department of Energy (DOE) operates anongoing research program known as ARM. For studiesof mixed-phase clouds, we exploited several datasetsacquired at the ARM SGP Calibration and RadiationTest Bed (CART) site located in northern Oklahoma.The ARM data considered in this study were acquiredfrom several ground-based instruments at the SGPCentral Facility (36.605°�, 97.486°W) on 16 March2002. Data from the ARM Millimeter-Wave Cloud Ra-dar (MMCR; Moran et al. 1997) on this day were usedto distinguish clear and cloudy periods, to locate cloudboundaries, and to identify periods of precipitation.MMCR radar reflectivity data are shown in Fig. 7. Re-trievals from the 12-channel Microwave RadiometerProfiler (MWRP; Gueldner and Spaenkuch 2001; Lil-jegren et al. 2001b) provided local atmospheric tem-perature, water vapor, and liquid water content profilesat the times of the AMSU-B overpasses. As a source ofmeteorological profiles, MWRP retrievals were pre-ferred over standard radiosonde data because of theirmuch higher sampling frequency. As compared withradiosondes, the accuracy of MWRP temperature pro-files is approximately 1°–2°C (rms), whereas MWRPwater vapor profiles are generally at least accurate to1–2 g m�3 (Liljegren et al. 2001b). For the period stud-ied, MWRP retrievals were available approximately ev-ery 15 min, whereas soundings were only available ap-proximately every 6 h. MWRP retrievals of LWP weresupplemented by LWP retrievals from the two-channelMicrowave Water Radiometer (MWR; Liljegren et al.2001a), which has a longer operational history and has

been more extensively validated. Retrievals of LWPfrom both the MWR and MWRP instruments for 16March 2002 are shown in Fig. 8.

Although the MMCR data shown in Fig. 7 exhibitperiods of both clear and cloudy conditions, two dis-tinct cloud layers are apparent. The higher cloud layerresides roughly between 6 and 10 km, whereas thelower cloud layer lies between 2 and 5 km. Analysis ofthese two cloud layers provides an important contextfor the ensuing analysis of the AMSU-B observationsof this system. The location of the higher cloud layersuggests that it was most likely cirrus. This conjecture issupported by Geostationary Operational Environmen-tal Satellite (GOES) retrievals (Minnis et al. 2001),which determined the cloud-top phase for the ARMSGP area during this day to be primarily ice. Additionalevidence for the lack of liquid in the upper cloud layeris provided by the ground-based MWRP and MWRinstruments. As shown in Fig. 8, retrievals of measur-able amounts of liquid (e.g., LWP greater than 0.03mm) for both instruments occur only between 1100 and2000 UTC. These times correspond closely to the ap-pearance and disappearance of the lower cloud layer,identifying that layer as the liquid-bearing layer. More-over, inspection of the MWRP liquid water content ver-tical profiles between 1100 and 2000 UTC indicates thatthe cloud liquid is heavily concentrated between thesurface and about 4 km. The consistent picture emerg-ing from all of these datasets is a two-layer cloud systemconsisting of either a liquid or mixed-phase cloud in themidtroposphere below a separated layer of thick cirrus.Last, the high peak radar reflectivity values observedin the lower cloud layer (often greater than 10 dBZ)

FIG. 7. Radar reflectivity recorded by the MMCR at the ARMSGP field site on 16 Mar 2002.

