An Overview of Satellite Measured Cloud Layer Structure and Cloud Optical and Microphysical Properties

Fu-Lung Chang Langley Research Center, NASA, Hampton, Virginia, USA [email protected]

Satellite observations are the only means of obtaining a continuous survey of cloud properties on global scales. Many climate and weather forecasting models have been evaluated using satellite-derived products. For clouds, the evaluation has been limited mainly to cloud amount, in part because conventional satellite remote sensing techniques cannot provide detailed information on cloud vertical structure, and in part because conventional satellite retrieval algorithms assume a single cloud layer in their retrievals. Increasing attention is being paid to the vertical structure of cloud fields. However, observations of cloud layer data on global scales are scarce and unreliable for model evaluation. The single-layer assumption does not emcompass the overlapping of cloud layers or the inhomogeneity of cloud vertical structure. In nature, overlapped upper-level cirrus and low-level stratus clouds occur frequently. As such, comparisons of cloud vertical structures derived from the satellite measurements and models remain a challenge. This article discusses some issues related to the single-layer assumption used in current satellite cloud remote sensing techniques. The assumption represents one major deficiency in satellite observations of cloud vertical structure. Some differences caused by different satellite retrieval algorithms and some improvements on enhanced satellite retrieval algorithms are discussed based on the research work by the author. Credible global cloud and radiation measurements are needed in order to evaluate and improve the performance of global climate and weather forecast models. The recent launch of the CloudSat radar and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) mission in a formation flight of the "A-Train" satellite constellation is expected to provide large-scale spatial and temporal observations on the vertical profile of cloud properties.

1. Introduction

Clouds play a dominant role in the Earth's climate and its changes. They strongly affect the balances of radiant energy and the water cycle. Accurate representation of clouds in climate models is one of the central issues in improving global climate modeling. Curently, satellite remote sensing is the only means of observing cloud and other climate variables on a global scale. Satellite observations have provided a wealth of information pertaining to cloud horizontal distribution and column-integrated optical properties. However, there is a dearth of information concerning the climatology of cloud vertical structure. Neglecting cloud vertical distribution can have a large impact on both the Earth's shortwave and longwave radiative fluxes (Wielicki et al., 1995, 1996; Chou et al., 1998, 1999). General circulation models (GCMs) simulate cloud fraction and cloud water/ice content at various levels. Using different assumptions of cloud overlap schemes in GCMs can lead to considerable variations in the cloud radiative forcing and modeled climate.

Prior to the Moderate-Resolution Imaging Spec-troradiometer (MODIS) instruments on board the NASA Terra and Aqua Earth Observing System (EOS) satellites, our knowledge of clouds had been gained primarily by means of the International Satellite Cloud Climatology Project (ISCCP) data products (Rossow et al., 1991; Rossow and Schiffer, 1999) using a series of polar-orbiting and geostationary satellite measurements made by international space instruments like the NOAA Advance Very High Resolution Radiometer (AVHRR) and the Geostationary Operational Environmental Satellite (GOES), the Japanese Geostationary Meteorological Satellite (GMS), and the European Geostationary Meteorological Satellite (METEOSAT). Due to the limited spectral channels provided by these satellite instruments, the retrievals of cloud properties have assumed single-layer clouds for retrieving either their bulk (e.g. cloud-column optical depth) or cloud-top (e.g. temperature, droplet effective radius) properties. There is little information available concerning the observations of cloud vertical structure.

The dearth of information on cloud vertical structure results in part from a lack of sensitivity to vertical structure in the current passive satellite sensors, as well as the inadequacy of the inversion algorithms employed in current satellite retrieval methods (Chang and Li, 2005a). For instance, cloud-top height data in the existing satellite products give either an effective cloud emission height, as in the ISCCP cloud data products, or the height of the uppermost cloud layer, as in the MODO6 of the MODIS cloud data products (King et al., 2003; Platnick et al., 2003). These cloud-top heights cannot reveal the vertical structure of clouds because their retrieval methods assume only a single-layer cloud. As such, comparison of the cloud vertical distributions derived from satellites and climate models has faced major difficulties.

