Upper tropospheric water vapor and clouds play an important role in Earth's climate, but knowledge of them, in particular diurnal variation in deep convective clouds, is limited. An essential variable to understand them is cloud ice water content. The Japanese Superconducting Submillimeter-Wave Limb-Emission Sounder (SMILES) on board the International Space Station (ISS) samples the atmosphere at different local times allowing the study of diurnal variability of atmospheric parameters. We describe a new ice cloud data set consisting of partial Ice Water Path and Ice Water Content. Preliminary comparisons with EOS-MLS, CloudSat-CPR and CALIOP-CALIPSO are presented. Then, the diurnal variation over land and over open ocean for partial ice water path is reported. Over land, a pronounced diurnal variation peaking strongly in the afternoon/early evening was found. Over the open ocean, little temporal dependence was encountered. This data set is publicly available for download in HDF5 format.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Clouds are the major source of uncertainty in the estimates of the radiative forcing of climate [IPCC, 2007]. Clouds have both a heating and cooling effect on the Earth-atmosphere system: a heating effect through absorption and reemission of thermal infrared radiation emitted below them, and a cooling effect through reflection of incoming sunlight. Whether the net effect is one of warming or cooling depends on many complex factors, such as cloud's altitude, horizontal extent, composition (ice versus liquid water), microphysical properties, as well as the incoming solar radiation. Particularly, the amount of ice clouds differ by orders of magnitude among climate models [e.g., Stephens et al., 1990; Penner, 2004; Li et al., 2005; Jiang et al., 2012; Li et al., 2012] due to poorly resolved dynamic processes that generate the ice clouds, the accuracy of the water vapor transport to and within the upper troposphere, the microphysical formation mechanism of the ice particles [Del Genio, 2001], and the assumptions made about the ice sedimentation rates [Wilson, 2000; Mitchell et al., 2008]. A review of the importance of ice crystals properties to climate prediction can be found in Baran .
 Visible and infrared observations of clouds have provided vast amount of information on cloud properties; however, these techniques are limited to thin clouds or to the topmost layer of clouds, hence, cannot provide much needed information on the internal structure of clouds, such as the cloud ice mass. Passive submillimeter satellite observations can penetrate through most ice clouds and provide cloud ice mass information primarily through scattering in the Rayleigh-Mie region so that cloud temperature, cloud and surface emission, and multiple scattering that affect other remote sensing techniques are relatively unimportant [e.g., Evans et al., 1998; Wu et al., 2005; Eriksson et al., 2007; Buehler et al., 2007; Jiménez et al., 2007; Buehler et al., 2012]. Furthermore, active instruments have been proven to measure cloud ice mass with high vertical resolution [Stephens et al., 2002; Winker et al., 2009; Kumagai et al., 2003].
 Measurements from the Earth Observing System (EOS) Microwave Limb Sounder (MLS), the CloudSat - Cloud Profiling RADAR (CPR) and the CALIPSO - Cloud Aerosol Lidar with Orthogonal Polarization (CALIOP) instruments (for more details about these instruments, see section 3) have provided a wealth of information about the atmospheric Ice Water Content (IWC) serving as guidance to improve the simulations and predictions of cloud-related processes [e.g, Li et al., 2005, 2012]. However, these instruments are all in the “A-train” constellation of satellites and observe essentially the same air mass within minutes of each other, conveying information only at two fixed local times each orbit (∼ 1:45 A.M./P.M.); the “A-train” constellation of satellites cannot resolve the diurnal cycle. In this study, we introduce the new partial Ice Water Path (pIWP) and IWC observations from the Superconducting Submillimeter-Wave Limb-Emission Sounder (SMILES).
