Abstract

The retrieval of ice cloud microphysical and optical properties from satellite-based infrared observation remains a challenging research topic, partly because of the sensitivity of observed infrared radiances to many surface and atmospheric parameters that vary on fine spatial and temporal scales. In this study, the sensitivity of an infrared-based ice cloud retrieval to effective cloud temperature is investigated, with a focus on the effects of cloud-top height and geometric thickness. To illustrate the sensitivity, the authors first simulate brightness temperatures at 8.5 and 11.0 μm using the discrete ordinates radiative transfer (DISORT) model for five cloud-top heights ranging from 8 to 16 km and for varying cloud geometric thicknesses of 1, 2, 3, and 5 km. The simulations are performed across a range of visible optical thicknesses from 0.1 to 10 and ice cloud effective diameters from 30 to 100 μm. Furthermore, the effective particle size and optical thickness of ice clouds are retrieved from the Moderate Resolution Imaging Spectroradiometer (MODIS) measurements on the basis of a lookup-table approach. Specifically, the infrared brightness temperatures are simulated from the collocated Atmospheric Infrared Sounder (AIRS) level-2 product at 28 atmospheric levels and prescribed ice cloud parameters. Variations of the retrieved effective particle size and optical thickness versus cloud-top height and geometric thickness are investigated. Results show that retrievals based on the 8.5- and 11.0-μm bispectral approach are most valid for cloud-top temperatures of less than 224 K with visible optical thickness values between 2 and 5. The present retrievals are also compared with the collection-5 MODIS level-2 ice cloud product.

1. Introduction

Ice clouds play an important role in regulating the earth’s radiation budget through their reflection of solar radiation and trapping of the earth’s outgoing infrared (IR) radiation. Determination of the net cloud radiative forcing of these clouds requires a global, diurnal climatology, which can most readily be achieved using satellite observation. At present, the parameterization of ice cloud properties is a primary cause of discrepancy in cloud radiative forcings calculated from various general circulation models (GCMs) and in model comparisons with flux measurements (Potter and Cess 2004; Webb et al. 2001). To increase confidence in climate prediction, global observations on fine spatial and temporal scales from satellite-based instruments are needed to retrieve ice cloud properties such as effective particle size and cloud optical thickness, which are essential to climate studies (Wielicki et al. 1995).

Methodologies have been developed for daytime retrievals of cloud properties from satellite measurements. A common approach is the bispectral technique developed by Nakajima and King (1990), which utilizes the reflectances at two bands: one visible, nonabsorbing band sensitive to cloud optical thickness and the other band a near-infrared wavelength at which water or ice particles are absorptive that is sensitive to the effective particle size. This approach is used for the Moderate Resolution Imaging Spectroradiometer (MODIS) operational cloud retrievals (Barnes et al. 1998; Platnick et al. 2003). MODIS 1.38-μm reflectance data have been used to retrieve optical thickness for radiatively important thin and subvisual cirrus (Dessler and Yang 2003; Meyer et al. 2004; Smith et al. 1998), while the capability of simultaneously retrieving effective size and optical thickness has been demonstrated with 1.38- and 1.88-μm reflectance measurements (Gao et al. 2004).

Satellite-observed IR radiances associated with a cloudy atmosphere are not only sensitive to effective particle size and cloud optical thickness, but also to underlying surface properties, the vertical profile of temperature, water vapor, and other absorbing gases, as well as cloud height and geometric thickness. Prior to any retrieval, a simulation of what a satellite is expected to observe at specific infrared wavelengths must be performed, and the simulation requires accurate knowledge of these atmospheric parameters. The inherent sensitivity of observed IR radiances to these parameters and their variability on fine spatial and temporal scales creates a challenge for robust retrieval schemes required to build a climatology of ice cloud properties not dependent on solar illumination, thus consistent for daytime and nighttime data.

Radiometric measurements from several hyperspectral and narrowband infrared sensors have been used to perform retrievals of ice cloud properties and sensitivity analyses to atmospheric and surface parameters. The spectral region of 750–1000 cm−1 (10.0–13.3 μm) has been studied extensively for sensitivity to ice particle effective size (Bantages et al. 1999; Chung et al. 2000). Wei et al. (2004) used Atmospheric Infrared Sounder (AIRS) observations to show that the slope of brightness temperatures (BTs) within this spectral region, along with data from the 1070–1135-cm−1 (∼8.8–9.3 μm) band, could be used to infer ice cloud optical thickness for clouds with visible optical thickness (τ0.55) less than 10. Li et al. (2005) applied a minimum-residual technique to AIRS observations at 790–960 cm−1 (∼10.4–12.6 μm) and 1080–1130 cm−1 to retrieve effective size and optical thickness. Similarly, Huang et al. (2004) used these spectral regions to simultaneously retrieve ice cloud properties from high-resolution interferometer sounder (HIS) data collected during the First International Satellite Cloud Climatology Project Regional Experiment–Arctic Cloud Experiment (FIRE-ACE) field campaign. Errors of about 10%–15% in these retrievals were demonstrated to be caused by error in assumed cloud temperature (height) for optically thick cloud and surface skin temperature for optically thin cloud. However, for clouds with τ0.55 greater than 5, IR brightness temperatures approach their asymptotic values, causing difficulty in the retrieval of ice cloud properties.

Other techniques have been developed to retrieve cloud properties from IR measurements. For example, Chiriaco et al. (2004) used observations at 8.65, 11.15, and 12.05 μm to retrieve the effective size of ice particles, and reduced uncertainty in the retrievals with depolarized lidar measurements by 20%–65%. Emissivities at 8.3 and 11.1 μm from the Television and Infrared Observation Satellite (TIROS-N) Operational Vertical Sounder have been used to retrieve effective size, independent of atmospheric state and with little sensitivity to cloud height or geometric thickness (Rädel et al. 2003; Stubenrauch et al. 1999). Baum et al. (2000) used theoretical calculations at 8.5 and 11.0 μm to infer effective size and optical thickness from MODIS Airborne Simulator (MAS) data, with consistency in optical thickness but discrepancy in effective size when compared with solar- and near-IR-based retrievals.

