Portal de Periódicos da CAPES

Partitioning CloudSat ice water content for comparison with upper tropospheric ice in global atmospheric models

2011; American Geophysical Union; Volume: 116; Issue: D19 Linguagem: Inglês

10.1029/2010jd015179

2156-2202

Wei‐Ting Chen, Christopher P. Woods, Jui‐Lin F. Li, Duane E. Waliser, J. Chern, Wei‐Kuo Tao, Jonathan H. Jiang, Adrian M. Tompkins,

Meteorological Phenomena and Simulations

Journal of Geophysical Research: AtmospheresVolume 116, Issue D19 Aerosol and CloudsFree Access Partitioning CloudSat ice water content for comparison with upper tropospheric ice in global atmospheric models Wei-Ting Chen, Wei-Ting Chen [email protected] Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorChristopher P. Woods, Christopher P. Woods Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USA Now at Aerospace Corporation, El Segundo, California, USA.Search for more papers by this authorJui-Lin F. Li, Jui-Lin F. Li Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorDuane E. Waliser, Duane E. Waliser Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorJiun-Dar Chern, Jiun-Dar Chern NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorWei-Kuo Tao, Wei-Kuo Tao NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorJonathan H. Jiang, Jonathan H. Jiang Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorAdrian M. Tompkins, Adrian M. Tompkins European Centre for Medium-Range Weather Forecasts, Reading, UK Also at Abdus Salam International Centre for Theoretical Physics, Trieste, Italy.Search for more papers by this author Wei-Ting Chen, Wei-Ting Chen [email protected] Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorChristopher P. Woods, Christopher P. Woods Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USA Now at Aerospace Corporation, El Segundo, California, USA.Search for more papers by this authorJui-Lin F. Li, Jui-Lin F. Li Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorDuane E. Waliser, Duane E. Waliser Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorJiun-Dar Chern, Jiun-Dar Chern NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorWei-Kuo Tao, Wei-Kuo Tao NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorJonathan H. Jiang, Jonathan H. Jiang Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorAdrian M. Tompkins, Adrian M. Tompkins European Centre for Medium-Range Weather Forecasts, Reading, UK Also at Abdus Salam International Centre for Theoretical Physics, Trieste, Italy.Search for more papers by this author First published: 12 October 2011 https://doi.org/10.1029/2010JD015179Citations: 28AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Abstract [1] The ice cloud estimates in current global models exhibit significant inconsistency, resulting in a significant amount of uncertainties in climate forecasting. Vertically resolved ice water content (IWC) is recently available from new satellite products, such as CloudSat, providing important observational constraints for evaluating the global models. To account for the varied nature of the model parameterization schemes, it is valuable to develop methods to distinguish the cloud versus precipitating ice components from the remotely sensed estimates in order to carry out meaningful model-data comparisons. The present study develops a new technique that partitions CloudSat total IWC into small and large ice hydrometeors, using the ice particle size distribution (PSD) parameters provided by the retrieval algorithm. The global statistics of CloudSat-retrieved PSD are analyzed for the filtered subsets on the basis of convection and precipitation flags to identify appropriate particle size separation. Results are compared with previous partitioning estimates and suggest that the small particles contribute to ∼25–45% of the global mean total IWC in the upper to middle troposphere. Sensitivity measures with respect to the PSD parameters and the retrieval algorithm are presented. The current estimates are applied to evaluate the IWC estimates from the European Centre for Medium-Range Weather Forecasts model and the finite-volume multiscale modeling framework model, pointing to specific areas of potential model improvements. These results are discussed in terms of applications to model diagnostics, providing implications for reducing the uncertainty in the model representation of cloud feedback and precipitation. Key Points A new method is developed to partition CloudSat IWC on the basis of particle size Data are compared to model output, pointing to areas for model improvements Data are crucial for evaluating and reducing uncertainty of climate models 1. Introduction [2] Ice clouds are an important modulator of the