 SMILES was launched in September 2009 and successfully attached to the front-side of the International Space Station (ISS). SMILES measured, on a time sharing basis, two of three frequency bands: 624.32–625.52GHz (band A), 625.12–626.32GHz (band B), and 649.12–650.32GHz (band C) from October 2009 to April 2010 [Kikuchi et al., 2010]. During this period, the SMILES antenna, with an instantaneous field of view of 0.09° (∼3 km at limb tangent) [Manabe et al., 2008, 2010], scanned the Earth's limb covering tangent heights between 10 and at least 60 km about 1600 times per day, with the objective of measuring emission lines of atmospheric minor constituents such as O3 and its isotopes, HCl, ClO, HO2, BrO, and HNO3. The radiation received was down-converted using two 4K cooled Superconductor-Insulator-Superconductor (SIS) mixers, and spectra were recorded by two sets of acousto-optical spectrometers (AOS). SMILES data has a latitudinal coverage from 65°N to 38°S covering all the local solar times. SMILES, as a payload of the ISS, which is in a 51.6° inclined orbit, made observations at local times that drifted ∼ 20 min earlier each day covering the entire diurnal cycle in a period of about 2 months. These measurements provide unique valuable information to potentially enable improvements in model representations of diurnal cycles in convection, known to be deficient in many scenarios [e.g., Tian et al., 2004; Lee et al., 2008].
 This paper is structured as follows. First, the retrieval methodology used to convert the SMILES measurements to relative humidity, IWC, and pIWP is presented followed by an analysis of the precision and systematic uncertainties. Then, we compare this new data to the MLS, CloudSat-CPR, and CALIPSO-CALIOP data sets. Finally, the cloud pIWP diurnal variation derived from SMILES is explored.
2 Retrieval Approach
 The fundamental measurement from which ice cloud retrievals derive is the cloud-induced radiance (Tcir). This quantity is defined as the difference between the measured radiance and the expected clear-sky radiance. Once the clear-sky radiance is known, we use simulated Tcir-pIWP and Tcir-IWC linear relationships to map it either to pIWP or IWC, respectively. Figure 1 outlines the SMILES ice cloud retrieval algorithm.
 The estimation of the clear-sky radiances starts with an inversion of SMILES spectra to determine the limb tangent pressure as well as the concentrations of O3 and HCl. The retrieval uses the optimal estimation technique as described by [Rodgers, 2000] using the hydrostatic equation as a constraint (in a similar manner as the retrieval described by [Livesey et al.2006]). The temperature information is taken from GEOS-5 [Rienecker et al., 2008], and the a priori for the O3and HCl is taken from the closest, geographically and in time, Microwave Limb Sounder (MLS) [Waters et al., 2006] scan (for more information on this data set, see section 3). Their respective a priori precision, their assumed random error, are 5 ppmv and 1 ppbv. GEOS-5 temperature, MLS O3, and HCl are chosen as a priori to use the most realistic atmospheric representation available.
 This inversion is performed in two phases. The first phase estimates these parameters over a subset of the vertical range with the purpose of deriving a height offset to mitigate the ISS pointing uncertainty. The subset radiances used are selected to have a minimum strength of 1K and a maximum strength of 60K to ensure that they are optically thin. The second phase uses the updated pointing heights and repeats the whole retrieval, now between 18 and 55km. The reason for the two-phase strategy is to ensure that the vertical range of radiances used for the pressure retrieval is consistent from scan to scan.
 Relative humidity (RHi) is inverted with a similar algorithm to that described by Read et al. . In this algorithm, relative humidity is retrieved directly from the water vapor continuum, modifying the water vapor Jacobians using the Goff-Gratch function [List, 1951], an empirical function that estimates the saturation water vapor pressure at a given temperature. Once relative humidity is known, the clear-sky radiance can be computed simulating the limb radiances using GEOS-5 temperature, the pressure, HCl, and O3retrieved values, as well as the retrieved relative humidity in values of up to 110% (i.e., any value greater than 110% is truncated). This RHi threshold was previously used by Wu et al.  and represents the uncertainty of RHi. The most useful frequencies to derive the clear-sky radiance are those away from strong spectral lines, the so-called “window channels.” Figure 2 shows the position of the window channels used in this study. These window channels were selected avoiding the influence of overlapping species and trying to find the most optically thin radiance at 100hPa; that is to say, where its optical depth is the smallest. Brightness temperature was computed inverting the Plank function.