Whether hyperspectral or narrowband satellite observations are used to retrieve ice cloud properties, the measured IR radiation, even at preferred window channels, is sensitive to not only the desired cloud properties but also to many atmospheric and surface parameters. Increasing accuracy in IR-based ice cloud property retrievals depends greatly on understanding the error associated with assumptions made about these atmospheric and surface parameters. Of significant importance, Hong et al. (2007) pointed out that common bispectral retrievals derived from IR measurements use precomputed lookup tables developed from one profile containing a cloud of fixed geometric thickness. It was then shown that not considering cloud geometric thickness could introduce either overestimation or underestimation in retrieved cloud properties. Similarly to Hong et al. (2007), we focus on the sensitivity of a bispectral retrieval technique utilizing two MODIS IR bands, 8.5 and 11.0 μm, to cloud-top height as well as geometric thickness. First, brightness temperatures at these two wavelengths are simulated in a standard tropical atmosphere for ice clouds with varying cloud-top height and geometric thickness to assess the impact of these a priori assumptions on our IR-based retrieval, and also to assess the sensitivity of our model (and robustness of bispectral IR models in general) to retrieve cloud optical thickness and effective particle size in terms of the cloud effective temperature.

In addition, a MODIS scene over the tropical Pacific Ocean is then chosen and retrievals of effective particle size and cloud optical thickness are performed for a case assuming a cloud-top height at the derived AIRS cloud-top pressure and a geometric thickness of 1 km (first case). Alternate retrievals are performed after adjusting the cloud-top height to the next highest and next lowest AIRS levels to assess the impact of varying cloud-top height on the retrieval. A similar case is carried out for clouds of varying geometric thicknesses, but a fixed cloud-top height. Retrieval results are intercompared, and also assessed relative to the MODIS level-2 operational cloud product.

This paper is organized as follows: Section 2 details the data and model used in this study. The results of the sensitivity analysis with varying cloud geometric height along with analysis of varying cloud geometric thickness are discussed in section 3, and retrieval cases are analyzed in section 4. The summary of this study is given in section 5.

2. Data and model

The radiometric measurements acquired by the Moderate Resolution Imaging Spectroradiometer and the Atmospheric Infrared Sounder on the Aqua satellite platform are used in this study. MODIS has 36 channels that cover key atmospheric bands ranging from 0.415 to 14.235 μm with varying spatial resolutions of 250, 500, and 1000 m and scans across a swath-width of 2330 km (Barnes et al. 1998). The level-1b radiance data at MODIS bands 29 and 31, which correspond to 8.5 and 11.0 μm, are used in this study to retrieve ice cloud properties. The results of the retrieval are then compared with the MODIS collection-5, level-2 cloud product (MYD06) effective size and optical thickness (King et al. 2006).

The AIRS version-4, level-2 standard retrieval (Chahine et al. 2002) product provides atmospheric temperature, pressure, ozone, and moisture profiles for 28 atmospheric levels, as well as for cloud-top pressure (CTP), which is derived from a relaxation method (Chahine 1974). The retrieved surface and atmospheric parameters are derived at a spatial resolution of 45 km × 45 km. At this resolution, thousands of 1-km MODIS pixels may be present within one AIRS pixel, thus requiring collocation of the datasets. It should be noted at the time of this study, the AIRS version 5 algorithms were not yet operational. With any improvements made to the AIRS sounding products, it should be expected that results from our retrieval model would likely improve (especially from more accurate temperature soundings in the upper troposphere where ice clouds are present).

To simulate the infrared radiances at top-of-atmosphere, we use the well-known discrete ordinates radiative transfer (DISORT) model (Stamnes et al. 1988). The DISORT requires the Legendre polynomial expansion of the phase function. To attenuate the Legendre polynomial expansion coefficients we use the δ-fit truncation method (Hu et al. 2000), which stems from the previous δ-M method (Wiscombe 1977), to truncate the scattering phase functions of ice clouds with various effective sizes.

The simulation of radiances at wavelengths of 8.5 and 11.0 μm requires accounting for the influence of absorption by molecules in the atmosphere (e.g., CO2, H2O vapor, O3, CH4, and N2O). Not accounting for gaseous absorption in the atmosphere could affect brightness temperature simulations at each wavelength by as much as a few kelvins, which would yield far different results in the present retrievals. In this study, the correlated k-distribution computational package developed by Kratz (1995) is used to account for the gaseous absorption in the atmosphere. As mentioned, profiles from the AIRS level-2 standard retrieval are used to determine pressure, temperature, water vapor, and ozone. For trace gases profiles, constant climatological values at the surface are used and scaled by the pressure at each atmospheric layer. With these profiles, the optical depth for each clear-sky atmospheric layer, at each band, is then computed for input into the radiative transfer model.

To be consistent with the MODIS collection-5 processing for level-2 cloud product, we use the bulk single-scattering properties reported in Baum et al. (2005a,b). The bulk ice particle models were derived from 1117 particle size distributions selected from five field campaigns studying both tropical and midlatitude cirrus clouds. For the selected MODIS bands, the single-scattering properties from Yang et al. (2005) are averaged into bulk single-scattering properties and weighted by the MODIS spectral response function for the following habit distribution: D < 60 μm, 100% droxtals; 60 μm < D < 1000 μm, 15% bullet rosettes, 50% solid columns, 35% plates; 1000 μm < D < 2000 μm, 45% solid columns, 45% hollow columns, 10% aggregates; 2000 μm < D, 97% bullet rosettes, 3% aggregates, where D is the maximum dimension. A detailed analysis for the computation of the bulk single-scattering properties is provided in Baum et al. (2005a,b).

