3.1.1 Meteorological conditions

To better understand the linkages between cloud properties and large-scale dynamics, meteorological fields are presented in Fig. 2. It shows temperature, specific humidity, wind field, and static stability at the lower (∼850 hPa) and upper (∼180 hPa) troposphere over SEA in four seasons: boreal spring (March, April, and May – MAM), summer (June, July, and August – JJA), autumn (September, October, and November – SON), and winter (December, January, and February – DJF). In the lower troposphere, relatively high and homogeneous temperatures are observed all year round (∼290 K), corresponding to the Indo-Pacific Warm Pool. However, temperatures drop in boreal winter (<285 K; Fig. 2a4) and raise in summer (>290 K; Fig. 2a2) over southeastern China and the South China Sea. In the upper troposphere, temperatures are relatively low at latitudes between 10^∘ S–10^∘ N (<213 K), while temperatures over South Asia in summer are 1–3 K higher than in other seasons. Similarly, high humidity (>10 g kg⁻¹ at lower troposphere and >0.03 g kg⁻¹ at upper troposphere) is located south of 10^∘ N latitude during spring and winter (Fig. 2b1, b4), while in autumn (Fig. 2b3), the humidity pattern is quite symmetric about the Equator. Also, air is especially moist over South Asia during summer than in other seasons at both the lower and upper troposphere. The summer high temperature and humidity over South Asia are related to the heating and convection over the Tibetan Plateau, which maintains a hot and humid upper troposphere, and the South Asian anticyclone (Yeh, 1982). The summer monsoon also helps transfer a large amount of moisture from the Indian Ocean to Asia (Fig. 2c2).

Figure 2(a1–a4) Temperature, (b1–b4) specific humidity, (c1–c4) wind field, and (d1–d4) static stability derived from the ECMWF-AUX data; shade for 850 hPa, contour for 180 hPa, grey vectors for wind field at 850 hPa, and red vectors for 180 hPa.

The seasonality of the wind field is evident (Fig. 2c1–c4). In the lower troposphere, the southwesterly wind flow brings warm and humid air to South Asia in summer when the ITCZ is located north of Equator, providing favorable conditions to form clouds and precipitation (summer monsoon). The wind direction in the upper troposphere is northeasterly, which is nearly opposite to that at the lower troposphere. In DJF, the ITCZ shifts to the Southern Hemisphere, and the wind flow reverses. The prevailing northeasterly flow near the lower troposphere (winter monsoon) is also opposite to the wind direction at the upper troposphere (southwesterly). However, the upper troposphere wind is much weaker in winter than in summer because the summer South Asian anticyclone above the Tibetan Plateau enhances the upper troposphere wind flow (Yeh, 1982). The spring and autumn are two transition seasons of summer and winter monsoonal flows.

The lower-troposphere static stability, LTSS = $({\mathit{\theta }}_{z=\mathrm{3}\mathrm{km}}-{\mathit{\theta }}_{z=\mathrm{0}})/\mathrm{3}\mathrm{km}$, and the upper-troposphere static stability, UTSS = $({\mathit{\theta }}_{z=\mathrm{tropopause}}-{\mathit{\theta }}_{z=\mathrm{tropopause}-\mathrm{3}\mathrm{km}})/\mathrm{3}\mathrm{km}$, are shown in Fig. 2d1–d4, where the θ is potential temperature in units of kelvin. The tropopause height is defined following the World Meteorological Organization; i.e., the lowest level where lapse rate is 2 ^∘C km⁻¹ or less, and the average lapse rate between this level and all higher levels within 2 km is smaller than 2 ^∘C km⁻¹ (Grise et al., 2010). Figure 2d1–d4 reveal that LTSS is usually smaller over land than over ocean. Small LTSS values (<4 K km⁻¹, yellow-green color) over ocean correspond to a wetter atmosphere (Fig. 2b). Relatively larger LTSS (>4 K km⁻¹) occurs in winter and spring such as over the East and South China seas. The LTSS has been proven to be an important parameter indicating low-level cloud formations. For instance, Klein and Hartmann (1993) showed that 1 ^∘C increase in stability is associated with a 6 % increase in stratus cloud area coverage. The spatial pattern of UTSS is similar to that of LTSS.

3.1.2 Occurrence of all clouds

This sections focuses on cloud spatial distributions over SEA. The horizontal occurrence frequency, defined as the ratio of total cloudy number to the total observation sample in each 5^∘ long × 2^∘ lat grid derived from the 2B-CLDCLASS-LIDAR data, is shown in the upper panels of Fig. 3. The zonal latitude–altitude cross sections, obtained by cloudy number in each 2^∘ lat × 250 m height cell divided by the observation sample in that cell, are displayed in the lower panels of Fig. 3.