 Figure 3 shows a typical example of the measured window channel radiance and the Tcir (the difference between the measured radiance profile and its corresponding clear-sky radiance). As shown, most of the points tend to bundle around zero Tcir with a ∼2K variation. Lines representing the 3σ– 2σ deviation are overlaid as an indication of the clear-sky limit. Points that fall outside this region are cloud occurrences. The 3σ – 2σ screening method is explained in detail by [Livesey et al., 2011]. In short, Tcir data was averaged iteratively rejecting 2σ outliers each iteration. After convergence, the 3σthreshold determined if a Tcir measurement is statistically significant. In other words, Tcir must be greater than the mean + 3σto be considered as a significant cloud hit.
 Figure 3 shows that clouds can enhance or reduce radiances with respect to the clear-sky radiance mostly depending on the tangent height measured, with positive Tcir at high altitudes and negative Tcir at low altitudes. This can be understood assuming a thin cloud acts only as a scattering medium, reflecting the negligible radiance from above the cloud (cosmic background radiation plus the downward radiation), and the radiance from below the cloud (the upward radiation) into the line of sight. Under this assumption, the measured radiance under the presence of this particular cloud will be almost constant, probably around 101K (a simple average of the 200K from the surface and the ∼3K from the cosmic background radiation) irrespective of the cloud height. Hence, at high altitudes, where the clear-sky radiance is smaller that the cloud radiance, the cloud will produce a positive Tcir, while at low altitudes the clear-sky radiance is greater than the cloud radiance producing a negative Tcir. In the interim altitudes, where the cloud radiance is similar to the clear-sky radiance, the lack of contrast between the two makes any ice cloud information difficult to obtain.
 At low tangent heights, in this case for pressures larger than ∼200hPa (approximately around 12.5km), the atmosphere becomes opaque (i.e., its optical depth becomes greater than 1) before the line of sight reaches the limb; therefore, although the tangent height of the measurement might be close to the surface, the actual cloud signal originates higher up in the atmosphere. In this scenario, vertically scanning does not provide much vertically resolved information on the cloud. Rather, each limb scan measures a different location thereby trading vertical sampling for horizontal sampling much like a nadir instrument viewing side to side sweeping. As such, negative Tcir conveys information that is more related to the pIWP, the ice partial column along the line of sight, than IWC, the amount of ice per unit volume at the measured altitude. At higher tangent heights, in this case for pressures smaller than 100hPa, where the atmosphere is optically thin, vertically scanning provides vertically resolved cloud information physically related to IWC. At the intermediate pressure levels, between 180 and 100hPa, the clouds lack contrast with the clear-sky background, and, as explained before, there is no cloud information in the measured radiance.
 Positive and negative Tcir have been mapped to IWC or pIWP, respectively, using radiative transfer simulations described in the following section. In summary, the entire SMILES data set has been processed: IWC, the density of ice at the measured tangent pressure, is available between 100 to 70hPa; and pIWP, the partial amount of ice along the SMILES line of sight above around 12.5km is available for pressures greater than 180hPa.
2.1 Radiative Transfer Simulations
 Once the SMILES Tcir are known, the next step to generating the ice cloud products is to use Tcir-pIWP and Tcir-IWC relationships derived using simulations from the 2-D Cloud-Sky Radiative Transfer (RT) model, a 2-D version of the model described by Wu  and Wu et al. . This model assumes spherical ice particles, computes the absorption and scattering cloud volume coefficients and phase functions assuming single scattering properties, and then solves the RT equation (neglecting polarization) iteratively (following [Wilheit et al., 1982; Yeh et al., 1990]) to include multiple scattering effects. Figure 4 shows an outline of these simulations.
 These simulations were performed on merged daily IWC fields from the CloudSat-CPR and the CALIPSO-CALIOP (for more information of these data sets see section 3). These merged fields were simply the largest IWC value between the two data sets at any given pressure level reported in a pressure grid of 24 levels per decade with an along-track resolution of half degree. This simple method allows us to exploit the sensitivity of CALIPSO-CALIOP for small ice particles that CloudSat-CPR cannot detect, as well as the higher dynamic range of CloudSat at IWC concentrations where CALIPSO-CALIOP measurements saturate. The goal was to produce a set of simulations which had realistic IWC concentrations and heterogeneity from which to produce Tcir–IWP and Tcir–IWC correlations. The atmospheric temperature and concentrations were taken from the nearest MLS profiles, and the particle size distribution (PSD) parametrization used was the McFarquhar and Heymsfield  parametrization derived from measurements during the Central Equatorial Pacific Experiment Campaign. Although, this PSD may be skewed due to impact shattering on the housing of probes [Korolev et al., 2011], and as shown by Field et al., this could result in an overestimation of the ice water content by the mass diameter relationship of around 20%–30%, this PSD was used to be consistent with the MLS and CloudSat retrieval algorithms. Note that these simulations take into account the SMILES antenna response function.