3. Brightness temperature sensitivity to cloud geometric height and geometric thickness

As stated, the brightness temperatures observed above an ice cloud at IR wavelengths are not only dependent on the cloud’s microphysical properties, but also on the effective temperature of the cloud itself. This effective temperature, in turn, depends on the cloud geometric height and geometrical thickness. Thus, when retrieving cloud optical and microphysical properties using our bispectral approach, placement of the cloud layer into the correct atmospheric layer(s) is essential for retrieval accuracy. In sections 3a and 3b, we quantitatively assess the influence that cloud geometric height and geometric thickness (cloud temperature) have on simulated IR brightness temperatures, respectively. The results are analyzed in terms of the impact these cloud property assumptions will have on our bispectral retrieval model.

a. Effect of cloud geometric height on BT simulations

We first investigate the sensitivity of simulated infrared brightness temperatures at top-of-atmosphere to cloud-top height (temperature) in a standard tropical atmosphere (McClatchey et al. 1972). It is these simulations on which our lookup tables are based, so it is necessary to first gauge the influence of cloud-top height. The standard tropical atmosphere contains 34 atmospheric computational layers. A single-layer cloud is inserted into five different layers with cloud-top heights ranging from 8 to 16 km, each with a constant cloud geometric thickness of 1 km. Brightness temperatures are simulated with DISORT for clouds with effective diameters of 30, 50, and 100 μm and visible optical thicknesses between 0.1 and 10.

Figures 1a, 1c, and 1e show simulated brightness temperatures at 8.5 and 11.0 μm for clouds with an effective diameter of 30, 50, and 100 μm, respectively, as a function of optical thickness. Dark lines are for 8.5-μm BTs, while lighter lines correspond to the 11.0-μm BTs. The cloud-top temperature ranges from 197 K at 16 km to 250 K at 8 km. For each curve, BTs at each wavelength are warm for optically thin cloud, and decrease with increasing optical thickness until they nearly reach their asymptotic limit where the BTs converge toward the actual cloud-top temperature.

Fig. 1.

(a),(c),(e) Simulated 8.5- and 11.0-μm BTs for five clouds of varying height and De of 30, 50, and 100 μm, respectively, as a function of optical thickness. (b),(d),(f) The 8.5–11.0-μm BT difference from (a), (c), and (e) for five clouds of varying height as a function of optical thickness. Each cloud layer is 1 km thick.

Fig. 1.

(a),(c),(e) Simulated 8.5- and 11.0-μm BTs for five clouds of varying height and De of 30, 50, and 100 μm, respectively, as a function of optical thickness. (b),(d),(f) The 8.5–11.0-μm BT difference from (a), (c), and (e) for five clouds of varying height as a function of optical thickness. Each cloud layer is 1 km thick.

Figures 1b, 1d, and 1f show the 8.5–11.0-μm brightness temperature difference (BTD8.5–11 for each cloud of varying height and effective sizes of 30, 50, and 100 μm as a function of optical thickness. The relationship between BTD8.5–11 and effective diameter is the foundation of our bispectral retrieval scheme. Comparing the simulations of clouds with effective diameters of 30, 50, and 100 μm for a height of 16 km, the BTD8.5–11 is largest for clouds with small effective diameters, illustrating good sensitivity of the model to smaller ice particles. However, for the clouds with diameters larger than ∼50 μm, BTD8.5–11 becomes much smaller as the model becomes less sensitive to larger ice particles. For each optical thickness value, BTD8.5–11 varies only slightly between clouds of equal height with effective diameters of 50 and 100 μm, in support of the prior statement.

Also, comparing the plotted curve of a cloud at 10 km and particles with a diameter of 30 μm with the curve of a cloud at 16 km but an effective diameter of 100 μm, we see that the BTD8.5–11 at each optical thickness are almost identical. Therefore, the retrieval of effective particle size in our retrieval model greatly depends on the assumption of cloud-top height (temperature), and that the assumption of one height cannot be given universally.

Figures 1b, 1d, and 1f also show that for optically thin clouds (τ0.55 < 2) and optically thick clouds (τ0.55 > 7), there is not much sensitivity to cloud-top height. For optically thin clouds, 11.0-μm transmissivity of the cloud is very high, and for optically thick clouds the emissivity is close to 1 for each wavelength, causing BTD8.5–11 to be very small. Even though the cloud-top height (temperature) assumption becomes less important, it would become difficult to infer ice cloud particle size and optical thickness for clouds that are too optically thin or thick using the BTD8.5–11 approach.

It is evident from Fig. 1 that the region of highest sensitivity for retrievals falls within an optical thickness range of τ0.55 = 2–5, where BTD8.5–11 is largest, but this range is narrowed as cloud height decreases or as effective size increases. Clouds with cloud-top temperatures colder than 224 K may be optimal for retrievals; however, for geometrically thicker clouds, this value may be colder as transmission from warmer layers within the cloud becomes an increasingly important factor, to be discussed later in section 3b. It is clear that if we insert our cloud layer too high (cold) or too low (warm), we may retrieve very different results through either increased or decreased sensitivity given by BTD8.5–11. Methods of improving the accuracy of this assumption are necessary for confident IR-based retrievals.