climate system [Lynch et al., 2002; Ramanathan and Collins, 1991; Stephens, 2005]. They significantly contribute to the radiation budget through both their shortwave albedo effects and longwave greenhouse effects [Chen et al., 2000; Comstock et al., 2002; Hartmann and Short, 1980; Liou, 1976; McFarquhar et al., 2000; Ramanathan et al., 1989; Ramanathan and Collins, 1991; Randall and Tjemkes, 1991]. Their generation and dissipation is closely connected to the hydrological cycles through associated processes, such as convection, latent heating, and precipitation [e.g., Baker, 1997; Del Genio and Kovari, 2002; Kahn et al., 2008; Krueger et al., 1995; Lilly, 1988; Kärcher and Ström, 2003; Luo and Rossow, 2004; Rossow and Schiffer, 1999]. Given their large radiative impacts, subtle variations of the spatial distribution, height, frequency of occurrence, thickness, and optical and microphysical properties of the ice clouds can cause substantial feedbacks to the climate [Fu and Liou, 1992; Hartmann and Doelling, 1991; Hartmann et al., 1992; Kiehl, 1994; Miller, 1997; Stephens et al., 1981; Zelinka and Hartmann, 2010]. Accurate characterizations of these properties are therefore crucial to understand the ice clouds' climatic influences, as well as their responses in a changing climate. [3] Climate sensitivity estimate for most models depends critically on the representation of clouds. It is very evident that large disagreement exists in the ice clouds represented in general circulation models (GCMs) [Li et al., 2005, 2007; Waliser et al., 2009]. This is illustrated in Figure 1 by the annual mean ice water path (IWP) estimates from fourteen GCMs contributing to the Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4). The differences can be nearly two orders of magnitude in some regions, and even for globally averaged values the disagreement between the lowest and highest model estimates can be a factor of 20 [Waliser et al., 2009]. It is imperative to reduce such levels of model uncertainty and bias for a quantity as fundamental and conceptually unambiguous as atmospheric ice mass, especially for the models that will provide future climate projections for the next IPCC assessment report. Figure 1Open in figure viewerPowerPoint Global maps of annual mean ice water path (in g m−2) in the 1970–1994 period from the 20th century simulations (scenario 20c3m) in the 14 GCMs contributing to IPCC AR4 (modified from Waliser et al. [2009]). [4] Efforts have been made to derive robust global observational constraints to facilitate the evaluation and improvement of the ice cloud representations in global models. The most recent and valuable additions to the observational resources are the vertically resolved ice cloud water content (IWC) retrieval estimates from radar/lidar/limb sounding, such as those from the CloudSat, the Cloud-Aerosol Lidar Infrared Pathfinder Satellite Observations (CALIPSO), and the Microwave Limb Sounder (MLS). These new satellite products represent a great leap forward in ice cloud observation, as the in situ measurements offer only sparse sampling [e.g., McFarquhar and Heymsfield, 1996; McFarquhar et al., 2000], and the global retrievals from passive nadir-viewing sensors (especially those using infrared, visible, and UV techniques) can exhibit large uncertainty when probing thick clouds with precipitation, mixed phased hydrometeors, and/or multilayer structures [Lin and Rossow, 1996; Stephens et al., 2002; Wu et al., 2006]. Although these data sets have proven beneficial in model comparisons and evaluation, important sampling and sensitivity issues should be carefully considered, as described in previous investigations [e.g., Li et al., 2007; Marchand et al., 2008; Waliser et al., 2009]. [5] An issue when performing model-data comparisons of IWC particularly germane to this investigation is the question of which component(s) of the frozen water mass are represented by the retrieval estimates, and how they relate to model representations. We note that, while it is understood that all ice particles are falling under the influence of gravity, cloud particles tend to be quasi-suspended or "floating" and will be referred to "cloud ice" to distinguish them from truly precipitating particles (i.e., snow and graupel). Given the variety of remote sensing instruments, algorithm sensitivities, and model parameterization techniques, significant complexities exist when attempting to carry out model-data comparisons. For example, the CloudSat Cloud Profiling Radar (CPR) detects nearly all frozen particles as radar reflectivity depends strongly on particle size, while the MLS IWC retrievals are more characteristic of cloud ice alone, owing to