FIG. 8. Retrievals of liquid water path for both the 2-channelMWR and 12-channel MWRP instruments at the ARM SGP fieldsite on 16 Mar 2002.


strongly suggest the presence of ice particles (Vive-kanandan et al. 1999). We, therefore, conclude that thelower cloud was mixed in phase. The radar data furtherindicate that, with the possible exception of a brief pe-riod at approximately 1800, this cloud system was notprecipitating. No measurable precipitation was ob-served at the SGP Central Facility surface station onthis day.

b. AMSU-B data

All of the AMSU-B observations from both theNOAA-K and -L satellites falling within a 1° latitude �1° longitude box centered on the SGP Central Facilityon 16 March 2002, were collected. A total of 89AMSU-B observations derived from four satellite over-passes were found and analyzed. Satellite overpassesoccurred at approximately 0105, 0810, 1325, and 1935UTC, which are labeled hereinafter (and in Fig. 7) as“A,” “B,” “C,” and “D,” respectively. According toarguments presented in the previous paragraph, over-pass A corresponds to a period of cirrus cover. Over-pass B occurred during a period of clear sky, just beforethe brief appearance of cirrus. Overpass C correspondsto a period during which both cloud layers were welldeveloped. Overpass D occurred after the disappear-ance of the lower cloud, but the cirrus layer was stillclearly evident. Because the lower cloud layer was ab-sent for overpasses A, B, and D, those overpasses arecollectively termed the “cirrus scenes.” Overpass C isthe sole satellite overpass during which the mixed-phase cloud was observed.

For each actual AMSU-B observation, correspond-ing clear-sky brightness temperatures were calculatedusing the radiative transfer model described in section3. Actual satellite zenith angles for each observationalong with temperature and water vapor profiles ob-tained from the MWRP were fed to the model to esti-mate the clear-sky brightness temperatures. A surfaceemissivity of 0.95 was assumed for all model calcula-tions. This value was selected because 1) it is well withinthe range of reported microwave surface emissivity val-ues over land (Felde and Pickle 1995) and 2) it pro-duced reasonable agreement between the observed andcalculated brightness temperatures for the three cirrusscenes at both 89 and 150 GHz. Studies based on mod-eling results (Muller et al. 1994) and airborne observa-tions (Wang et al. 1998) indicate that cirrus-inducedbrightness temperature depressions at 89 GHz are typi-cally less than a few kelvin. For the purpose of estimat-ing surface emissivity, we neglected the radiative effectsof cirrus entirely. For each observation and AMSU-Bchannel, �TB provides a quantitative measure of thenet radiative effect of the clouds within the AMSU-Bfield of view.

Comparisons of observed AMSU-B brightness tem-peratures with corresponding calculated clear-sky val-ues are presented as scatterplots in Fig. 9. Unique plot-ting symbols (defined in the figure caption) distinguish

FIG. 9. Comparison of observed and model-calculatedAMSU-B brightness temperatures at the ARM SGP site for 16Mar 2002. Data for mixed-phase cloud overpass at 1325 UTC areplotted with the � symbol. Results for cirrus-scene overpasses areplotted with the � symbol.


the AMSU-B observations made during overpass Cfrom the other three overpasses. The �TB statistics(both means and standard deviations) were calculatedseparately for overpass C and the other three over-passes and are listed in Table 4.

c. Analysis

For the 89- and 150-GHz channels, absolute devia-tions between the observed and calculated clear-skybrightness temperatures in the three cirrus scenes aretypically only a few kelvin. (For these channels, theassumed surface emissivity value directly affects thesystematic bias between the observed brightness tem-peratures and corresponding calculated clear-skybrightness temperatures, but has no effect on the bias of�TB values for overpass C relative to the three cirrusscenes.) At 89 GHz, �TB statistics for overpass C gen-erally agree with values for the cirrus scenes. However,for the 150-GHz channel, an overall �TB bias of 4.6 Kis observed between overpass C and the other threeoverpasses (as seen in Table 4), with a maximum �TB

value of about 7 K. Qualitatively, the stronger effect ofthe mixed-phase cloud on the 150-GHz channel relativeto the 89-GHz channel agrees with modeling resultspresented in section 4.