 For each cloudy scene, the RT model was run under clear-sky conditions to compute the corresponding clear-sky radiances. Then, the pairing of the known pIWP or IWC and the Tcir was used to derive the Tcir-pIWP and Tcir-IWC relationships. As an example, Figure 5 shows a typical Tcir-pIWP scatter plot at 1000hPa for simulations of band A B, and C, using data corresponding to 15 March 2010. In addition, a linear fit is shown. This fit is used to characterize the Tcir-pIWP relationship at this tangent pressure level and for this particular day; the linear Pearson correlation coefficient and the slope 1-sigma uncertainty for the three bands was around −0.95 and 0.016, respectively. Scatter plots at different tangent pressure levels were analyzed to find this relationship at different tangent pressures. This analysis was carried-out for ∼30 days sparsely distributed throughout the entire SMILES mission, and no seasonal dependence was found.
 Given the lack of seasonality, an average of the daily corresponding fits, for each pressure level, and their standard deviation were used to characterize the data. Figure 6 displays the mean value and standard deviation for several pressure levels. As the pressure decreases, the standard deviation increases; above 180hPa, the standard deviation is so big that no useful relationship can be found. These values were used to map the measured Tcir onto the corresponding pIWP; the standard deviation of these fits is a measure of the uncertainty in pIWP due to vertical and along-track inhomogeneities as well as the pIWP missing contributions that are below the penetration depth of the signal.
 A similar analysis was performed to find relationships between Tcir and IWC. In this case, the Tcir-IWC relationships were restricted to Tcir measured at tangent pressures between 100 and 70hPa, where most of the Tcir are positive. Furthermore, the Tcir used were not a single measurement but rather the mean value of a defined atmospheric volume (275km along the track and 3.3km in the vertical). As with pIWP, no seasonality was found during the analysis of the ∼30 days spread throughout the entire SMILES mission. Hence, as before, for each pressure level, an average of the corresponding fits and their standard deviation was used to characterize the data. In this case, the spread is caused by cloud inhomogeneities along the line of sight.
2.2 Tcir Error Assessment
 The total error in the calculated Tcir is given by a sum of random and systematic errors. The random errors are determined by the noise in the SMILES measurements, in this case less than 0.7K for a single AOS channel [Kikuchi et al., 2010]. And the systematic errors arise from uncertainties in the forward model, instrumental issues, and retrieval approximations.
 The random errors plus some systematic errors in the derived Tcir can be estimated empirically from the data using the 3σ-2σmethod used to find the clear-sky limit. The systematic errors captured in this error estimate are the uncertainties in the temperature and in the gas concentrations retrievals used as part of the Tcir calculation. Note that other sources of systematic errors affecting the Tcir estimates, such as spectral and radiometric calibration issues, and pointing deficiencies, still need to be investigated, but are presumably negligible compared with the rest of the systematic uncertainties (see section 2.3). Figure 7 shows the Tcir bias and precision for a typical day. As shown, the bias (the mean) is less than 0.5K while the precision (1σvariation) is around 1K for pressure levels between 1000 and 200hPa, and less than 0.5K for pressures smaller than 200hPa.
 Figure 8 shows how these errors propagate into the retrieved pIWP and IWC for band A (bands B and C are similar). These propagations were performed updating the Tcir-pIWP and the Tcir-IWC relationships discussed in section 2.1. For pIWP, the error varies from less than ∼10% to up to 40%. For IWC, the error varies from less than ∼10% to up to 80%, depending on the induce Tcir value. Maximum relative errors occur when the Tcir approaches zero, that is to say, for cloud free scenes. The absolute errors give a sense of the minimum ice detection threshold and are also specified in Figure 8.