difficulties of the microwave radiometer in penetrating thicker clouds. [6] With the consideration of computational efficiency, most GCMs use simple ice representations that typically divide the total frozen condensate into an amount that remains suspended in the atmosphere and an amount that precipitates. Precipitation is assumed in the model to fall either instantly onto the surface or with sedimentation considered in one time step. The difficulties that arise for comparing CloudSat ice contents with GCMs with simplified ice microphysics are somewhat mitigated for models with a multispecies microphysics scheme, such as the Goddard finite-volume multiscale modeling framework (fvMMF) [Tao and Simpson, 1993; Tao et al., 2003, 2009]. In such models, both suspended and precipitating forms of ice may coexist at any model grid point without instantaneous fallout. Therefore, it is imperative to develop a method to categorize the ice mass estimated by CloudSat into portions of suspended, small particles (cloud ice) and precipitating, large particles (snow and graupel). Such distinction allows one to make more meaningful comparisons to global models with different types of ice representation in observations. [7] As a preparatory step for making a distinction between cloud versus precipitating IWC, Waliser et al. [2009] filtered out retrievals that were flagged by CloudSat 2B-CLDCLASS algorithm as either exhibiting surface precipitation or the convective cloud types, both of which would be associated with significant amounts of larger falling hydrometeors. The subset of the nonprecipitating and nonconvective (NP and NC) cases was used as a judicious estimate of the suspended ice particles. With the NP and NC constraints applied, the subsampled CloudSat IWC estimates strongly resembles the IWC values estimated by MLS (version 2.2) [Livesey et al., 2007], which is more representative of only the amount of cloud ice. The NP and NC filtering of CloudSat IWC estimates (hereafter referred to as NPC) serves as an initial guide for assessing the cloud ice constituent represented in GCMs. Figure 2 shows the monthly average of total IWC (IWCTOTAL) retrieved by CloudSat Radar-Visible Optical Depth Cloud Water Content retrieval algorithm (2B-CWC-RVOD, v5.1, R04) in August 2006, and the filtered IWC for precipitating or convective (PoC) cases (IWCPoC) and NPC cases (IWCNPC). The values at 350 hPa (Figure 2a) and the zonal mean profile (Figure 2b) suggest that, on a monthly basis, the cloud ice species (IWCNPC) contributes to around 30–60% of the total atmospheric ice mass. However, a filtering based on qualitative information/flags is subject to limitations and caveats. Precipitating and nonprecipitating hydrometeors can coexist at some levels in the column with surface precipitation and/or convective activity. Moreover, not all precipitation in the column will reach surface. Therefore, the PoC cases may still include some cloud ice mass in the precipitating/convective cloud column, while the NPC cases may contain precipitating columns without the presence of surface precipitation owing to evaporation. Moreover, similar subsampling might need to be applied to the model estimates before carrying out model-data comparison. Figure 2Open in figure viewerPowerPoint Monthly mean total IWC (IWCTOTAL) and the partitioned IWC based on the filtering technique in the work of Waliser et al. [2009] for precipitating or convection cases (IWCPoC), and nonprecipitating and nonconvection cases (IWCNPC) from CloudSat RVOD retrievals in August 2006 (in mg m−3), with (a) the values at 350 hPa plotted at a 8° × 4° resolution and (b) the zonal altitude mean plotted at a resolution of 4° × 25 hPa. [8] The present study aims to develop a technique that distinguishes atmospheric ice species based on a quantitative, microphysical measure, the ice particle size distribution (PSD) provided by the CloudSat algorithm for each IWC retrieval. The new method separates the amount of ice water mass between particles smaller and larger than a selected particle size threshold; the small-sized particles are considered as representative of the suspended cloud ice, while the large ice are deemed precipitating hydrometeors, regardless of the presence of surface precipitation or cloud type. Vertical distributions of cloud ice and precipitating ice mass estimates can therefore be derived in each CloudSat profile. Section 2 describes the CloudSat IWC retrieval algorithm, the methodology for partitioning the ice mass, and the analysis for selecting the appropriate size threshold. Using an ice particle size threshold to distinguish between small ice particles and larger precipitating hydrometeors, the size-partitioned CloudSat IWC estimates are presented in section 3. Sensitivity metrics to the input PSD parameters, and the IWC retrieval algorithm is also discussed. In section 4, the partitioned CloudSat IWC estimates are compared with model representations from the European Centre for Medium-Range Weather Forecasts (ECMWF) model and the fvMMF model. Finally, a summary is provided in section 5 along with a discussion of future development and applications. 