Because of the strong frequency dependence of thesingle scattering properties of ice particles (as shown inFig. 2), cirrus-induced brightness temperature depres-sions are potentially much greater for the water vaporchannels than for the window channels (Muller et al.1994; Wang et al. 1998). For these channels more thanfor the window channels, absorption and emission bywater vapor in and above any cirrus cloud would reducethe brightness temperature depression caused by thecloud. Thus, cirrus-induced brightness temperature de-pressions for the water vapor channels depend on theposition of the ice layer relative to each of the watervapor channels’ weighting functions. For example, ifthe cirrus layer resides well above the peaks of theweighting functions for all three water vapor channels,cirrus-induced brightness temperature depressions willbe approximately the same for all three channels. Inconditions where the cirrus layer is not well above thewater vapor channel weighting functions, the maskingeffect of water vapor (which tends to reduce brightnesstemperature depression) in and above the cirrus layerwill be strongest for the 183 � 1-GHz channel andweakest for the 183 � 7-GHz channel. Thus, the 183 �

7-GHz channel should generally be the most cirrus-sensitive AMSU-B channel. The lack of any appre-ciable brightness temperature depressions (i.e., largerthan a few kelvin) at 183 � 7 GHz for the three cirrusscenes indicates that the cirrus layer had negligible ef-fect on the AMSU-B data analyzed in this paper.

Retrievals of temperature and water vapor profilesfrom the ground-based MWRP were used in model cal-culations instead of radiosonde data because of themuch greater sampling frequency. (Any benefits thatmight be gained from using radiosonde data for thispurpose would be greatly outweighed by large errorsencountered because of the lack of synchronicity be-tween the radiosonde measurements and the AMSU-Bobservations. Moreover, the much higher vertical reso-lution offered by the radiosonde measurements is rela-tively unimportant for simulating millimeter-wavebrightness temperatures, which respond to water vaporaveraged over thick layers.) MWRP water vapor pro-files exhibit relative errors that increase with altitude(Liljegren 2004). This effect, combined with the watervapor channels’ weighting functions (as shown in Fig.3), implies that water vapor profile uncertainties shouldtypically produce the largest errors in calculated bright-ness temperatures for the 183 � 1-GHz channel and thesmallest errors in calculations for the 183 � 7-GHzchannel. This expectation is confirmed in the three cir-rus scenes (plotted in Fig. 9 with the � symbol). Thischaracteristic of the ancillary data implies that smallercloud-induced �TB values should be resolvable for the183 � 7-GHz channel than for the 183 � 1- and 183 �3-GHz channels.

Comparing the water vapor channels only, the resultspresented graphically in Fig. 9 and statistically in Table4 show that the mixed-phase cloud present during over-pass C produces the most obvious effect at 183 � 7GHz. For this channel, the overall bias between �TB

values for overpass C and the three cirrus-scene over-passes of 3.2 K is nearly 4 times as large as the standarddeviation of the cirrus-scene �TB values. The maximum�TB value of approximately 6 K is nearly 2 times aslarge as the largest value for the three cirrus scenes. Incontrast, for both the 183 � 1- and 183 � 3-GHz chan-nels, �TB values for overpass C fall well within therange of values for the other three overpasses.

The results of this case study qualitatively agree withthe modeling results presented in section 4 in that 1)brightness temperature sensitivities to both LWP and

TABLE 4. Comparison of AMSU-B �TB statistics at ARM SGP site on 16 Mar 2002 for 1) mixed-phase cloud observations(overpass C) and 2) three cirrus scenes (overpasses A, B, and D).


89 150 183 � 1 183 � 3 183 � 7

�T CB �1.10 � 0.78 �3.57 � 1.44 �2.43 � 1.82 2.52 � 1.09 �3.52 � 1.27

�T ABDB 1.71 � 2.29 1.00 � 1.84 �3.09 � 1.97 �0.15 � 1.53 �0.32 � 0.87


IWP are stronger at 150 than at 89 GHz, and 2) amongthe three water vapor channels, the largest brightnesstemperature depressions occur for the 183 � 7-GHzchannel. These results also demonstrate empiricallythat real mixed-phase clouds over land can producemeasurable effects on AMSU-B measurements.