b. Effect of cloud geometric thickness on BT simulations

Our model is used to simulate 8.5- and 11-μm BTs in the standard atmosphere for clouds with geometric thicknesses of 0.5, 1, 2, 3, and 5 km (ΔT of 3, 6, 13, 20, and 32 K, respectively) for optical thicknesses from 0.1 to 10, to investigate the sensitivity of the BTD8.5–11 retrieval approach to cloud geometric thickness. Figures 2a, 2c, and 2e show the results for simulations of a cloud top at 12 km (T = 224 K) and effective diameters of 30, 50, and 100 μm, respectively, as a function of optical thickness. The simulated BTs of the 0.5-km-thick cloud are taken as truth and subtracted from the simulated BTs of each thickness (e.g., ΔZ BTD = BT5.0km − BT0.5km) for the figures. The simulations show that as ΔT increases with cloud geometric thickness, the ΔZ BTD also increases. For example, for a cloud with τ0.55 of 4.5 and a De of 30 μm at a height of 12 km, there is an 8-K difference in BTs at 11.0 μm if the cloud is 5 km thick compared to if it were only 0.5 km thick. The BTs are also higher at 8.5 μm for thicker clouds, although to a lesser degree than 11.0 μm because of less absorption at 8.5 μm for τ0.55 less than 6. Physically, as a cloud layer extends into lower (warmer) regions of the atmosphere, the IR radiance at each wavelength is emitted from ice crystals in warmer regions, which increases the effective emitting temperature of the cloud. Thus, BTs at 8.5 and 11.0 μm will be higher for a geometrically thick cloud (large ΔT) when compared with a geometrically thinner cloud (small ΔT), with all other atmospheric parameters and cloud properties being equal. With this difference potentially being 8 K or even more, the impact on retrieved cloud properties will be significant depending on the assumed cloud thickness. However, the ΔZ BTD for each geometric thickness does not show much sensitivity as the effective diameter is increased from 30 to 100 μm.

Fig. 2.

(a),(c),(e) Simulated 8.5- and 11.0-μm BTs for clouds with geometric thickness of 1, 2, 3, and 5 km minus the simulated BT for a cloud with geometric thickness of 0.5 km as a function of optical thickness. (b),(d),(f) The 8.5- and 11.0-μm BTD from (a), (c), and (e). Simulations are performed for clouds with effective diameters of 30, 50, and 100 μm and a cloud-top height of 12 km.

Fig. 2.

(a),(c),(e) Simulated 8.5- and 11.0-μm BTs for clouds with geometric thickness of 1, 2, 3, and 5 km minus the simulated BT for a cloud with geometric thickness of 0.5 km as a function of optical thickness. (b),(d),(f) The 8.5- and 11.0-μm BTD from (a), (c), and (e). Simulations are performed for clouds with effective diameters of 30, 50, and 100 μm and a cloud-top height of 12 km.

Although we may have a better understanding of how cloud geometric thickness affects the transmission of IR radiance through the cloud layer, a more direct assessment of how cloud geometric thickness may affect our bispectral retrieval is to gauge its impact on the BTD8.5–11 itself. Figures 2b, 2d, and 2f show the BTD8.5–11 for each geometric thickness given in Figs. 2a, 2c, and 2e, respectively. Again, the BTD8.5–11 for each thickness is differenced from the 0.5-km BTD8.5–11Z BTD8.5–11). As geometric thickness increases from 1 to 5 km, ΔZ BTD8.5–11 values increase from 0.5 to 1.5 K. Although this decreases with increasing effective particle size, the differences are still large enough to impact a retrieval using this bispectral approach. This is apparent when analyzing the bispectral plots in section 4.

Figure 3 is the same as Fig. 2, except it is for a cloud-top height of 15 km (203 K). For Figs. 3a, 3c, and 3e, there are still large differences in BTs between the various geometric thicknesses, and these are comparable to the previous simulations at 12 km. However, Figs. 3b, 3d, and 3f, show much larger ΔZ BTD8.5–11 values for the various geometric thicknesses, on the order of 0.5 K or greater.

Fig. 3.

As in Fig. 2, but for a cloud-top height of 15 km.

Fig. 3.

As in Fig. 2, but for a cloud-top height of 15 km.

This result indicates that the cloud geometric thickness assumption becomes perhaps even more important as cloud-top temperature increases (lower cloud tops) while ΔT remains nearly constant, as the sensitivity of our model to retrieve cloud optical thickness and effective particle size depends strongly on the BTD8.5–11. Not only does this illustrate the importance of placing the cloud in the correct atmospheric layers to accurately create bispectral lookup tables, but it may also mean increased difficulties in retrieving cloud properties because of decreased sensitivity to optical thickness and effective particle size in relatively warmer ice clouds.

4. Effective cloud temperature impact on IR-based retrieval

a. Impact of cloud geometric height on IR-based retrieval

For the ice cloud retrieval case study where cloud geometric height is varied, a MODIS granule containing a region of a single-layer ice cloud over the tropical Pacific Ocean is chosen and assumed to be homogeneous (Fig. 4). Figures 5a and 5b show corresponding plots of MODIS-retrieved optical thickness and effective radius, respectively, for the case area. All of the pixels within this scene are classified as “ice” by the MODIS 1-km Quality Assurance phase detection algorithm (approximately 1100 pixels). The MODIS optical thickness range falls within our most sensitive region from the theoretical simulation where BTD8.5–11 is largest, approximately between 2 and 5. The MODIS effective radius is generally below 30 μm, which is on the threshold of low-to-moderate sensitivity.

Fig. 4.

False color image (0.65 μm in red, 2.13 μm in green, and 11.0 μm in blue) of MODIS case area from granule MYD021KM.A2005354.0140, approximately 1100 pixels. Box is centered at 3.05°N, 174.6°W over the Pacific Ocean on 20 Dec 2005.

Fig. 4.

False color image (0.65 μm in red, 2.13 μm in green, and 11.0 μm in blue) of MODIS case area from granule MYD021KM.A2005354.0140, approximately 1100 pixels. Box is centered at 3.05°N, 174.6°W over the Pacific Ocean on 20 Dec 2005.

Fig. 5.