2. Data and Methodology CloudSat 2B-CWC-RVOD Retrieval Algorithm [9] CloudSat is one of the five satellites in the A-Train constellation that makes equatorial passes at approximately 01:30 and 13:30 local time. A vertical profile of radar reflectivity factor (Ze) is measured by the 94 GHz CPR at a vertical resolution of 240 m between the surface and 30 km altitude, during each of the 160 ms measurement intervals. The footprint size is around 1.3 km across track and 1.8 km along track. IWC analyzed in the present study is retrieved from the CloudSat 2B-CWC-RVOD algorithm. This algorithm is a modification of the Radar Only algorithm by Austin et al. [2009] (an earlier version is described by Benedetti et al. [2003]). Only a brief description of the algorithm is provided here, and for details the readers can refer to Austin et al. [2009] and Heymsfield et al. [2008]. [10] The forward model in the retrieval algorithm assumes the ice particles to be spheres with a lognormal particle size distribution (PSD) (N(D)); that is, where D represents the diameter of an equivalent mass sphere, Dg is the geometric mean diameter of the ice particles, NT is the total ice particle number concentration, and σlog is the distribution width parameter. By integrating the third moment of the PSD over all possible ice particle sizes assuming a constant ice density (ρi = 917 kg m−3), a simple expression can be derived to express the total IWC with the three size distribution parameters (NT, Dg and σlog), [11] The retrieval algorithm obtains an optimized solution by minimizing the difference between the vector of measured reflectivities and the vector of modeled reflectivities derived from the forward model. The optimization iteration is initialized with an a priori PSD, of which the values Dg and σlog are specified by the temperature dependences obtained from synoptic probe data [Austin et al., 2009], with the temperature information obtained from ECMWF operational analyses. A Ze-IWC relationship from Liu and Illingworth [2000] is applied to solve for the a priori value of NT, using the measured reflectivity and a priori values Dg and σlog as inputs; therefore both temperature and reflectivity are taken into account in determining the a priori PSD. Retrieved IWC and PSD values are obtained by testing for convergence of the iterative solution. [12] Several ice cloud microphysical retrieval algorithms are compared in the work of Heymsfield et al. [2008], using simulated reflectivity and optical depth values based on cloud probe measurements. The mean retrieved-to-measured ratio for IWC from the CloudSat RVOD algorithm is found to be 1.27 ± 0.78 when equivalent radar reflectivity is smaller than −28 dBZe. While most of the IWC retrievals are within ± 25% of the true value, the algorithm exhibits high bias of over 50% when IWC is less than ∼100 mg m−3, with some of the biases related to the potential errors in the measured extinction for small ice crystals in the probe data; therefore the estimated systematic error for IWC is likely ± 40% [Heymsfield et al., 2008]. Thin ice clouds are "missed" by CloudSat owing to the instrumental detection limit (∼0.5 mg m−3). On the basis of the comparison with MLS IWC retrievals in the work of Waliser et al. [2009] and Wu et al. [2009], the amount of thin cirrus (IWC < 2 g m−3) missed by CloudSat is estimated to be less than 10% in terms of mass. We also note that in many global models the thin cirrus clouds are not represented or resolved by parameterization. For these global models the IWC estimates presented in this study can be considered relevant and useful for model evaluation. The upcoming multisensor retrieval data sets [e.g., Delanoë and Hogan, 2010], which take advantages of the different sensitivity of each instrument to particle size and cloud optical depth, are expected to provide better observational constraints for models that are able to represent thin cirrus clouds, and this will be the subject of future studies. Partitioning CloudSat IWC on the Basis of PSD [13] The CloudSat total IWC (IWCTOTAL) can be partitioned into portions of particles smaller and larger than a specific cutoff diameter (Dc). By representing the ice particles as equivalent spheres with