6. Conclusions

A wide range of geophysical parameters affect ob-servations from satellite-based millimeter-wave radi-ometers (such as AMSU-B), including surface tempera-ture and emissivity, atmospheric temperature and wa-ter vapor profiles, clouds, and precipitation. Thus, theusefulness of each of the AMSU-B channels for study-ing mixed-phase clouds depends not only on the sensi-tivity of brightness temperatures to mixed-phase cloudproperties, but also the availability of precise ancillarydata. The 89- and 150-GHz channels are rather insen-sitive to atmospheric temperature and water vapor, buthighly sensitive to surface temperature and emissivity.Conversely, the three AMSU-B water vapor channelsat 183 � 1, 3, and 7 GHz, are relatively insensitive tothe surface but are much more sensitive to the atmo-spheric temperature and water vapor profiles. In termsof single scattering properties only, the water vaporchannels are more sensitive to both liquid and ice par-ticles than are the 89- and 150-GHz channels.

A case study was presented to demonstrate the po-tential of AMSU-B observations for studying mixed-phase clouds over land. A key aspect of this study wasthe availability of temperature and water vapor profiles(from a ground-based remote sensing instrument)matched in time and space with the AMSU-B observa-tions. Comparisons of observed and calculated clear-sky AMSU-B brightness temperatures at the ARMSGP site demonstrate that a mixed-phase cloud pro-duced measurable brightness temperature depressionsat the 150- and 183 � 7-GHz channels.

Retrieval algorithms designed to exploit millimeter-wave observations for cloud studies have so far onlyconsidered single-phase systems, that is, pure liquidclouds or pure ice clouds. Mixed-phase cloud simula-tion results presented here indicate that both phases ofwater can produce significant effects on the observedbrightness temperatures for all AMSU-B channels.Therefore, retrieval algorithms developed to retrieveproperties (such as IWP) of one particular phase shouldnot be applied to mixed-phase clouds.

The development of mixed-phase cloud classificationor retrieval schemes based on AMSU-B data might fol-low one of three approaches. In the first approach, nec-essary ancillary data would be provided by other satel-lite-based remote sensing instruments. This is the ap-proach taken by Rosenkranz (2001), who developed aretrieval algorithm for liquid-only clouds based onAMSU-A and AMSU-B observations. This approach

might also benefit from thermal infrared observations,which are used routinely for determining cloud-topheight. The second approach would involve the use ofwater vapor and temperature profiles generated by nu-merical weather prediction models. Further work isneeded to assess the classification or retrieval errorsthat would result from the use of these models. Last,the third alternative would be to combine AMSU-Bdata with observations from ground-based remote sens-ing instruments. Although such an approach wouldonly yield retrieval results over limited regions, ground-based instruments (such as the MWRP) often providemeasurements that are not feasible from space-basedplatforms.

Acknowledgments. Portions of this research weresponsored by the National Science Foundation throughan interagency agreement in response to requirementsand funding by the Federal Aviation Administration’sAviation Weather Research Program. The views ex-pressed are those of the authors and do not necessarilyrepresent the official policy or position of the U.S. Gov-ernment. Data used in this study were obtained fromthe Atmospheric Radiation Measurement (ARM) Pro-gram sponsored by the U.S. Department of Energy,Office of Science, Office of Biological and Environ-mental Research, Environmental Sciences Division.

REFERENCES

Baum, B. A., P. F. Soulen, K. I. Strabala, M. D. King, S. A.Ackerman, W. P. Menzel, and P. Yang, 2000: Remote sensingof cloud properties using MODIS airborne simulator imageryduring SUCCESS. II. Cloud thermodynamic phase. J. Geo-phys. Res., 105, 11 781–11 792.

Burns, B. A., X. Wu, and G. R. Diak, 1997: Effects of precipita-tion and cloud ice on brightness temperatures in AMSUmoisture channels. IEEE Trans. Geosci. Remote Sens., 35,1429–1437.