(a) MODIS-retrieved cloud optical thickness, (b) MODIS-retrieved effective radius, and (c) collocated AIRS pixels within the scene, labeled with the granule index number.

Fig. 5.

(a) MODIS-retrieved cloud optical thickness, (b) MODIS-retrieved effective radius, and (c) collocated AIRS pixels within the scene, labeled with the granule index number.

Two collocated AIRS profiles are used for atmospheric profile data, as well as for cloud-top pressure, which determines the atmospheric layer containing cloud. Figure 5c shows the coverage of AIRS pixels 549 and 550 in the case area, corresponding to the approximately 1100 MODIS pixels. Since the AIRS CTP usually places the cloud top between two AIRS standard output levels, the atmosphere between 100 and 300 hPa (where pure ice cloud is likely to occur) is interpolated from 5 to 20 atmospheric levels. With the finer vertical resolution in the profile, cloud properties are retrieved for a greater variety of cloud-top heights and geometric thicknesses (section 4b).

The first retrieval initially places the cloud at the AIRS specified cloud-top pressure. The derived CTP for pixels 549 and 550 are 126 and 148 hPa, respectively. For the second and third retrievals, the cloud-top heights are adjusted to the next highest and next lowest AIRS levels (100 and 150 hPa, respectively) from the original 28 AIRS output levels. The cloud-layer thickness for the retrievals is fixed at 1 km and is not isothermal.

Figure 6 shows the IR bispectral plots representing the lookup tables for each retrieval, which result from simulating IR brightness temperatures with the profile data from AIRS pixels 549 and 550 using DISORT. The IR brightness temperature observations from each MODIS pixel are plotted on the figures. For a given cloud-top-level simulation, the cloud for each pixel is placed at the same level so that differences in the lookup tables are attributed to differences in the atmospheric profiles between the two AIRS pixels. Table 1 lists the differences in cloud and profile data for each AIRS pixel for all three retrievals. Differences in the skin temperature and cloud-top and bottom temperatures are fairly significant between the pixels in each case. However, independent retrievals from each AIRS pixel area did not show that one pixel performed better than the other when compared to the MODIS operational cloud product. This could be due to the fact that for moderately optically thick clouds, the retrievals may be more sensitive to cloud-top temperature than to the atmospheric profile or skin temperature, similar to Huang et al. (2004).

Fig. 6.

IR bispectral plots for AIRS pixels 549 and 550, respectively, at the (a),(b) AIRS 100-hPa pressure level, (c),(d) AIRS-derived cloud-top pressure level, and (e),(f) AIRS 150-hPa pressure level. Cloud geometric thickness is fixed at 1 km. MODIS IR observations are overplotted. Please note different scale for (a) and (b).

Fig. 6.

IR bispectral plots for AIRS pixels 549 and 550, respectively, at the (a),(b) AIRS 100-hPa pressure level, (c),(d) AIRS-derived cloud-top pressure level, and (e),(f) AIRS 150-hPa pressure level. Cloud geometric thickness is fixed at 1 km. MODIS IR observations are overplotted. Please note different scale for (a) and (b).

Table 1.

Cloud and profile data for three retrievals with varied cloud-top height. The varying typefaces differentiate the three retrieval cases.

Cloud and profile data for three retrievals with varied cloud-top height. The varying typefaces differentiate the three retrieval cases.
Cloud and profile data for three retrievals with varied cloud-top height. The varying typefaces differentiate the three retrieval cases.

Comparison of the IR bispectral plots among the three retrievals reveals large shifts toward lower BTD8.5–11 and toward warmer 11.0-μm BTs as the CTP decreases from 100 (Figs. 6a,b) to 150 hPa (Figs. 6e,f). This is an expected result, as our sensitivity analysis showed smaller BTD8.5–11 values for the warmer cloud-top temperatures. Also, as the cloud-top temperature and base temperature become warmer, the 11.0-μm BT should increase more than at 8.5 μm where absorption is less. The MODIS IR observations overplotted in Fig. 6 also show the influence that the cloud-top temperature assumption has on this retrieval case. For the coldest cloud top at 100 hPa, the MODIS observation is plotted in the region with little sensitivity to effective particle size. As the cloud-top temperature increases for the AIRS CTP and 150-hPa CTP cases, the plotted lines shift to the right and down in the table, and the MODIS IR observations then become plotted in regions of higher sensitivity.

Figure 7 shows the geolocated results of the retrievals for optical thickness and effective size, with corresponding scatterplots in Fig. 8. At 100 hPa, the optical thickness underestimates MODIS, suggesting that the simulated 11-μm BT is too cold. Optical thickness retrievals perform better for the cases with warmer cloud tops, although they slightly underestimate those derived from MODIS. From the evaluation of results for effective particle size, we can see the influence of the colder cloud top on the 100-hPa retrieval. Although many retrieved pixels lay close to a 1-to-1 line with the MODIS-retrieved effective size, several pixels are shown to have values different from MODIS. These pixels correspond to the MODIS IR observations overplotted in the region that tend to show little sensitivity to effective size in the bispectral plots. Again, as cloud-top temperatures become warmer, the MODIS IR observations allow for a more consistent retrieval of effective particle size. However, compared to the archived level-2 MODIS retrievals, our retrieved effective particle size tends to be underestimated.

Fig. 7.

Geolocated retrieved (left) optical thickness and (right) effective radius of case area for a cloud at the (a),(b) AIRS 100-hPa pressure level, (c),(d) AIRS-derived cloud-top pressure level, and (e),(f) AIRS 150-hPa pressure level.

Fig. 7.

Geolocated retrieved (left) optical thickness and (right) effective radius of case area for a cloud at the (a),(b) AIRS 100-hPa pressure level, (c),(d) AIRS-derived cloud-top pressure level, and (e),(f) AIRS 150-hPa pressure level.

Fig. 8.