Over SEA, the annual mean cloud frequency is about 81.4 %, being smaller in winter and larger in summer (Table 3). As expected, seasonal variations in cloud occurrence are generally associated with the movement of large humidity, warm temperature, and low stability in the lower and upper troposphere (Figs. 2, 3). Clouds frequently occur in Southeast Asia during the summer monsoon season, while their frequency shifts to the Malaysia and Indonesia regions as the summer monsoon retreats and the ITCZ shifts southward. The cross sections (Fig. 3b1–b4) display prevailing high-level clouds located at around 10–15 km, matching to the ubiquitous nature of cirrus over the warm pool regions (Sassen et al., 2008). These ice clouds have their largest frequencies centered ∼12^∘ N in summer but shift to south of the Equator in winter, i.e., moving with the ITCZ and the monsoon.

While it is clear that clouds favor warm, humid, and unstable conditions, large cloud occurrences are found in other conditions in the region. Frequent cloud occurrence (∼70 %) is also observed over southeastern China and the East China Sea during winter when the atmosphere is cold, dry, and stable (Figs. 2, 3a4). These winter clouds north of 10^∘ N usually have low cloud heights (<5 km) with little cirrus above as seen from the cross section (Fig. 3b4).

Figure 3Cloud occurrence frequency derived from 2B-CLDCLASS-LIDAR, for (a1–a4) horizonal distribution and for (b1–b4) zonal latitude–altitude cross section.

While Fig. 3b displays a vertical cross section of cloud occurrence frequencies, it says little about cloud overlap frequencies. As shown in Yuan and Oreopoulos (2013), low-level clouds have a high chance of being overlapped by upper clouds in the warm pool region. In the next section, we will examine cloud overlap with a focus on cloud phase.

3.1.3 Occurrence of cloud phase

As stated in Sect. 2.2, we classify clouds according to cloud phase and cloud layer in five main groups: ice-only, liquid-only, mixed-only, ice-above-liquid, and ice-above-mixed clouds. Each group contains both single and multiple layers of the same phase. Our analysis (Table 3) shows that one-layer–one-phase clouds have much larger frequency than multilayer same-phase clouds. For example, multilayer ice-only clouds (∼8.6 %) occur less frequently than one-layer–ice-only clouds (20 %). Liquid-only clouds mostly form in a single layer (14 %), and the frequency of multilayer liquid-only clouds is only 2 %. A careful comparison between single and multiple layers of the same-phase clouds shows no significant difference in the properties that we are interpreting, which justifies our simpler classification.

Figure 4 shows horizontal and vertical distributions of the five cloud groups as defined in Fig. 1. The mean occurrence frequency of each cloud class in four seasons over ocean and land is summarized in Table 3. The five cloud groups display visible differences in both of their mean frequencies and spatial distributions. Ice-only clouds (Fig. 4a) occur most frequently (∼28.6 %) among all cloud classes. These clouds widely spread in the tropical belt and prefer the locations north of the Equator in summer but move to the south in winter. Ice-only clouds mainly are located at high altitudes between 10 and 15 km (Fig. 4b), corresponding to the prevalent tropical cirrus discussed in many other studies (Hong and Liu, 2015; Reid et al., 2013; Sassen et al., 2008).

Figure 4Occurrence frequency of the five cloud groups derived from 2B-CLDCLASS-LIDAR; (a) for horizonal distribution and (b) for zonal latitude–altitude cross section.

Table 3Cloud occurrence frequency (%) derived from 2B-CLDCLASS-LIDAR. In the section ocean vs. land, the number outside parentheses represents both land and ocean. The first number in parentheses is for ocean, and the second one is for land. In the section single vs. multiple cloud layers, the first number in parentheses represents single layer, and the second number is for multiple layer.

^a The single and multiple layers represent the overlying ice clouds only. ^b All single-layer clouds are the summation of single-layer–ice-only (20.0 %), liquid-only (14.0 %), and mixed-only (6.6 %) clouds, while all multilayer clouds include multilayer ice-only (8.6 %), liquid-only (2.0 %), mixed-only (0.1 %), ice-above-liquid (20.1 %), ice-above-mixed (9.3 %), and other (0.74 %) clouds.