Cober, S. G., G. A. Isaac, A. V. Korolev, and J. W. Strapp, 2001:Assessing cloud-phase conditions. J. Appl. Meteor., 40, 1967–1983.

Deeter, M. N., and K. F. Evans, 1998: A hybrid Eddington-singlescattering radiative transfer model for computing radiancesfrom thermally emitting atmospheres. J. Quant. Spectrosc.Radiat. Transfer, 60, 635–648.

——, and ——, 2000: A novel ice-cloud retrieval algorithm basedon the Millimeter-wave Imaging Radiometer (MIR) 150- and220-GHz channels. J. Appl. Meteor., 39, 623–633.

Ellingson, R. G., J. Ellis, and S. Fels, 1991: The intercomparison ofradiation codes used in climate models: Long wave results. J.Geophys. Res., 96, 8929–8953.

Evans, K. F., and J. Vivekanandan, 1990: Multiparameter radarand microwave radiative transfer modeling of nonsphericalatmospheric ice particles. IEEE Trans. Geosci. Remote Sens.,38, 423–437.

——, and G. L. Stephens, 1995: Microwave radiative transferthrough clouds composed of realistically shaped ice crystals.Part I: Single scattering properties. J. Atmos. Sci., 52, 2041–2057.

——, S. J. Walter, A. J. Heymsfield, and M. N. Deeter, 1998:Modeling of submillimeter passive remote sensing of cirrusclouds. J. Appl. Meteor., 37, 184–205.

Felde, G. W., and J. D. Pickle, 1995: Retrieval of 91 and 150 GHz


earth surface emissivities. J. Geophys. Res., 100, 20 855–20866.

Fleishauer, R. P., V. E. Larson, and T. H. Vonder Haar, 2002:Observed microphysical structure of midlevel, mixed-phaseclouds. J. Atmos. Sci., 59, 1779–1804.

Gasiewski, A. J., 1993: Microwave radiative transfer in hydro-meteors. Atmospheric Remote Sensing by Microwave Radi-ometry, M. A. Janssen, Ed., John Wiley and Sons, 572 pp.

Goody, R. L., and Y. L. Yung, 1995: Atmospheric Radiation:Theoretical Basis. 2d ed. Oxford University Press, 519 pp.

Greenwald, T. J., C. L. Combs, A. S. Jones, D. L. Randel, andT. H. Vonder Haar, 1997: Further developments in estimat-ing cloud liquid water over land using microwave and infra-red satellite measurements. J. Appl. Meteor., 36, 389–405.

——, S. A. Christopher, J. Chou, and J. C. Liljegren, 1999: Inter-comparison of cloud liquid water path derived from theGOES 9 imager and ground based microwave radiometersfor continental stratocumulus. J. Geophys. Res., 104, 9251–9260.

Gueldner, J., and D. Spaenkuch, 2001: Remote sensing of thethermodynamic state of the atmospheric boundary layer bymicrowave radiometry. J. Atmos. Oceanic Technol., 18, 925–933.

Isaacs, R. G., and G. Deblonde, 1987: Millimeter wave moisturesounding: The effect of clouds. Radio Sci., 22, 367–377.

Jet Propulsion Laboratory, 2000: AIRS project algorithm theo-retical basis document, level 1b, part 3: Microwave instru-ments, version 2.1. California Institute of Technology Publi-cation JPL D-17005, 59 pp.

Jones, A. S., and T. H. Vonder Haar, 1990: Passive microwaveremote sensing of cloud liquid water over land regions. J.Geophys. Res., 95, 16 673–16 683.

Liljegren, J. C., 2004: Improved retrievals of temperature andwater vapor profiles with a twelve-channel radiometer. Pre-prints, Eighth Symp. on Integrated Observing and Assimila-tion Systems for Atmosphere, Oceans, and Land Surface(IOAS-AOLS), Seattle, WA, Amer. Meteor. Soc., CD-ROM,4.7.