2.3 pIWP and IWC Systematic Uncertainties
 In addition to the errors discussed in the previous section, the total error in pIWP and IWC needs to include more systematic uncertainties, such as cloud inhomogeneity, water vapor spectroscopy uncertainties, uncertainties due to different PSDs, and uncertainties in the particle shape. The impact of these systematic uncertainties are summarized in Table 1.
Table 1. Estimated Systematic Uncertainties for pIWP and for IWC
 The uncertainty given by cloud inhomogeneity was estimated from the standard deviation of the fits discussed in section 2.1. These fits were obtained using realistic distributions of ice clouds data from CloudSat-CPR and CALIPSO-CALIOP and, as such, the standard deviation in these fits represent the variation due to cloud spatial variability as well as temporal variability, at least in the sense that several days were used to estimate those fits.
 To investigate the error associated with the uncertainties in the water vapor spectroscopy, its parameters were perturbed by 30% and the Tcir-pIWP and Tcir-IWC relationships were recomputed. The errors were found comparing the Tcir-pIWP and Tcir-IWC relationships found using the perturbed spectroscopy parameters against the unperturbed PSD.
 Furthermore, the Tcir-pIWP and Tcir-IWP relationships were recomputed using different PSD parameterization schemes: the one given by Field et al.  and those described by Donovan and Lammeren  and Heymsfield et al. . When using the first two, the particle maximum dimension was converted to the mass equivalent diameter using the mass-diameter relationship described by Cotton et al. . Unfortunately, the Donovan and Lammeren  and Heymsfield et al.  parameterization schemes are limited to temperatures greater than −50 °C and therefore the errors associated to different particle size distribution can only be computed for tangent pressures between 1000 to ∼220 hPa. This pressure range only matches the pressure levels of the Tcir-pIWP relationships, so, for these PSDs, their corresponding error was only computed for pIWP.
 To investigate the uncertainty due to particle shape or habit, a forward model capable of simulating polarized scattered radiance is needed. That is to say, to propagate the four elements of the Stokes vector through the appropriate extinction and scattering phase matrices. For instance, the Atmospheric Radiative Transfer Simulator (ARTS) [Buehler et al., 2005; Eriksson et al., 2011a] could be coupled with a T-matrix code (such as the one described by Mishchenko and Travis ), to estimate this error; however, this is outside the scope of this study. Yet, as discussed by Eriksson et al. [2011b], an instrument that covers the left-and-right-hand circular components of the polarization effects, such as SMILES, should be less affected by the naturally occurring cloud complex ice particle shapes. Considering this, we assume an error of 20% for this uncertainty, as derived by Wu et al.  for the MLS instrument (see section 3 for more details about this instrument). As discussed by Eriksson et al. [2011b], as the MLS instrument measures the horizontal and vertical linear components, it should be more affected by the non-spherical particle shapes than SMILES. Presumably, even considering that the SMILES window channels will be more sensitive to smaller particles, which have larger polarization effects, our estimate is in the right magnitude.
3 Comparison With Other Data Sets
 Comparisons of a monthly means were made with those of the MLS, CloudSat-CPR and CALIPSO-CALIOP instruments.
 MLS measures thermal microwave limb emission in five spectral regions from 115GHz to 2.5THz, including a 640GHz region with similar coverage as SMILES. It was launched in July 2004 on board the Aura spacecraft. MLS scans the limb from the ground to about 95km roughly 3500 times per day. It covers between 82°S and 82°N providing near global coverage. At most latitudes, about half of these measurements are made around 1:45 P.M. and the other half around 1:45 A.M., except around the poles where the measurements change between daytime and nighttime conditions or vice versa [Waters et al., 2006].