Download Print Version | Download XLSX

In contrast, liquid-only clouds (Fig. 4a) are widely distributed over southeastern China and the East China Sea and show large seasonality. These liquid-only clouds mostly have their cloud tops lower than 3 km (Fig. 4b). They are the so called “Chinese stratus” by Klein and Hartmann (1993) associated with lower-troposphere cold and dry air and large LTSS (Fig. 2). Elsewhere, liquid-only clouds have very small frequencies (<10 %). The annual mean frequency of liquid-only clouds is ∼16.0 %.

Ice-above-liquid clouds have an annual mean occurrence frequency of ∼20.1 % (Fig. 4a). Ice layers located at 10–15 km cover the underlying liquid clouds mostly with heights below 3 km (Fig. 4b). These clouds occur frequently over southern China and Southeast Asia during the summer monsoon season and move to western Pacific Ocean and Malaysia–Indonesia regions in winter.

The areas where widely distributed ice-only and ice-above-liquid clouds are also associated with relatively frequent mixed-only (annual frequency ∼6.7 %) and ice-above-mixed clouds (annual frequency ∼9.3 %) (Table 3) such as in Southeast Asia in summer and the Malaysia–Indonesia region in winter. Mixed-only clouds are mature convective clouds as seen from their cross sections (Fig. 4b), which extend from near surface up to above 15 km. Nevertheless, near 30^∘ N from fall through spring where liquid-only clouds dominate, there exist relatively frequent mixed-only clouds with their top below 10 km. The ice-above-mixed clouds, frequently occurring at 6 km, are more likely under development with mixed-layer tops reaching to around 10 km. If the mixed layers develop higher, they would merge with the overlying ice clouds and are classified into the mixed-only cloud class. Overall, liquid-only clouds are associated with high LTSS and lower temperature in the lower troposphere, agreeing with the relationship of low-level clouds with stability (Klein and Hartmann, 1993; Li et al., 2014). In contrast, ice-only, ice-above-liquid, ice-above-mixed, and mixed-only clouds, collectively named as “ice-contained clouds”, favor a humid, warm, and unstable environment.

Figure 5 further summarizes the mean and standard deviation of the meteorological variables discussed in Fig. 2 from the ECMWF-AUX product for the five cloud groups in summer and winter seasons over ocean (to avoid the low static stability over land). In the lower troposphere, all cloud groups in summer tend to have smaller LTSS and higher temperature and humidity than in winter (Fig. 5a, b). The standard deviations in summer are also smaller, being consistent with a more homogeneous spatial pattern of the meteorological fields (Fig. 2). In winter the liquid-only clouds tend to have a much smaller humidity and colder temperature corresponding to the occurrence of the Chinese stratus (Klein and Hartmann, 1993). For those ice-contained clouds, they are still located in a relatively warm, moist, and unstable atmosphere, but their standard deviations are much larger than that in summer, agreeing with the less homogeneous spatial pattern of meteorology in winter (Fig. 2).

Figure 5Mean and standard deviation of meteorological variables over ocean for the five cloud groups.

Download

In the upper troposphere (Fig. 5c, d), the relationship between the five cloud types and meteorology is similar in both summer and winter, with liquid-only clouds deviating from ice-contained clouds. The ice-contained clouds relate to smaller UTSS as reported in Li et al. (2014). Also, ice-only and ice-above-liquid clouds share very similar upper tropospheric meteorology as their mean and standard deviations are nearly the same, which is not surprising because the low stability and high moisture are essential to maintain cirrus (Christensen et al., 2013; Li et al., 2014). However, the specific humidity in the upper troposphere of ice-above-mixed clouds is larger than both ice-only and ice-above-liquid clouds (Fig. 5c, d), which may reveal that convection below brings moisture to the upper troposphere.

3.1.4 Distributions of cloud phase properties

Figure 6 presents the probability distribution function (PDF) of cloud properties including cloud top for all cloud phases and base, geometric thickness, τ, and R_e for ice layers from the 2C-ICE product. The averages of the cloud properties in the four seasons are summarized in Table 4.

Figure 6Annual PDFs of (a) cloud top, (b) base, (c) geometric thickness, (d) optical depth, and (e) effective radius for ice clouds including both ocean and land. The vertical dashed lines in (a)–(e) indicate the median values of the PDFs; (f) cloud top for liquid-only cloud and the liquid below ice and (g) cloud top for mixed-only and the mixed cloud below ice. In (f) and (g), solid and dashed curves are for ocean and land, respectively. Cloud top and base bins adopt an interval of 0.5 km. Ice τ and R_e bins use an interval of 0.1 on a log scale and 1 µm, respectively.

