Climatology and Trend of Tropical Cloud Cover Using Gridded Satellite (GridSat) Data

Climatology and trend of different types of clouds over the tropics are studied using 22 year long (1998-2019) Gridded Satellite (GridSat) data. Brightness temperature in window and water vapor channel is used as the proxy for the cloud top altitude. Threshold and bispectral methods are used to classify clouds depending on their cloud top altitude. Clouds are classied into Low level clouds, Mid level clouds, Deep clouds, Very Deep clouds (VDC), and Semi Transparent Cirrus (STC) clouds. Climatology of the spatial distribution of each cloud type over the tropics is examined.Tropical mean of occurrence of different cloud types show a steady declining trend with a value of -0.18% /decade, -0.06% /decade, -2.12% /decade, -2.29% /decade for mid level clouds, deep clouds, STC, total clouds respectively. Low level cloud shows a steady increasing trend of 0.08% /decade. Interestingly, VDC shows a steady declining trend up to 2011, and thereafter it shows a signicant increasing trend of 0.1% /decade. Though the spatial distribution of total cloud cover generally shows a negative trend, the Western equatorial Pacic Ocean, Indian subcontinent, Indian Ocean, and Saharan desert region show a positive trend. Though low level clouds show an increasing trend, the regions of abundant low clouds show a negative trend. VDC show a declining trend over the western Pacic region, whereas other prominent VDC regions show a positive trend.


Introduction
As an important component of the climate system, clouds signi cantly affect the hydrological cycle and energy budget of the earth-atmospheric system. They do exist on a wide range of time and space scales. Cloud cover and its temporal evolution predominantly drive the variability of atmospheric re ectivity that determines the amount of solar radiation reaching the Earth's surface (Danso et al., 2019). Through various dynamic and thermodynamic processes, clouds also have signi cant feedbacks on the atmospheric circulation and climate (Stephens, 2005;Bony et al., 2006;Huang et al, 2015). Because of the disagreement among climate models and observational datasets, perhaps, clouds are the largest uncertainty in our understanding of climate change (Dufresne and Bony, 2008). A large degree of uncertainty in global climate models (GCMs) due to the inaccuracy of cloud representation is the main reason for the uncertainty in climate sensitivity estimates and climate change predictions (Collins et al. 2013; Bony et al. 2015). To reduce this uncertainty, long-term and high-resolution cloud observations from both surface and satellite-based remote sensing are highly essential. These observations could be used to formulate a better cloud parametrization and thereby improve existing models or develop a better climate model with higher spatiotemporal resolution (Huang et al, 2015).
Long term observations and climate models revealed that the cloud properties like cloud amount, height, thickness, geographical distribution, and morphology are being changed in the warming climate. Warren et al., (2007)  Antarctica, and Europe) decreased, whereas cloud cover over ocean areas (especially the Indian and Paci c Oceans) increased. Also, Warren et al., (2007) reported a declining trend of cirrus cloud cover over all continents. Eastman and Warren (2013) reported a declining trend in total cloud cover of the order of 0.4% /decade and this is due to the declining trend of middle latitude's high and middle level clouds. Hong et al., (2008) examined the occurrence of mean tropical deep convective clouds and found a slightly decreasing trend with − 0.016% /decade in 1999-2005 while the mean convective overshooting has a distinct decreasing trend with − 0.142% /decade. Using surface observation of upper-level clouds from 1952 to 1997, Norris (2005) found an overall decline of cloud cover over the tropical ocean in general and the Indo-Paci c region in particular. However, Dai et al (2006) reported an increasing trend of total cloudiness over US by using surface and satellite observation from 1976 to 2004. Kaiser (1998;2000) studied the trends in Chinese total cloud amount for the period 1951-1994 by using a database of 6-hourly weather observations and found that a decreasing trend in both midday and midnight cloud amount over much of China with a statistically signi cant decreases of 1-3% sky cover per decade.
The availability of long-term and homogeneous global observations of clouds is one of the main di culties to examine the climatology of the global cloud cover. Even though surface observations of cloudiness may enable the study of climatology and trends, these are available only at selected meteorological stations and their time series are often inhomogeneous (Karl and Steurer, 1990). However, the present day uses a long time series of observations by different satellite sensors to examine the climatology of global cloud cover and properties. The International Satellite Cloud Climatology Project (ISCCP; Schiffer and Rossow 1983; Rossow and Schiffer 1999) leading the satellite based study of the climatology of cloud cover and cloud properties globally since 1983. The ISCCP uses the spectral channels common to the operational weather satellites-the 0.5-µm visible channel, the 6.7-µm infrared water vapor channel, and the 11-µm longwave infrared window channel. The IR channel senses Earth's surface under clear-sky conditions, cloud-top temperatures of thick clouds, and a combination of cloud and surface for optically thin clouds or broken clouds within a pixel, and the visible channel also provides information on clouds and the surface. The water vapor channel is sensitive to humidity in the upper troposphere. Cloud information is also available from reanalysis products, though it tends to The present study attempted to examine the climatology and the recent trend in the occurrence of different cloud types over the tropics by using a 22-year long brightness temperature data obtained from Gridsat for a period of 1998 to 2019. Section 2 describes the data and method of analysis which followed by results and discussions in Sect. 3. Section 4 summarizes the present study.