——, E. E. Clothiaux, G. G. Mace, S. Kato, and X. Dong, 2001a:A new retrieval for cloud liquid water path using a ground-based microwave radiometer and measurements of cloudtemperature. J. Geophys. Res., 106, 14 485–14 500.

——, B. M. Lesht, S. Kato, and E. E. Clothiaux, 2001b: Evaluationof profiles of temperature, water vapor, and cloud liquid wa-ter from a new microwave radiometer. Proc. 11th ARM Sci-ence Team Meeting, Atlanta, GA, U.S. Department of En-ergy.

Liou, K. N., 1992: Radiation and Cloud Processes in the Atmo-sphere. Oxford University Press, 487 pp.

Liu, G., and J. A. Curry, 2000: Determination of ice water pathand mass median particle size using multichannel microwavemeasurements. J. Appl. Meteor., 39, 1318–1329.

Minnis, P., W. L. Smith Jr., and D. F. Young, 2001: Cloud macro-and microphysical properties derived from GOES over theARM SGP domain. Proc. 11th ARM Science Team Meeting,Atlanta, GA, U.S. Department of Energy.

Moran, K. P., B. E. Martner, D. C. Welsh, D. A. Merritt, M. J.Post, and T. Uttal, 1997: ARM’s cloud-profiling radar. Pre-prints, 28th Conf. on Radar Meteorology, Austin, TX, Amer.Meteor. Soc., 296–297.

Muller, B. M., H. E. Fuelberg, and X. Xiang, 1994: Simulations ofthe effects of water vapor, cloud liquid water, and ice onAMSU moisture channel brightness temperatures. J. Appl.Meteor., 33, 1133–1154.

Politovich, M. K., 1989: Aircraft icing caused by large supercooleddroplets. J. Appl. Meteor., 28, 856–868.

Ray, P. S., 1972: Broadband complex refractive indices of ice andwater. Appl. Opt., 11, 1836–1844.

Riley, R. K., 1998: Mixed-phase icing conditions: A review. U.S.Dept. of Transportation Rep. DOT/FAA/AR-98/76, 45 pp.

Rosenkranz, P. W., 2001: Retrieval of temperature and moistureprofiles from AMSU-A and AMSU-B measurements. IEEETrans. Geosci. Remote Sens., 39, 2429–2435.

Vivekanandan, J., L. Li, L. Tsang, and C. Chan, 1997: Microwaveradiometric technique to retrieve vapor, liquid and ice: PartII—Joint studies of radiometer and radar in winter clouds.IEEE Trans. Geosci. Remote Sens., 35, 237–247.

——, B. E. Martner, M. K. Politovich, and G. Zhang, 1999: Re-trieval of atmospheric liquid and ice characteristics usingdual-wavelength radar observations. IEEE Trans. Geosci.Remote Sens., 37, 2325–2333.

Wang, J. R., P. Racette, J. D. Spinhirne, K. F. Evans, and W. D.Hart, 1998: Observations of cirrus clouds with airborne MIR,CLS, and MAS during SUCCESS. Geophys. Res. Lett., 25,1145–1148.

Westwater, E. R., 1993: Ground-based microwave remote sensingof meteorological variables. Atmospheric Remote Sensing byMicrowave Radiometry, M. A. Janssen, Ed., John Wiley andSons, 572 pp.

Zhang, G., and J. Vivekanandan, 1999: Microwave radiation frommixed phase cloud and parameter retrievals. Proc. Int. Geo-science and Remote Sensing Symp., Hamburg, Germany,IEEE, 2182–2184.

Zhao, L., and F. Weng, 2002: Retrieval of ice cloud parametersusing the Advanced Microwave Sounding Unit. J. Appl. Me-teor., 41, 384–394.


AMSU-B Observations of Mixed-Phase Clouds over Land

Documents

Transcript of AMSU-B Observations of Mixed-Phase Clouds over Land