Download

Table 4Average of cloud properties for the five cloud groups over ocean and land.

^a T for cloud top. ^b B for cloud base. ^c G for geometric thickness. Subscripts: i for ice, w for water, and m for mixed.

Download Print Version | Download XLSX

The properties of ice layers in the three categories – ice-only, ice-above-liquid, and ice-above-mixed clouds – are displayed in Fig. 6a–e. We combine both land and ocean data to investigate the distributions of ice layer properties, because their PDFs display similar shapes between that over land and ocean (figure not shown); however, their averages are separately summarized in Table 4. The three categories of ice layers share many similarities in their PDFs: the modes of ice top PDFs (∼16 km) and base (∼12.5 km) slightly greater than their means and medians (Table 4 and vertical dashed lines in Fig. 6). The modes of geometrical thickness PDFs are around 1 km, which are smaller than their mean and median values of 2–4 km (Fig. 6c and Table 4). These statistics demonstrate that the distributions of ice clouds skew to higher locations and thinner thickness, corresponding well to the properties of cirrus near the tropopause (Haladay and Stephens, 2009; McFarquhar et al., 2000). In addition, the τ PDFs show two modes for the three types of ice layers and the R_e PDFs with their modes slightly greater than 15 µm.

There also exist differences between the three groups of ice layers. For example, ice-only clouds tend to locate 0.4–1.4 km lower over land than ocean, and the lower location may allow more moisture to feed into ice clouds, which may explain why ice clouds over land have mean τ values of 0.6–1.6 and R_e of 3–5 µm larger than those over ocean (Table 4). Compared to ice-only clouds, the ice layers above liquid or mixed clouds show much less land–ocean contrast in these properties (Table 4). Also, the ice layers above liquid clouds contain about 73.8 % of samples with geometrical thickness <3.0 km and about 91.5 % of samples with ice τ<3.0 – the threshold often used to define cirrus (Sassen et al., 2008) (Fig. 6c, d). In contrast, ice-only clouds are thicker with larger means, medians, and R_e (Table 4), which are contributed by more ice-only cloud samples with geometrical thickness >3.5 km, τ>1.6, and R_e>60 µm (Fig. 6). Due to the fact that lidar signals are attenuated when cloud τ>3.0 and the radar fails in detecting shallow clouds, some ice-above-liquid clouds (e.g., thick ice over shallow liquid) could be classified into the ice-only group, leading to some sampling biases to the mean ice τ and R_e. However, in the ice τ range of 1.6–3.0 where CALIPSO can penetrate the cloud, ice-only clouds have a total frequency of ∼0.11 in this τ range, being higher than the frequency (∼0.07) of the other two groups and demonstrating a higher probability of ice-only clouds being thicker.

For liquid or mixed layers, we only focus on cloud tops, because the determination of cloud base suffers from larger uncertainties than cloud top due to the limitation of instruments; i.e., CloudSat radar has difficulty in distinguishing the cloud base near Earth's surface, and the CALIPSO lidar signal is attenuated by clouds with optical thickness greater than 3 (Hu et al., 2009; Stephens et al., 2008). For the liquid top PDFs (Fig. 6f), there are two modes for the liquid below ice clouds (green). One mode is located at 1 km for ocean (2–3 km for land), and the other is located at ∼6 km for both ocean and land. We further obtain the spatial distributions of the liquid below ice clouds with a liquid top greater than and lower than 5 km (figure not shown). It is shown that liquid clouds with a top <5 km are widely distributed over SEA, while those with a liquid top >5 km are more concentrated in locations with a latitude <10^∘, corresponding to a more unstable environment as shown in Fig. 2. In another words, greater moisture and small LTSS allow liquid clouds to develop deeper. For liquid-only clouds, they have a much higher frequency at the PDF mode of 1 km, and the second mode is not evident. The averaged top value of the liquid below ice clouds is higher than liquid-only clouds (Table 4).

For the mixed layers below ice clouds (Fig. 6g), the mode of cloud top PDF (cyan) is at around 6 km, and the mean is about 8.5 km, with the values over ocean about several hundred meters higher than that over land (Table 4). Note that as the mixed layers develop deeper, e.g., top >10 km, and merge with the upper ice layers, these clouds would be grouped as mixed-only clouds. For mixed-only clouds, the primary mode of cloud top PDF is at around 16 km, and the secondary mode is at around 6 km. The primary mode is much higher over ocean than over land, while the secondary mode is higher over land, indicating that more mixed clouds over land are under development around 13:30 local time. This agrees with the results in Nesbitt and Zipser (2003) that convective clouds keep developing in the early afternoon and reaching toward an intensity maximum in the late afternoon over land, while diurnal variation in convection intensity is insignificant over ocean.