Knowledge of the vertical distribution of clouds is in demand in all aspects of model simulations. For instance, in testing cloud overlap schemes in GCMs, it is necessary to understand not only the horizontal distribution of clouds but also their vertical distribution. In testing the results of cloud-resolving models (CRMs), it is necessary to have observations of cloud altitude distribution at high spatial resolution. Also, in testing chemical transport models that include the conversions of SO2 into sulfate particles within cloud layers, accurate cloud layering information is required.

In 2006, the National Aeronautics and Space Administration (NASA) of the United States launched two active satellite sensors: the Cloud-Sat radar (Stephens et al., 2002) and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO; Winker et al., 2002). The combined CloudSat and CALIPSO mission is a satellite experiment primarily designed to provide measurements of the vertical structure of clouds from space. The launches of CloudSat and CALIPSO join the Afternoon 'A-Train' Satellite Constellation in a tight formation flight. The A-Train formation flight consists of six satellites flying in close proximity. Given that both sensors provide only a nadir-pointing view of the Earth along the satellite tracks, a period of observation is required to accumulate enough cloud samples to develop a credible climatology of cloud layering data on global scales. The important measurements from CloudSat and CALIPSO are the vertical profiles of cloud liquid and ice water contents and related cloud physical and radiative properties, which are unavailable on the global scale but needed for the quantitative evaluation of global models.

2. Cloud Problems in the Climate


Clouds are the ever-changing features in the Earth's atmosphere as seen from space. They not only act as sources and sinks of the global water cycle, but also dominate the solar energy flows into the climate system while at the same time redistributing the diabatic heating in the system (Webster and Stephens, 1984; Stephens et al., 2002). Current modeling activities in accurately capturing the change of clouds due to climate warming have been slow. Little progress has been made in studying cloud feedbacks, as discussed in a critical review by Stephens (2005). As clouds can redistribute the energy and water in the climate system, even a small disturbance within them can have a large impact on the Earth's climate.

State-of-the-art GCMs still cannot convincingly simulate cloud processes (Rossow and Schiffer, 1991; Mitchell, 1993a,b). Comparisons of GCMs from around the world (Cess et al., 1990; Arking, 1991) have documented how sensitive the GCM simulations are to the different assumptions made about clouds. The differences lead to very different top-of-atmosphere (TOA) radiative forcing and climate responses. Most GCMs show a tendency for cloud cover to decrease and cloud height to increase with warming, especially for low and middle-level clouds. Radiative-convective models built on observations of increasing cloud liquid water with temperature suggest a negative cloud optical depth feedback (Somerville and Remer, 1984; Betts and Harshvardhan, 1987). On the other hand, satellite cloud retrievals suggest a tendency for low-cloud optical depth to decrease with warming (Tselioudis et al., 1992; Chang and Coakley, 2007). This is contrary to the model tendencies. GCMs can produce either a positive or negative cloud optical feedback, depending on their cloud parametrizations. Current models cannot convincingly determine whether cloudiness will increase or decrease as the climate warms. The shortcomings are physical reliable cloud observations that adequately constrain the treatment of cloud processes in GCMs. Such observations are crucial for determining cloud forcing, cloud variability, and cloud change as the climate warms.

3. Current Assessment of the Satellite Measured Clouds

The most widely used satellite cloud products for current assessment of GCM clouds are from the ISCCP of the World Climate Research Programme. The ISCCP provides a long history (since 1983) of global observations of cloud horizontal coverage (cloud amount), cloud top height (temperature), and cloud optical depth. Rossow and Schiffer (1999) reported that the annual-mean global average cloud amount was 63% based on the C-series of the ISCCP data, rising to 68% with the new D-series of the ISCCP products. Stowe et al. (2002) and Jacobowitz et al. (2003) compiled a 10-year (1985-1994) cloud climatology from the CLAVR (Clouds from AVHRR) dataset of the NOAA AVHRR Pathfinder Atmosphere Project (PATMOS), and found a rather constant, 48%-52% monthly-mean global cloud amount. The different cloud amounts between ISCCP and CLAVR are attributed to the different cloud detection techniques used by the two algorithms. Stowe et al. (2002) suggest that CLAVR tends to be more conservative in preserving the overcast radiances, so a large portion of variable cloudy pixels were classified as mixed pixels, which were assigned a cloud cover of 50%. On the other hand, ISCCP tends to treat the mixed pixels as completely overcast with 100% cloud coverage. As the majority of the mixed pixels were not fully covered by clouds, the ISCCP-derived cloud amounts tend to be overestimated (Chang and Coakley, 1993; Coakley et al., 2005).