Data And Methodology
This study is uses 22 years of Gridded Satellite (GridSat) data (Knapp et al., 2011) from 1998 to 2019 to examine the climatology and the trend of different cloud types over the tropics. GridSat data are derived from the ISCCP B1 data (Knapp, 2008b) having similar spatial, temporal, and spectral features to the Hurricane Satellite (HURSAT) dataset (Knapp and Kossin 2007), but at a global scale. The spatial resolution of the GridSat data is at an equal area grid of 0.07° latitude (~ 8 km at the equator) and spans the globe in longitude and ranges from 70°S to 70°N in latitude. The spatial and temporal coverage of the satellites contributing to ISCCP B1 is provided in Fig. 1 of Knapp et al., (2011). The data derive from fulldisk images of these satellites whose scans are closest to the synoptic times 0000, 0300,. . ., 2100 UTC.
GridSat data provide observations of the infrared window and visible channels at 11 and 0.6 µm, respectively during the entire period of records. Whereas, the infrared water vapor channel near 6.7 µm is available since 1998, on a global basis. Since both the infrared window and water vapor channel data is available from 1998, the data period of this study is started from 1998 onwards. Data are calibrated and stored in GridSat les as brightness temperature (T b ) for longwave channels and re ectance for the visible channel. GridSat data attempted to reduce intersatellite differences by intersatellite normalization and also performs temporal normalization via calibration against High-Resolution Infrared Radiation Sounder (HIRS) during the GridSat period of record. GridSat IR calibration uncertainty is less than 0.5 K for anyone satellite with a very stable temporal uncertainty that is less than 0.1 K decade − 1 (Knapp et al., 2011).
In this study, the brightness temperature (T b ) data in the water vapor (WVIR) and thermal window (TIR) channel obtained from GridSat used as the proxy for cloud top altitude (Roca et al., 2002;Rajeev et al., 2008) to classify the different types of clouds in terms of its altitude of occurrence. The clouds are classi ed using the threshold criteria followed by Muhsin et al., (2019) and Roca et al., (2002). Table 1 summarizes the thresholds used in this study. The cloud top altitudes identi ed using these T b thresholds are validated by Meenu et al. (2010) with those obtained from CALIPSO observations and showed a fairly good agreement between the two. The clouds are classi ed into Low level, Middle level, Deep, Very deep, and Semi Transparent Cirrus (STC) clouds. STCs and low level clouds are obtained by using bispectral methods (Roca et al., 2002) and other clouds are classi ed using threshold methods (Muhsin et al., 2018). All the cloud types except STC are optically thick and T b from the window channel is su cient to determine the cloud top altitude. Whereas, in the case of STC, the radiance observed in the window channel does not correspond only to the cloud top, but weighted by the radiation emitted from the altitudes below. This causes the T b observed in the window channel through a thin cloud to be larger than the actual T b of the cloud top, leading to erroneous identi cation of such cirrus as warm clouds. T b from the water vapor channel is highly sensitive to the clouds and humidity content in the upper troposphere (Soden and Bretherton, 1993), and hence a combination of T b observed in the window and water vapor channel can be used to detect STC (Desbois et al., 1982). The occurrence of each cloud type is calculated for each le (every three hours) from 1998 to 2019 over the tropics. These cloud occurrences are later used to obtain the climatology as well as the trend of each cloud type over the entire tropics.    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 Climatological mean and standard deviation of brightness temperature in the window (TIR) and Water vapor (WVIR) channel obtained from GridSat data during 1998-2019. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 2
Climatology of the spatial distribution of different types of tropical cloud cover derived from GridSat data from 1998 to 2019. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 3
Variation of monthly mean cloud cover over the entire tropics (30˚ S -30˚ N) from 1998 to 2019 obtained from GridSat data. Red and blue lines are the best t lines obtained from the linear regression model and these are above 95% signi cant level.

Figure 4
Spatial distribution of trend of different clouds over the tropics obtained from the linear regression model. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.