3.1.5 Spatial heterogeneity of cloud phase

Section 3.1.1–3.1.4 discussed the macrophysical properties including spatial distributions, cloud thickness, top and base heights for five cloud groups, and their relationship with meteorology based only on CC data. In this section, by combining the MODIS and 2B-CLDCLASS-LIDAR data, we investigate how the spatial heterogeneity index (H_σ) relates to cloud phase. The H_σ not only is closely related to cloud micro- and macrophysical properties but also affects the accuracy of cloud retrievals from passive sensors and radiative transfer modeling (Ham et al., 2015; Zhang and Platnick, 2011). Only ocean data are considered to avoid complications with the effects of land surface heterogeneity on interpreting results.

The H_σ PDFs for the CC clear sky and the five cloud groups are shown in Fig. 7a. The PDF of CC clear sky has a sharp peak at H_σ∼0.01, indicating that clear sky is usually spatially homogeneous. For cloudy sky, liquid-only clouds are the most heterogeneous among all cloud groups, with H_σ ranging from 0.01 to 1, with a peak of ∼0.5. The ice-only clouds in contrast are homogeneous, as the PDF has a peak close to that of clear sky. This suggests that the biases in retrieved optical and microphysical properties of clouds from passive sensors caused by the plane-parallel assumption will be larger for water clouds compared to ice clouds in the SEA region. Indeed, MODIS liquid R_e differences retrieved from three wavelengths (1.6, 2.1, and 3.7 µm) are especially large over this region (up to 10 µm) (Zhang and Platnick, 2011; Fu et al., 2019). Care is therefore recommended when using the MODIS cloud microphysical products in this region, such as in interpreting the cloud–aerosol relationship over SEA (e.g., Ross et al., 2018).

Figure 7(a) Annual PDFs of spatial heterogeneity for CC clear sky and the five cloud groups; panel (b) is the same as (a) but for CDF; (c) liquid cloud fraction in the 5 km × 5 km surrounding of the collocated CC–MODIS pixel derived from the MYD06 IR cloud phase retrievals. H_σ bin interval is 0.01 on a log scale.

Download

For ice-above-liquid clouds, the PDF curve moves slightly to a smaller H_σ region compared to liquid-only clouds, as the overlying ice clouds have a spatial smoothing effect on the radiation emerging from the liquid clouds below. However, due to the small optical thickness of the overlying ice clouds, radiation from underlying liquid clouds dominate (Sect. 3.1.4). The H_σ PDF of ice-above-mixed clouds is similar to that of ice-only clouds but with some samples having H_σ smaller than for clear sky (e.g., H_σ<0.01). This feature is more obvious for the mixed-only clouds, which have about 50 % of their samples with H_σ<0.01 (Fig. 7b), indicating that these clouds are extremely homogeneous. These cases of very smooth ice-above-mixed and mixed-only clouds correspond to high mixed-layer tops and large reflectance at 0.645 µm (discussed in next section), which are associated with deep convection. These clouds are locally homogeneous and hence favor the plane-parallel assumption in radiation computation (Ham et al., 2015).

Note that CC clear sky and ice-only, ice-above-mixed, and mixed-only clouds are usually homogeneous, but there exist some heterogeneous cases. The PDF of clear sky has a long tail of H_σ values extending up to 1 (Fig. 7a) and has about 20 % of samples with H_σ values greater than 0.1 (Fig. 7b). Mismatch of pixels in collocation or difference in spatial resolutions of CC (1.8 km × 1.4 km) and MODIS (1 km) can contribute uncertainties to the H_σ of CC clear sky. Yet, with a focus on the clear-sky pixels on which both CC and MODIS agree, they behave nearly the same as those of CC clear sky, as shown in Fig. 7a, indicating that the long tail is due to the significant amount of misdetection of small subpixel clouds by both MODIS and CC. Indeed, many small liquid clouds with sizes ranging from a few tens to hundreds of meters can go undetected by MODIS (Zhao and Di Girolamo, 2006). We revisit the MODIS and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data (15 m resolution) used in Zhao and Di Girolamo (2006) over the tropical western Atlantic. The long tail of the H_σ PDF is significantly reduced, i.e., frequency change from 0.3 % to 0.1 % at H_σ∼0.1, when the ASTER data are applied to exclude the MODIS clear-sky pixels that contain ASTER-reported clouds. This further affirms that the undetected clouds in MODIS and CC clear-sky pixels contribute to large H_σ values, which may at least impact 20 % of clear-sky samples whose H_σ>0.1 (Fig. 7a, b).