Scatterplots comparing retrieved values of (left) optical thickness and (right) effective radius with MODIS MYD06 retrieved values corresponding to the plots in Fig. 7. The standard deviation σ and correlation coefficient r are shown for each plot.

Fig. 8.

Scatterplots comparing retrieved values of (left) optical thickness and (right) effective radius with MODIS MYD06 retrieved values corresponding to the plots in Fig. 7. The standard deviation σ and correlation coefficient r are shown for each plot.

b. Impact of cloud geometric thickness on IR-based retrieval

While assessing the impact of cloud geometric thickness on our bispectral retrieval, the same model and data from section 4a are applied. In this case, we use the AIRS-derived CTP as a fixed cloud height instead of altering the cloud-top height as in section 4a. Instead, cirrus cloud effective size and optical thickness retrievals are performed for cloud geometric thicknesses of 1, 3, and 5 km (ΔT of ∼10, 20, and 35 K). Table 2 lists the atmospheric and cloud temperature properties for each AIRS pixel and retrieval case.

Table 2.

Cloud and profile data for three retrievals with varied cloud geometric thickness.

Cloud and profile data for three retrievals with varied cloud geometric thickness.
Cloud and profile data for three retrievals with varied cloud geometric thickness.

Figure 9 shows the IR bispectral plots, which represent the lookup tables associated with each retrieval where cloud geometric thickness is varied. Comparing the figures first for each individual retrieval case, the plots only shift slightly between AIRS pixels 549 and 550, due to the few Kelvin differences in cloud-top temperature and skin temperature. However, comparing the plots among the different retrievals reveals the influence of altering cloud geometric thickness. The main impact is that the BTD8.5–11, drawn for each constant effective diameter, decreases with increasing geometric thickness (or increasing cloud effective temperature). For the 1-km case, the maximum BTD8.5–11 along the 10-μm diameter line is close to 21 and 20 K for pixels 549 and 550, respectively. In the 5-km case, these values are decreased to 18 and 16 K, respectively. Physically, as a cloud with moderate optical thickness extends into warmer layers of the atmosphere, the cloud gains a higher effective emitting temperature at 11.0 μm where absorption is greater, as compared to 8.5 μm where scattering is more prominent. This translates into higher 11.0-μm transmittance (and BT) while there is little change in the 8.5-μm transmittance. Thus, as cloud geometric thickness increases, the 8.5- and 11.0-μm BTs converge, decreasing the BTD8.5–11 values as well as the sensitivity of the model to cloud optical thickness and effective particle size.

Fig. 9.

IR bispectral plots for AIRS pixels 549 and 550, respectively, for (a),(b) a 1-km-thick cloud, (c),(d) 3-km-thick cloud, and (e),(f) 5-km-thick cloud at the AIRS-derived cloud-top pressure level. MODIS IR observations are overplotted.

Fig. 9.

IR bispectral plots for AIRS pixels 549 and 550, respectively, for (a),(b) a 1-km-thick cloud, (c),(d) 3-km-thick cloud, and (e),(f) 5-km-thick cloud at the AIRS-derived cloud-top pressure level. MODIS IR observations are overplotted.

Figure 10 shows the geolocated results for each retrieval case along with corresponding scatterplots in Fig. 11. Optical thickness retrievals are fairly consistent among the retrievals, and compare well with the MODIS retrievals. For the cloud with 1-km thickness, retrieved optical thickness underestimates MODIS, but has a very high correlation. As the cloud layer thickens, the optical thickness retrieval values increase, but they then begin to overestimate MODIS in the most optically thick regions. Effective particle size retrievals also underestimate MODIS, with the most error occurring with the 5-km-thick cloud where standard deviations are highest. In each retrieval case, shifting the bispectral plot to the left (colder 11.0-μm BTs) by decreasing the cloud-top temperature may improve the effective size retrieval. Although nonlinearity to the simulated BTs also exists because of skin temperature error and gaseous absorption at the IR wavelengths. Thus, further investigation is warranted.

Fig. 10.

Geolocated retrieved optical thickness and effective radius of case area for a cloud with (a),(b) 1-km thickness, (c),(d) 3-km thickness, and (e),(f) 5-km thickness at the AIRS-derived cloud-top pressure level.

Fig. 10.

Geolocated retrieved optical thickness and effective radius of case area for a cloud with (a),(b) 1-km thickness, (c),(d) 3-km thickness, and (e),(f) 5-km thickness at the AIRS-derived cloud-top pressure level.

Fig. 11.

Scatterplots comparing retrieved values of optical thickness and effective radius with MODIS MYD06 retrieved values corresponding to the plots in Fig. 10. The standard deviation σ and correlation coefficient r are shown for each plot.

Fig. 11.

Scatterplots comparing retrieved values of optical thickness and effective radius with MODIS MYD06 retrieved values corresponding to the plots in Fig. 10. The standard deviation σ and correlation coefficient r are shown for each plot.

5. Summary

Ice cloud–top height and cloud geometric thickness are important for the simulation of infrared brightness temperatures at the top of the atmosphere. The simulations are necessary for the retrieval of ice cloud effective particle size and optical thickness. Through simulations in a standard tropical atmosphere, we have shown that the maximum sensitivity for retrievals utilizing the BTD8.5–11 bispectral approach is for cloud-top temperatures colder than 224 K. If a cloud is too optically thin (τ0.55 < 1), then transmission of upwelling radiance minimizes BTD8.5–11. Similarly, for ice clouds with optical thicknesses greater than 7 or effective diameters greater than 50 μm, the cloud will radiate as a blackbody and brightness temperatures between each channel converge to their asymptotic values. However, adequate sensitivity in brightness temperature simulations exists for clouds with an optical thickness range of 2–5 and for effective particle sizes less than 50 μm. Furthermore, by increasing the geometric thickness from 0.5 to 5 km, we have shown that simulated BTs can increase by as much as 8 K and decrease the BTD8.5–11 by as much as 2 K. Thus, consideration of these parameters is essential for accurately retrieving ice cloud properties.