 Because the penetration depth of passive sensors, such as SMILES and MLS, depends on the frequency used, in this study, we consider the MLS pIWP (version 3.3) from the 640GHz radiometer to better match the SMILES observations (the standard MLS pIWP have an altitude base of ∼6km while the MLS-640GHz and SMILES pIWP have one at ∼12.5km). For the IWC comparisons, we show both the standard and the MLS-640GHz products to compare the validated MLS-IWC product as well as the IWC derived with SMILES-like frequencies. Note that, although MLS-640GHz and SMILES pIWP and IWC are derived from similar frequencies, they used different retrieval algorithms. In addition, the MLS-640GHz products do not have a published assessment of their quality.
 CloudSat is the first spaceborne Cloud Profiling RADAR (CPR) launched in April 2006 [Stephens et al., 2002]. It measures the power backscattered by clouds as a function of distance. This satellite orbits in formation minutes ahead of the Aura spacecraft. The effective dimensions of a single measurement are approximately 1.3km cross-track and 1.7km along-track. Each CPR profile consists of 125 levels that are ∼240m apart from the ground to ∼28km, although the CPR resolution is approximately 500m. Here we use the IWC profiles from the 2B-CWC-RO R04 product.
 The Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) instrument is a two wavelength polarization sensitive LIDAR that provides high resolution profiles of aerosols and clouds [Winker et al., 2009]. It was launched in April 2006, alongside CloudSat, on board the CALIPSO platform. This platform orbits in formation with Aura, as CloudSat also does.
 pIWP were derived from the CloudSat-CPR and CALIPSO-CALIOP IWC data sets assuming a base altitude of 12.5km to match the penetration depth of the passive sensors, the altitude at which most atmospheric scenarios the optical depth window channels become optically thick.
3.1 Geographical Distribution
 Figure 9 shows a monthly mean of pIWP for January 2010 for SMILES, MLS, CloudSat-CPR, and CALIPSO-CALIOP data. For SMILES, two monthly means are shown, one which represents a simple average of all the data (a merge of the three bands) with no diurnal screening, and one showing the SMILES measurements made around 1:45 A.M. and 1:45 P.M. Figure 9 also displays the percentage difference between the data sets. To compute it, all the SMILES data were used in order to maximize the number of SMILES points per latitude-longitude bin. Hence, direct comparison with the other measurements needs to keep diurnal sampling issues in mind. As can be seen, the four data sets show relatively high values around the inter-tropical convergence zone (ITCZ), agreeing well in the pIWP geographical pattern. The data sets do not agree entirely in magnitude, but this is expected; the instruments are sensitive to different particle sizes and different assumptions are made in each retrieval. Also, cloud inhomogeneity, variability, and the different sampling of the diurnal cycle, will all contribute to disagreement in pIWP. In general, SMILES pIWP lies between 100% smaller and 100% greater than MLS-640GHz, CloudSat-CPR, and CALIPSO-CALIOP data, depending on location.
 Figure 10 shows a monthly mean of IWC at around 100hPa for SMILES (bands A, B, and C), MLS-640GHz, and CALIPSO-CALIOP data. CloudSat-CPR data is not shown because it is not sensitive to ice clouds at this pressure level. Again, the three data sets agree well in the IWC geographical pattern, with high values in the ITCZ. As for pIWP, SMILES IWC is around 100% smaller/greater than the MLS and CALIPSO-CALIOP data, respectively, depending on location.
4 Cloud Ice Water Path Diurnal Variation
 To further explore this new data set, the entire pIWP data was binned by time and separated into land and open ocean regimes to study their respective ice cloud diurnal variations. This diurnal variation is linked to the diurnal variation of cloudiness, deep convection and humidity, with all these diurnal cycles associated with the 24 h variation of the solar forcing. As such, these are fundamental components of the variability of the global climate system.
 Many studies have shown that over land, these diurnal cycles peak in the afternoon and early evening while, over the ocean, they peak around the early morning [e.g., Janowiak et al., 1994; Chen and Houze, 1997; Tian et al., 2004; Hong et al., 2006; Yang et al., 2001; Eriksson et al., 2010]. Figure 11 shows the land and open ocean regimes used in this study as well as their respective diurnal deviation from the mean.