To further investigate this last point, we calculate the MODIS liquid cloud fraction, defined as the ratio of liquid cloud samples based on the MYD06 product to the total 25 pixels in a 5 km by 5 km surrounding of the collocated CC–MODIS pixel. As shown in Fig. 7c, as H_σ increases, a larger fraction of MODIS liquid clouds is observed around the CC clear sky and ice-only, mixed-only, and ice-above-mixed clouds. This indicates that the heterogeneous pixels could also be due to undetected liquid clouds in the subpixels.

Figure 8 shows the mean H_σ for CC clear sky, the five cloud groups, and all sky that includes both clear and cloudy sky. As expected, ice-only, ice-above-mixed, and mixed-only clouds are homogeneous everywhere ($H_{\mathit{\sigma }}<\sim \mathrm{0.05}$), with some relatively large values (H_σ∼0.1) in the liquid-only-cloud prevailing regions such as the East China Sea in winter (Figs. 4, 8b4, d4, f4). As liquid-only clouds are the most heterogenous, they show the largest spatial H_σ values over SEA (Fig. 8c1–c4). Also, the H_σ values over the East China Sea in spring, fall, and winter are relatively smaller than the H_σ in other regions, implying that the Chinese stratus that favor dry and stable meteorological conditions are less heterogenous than other liquid-only clouds (Fig. 8c1–c4). Ice-above-liquid clouds (Fig. 8e1–e4) are smoother than liquid-only clouds, and their relatively small values tend to coincide with frequent ice-above-liquid cloud occurrence that are associated with the monsoon and ITCZ, such as in the Malaysia–Indonesia region in winter or the North Indian Ocean in summer (Fig. 4). The H_σ pattern of CC clear sky (Fig. 8a1–a4) shows smaller values than ice-above-liquid clouds but larger values than ice-only, ice-above-mixed, and mixed-only clouds. Also, the places with large H_σ values of CC clear sky are consistent with those with frequent occurrence of ice-above-liquid or liquid-only clouds, which in turn indicates the high chance of undetected liquid clouds increasing the subpixel variability.

Figure 8Spatial distributions of H_σ for clear sky, the five cloud groups, and all sky from panels (1) to (4) for MAM, JJA, SON, and DJF, respectively.

For all sky (Fig. 8g1–g4), the small H_σ values occur north of the Equator in summer, including the Indian Ocean and South China Sea. The pattern is quite symmetric about the Equator in fall and moves south of the Equator in winter – consistent with the shift of the monsoon and ITCZ. This is because the small H_σ values are primarily contributed by ice-only clouds due to their large occurrence frequency (Fig. 4) and spatial homogeneous features. In contrast, the pattern of large H_σ agrees more with that of liquid-only cloud occurrence (Fig. 4).

Overall, liquid clouds are spatially heterogeneous over SEA, whereas ice-only and mixed clouds are usually homogeneous. Due to the smoothness of the overlying ice clouds, ice-above-liquid clouds are less heterogenous than liquid-only clouds, but their H_σ values are still large because overlying ice clouds are optically thin and the emerging radiance from underlying liquid clouds dominates. Clear sky is smooth with H_σ∼0.01, but undetected liquid clouds increase its subpixel variability. The seasonal variations in all-sky H_σ spatial patterns are in accordance with cloud movements associated with the monsoon and ITCZ.

3.1.6 Spectral radiative feature

This section examines the spectral radiance at the TOA observed by MODIS for the CC clear sky and the five cloud groups defined in Fig. 1 to investigate how radiative features observed at the TOA relate to cloud phase. Similar to Sect. 3.1.5, only ocean data are adopted. The averages of the reflectance (R) and the brightness temperature (BT) (not weighted by cloud occurrence frequency) for each cloud group are shown over SEA in Fig. 9 and summarized in Table 5, while the PDFs and the cumulative distribution functions (CDFs) are displayed in Fig. 10.

Figure 9Average reflectance and brightness temperatures over ocean for clear sky and the five cloud groups, including all solar zenith angles. The sold lines represent annual mean, and symbols denote seasonal averages.

Download

Table 5Average reflectance at 0.645, 1.375, 1.64 and 2.13 µm (unit: %) and brightness temperature at 8.55 and 11.03 µm (unit: K).