In our case study, we examined a scene of a single-layer cirrus cloud over the tropical Pacific Ocean. AIRS-retrieved atmospheric, surface, and cloud-top properties over the case area were used to simulate IR brightness temperatures at 8.5 and 11.0 μm necessary for the construction of lookup tables used for retrieving cirrus cloud properties from collocated MODIS IR observations. It was shown that placement of the cloud-top height and the assumption of geometric thickness both have a large effect on the simulations and lookup tables, and thus impact the retrieval of cirrus properties. Therefore, constraining both of these parameters could improve the cloud retrieval results compared to the MODIS level-2 cloud product.

Applicability of this model to retrievals is limited to single-layer cirrus scenes over an ocean surface. In nature these types of scenes are usually contaminated by overlapping clouds, and occur over all surface types. It is also limited by the fact that each scene contains a cloud with a unique cloud-top height and geometric thickness, therefore requiring on-the-fly lookup table computations to be used for retrieval. Future work should investigate the potential of IR-based retrievals in the case of multilayered clouds over all surface types, with the added requirement of precomputed lookup libraries to improve the computational efficiency of the model. Other sensitivities to the forward model that would affect simulated BTs, such as water vapor or trace gases, also have grounds for investigation to fully understand the errors associated with retrieving cloud optical and microphysical properties in the IR.

Acknowledgments

This study is partly supported by the National Science Foundation Physical Meteorological Program managed by Dr. Andrew Detweiler (ATM-0239605), and by a NASA research grant (NNG04GL24G) from the Radiation Sciences Program managed by Dr. Hal Maring and NASA MODIS program managed by Dr. Paula Bontempi.