diameter D and constant ice density ρi, the partitioned IWC of ice particles larger than Dc (IWC>Dc) is derived by integrating the third moment of the lognormal PSD with Dc as the lower limit, [14] The partitioned IWC of ice particles smaller than Dc (IWC Dc from IWCTOTAL. Figure 3 shows a sample lognormal PSD (red line) and its corresponding mass distribution (black line). An example of cutoff diameter Dc = 150 μm (dotted line) is used to demonstrate the partitioning of IWC 150 (blue area). Figure 3Open in figure viewerPowerPoint A sample lognormal ice number distribution (red curve; left ordinate in log scale), and the corresponding mass distribution (black curve; right ordinate in linear scale). The dotted line represents the cutoff diameter for IWC partitioning (Dc = 150 μm as an example). The partial integrals of the mass distribution for particles smaller and larger than Dc correspond to IWC 150 (blue area), respectively. [15] Selecting the appropriate cutoff diameter for the present methodology is a challenging task, as the size separation between nonprecipitating and precipitating ice particles is not definitive. For example, Morrison and Gettelman [2008] used 200 μm as the threshold size (for the particle maximum length) for converting cloud ice to snow by autoconversion in the GCM bulk cloud scheme. However, on the basis of the analysis of ground-based Doppler radar data, Deng and Mace [2008] found that the mass-weighted fall velocity of cirrus diagnosed from the radar agrees well with sedimentation rate of precipitating ice in the Morrison and Gettelman parameterization, suggesting that all upper tropospheric ice might be considered as precipitation. To identify the likely size separation for floating and precipitating ice species, lights can be shed by investigating the differences of the retrieved PSD between the PoC and NPC filtered cases defined by the surface precipitation and convection flags [Waliser et al., 2009] (also see section 1). Figure 4 shows the temperature (T) dependence of the effective radius (Re), NT, and σlog retrieved from all CloudSat measurements (ALL), and the PoC and NPC cases. The ambient temperature data are taken from collocated ECMWF operational analysis, as in the retrieval algorithm. The mean values and the standard deviations of the three PSD parameters were computed from global observations at 2 K temperature bins. Regional statistics over land and ocean and various latitude bands are subtly different from the global statistics and therefore not shown. The retrieved Re and σlog (Figures 4a and 4c) smoothly increases with T; the values are similar between the PoC and NPC subsets when T is below 240 K. In warmer temperatures the two filtered cases gradually differ: PoC cases show stronger temperature dependence and exhibit larger particle size and PSD width than the NPC cases. The retrieved NT between PoC and NPC cases differ more in lower temperatures and converge when temperature becomes higher, and the PoC cases have higher mean NT in all temperatures. Figure 4Open in figure viewerPowerPoint Temperature dependence of (a) effective radius (Re), (b) ice number concentration, (c) size distribution width parameters (σlog), and (d) probability density function of in-cloud temperature. The means and the standard deviations are calculated in 2 K temperature bins for all in-cloud cases (ALL, blue line), PoC cases (yellow line), and NPC cases (red line) from CloudSat RVOD retrievals in August 2006. [16] To further examine the size separation, the ice mass distributions at six representative temperature bins are plotted in Figure 5 for the PoC (solid yellow line) and NPC (solid red line) subset; each lognormal curve corresponds to the mean value of the PSD parameters calculated in Figure 4 at the selected temperature. The dashed line shows the volume (mass) mean diameter () of each distribution, defined as Figure 5Open in figure viewerPowerPoint Lognormal ice mass distribution for the CloudSat PoC cases (yellow solid line) and NPC cases (red solid line) with the mean values of the PSD parameters in Figure 4 for temperature bins of (a) 209–211 K, (b) 229–231 K, (c) 245–247 K, (d) 255–257 K, (e) 261–263 K, and (f) 271–273 K. The dashed line shows the volume mean diameter () of each distribution. Note the difference in the vertical axes. [17] For T < 240 K (Figures 5a and 5b), the size ranges of the PoC and NPC distribution are narrow (50–200 μm) and mostly overlap, both with the value of around 100 μm, smaller than the size in higher-temperature bins. As discussed in section 1, the PoC cases can contain a certain amount of cloud ice in the column, and most of these small particles in the upper most troposphere