 As can be seen, there is a pronounced diurnal cycle over land peaking strongly in the afternoon/early evening. The timing of its maximum is similar to those of deep convective clouds and high cold clouds reported, either for tropical South Africa, tropical South America, or Australia, by Hong et al., . These cycles were derived with the Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar (PR), e.g., [Kummerow et al., 1998] and by the Visible and Infrared Scanner (VIRS). Over the open ocean, SMILES data show little temporal dependence and a multipeak structure.
5 Data File Format
 All the SMILES IWC and pIWP data described here are available from ftp://mls.jpl. nasa.gov/pub/outgoing/smiles. The data are stored in the standard HDF version 5 on a one-day granularity, and named according to SMILES_icecloud_ yyyy-mm-dd_vXX_XX.h5, where yyyy, mm, and dd are the calendar year, month and day, respectively, and where XX_XX is the version number. Each file contains two swaths, IWC and pIWP and each of them contains three more swaths “A”, “B”, and “C”, which correspond to the data obtained from each of the SMILES bands.
 Each of these A/B/C swaths contains data fields called “val” and “precision”, which describe the value and the precision of the data, either in mg/m3for IWC, or in g/m2for pIWP. In addition to these fields, the geolocation information is given by the fields “latitude” and “longitude”; the universal time and date are given by the fields “ut” and “date”; and the local time and local date are saved in the fields, “localtime” and “localdate”.
 Furthermore, the IWC and RHi swaths include the field “pressure” in hPa, which indicates the pressure surface of the measurement. The RHi field also contains temperature in kelvins, water vapor and water vapor precision in ppmv and radiance χ2(chisq) fit value. In the near future, there will be another SMILES upper troposphere - lower stratosphere (UTLS) humidity product developed by the SMILES level-2 research processor in the National Institute of Information and Communications Technology (NICT).
 In addition to all these fields, each of the A/B/C swaths contain a “status” field. This is merely a flag indicating if that specific SMILES band was used that particular day. Values of “status” equal to zero indicate that there is no data in that swath.
6 Summary and Discussions
 After retrieving O3, HCl, pressure, and RHi, we have estimated the clear-sky radiance for the entire SMILES mission. This clear-sky radiance is simply the forward model run, using GEOS-5 temperature, the previously retrieved O3, HCl, pressure, and RHi values, but truncating the RHi at a value of 110%. Once the clear-sky radiance was known, we subtracted it from the measured radiances to compute the Tcir.
 These Tcir were mapped onto pIWP or IWC using Tcir-pIWP or Tcir-IWC relationships found using simulations of the 2-D Cloud-Sky Radiative Transfer model. These simulations were driven with a CloudSat-CPR and CALIPSO-CALIOP IWC merged fields, and the nearest MLS temperature and gas concentrations conditions. pIWP was derived from mostly negative Tcir measured between 1000hPa and 180hPa, while IWC was derived from 100hPa to 70hPa from mostly positive cloud induced radiances. Between these pressure ranges, the lack of contrast between clouds and the clear-sky background makes any cloud information difficult to infer. Systematic uncertainties affecting these products were investigated; the total systematic error was found to vary from 50% to 110% depending on the tangent pressure.
 pIWP and IWC visual inspection comparisons with the MLS, CloudSat-CPR, and CALIPSO-CALIOP data sets were found to agree well in the geographical pattern, displaying relatively high values in the ITCZ. Although no exact agreement was found in the magnitude with SMILES pIWP and IWC smaller/greater than the MLS, CloudSat-CPR and CALIPSO-CALIOP data, depending on the location, this was somewhat expected due to the different particle size sensitivity of each instrument, the different assumptions in each retrieval, and the diurnal nature of the SMILES data.
 After the pIWP was binned by time and region, the diurnal deviation from the mean was computed for land and open ocean conditions. Over land, the expected diurnal variation was found with a pronounced diurnal variation peaking strongly in the afternoon/early evening. Over the open ocean, little temporal dependence was found.
 This data set is publicly available for download in HDF5 format. Future analysis of this data set, in combination with other observations such as the TRMM precipitation radar, can greatly improve our understanding of diurnal variations of cloud properties.
 The research described in this paper was carried out by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. JEM/SMILES mission is a joint project of Japan Aerospace Exploration Agency (JAXA) and National Institute of Information and Communications Technology (NICT).