3.1. Cloud top height and optical depth

The ISCCP cloud top height is retrieved using the infrared (IR) window measurement at the 11 ¡m spectral channel, which is a common channel available from all weather satellite instruments, including AVHRR, GOES, VIRS, GMS, and METEOSAT. There are pros and cons concerning the use of the 11 ¡m radiance as a measure of cloud top height (Menzel et al., 1992; Jin et al., 1996; Chang and Li, 2005a). A major concern with the 11 ¡m channel is that the 11 ¡m brightness temperature is not a good measure of cloud top location for semitransparent clouds such as cirrus (Liou, 1986).

The MODIS instrument provides 36 channels of high-resolution imagery data (King et al., 2003; Platnick et al., 2003). The MODIS-measured radiances facilitate the use of the CO2-slicing method to determine cloud layer pressures (Chahine, 1974; Smith and Platt, 1978; Wylie and Menzel, 1989). The CO2-slicing method is based on multiple sounding channels in the 15 ¡m CO2-absorbing band. The method is suitable for retrieving cloud layer pressures for mid-level to upper-level clouds. One advantage of the CO2-slicing method is its ability to determine the altitudes for semitransparent cirrus clouds. The disadvantage of this method is that its application is limited to mid-level and high-level clouds, but not the low-level clouds because of their low signal-to-noise ratios (Wielicki and Coakley, 1981). Thus, the MODIS operational cloud product (MODO6) uses the CO2-slicing method to retrieve cloud top pressure above 700 hPa, but uses the 11 ¡m channel to determine the altitude for low clouds below 700 hPa (Menzel et al., 2002).

A new method recently developed by Chang and Li (2005a) combines the MODIS mul-tispectral CO2-slicing method with conventional visible and IR window methods to determine the cloud top heights for single-layer and overlapped clouds. Applying the new method to near-global (polar areas excluded) Terra/MODIS data in 2001, they reported a bimodal distribution of cloud top height, with clouds lying predominantly in a high-cloud regime (200-350 hPa) and a low-cloud regime (650-800hPa). The bimodal distribution shows a minimum in cloudiness between 500 and 600 hPa.

To illustrate the differences of the three methods, Fig. 1 shows comparisons of the frequency of cloud top pressure and cloud optical depth retrieved by the new method of Chang and Li (2005a) [Fig. 1(a)], the MODIS

Figure 1. Comparisons of cloud top pressure and cloud optical depth derived by applying three different retrieval algorithms: (a) Chang and Li (2005a), (b) MODIS (MODO6), and (c) simulation of the ISCCP algorithm to one day (6 March 2003) of the MODIS radiance data.

operational method (MODO6 Collection 004) [Fig. 1(b)], and a simulation of the ISCCP method [Fig. 1(c)]. The figure shows results obtained by applying three different methods to the same MODIS radiance data for 6 March 2003, from 60°S to 60°N. The three methods resulted in significantly different cloud layer structures. As discussed by Chang and Li (2005a), their new method deals with both single-layer and overlapped clouds whereas the MODIS operational product ignores the underlying low clouds when overlapped by upper-level cirrus. As suggested by Chang and Li (2005b), the MODIS operational product (M0D06) [Fig. 1(b)] underestimates the occurrence of low-level clouds, especially in regions with abundant cirrus clouds, like the tropics and midlatitude storm tracks. This is because when high-level clouds are retrieved by the MODIS C02-slicing method, the M0D06 product ignores any potential low clouds. On the other hand, the simulated ISCCP method based on the 11 ¡m channel places high-level cirrus clouds at lower altitudes. For cirrus overlapping low clouds, the ISCCP method retrieves a biased single-layer mid-level cloud. The biased mid location depends on the opacity of the cirrus and underlying low-level cloud.

3.2. Cirrus overlapping lower clouds

Cirrus clouds often overlap lower-level clouds. Surface observations from ships rarely report cirrus clouds existing alone (Warren et al., 1985). When cirrus clouds are present, most satellite retrieval methods ignore the lower clouds beneath the cirrus by assuming only the presence of cirrus clouds. As a result, the radiances from the cirrus-overlapped lower clouds can influence satellite-measured signals through cirrus. For instance, the underlying low-cloud optical depth can be misinterpreted as being associated with the upper cirrus.

