Figure 1 shows the mean and standard deviation of 22 yearlong brightness temperature (Tb) in the window (~ 10.5 µm) and the water vapor (~ 6.7 µm) channel over the tropics. It should be noted that there are discontinuities that can be seen in the data at the borders between areas covered by different satellites and this is more clear in the water vapor channel. Since Tb is the proxy for the clouds, this figure depicts the climatology of clouds over this region. The larger value of Tb in the window channel corresponds to clear sky or low level clouds and the smaller values represent the presence of clouds with higher cloud top altitude. Mean Tb in the window channel varies from 255 K to 295 K and the spatial pattern shows the typical cloud climatology over the tropics with smaller values of Tb at ascending limbs of walker circulation and in the regions of Inter Tropical Convergent Zone (ITCZ) indicating the presence of deeper clouds. The higher values of Tb are shown at the subtropical high pressure zones over the oceans and the subtropical deserts over land. This indicates that these regions are either cloud free region or there exist low level clouds. The outlines of the Andes Mountains, Himalayas, and Tibetan Plateau are also seen in the mean window channel Tb plot and these are likely the surface emissions caused by high-altitude mountains. Mean Tb from water vapor channel is also shows a similar spatial pattern as that of Tb in window channel with values varies from 230 K to 255 K. Generally, Tb in the Water vapor channel is taken as a proxy of concentration of water vapor in the upper troposphere. A low value of the Tb implies a higher concentration of water vapor in the upper troposphere. The standard deviation of both Tb shows that the regions of larger deviation of Tb observed are the regions having smaller mean Tb values and the regions of smaller deviation of Tb are the regions of larger values of mean Tb exist. This indicates that the regions where large values of mean Tb are observed always exist either cloud free conditions or only low level clouds. Similarly, the regions with smaller mean Tb values (large deviation in Tb) indicate that the cloud occurrence varies significantly. This may be due to seasonal or inter annual or decadal variations of cloud occurrence.
Figure 2 shows the climatology of the spatial distribution of occurrence of different types of clouds derived from Tb by employing the threshold conditions and bispectral method, detailed in the previous section. The occurrence of a particular cloud type means the relative percentage of occurrence of that cloud type with respect to total observations including clear sky condition. The spatial distribution of the total cloud is obtained by summing up all clouds derived from the Tb. Clouds are present nearly all the time over the ascending limb of the walker circulation, namely the western Pacific and Indonesian region, Central Africa, and Northern South America. Nearly 80% of the time clouds are present along the ITCZ regions and the southwestern Pacific Ocean. The cloud occurrence at the northern hemispheric subtropics is comparatively less (50–60 %) and the regions of southern hemispheric eastern Pacific and the Atlantic Ocean exhibit very less cloud occurrence (~ 20%). Comparatively large occurrence (~ 20%) of low level clouds are observed near the coast of Chile and moderate occurrence (~ 10%) of low level clouds are observed near the Atlantic coast of southern Africa, Southern Indian Ocean, Australia, and eastern China. Northern hemispheric desert region and Pacific coast of Mexica are showing a lesser (~ 5%) occurrence of low level clouds. The occurrence of Mid level clouds, Deep clouds, and very deep clouds show a similar spatial pattern with different occurrence values with large values for mid level clouds and least values for very deep clouds. The spatial occurrence of STC also shows an almost similar pattern as that of mid level clouds with a wider area of occurrence.