Download Print Version | Download XLSX

The PDF of R_0.645 for clear sky shows a narrow peak at R_0.645∼0.02 (Fig. 10a1). Ice-only cloud shows its PDF peaks near the clear-sky peak, and its CDF shows that about 80 % of its samples have R_0.645<0.2, proving the thin features of ice-only clouds over SEA and agreeing with their small optical depths in Fig. 6d. Similarly, for liquid-only cloud, the PDF also shows its mode near that of clear sky. A large fraction of liquid-only clouds are optically thin clouds (e.g., more than 75 % of liquid-only clouds with R_0.645<0.2), being consistent with the findings of Leahy et al. (2012) that the thin marine low-cloud fraction is greater than 80 % for the SEA region. Ice-above-liquid clouds have a larger mean R_0.645 than either ice-only or liquid-only clouds in the column (Table 5 and Fig. 9), but ice-above-liquid clouds still contain more than 60 % of samples with R_0.645<0.2, further demonstrating the ubiquity of thin clouds over SEA. Many of these thin clouds often go undetected by MODIS as shown in Table 2, where around 11 % of CC ice-above-liquid cloud and 30 % of CC liquid-only and ice-only cloudy samples are reported to be clear sky by MODIS. The largest average R_0.645 is seen in mixed-only clouds (∼0.59), followed by the ice-above-mixed clouds (∼0.45). Their PDFs are broad and flat, but the frequencies at R_0.645>0.4 are evident as displayed in Fig. 10a1. Their CDFs reveal that 60 % of ice-above-mixed and 80 % of mixed-only clouds have R_0.645>0.4, indicating that these clouds are geometrically deep (consistent with Fig. 4) and optically thick.

Figure 10The annual PDFs and CDFs of reflectance at 0.645 µm, reflectance ratio, BT at 11 µm, and BTD between 8.5 and 11 µm. The intervals for reflectance, reflectance ratio, BT, and BTD are 0.01, 0.01, 1, and 0.1 K, respectively.

Download

At 1.64 µm, the average R_1.64 of ice-above-liquid clouds is nearly same as liquid-only clouds (∼0.13) (Table 5 and Fig. 9). Both these cloud groups are more reflective than ice-only clouds (average R_1.64∼0.06). The average R_1.64 of ice-above-mixed and mixed-only clouds is about 0.20, implying that ice-above-mixed clouds are optically thick enough to reach the asymptotic reflectance, and mixed-only clouds do not increase R_1.64. Similarly, the average R_2.13 of both mixed-only and ice-above-mixed clouds is also nearly equal (0.11), which in turn demonstrates their large thickness. The curve of R_2.13 across different cloud groups (Fig. 9) shows a similar shape with that of R_1.64 but with a smaller magnitude, owing to the larger imaginary part of refractive index of water and ice at 2.13 µm than at 1.64 µm. Particularly, the imaginary part of ice refractive index is larger than water at both wavelengths. Hence, when examining the reflectance ratio of SWIR (1.64 or 2.13 µm) to the visible (0.645 µm), we would expect that liquid-only clouds show a larger ratio than any other cloud groups. Because the Aqua MODIS 1.6 µm band has many dead detectors (King et al., 2013), we display the PDF and CDF for reflectance ratios from 2.13 to 0.645 µm ($\frac{R_{\mathrm{2.13}}}{R_{\mathrm{0.645}}}$) to further emphasize the spectral features of different cloud phases (Fig. 10b1 and b2). As expected, the PDF of liquid-only cloud extends to large reflectance ratio regions, and the same is true for ice-above-liquid clouds (Fig. 10b1). Moreover, there are about 60 % of liquid-only clouds and 50 % of ice-above-liquid clouds with a reflectance ratio greater than 0.4 (Fig. 10b2), indicating that the reflectance ratio of ice-above-liquid clouds is in accord with low-level liquid clouds because of the thin features of the overlying ice clouds as discussed in Sect. 3.1.4 and 3.1.5. In contrast, ice-only clouds have very small frequencies when the reflectance ratio is greater than 0.4 (Fig. 10b1), and the same is true for mixed-only and ice-above-mixed clouds. For CC clear sky, the PDF shows its mode at the ratio of ∼0.1 with its width ranging from 0.0 to 0.4. We also note that as the reflectance ratio of CC clear sky becomes larger, the corresponding H_σ increases as well, indicating that the undetected liquid clouds in clear-sky pixels (Sect. 3.1.5) also enlarge the SWIR to the visible reflectance ratio.

