2. IMPACTS OF OVERLAPPING CLOUDS ON SATELLITE-BASED GLOBAL
2.1 Background
Satellite remote sensing is the only means to provide cloud climatology on a global scale (Rossow and Schiffer 1999). However, overlapping clouds have posed a major challenge in interpreting satellite data properly. For instance, traditional satellite imagers and sounders only see the uppermost cloud top in each vertical column. Passive satellite remote sensors provide information on cloud top height but often have some systematic errors, especially for optically thin cirrus clouds (Minnis et al. 1993; Rossow and Schiffer 1999; Naud et al. 2007). Additionally, some studies using passive remote sensing techniques assume all clouds are homogeneous and single-layered. One major problem with this assumption is that multi-layer clouds have been frequently observed by surface observers, radiosonde, and in-situ aircraft measurements, as well as satellite observations (Hahn and Warren 1999; Poore et al. 1995; Tian and Curry 1989; Verlined et al. 2004; Heidinger and Pavolonis 2005; Wang and Dessler 2006; Joiner et al. 2010). To improve our understanding of the multi-layer clouds, some approaches have been proposed to detect multi-layer clouds (Nasiri and Baum 2004; Chang and Li 2005).
Cloud overlap can cause large biases in the satellite retrievals of many cloud properties including cloud top height, thermodynamic phase, and radiative properties (Minnis et al. 1993; Baum and Wielicki 1994; Cho et al. 2009; Huang et al. 2005). Multi-layer clouds sometimes cause errors in cloud height retrievals that depend on specific algorithm and cloud properties (Naud et al. 2007). A thin cirrus cloud over a lower level water cloud is one of the most problematic overlapping cloud scenarios
for global cloud property retrievals, particularly for the retrieval of effective radius and optical depth (Nasiri 2004; Chang and Li 2005; Huang et al. 2005). In the case of high ice clouds over low water clouds, one of the greatest impediments to accurately determine cloud ice mass for a given atmospheric profile is the influence of the underlying liquid water clouds on the radiances observed in the visible and near- infrared wave-lengths at TOA (Minnis et al. 1993; Chang and Li 2005). Additionally, retrievals of cloud water path tend to be biased when an ice cloud overlaps a liquid water cloud (Minnis et al. 2007).
Clouds at different levels can have different radiative impacts. For instance, thick low clouds can reflect a significant amount of incoming SW solar radiation back to space, and thin high clouds can reduce outgoing LW radiation. Multi-layer clouds therefore behave radiatively differently compared to single-layer clouds. Cloud radia- tive effects (CRE) of high-level clouds depend on a number of factors, including cloud fraction, cloud top temperature, cloud optical properties, and cloud particle habit (Rossow and lacis 1990; Garrett et al. 2009; McFarquhar et al. 2002). Generally, the net CRE of high-level thin cirrus clouds is positive at TOA (McFarquhar et al. 2000), while the net CRE of thick anvils with visible optical depth larger than about 10 could be negative (Jensen et al. 1994). Low-level clouds typically have a cooling effect, because they can reflect more SW radiation to space and have a relatively small impact on the LW radiation (Chen et al. 2000; Dong et al. 2003). The impact of mid-level clouds depends on the strengths of these two radiative effects of high- and low-level clouds (Hartmann et al. 1992; Zhang et al. 2005; Zelinka et al. 2012). Cloud vertical morphology contributes a major uncertainty in the analysis of satellite data used for climate studies (Heidinger and Pavolonis 2005; Wang and Dessler 2006; Li et al. 2011; Subrahmanyam and Kumar 2011). The ISCCP (Rossow
frared (IR, ∼11 µm) radiances from the imaging radiometers on the international constellation of weather satellites since 1983. The ISCCP classification of clouds at high-level, mid-level, and low-level using cloud top pressure is widely used in the atmospheric community, even though there exist some uncertainties (Jin et al. 1996; Wylie et al. 2005; Rossow and Zhang 2010).
ISCCP builds a cloud climatology by relating the observed radiances to cloud radiative properties. Cloud top temperature is first retrieved by assuming that all clouds are opaque. Cloud top pressure is then determined from cloud top temperature using a profile of atmospheric temperature with pressure. If the cloud is opaque, it radiates likes a blackbody hence the emission temperature is equivalent to the cloud top temperature. However, if the cloud is optically thin, then the emission will appear to be larger than that for the cloud top temperature due to contamination by the radiation emitted by the warmer atmosphere and surface below. For clouds with visible optical depth between 2 and 6, the cloud top temperature is corrected by decreasing cloud top temperature (or increasing cloud top height) as a function of optical depth. This correction produces overestimations of cloud top temperature and pressure. During nighttime, however, semitransparent cirrus clouds may be falsely identified as mid-level clouds by not being able to use visible optical depth to correct cloud height. This leads to the general underestimation and overestimation of high- and mid-level clouds amounts in the ISCCP dataset, respectively (Jin et al. 1996; Rossow and Zhang 2010).
In contrast, the TOVS-B (Stubenrauch et al. 2006) dataset uses the IR ra- diances together with microwave observations and is more sensitive to high clouds. Similarly, the HIRS, which uses the CO2 slicing analysis (Wylie and Menzel 1989;
Wylie et al. 1994), also has better sensitivity to cirrus clouds. While the TOVS-B and HIRS are capable of detecting cirrus clouds, they do not detect low clouds ob-
scured by the high clouds. For instance, in the case of thin cirrus overlying mid- or low-level clouds, TOVS-B or HIRS provides the properties of the cirrus, whereas the use of visible channel by ISCCP leads to inference of a mid-level cloud. Discrepancies in the global coverage between ISCCP and TOVS-B (Stubenrauch et al. 2006) and between ISCCP and HIRS (Jin et al. 1996; Wylie et al. 2005) have been explained by differences in temperature profiles, horizontal heterogeneities (partial cloud cover) and vertical heterogeneities (multi-layer clouds).
In this chapter, we investigate the impacts of overlapping clouds in global statistics at different levels using the combined lidar and radar observations aboard the “A-Train” satellites. Comparisons of cloud amounts at different levels with the ISCCP dataset are also made for the same time period.