REFERENCES

REFERENCES
Bantages
,
R. J.
,
J. E.
Russel
, and
J. D.
Haigh
,
1999
:
Cirrus cloud top-of-atmosphere radiance spectra in the thermal infrared.
J. Quant. Spectrosc. Radiat. Transfer
,
63
,
487
498
.
Barnes
,
W. L.
,
T. S.
Pagano
, and
V. V.
Salomonson
,
1998
:
Prelaunch characteristics of the Moderate Resolution Imaging Spectroradiometer (MODIS) on EOS-AM1.
IEEE Trans. Geosci. Remote Sens.
,
36
,
1088
1100
.
Baum
,
B. A.
,
D. P.
Kratz
,
P.
Yang
,
S. C.
Ou
,
Y-X.
Hu
,
P. F.
Soulen
, and
S-C.
Tsay
,
2000
:
Remote sensing of cloud properties using MODIS airborne simulator imagery during SUCCESS 1. Data and models.
J. Geophys. Res.
,
105
,
11767
11780
.
Baum
,
B. A.
,
A. J.
Heymsfield
,
P.
Yang
, and
S. T.
Bedka
,
2005a
:
Bulk scattering properties for the remote sensing of ice clouds. Part I: Microphysical data and models.
J. Appl. Meteor.
,
44
,
1885
1895
.
Baum
,
B. A.
,
P.
Yang
,
A. J.
Heymsfield
,
S.
Platnick
,
M. D.
King
,
Y-X.
Hu
, and
S. T.
Bedka
,
2005b
:
Bulk scattering properties for the remote sensing of ice clouds. Part II: Narrowband models.
J. Appl. Meteor.
,
44
,
1896
1911
.
Chahine
,
M. T.
,
1974
:
Remote sounding of cloud atmospheres. I. The single cloud layer.
J. Atmos. Sci.
,
31
,
233
243
.
Chahine
,
M. T.
, and
Coauthors
,
2002
:
AIRS-Team retrieval for core products and geophysical parameters.
AIRS Algorithm Theoretical Basis Doc. JPL D-17006. [Available online at http://eospso.gsfc.nasa.gov/eos_homepage/for_scientists/atbd/docs/AIRS/atbd-airs-L2.pdf.]
.
Chiriaco
,
M.
,
H.
Chepfer
,
V.
Noel
,
A.
Delaval
,
M.
Haeffelin
,
P.
Dubuisson
, and
P.
Yang
,
2004
:
Improving retrievals of cirrus cloud particle size coupling lidar and three-channel radiometric techniques.
Mon. Wea. Rev.
,
132
,
1684
1700
.
Chung
,
S.
,
S.
Ackerman
,
P. F.
van Delst
, and
W. P.
Menzel
,
2000
:
Model calculations and interferometer measurements of ice-cloud characteristics.
J. Appl. Meteor.
,
39
,
634
644
.
Dessler
,
A. E.
, and
P.
Yang
,
2003
:
The distribution of tropical thin cirrus clouds inferred from Terra MODIS data.
J. Climate
,
16
,
1241
1248
.
Gao
,
B-C.
,
K.
Meyer
, and
P.
Yang
,
2004
:
A new concept on remote sensing of cirrus optical depth and effective ice particle size using strong water vapor absorption channels near 1.38 and 1.88 μm.
IEEE Trans. Geosci. Remote Sens.
,
42
,
1891
1899
.
Hong
,
G.
,
P.
Yang
,
H-L.
Huang
,
B. A.
Baum
,
Y-X.
Hu
, and
S.
Platnick
,
2007
:
The sensitivity of ice optical and microphysical passive satellite retrievals to cloud geometric thickness.
IEEE Trans. Geosci. Remote Sens.
,
45
,
1315
1323
.
Hu
,
Y-X.
,
B.
Wielicki
,
B.
Lin
,
G.
Gibson
,
S-C.
Tsay
,
K.
Stamnes
, and
T.
Wong
,
2000
:
Delta-fit: A fast and accurate treatment of particle scattering phase functions with weighted singular-value decomposition least-square fitting.
J. Quant. Spectrosc. Radiat. Transfer
,
65
,
681
690
.
Huang
,
H-L.
,
P.
Yang
,
H.
Wei
,
B. A.
Baum
,
Y-X.
Hu
,
P.
Antonelli
, and
S. E.
Ackerman
,
2004
:
Inference of ice cloud properties from high spectral resolution infrared observations.
IEEE Trans. Geosci. Remote Sens.
,
42
,
842
853
.
King
,
M. D.
,
S.
Platnick
,
P. A.
Hubanks
,
G. T.
Arnold
,
E. G.
Moody
,
G.
Wind
, and
B.
Wind
,
2006
:
Collection 005 change summary for the MODIS cloud optical property (06_OD) algorithm.
.
Kratz
,
D. P.
,
1995
:
The correlated k-distribution technique as applied to the AVHRR channels.
J. Quant. Spectrosc. Radiat. Transfer
,
53
,
501
517
.
Li
,
J.
, and
Coauthors
,
2005
:
Retrieval of cloud microphysical properties from MODIS and AIRS.
J. Appl. Meteor.
,
44
,
1526
1543
.
McClatchey
,
R. A.
,
R. W.
Fenn
,
J. E. A.
Selby
,
F. E.
Volz
, and
J. S.
Garing
,
1972
:
Optical properties of the atmosphere.
Air Force Cambridge Research Laboratory Research Paper 411, AFCRL-72-0497, 108 pp
.
Meyer
,
K.
,
P.
Yang
, and
B-C.
Gao
,
2004
:
Optical thickness of tropical cirrus clouds derived from MODIS 0.66- and 1.375-μm channels.
IEEE Trans. Geosci. Remote Sens.
,
42
,
833
841
.
Nakajima
,
T.
, and
M. D.
King
,
1990
:
Determination of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements. Part I: Theory.
J. Atmos. Sci.
,
47
,
1878
1893
.
Platnick
,
S.
,
M. D.
King
,
S. A.
Ackerman
,
W. P.
Menzel
,
B. A.
Baum
,
J. C.
Riédi
, and
R. A.
Frey
,
2003
:
The MODIS cloud products: Algorithms and examples from Terra.
IEEE Trans. Geosci. Remote Sens.
,
41
,
459
473
.
Potter
,
G. L.
, and
R. D.
Cess
,
2004
:
Testing the impact of clouds on the radiation budgets of 19 atmospheric general circulation models.
J. Geophys. Res.
,
109
,
D02106
.
doi:10.1029/2003JD004018
.
Rädel
,
G.
,
C. J.
Stubenrauch
,
R.
Holz
, and
D. L.
Mitchell
,
2003
:
Retrieval of effective ice crystal size in the infrared: Sensitivity study and global measurements from TIROS-N Operational Vertical Sounder.
J. Geophys. Res.
,
108
,
4281
.
doi:10.1029/2002JD002801
.
Smith
,
W. L.
,
S.
Ackerman
,
H.
Revercomb
,
H.
Huang
,
D. H.
DeSlover
,
W.
Feltz
,
L.
Gumley
, and
A.
Collard
,
1998
:
Infrared spectral absorption of nearly invisible cirrus clouds.
Geophys. Res. Lett.
,
25
,
1137
1140
.
Stamnes
,
K.
,
S-C.
Tsay
,
W.
Wiscombe
, and
K.
Jayaweera
,
1988
:
A numerically stable algorithm for discrete-ordinate-method radiative transfer in multiple scattering and emitting layered media.
Appl. Opt.
,
27
,
2502
2509
.
Stubenrauch
,
C. J.
,
R.
Holz
,
A.
Chedin
,
D. L.
Mitchell
, and
A. J.
Baran
,
1999
:
Retrieval of cirrus ice crystal sizes from 8.3 and 11.1 μm emissivities determined by the improved initialization inversion of TIROS-N operational vertical sounder observations.
J. Geophys. Res.
,
104
,
793
808
.
Webb
,
M. C.
,
C.
Senior
,
S.
Bony
, and
J-J.
Morctrette
,
2001
:
Combining ERBE and ISCCP data to assess clouds in the Hadley Centre, ECMWF and LMD atmospheric climate models.
Climate Dyn.
,
17
,
905
922
.
Wei
,
H.
,
P.
Yang
,
J.
Li
,
B. A.
Baum
,
H-L.
Huang
,
S.
Platnick
,
Y-X.
Hu
, and
L.
Strow
,
2004
:
Retrieval of semitransparent ice cloud optical thickness from Atmospheric Infrared Sounder (AIRS) measurements.
IEEE Trans. Geosci. Remote Sens.
,
42
,
2254
2267
.
Wielicki
,
B. A.
,
E. F.
Harrison
,
R. D.
Cess
,
M. D.
King
, and
D. A.
Randall
,
1995
:
Mission to planet Earth: Role of clouds and radiation in climate.
Bull. Amer. Meteor. Soc.
,
76
,
2125
2153
.
Wiscombe
,
W.
,
1977
:
The Delta–M method: Rapid yet accurate radiative flux calculations for strongly asymmetric phase functions.
J. Atmos. Sci.
,
34
,
1408
1422
.
Yang
,
P.
,
H.
Wei
,
H. L.
Huang
,
B. A.
Baum
,
Y-X.
Hu
,
M. I.
Mishchenko
, and
Q.
Fu
,
2005
:
Scattering and absorption property database of various nonspherical ice particles in the infrared and far-infrared spectral region.
Appl. Opt.
,
44
,
5512
5523
.

Footnotes

Corresponding author address: Kevin J. Garrett, Dept. of Atmospheric Sciences, 3150 TAMU, College Station, TX 77843. Email: kjgarrett@tamu.edu