At 1.375 µm, liquid-only cloud and clear sky show near-zero average reflectance as the photons are nearly all absorbed by water vapor (Table 5). The mixed-only clouds have the largest average reflectance of 0.225 compared to other cloud groups, because these clouds are much higher and thicker as shown in Fig. 4 and Table 4. Also, the average R_1.375 of ice-only cloud (∼0.06) is greater than that of ice-above-liquid cloud (∼0.04). Because the specific humidity at the upper troposphere is nearly the same for both cloud groups (Fig. 5), the larger R_1.375 of ice-only clouds demonstrates an average larger thickness than for the ice layers above liquid clouds. This agrees with the results from the CC observations as discussed in Sect. 3.1.4 (Table 4).

At the IR, the average BT_8.5 and BT₁₁ have similar magnitudes (Fig. 9 and Table 5). Only the PDF and CDF of BT₁₁ are shown in Fig. 10c1 and c2, respectively. As expected, the average BT₁₁ of mixed-only clouds is lowest (∼234.3 K), followed by ice-above-mixed clouds (average BT₁₁ of ∼ 247.7 K). The widths of BT₁₁ PDFs of mixed-only and ice-above-mixed clouds are broad, but the frequencies are low when BT₁₁>260 K, with about 20 % of samples at that BT₁₁ region (Fig. 10c2). The average BT₁₁ (∼ 289.8 K) of liquid-only cloud is only slightly smaller than that of clear sky (∼295.2 K) due to the low liquid cloud top (mode ∼ 1 km, Fig. 6f) and thin features. For ice-above-liquid cloud, the average BT₁₁ (∼276.8 K) is slightly larger than ice-only cloud (∼272.4 K). This may due to the fact that the ice layers above liquid clouds do not absorb as strongly as ice-only clouds because the former is averagely thinner than the latter over ocean (Table 4). Also, the peaks of BT₁₁ PDFs of ice-only and ice-above-liquid clouds are close to that of clear sky, and 50 % of samples of these two cloud groups are with BT₁₁>280 K (Fig. 10c2), demonstrating the thin features of these clouds, which agree with the conclusions from the CC data (Fig. 6) and R_0.645 analysis.

Another notable feature of the BT shown in Fig. 9 is that the BT_8.5 is averagely larger than the average BT₁₁ for ice-only, ice-above-liquid, ice-above-mixed, and mixed-only clouds, i.e., ice-contained clouds, and vice versa for clear sky and liquid-only clouds. This is because in clear sky, absorption by the atmosphere at 8.5 µm is slightly greater than at 11 µm, and hence, negative BTD between 8.5 and 11 µm (BT_8.5−BT₁₁) is observed (Fig. 10d1). In cloudy sky, cloud absorption at 11 µm is larger than at 8.5 µm (Wolters et al., 2008), which decreases the BT₁₁. The absorption difference between 11 and 8.5 µm is small for water, which explains why the PDF of liquid-only clouds is located close to that of clear sky but at larger BTD regions (Fig. 10d1). Ice clouds have a larger absorption difference between 11 and 8.5 µm than liquid clouds (Wolters et al., 2008). This explains positive BTD values for the ice-contained clouds. However, ice-only and ice-above-liquid clouds have their BTD PDFs peak at around −1.4 K. The negative BTDs of these clouds also indicate their thin features over the SEA regions so that the absorption by ice clouds is insufficiently significant to produce positive BTD.

For seasonal variations (Fig. 9 and Table 5), the change in reflectance and brightness temperature of each cloud group is associated with cloud occurrence, thickness and cloud top. For example, ice-only clouds have a larger average R_0.645 and smaller BT in summer due to the fact that these clouds occur more frequently and are thicker in this season. In contrast, liquid-only clouds over ocean have similar cloud tops but occur more frequently in winter, and they show greater average R_0.645 in winter than in summer.

In summary, mixed-only and ice-above-mixed clouds are bright in the visible (i.e., large R_0.645) and cold in the infrared (i.e., low BT). These clouds also have a small reflectance ratio and positive BTD between 8.5 and 11 µm. Although liquid-only and ice-only clouds have similar R_0.645, liquid-only clouds tend to have relatively larger BT₁₁ and a larger reflectance ratio than ice-only clouds. Ice-above-liquid clouds show slightly larger R_0.645 than either liquid-only or ice-only clouds, with a reflectance ratio similar to liquid-only clouds but BT₁₁ and BTD closer to ice-only clouds. The spectral radiative features of ice-only, liquid-only, and ice-above-liquid clouds also demonstrate widespread thin clouds over SEA.