Figure 3 shows the 22-year variation of the monthly mean occurrence of different clouds over the entire tropics (30˚ S – 30˚ N). The red and blue lines are the best fit lines determined using the linear regression model and these are above 95% significant level. The occurrence value of each cloud type is with respect to the total observations including clear sky conditions. The occurrence of tropical low level clouds is 1–4% with maximum and minimum occurrence during June-July and February-March months respectively. It shows a positive trend of occurrence with a value of 0.08% /decade. Whereas, Mid level clouds show a declining trend of occurrence with a value of -0.18% /decade. It shows a clear seasonal pattern with maximum occurrence up to 12% during the winter season. The deep cloud occurrence is 5.5–7 % with a maximum occurrence during the winter season. This cloud type also shows a decreasing trend of -0.06% /decade. Very deep clouds (cloud top altitude above 14 km, including overshooting clouds) occur 0.6 to 0.8 % of the time and do not show a clear seasonal pattern. Unlike other cloud types, VDC shows a peculiar trend of occurrence. It shows a decreasing trend up to 2011 and suddenly the trend becomes positive with a value of 0.1% /decade. This is a significant trend value when compared to its total occurrence and it means that VDC occurrence is increasing 1.3% every year. STC took a major contribution towards the total cloudiness of the tropics with an occurrence frequency of 36–48 %. This cloud also has a negative trend line with a value of -2.12% /decade. Over the tropics, the occurrence of the total cloud is 54–66% and it has a steady decreasing trend with a value of -2.29% /decade. This large value of the decreasing trend is majorly contributed by the STC trend. The decreasing trend of total cloud cover and STC is already reported by Warren et al. (2007) and Dim et al. (2011).
In order to examine how these mean tropical trends of different clouds are spatially distributed, the trend at each pixel of satellite observation is calculated and presented in Fig. 4. As mentioned in Fig. 1, the discontinuities at the borders between areas covered by different satellites are clearer, especially in the subplots of total clouds and STC. Though total cloud cover generally shows a negative trend, the western equatorial Pacific Ocean, Indian subcontinent, Indian Ocean, and Saharan desert region show a positive trend. Southern hemispheric Pacific Ocean loses its cloud cover at the rate of ~ 0.1% / decade. Low level clouds generally show an increasing trend. Whereas, subtropics of the Indian Ocean region and the Pacific coast of the Mexica shows a negative trend of low cloud cover. Interestingly, these are the pockets of abundant low level clouds. Mid level and deep clouds show an almost similar pattern of cloud cover trend with the positive trend at western and northeastern Pacific Ocean, North Africa, and the Indian region. STC shows a similar spatial pattern as that of the total cloud. Though STC shows a declining trend in general, it shows an increasing trend across the longitude region of the Indian subcontinent. Pandit et al. (2015) also reported an increasing trend of the sub-visible cirrus clouds over the Indian region by using 16 years of lidar observations. VDC shows a decreasing trend over the western Pacific region, whereas the other prominent VDC regions show a positive trend. Interestingly, the western equatorial Indian Ocean shows a positive trend though it is not a prominent VDC occurring region.
Eastman and Warren (2013) reported a declining trend of global total cloud cover of the order of 0.4% /decade and this is due to the declining trend of high and middle level clouds in the midlatitude. Also, they reported that the declining trend of zonal cloud cover indicates the poleward shifts of the jet streams in both hemispheres. The present study also shows a declining trend of total cloud cover over the tropics but for a larger magnitude. Hong et al. (2008) studied the trend of tropical deep convective and overshooting clouds during 1999–2005 and found that both are in the decreasing trend. Though the present analysis also in consistent with this result during the period, it shows an increasing trend from 2011. Wylie et al. (2005) reported a small but statistically significant increase of high clouds (cloud top altitude < 440 mb) over the tropics and the northern hemisphere by using NOAA High Resolution Infrared Radiometer Sounder (HIRS) polar-orbiting satellite data from 1979 to 2001. Norris (2005) studied the spatial pattern of linear trends in upper level cloud cover using surface observations over the tropical Indo-Pacific Ocean during 1952–1997. They reported that upper-level cloud cover increased by about 4% over the central equatorial South Pacific and decreased by about 4–6% over the adjacent subtropics, the western Pacific, and the equatorial Indian Ocean. The present study also observed a similar trend over the western Pacific and the equatorial Indian Ocean. However, the other two regions show an exactly opposite trend in high cloud cover. Norris (2005) also mentioned that the calculated trend is a linear trend and the trend direction may change in the future. Norris et al. (2016) reported that cloud amount is increased over the northwest and southwest tropical Pacific Ocean, the northwest Indian Ocean, and north of the Equator in the Pacific and Atlantic oceans. The present study also showing a very similar cloud cover trend at all the above mentioned regions except for the southwest tropical Pacific Ocean which shows a decreasing trend in this study.