Figure 2 shows a case study for a cloud field containing cirrus-and-stratus overlapped systems that were observed on 2 April 2001, 1715(UTC), when Terra MODIS passed

Figure 2. Comparisons of the cloud top pressures and cloud optical depths derived from (a) the method of Chang and Li (2005a), (b) MODIS operational products (MOD06), and (c) a simulation of the ISCCP algorithm for a cirrus layer over stratus clouds observed for a (500 km)2 area centered at the ARM Central Facility site in northern Okhlahoma.

Figure 2. Comparisons of the cloud top pressures and cloud optical depths derived from (a) the method of Chang and Li (2005a), (b) MODIS operational products (MOD06), and (c) a simulation of the ISCCP algorithm for a cirrus layer over stratus clouds observed for a (500 km)2 area centered at the ARM Central Facility site in northern Okhlahoma.

over the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Program's central facility ground site in northern Oklahoma (Ackerman and Stokes, 2003). The figure shows comparisons of the joint histogram of cloud top pressure and cloud optical depth retrieved using the three methods. As confirmed by the ground-based radar and lidar measurements, the method of Chang and Li (2005a) [Fig. 2(a)] retrieved an upper-level cirrus layer with small optical depths and a low-level stratus layer with larger optical depths. The MODIS product (MOD06 Collection 004) [Fig. 2(b)] also retrieves the upper-level cirrus layer, but no low-level cloud. It also misinterprets the overlapped system as an optically thick layer of upper-level clouds. The simulated ISCCP method [Fig. 2(c)] misinterprets the overlapped system as a mid-to-low-level cloud system which does not separate the upper and lower clouds. The single-layer assumption caused the disappearance of the underlying low-level clouds in the MODIS product and the mixture of the upper-level cirrus with the lower-level stratus cloud into a single midlevel cloud in the simulated ISCCP retrievals. The discrepancies among the three cloud layer pressures as shown in Fig. 2 can have a significant impact on the evaluation of the vertical distribution of clouds generated by models. The differing resultant optical depths for upper-level clouds can also have a significant impact on cloud forcing at TOA and the heating profile within the atmosphere.

3.3. The bimodal distribution of cloud top altitudes

Chang and Li (2005b) suggest a distinct bimodal distribution of cloud top altitudes when examining the Terra MODIS data (excluding height-latitude regions) for January, April, July, and October 2001. Their algorithm discriminates high clouds, defined by a cloud top pressure <500 hPb, into three categories: (1) High1 for single-layer cirrus, (2) High2 for overlapped cirrus with underlying low cloud, and (3) High3 for optically thick high clouds that cannot be determined for the overlapping situation. Table 1 gives the different categories of single-layer and overlapped high clouds, along with two categories of low clouds defined by a cloud top pressure >500 hPa, where Low1 is for the single-layer low cloud not masked by any upper cloud and Low2 is for the overlapped low clouds co-occurring with High2.

Figure 3 shows the frequency of occurrence of cloud top pressures obtained for the single-layer and overlapped categories that were obtained over ocean (upper subpanels) and over land (bottom subpanels) for each of the four months in 2001. It is noted that because of the challenges in dealing with broken clouds that do not fill a partly cloudy imager pixels, the frequency distributions of cloud top pressures as shown in Fig. 3 are obtained only for each of the 5 km regions that are overcast. The 5 km overcast regions are determined using the MODIS Collection 004 cloud mask products (MOD35; Ackerman et al., 1998). The monthly-mean overcast cloud

Table 1. Classification of single-layer, overlap, and high thick clouds and the associated cloud-top pressure (Pc) and 11 /m cloud emissivity (ec).

Cloud type Cloud properties

High1 Single-layer cirrus (Pc < 500 hPa and sc < 0.85)

High2 Overlapped cirrus (Pc < 500 hPa and £c < 0.85) with Low2

High3 High thick cloud (Pc < 500 hPa and ec > 0.85)

Low1 Single-layer low cloud (Pc > 500 hPa)

Low2 Overlapped low cloud (Pc > 500 hPa) with High2

Single-layered clouds: High1 and Low1 o High2 (overlapped cirrus) A Low2 (overlapped low)

Solar Power

Solar Power

Start Saving On Your Electricity Bills Using The Power of the Sun And Other Natural Resources!

Get My Free Ebook

Post a comment