are probably the cloud ice species. For T > 240 K (Figures 5a–5f), the mean diameter of the NPC cases remains around 100 μm, although the width of the PSD becomes larger in warmer temperatures. Some of the larger particles in the NPC distribution are likely precipitating ice hydrometeors that are not filtered out by the CloudSat surface precipitation flag. The mean diameter of the PoC cases, however, increases with T to as large as 150 μm for T = 271–273 K (Figure 5f), also with strong dispersion. [18] On the basis of the variation of the mean mass distribution, the size separation of the cloud ice and precipitating ice likely falls between 100 and 200 μm, at least on a global mean basis. This range encompasses the values commonly adopted in GCM cloud parameterization [e.g., Ryan, 2000; Morrison and Gettelman, 2008]. We acknowledge that such size threshold can depend strongly on many variables such as cloud type, updraft velocity, and/or the aspect ratio of the ice species, and therefore the threshold can vary drastically among regions and seasons. Detailed investigation will require collocated information from model analysis and the other remotely sensed measurements, and will be the subject of future studies. For the purpose of this study, we aim to provide a first-order estimate of the partitioned CloudSat IWC for the evaluation of the global models. Therefore we report three sets of estimates in section 3, each with a specific, globally constant Dc value, to cover the most possible range of the threshold values identified in the above global statistics and adopted in models. These results, together with those based on the filtering method, represent the lower and upper limits of the partitioned IWC estimates. We also note that, as the technique is under development and is subject to uncertainty associated with the retrieval algorithm, the current results should be taken more qualitatively than quantitatively. 3. Partitioned CloudSat IWC Estimates Monthly Mean Size-Partitioned CloudSat IWC Data [19] The partitioned IWC estimates with Dc = 100, 150, and 175 μm are shown and discussed in this section. In Figures 6, 7, 10, 12, and 13 the IWC fields are spatially averaged into 8° longitude by 4° latitude grids in the global maps and 4° latitude by 25 hPa grids in the zonal mean profiles. This is a compromise between a grid size that is comparable to GCM model output, adequate to capture monthly mean variability, and provide adequate samples for each box given CloudSat's nadir-only sampling. We also note that, as the retrieval algorithm does make a priori assumptions, the partitioned IWC presented here represents the interpretation specifically based on the RVOD products. The partitioned IWC will be affected the retrieved PSD parameters and the algorithm employed to retrieve ice PSD parameters; sensitivity analyses to the PSD parameters and the retrieval products are presented in section 3.2. However, we would like to stress that model estimates of ice water path (IWP) vary by two orders of magnitude (e.g., Figure 1), and thus having a quantitative estimate, such as that described here, has significant value despite its shortcomings and uncertainties. Moreover, it prototypes a useful technique that can be elaborated on with the development of increasingly sophisticated instruments and algorithms. Figure 6Open in figure viewerPowerPoint Monthly mean IWC>Dc and IWC<Dc from CloudSat RVOD retrievals in August 2006 (in mg m−3): (a) values at 350 hPa plotted at a 8° × 4° resolution, with a cutoff diameter (Dc) of 100 μm; (b) zonal altitude mean plotted at a resolution of 4° × 25 hPa, with Dc = 100 μm; (c and d) similar to Figures 6a and 6b but with Dc = 150 μm; (e and f) similar to Figures 6a and 6b but with Dc = 175 μm. Figure 7Open in figure viewerPowerPoint Monthly zonal altitude mean ratio of CloudSat small (nonprecipitating) ice IWC based on (a) IWC<100, (b) IWC<150, (c) IWC Dc) and smaller (IWC Dc in the tropical upper troposphere peaks at 300–400 hPa, while IWC<Dc peaks at a higher altitude between 250 and 300 hPa. IWC<Dc fields at 350 hPa with the three cutoff diameter all exhibit high values over areas with frequent and intense convective activities, such as the tropical western Pacific, Central America, and central Africa; the maximum IWC<Dc is around 6 mg m−3 for Dc = 100 μm, 13 mg m−3 for Dc = 150 μm, and 20 mg m−3 for Dc = 175 μm. [21] Comparisons can be made between the PSD partitioned IWC (Figure 6) and the filtered IWC based on surface precipitation and convection flags (Figure 2). Despite the different methods, the patterns of the spatial distribution of the partitioned IWC are consistent between the two estimates both vert