Spatio-Temporal Distribution of Aerosol and Cloud Properties over Sindh Using MODIS Satellite Data and a HYSPLIT Model

In this study, aerosols spatial, seasonal and temporal variations over Sindh, Pakistan were analyzed which can lead to variations in the microphysics of clouds as well. All cloud optical properties were analyzed using Moderate Resolution Imaging Spectroradiometer (MODIS) data for 12 years from 2001 to 2013. We also monitored origin and movements of air masses that bring aerosol particles and may be considered as the natural source of aerosol particles in the region. For this purpose, the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model was used to make trajectories of these air masses from their sources. Aerosol optical depth (AOD) high values were observed in summer during the monsoon period (June–August). The highest AOD values in July were recorded ranges from 0.41 to 1.46. In addition, low AOD values were found in winter season (December–February) particularly in December, ranges from 0.16 to 0.69. We then analyzed the relationship between AOD and Ångström exponent that is a good indicator of the size of an aerosol particle. We further described the relationships of AOD and four cloud parameters, namely water vapor (WV), cloud fraction (CF), cloud top temperature (CTT) and cloud top pressure (CTP) by producing regional correlation maps of their data values. The analyses showed negative correlation between AOD and Ångström exponent especially in central and western Sindh. The correlation between AOD and WV was throughout positive with high correlation values > 0.74 in whole Sindh except eastern most arid strip of the Thar Desert in the region. The correlation between AOD and CF was positive in southern Sindh and goes to negative in northern Sindh. AOD showed a positive correlation with CTP and CTT in northern Sindh and a negative correlation in southern Sindh. All these correlations were observed to be dependent on the meteorological conditions for all of the ten sites investigated.


INTRODUCTION
Aerosols are minute solid and liquid particles suspended in the atmosphere.These may occur in the form of natural dust particles, volcanic dust, mist, clouds, fog, sea salts, anthropogenic aerosols, oceanic sulphates, industrial sulphates, soot (Black carbon), organic particles and some toxic pollutants after various physical and chemical processes that have several effects on our atmosphere ( Gupta et al., 2013).
Interactions between aerosols and clouds have become a key topic of scientific research due to the importance of clouds in controlling climate (Mahowald and Kiehl, 2003;Alam et al., 2010Alam et al., , 2014)).Aerosols and clouds play a very significant role in the climatic system (Balakrishnaiah et al., 2012;Alam et al., 2014).Aerosols demonstrate both spatial and temporal variations, which may lead to variations in the cloud optical properties (Alam et al., 2010).Their ability to absorb or scatter solar radiations have direct impact to increase or decrease the temperature of the atmosphere (Ichuku et al., 2004;IPCC, 2013).In addition, their indirect affect modifies the cloud lifetime, albedo, and the precipitation by changing the size and density of cloud droplets in the atmosphere (Lohmann., 2002;Alam et al., 2010;Balakrishnaiah et al., 2012).In addition, as smaller cloud droplets are less efficient to produce precipitation, an increased aerosol population will cause to a longer cloud life.Cloud formation requires super-saturation with aerosol that also explains the importance of aerosols in our atmosphere (Alam et al., 2010).These particles also emitted and reflected radiation from the earth in different directions depending on their physical and chemical properties (Ichuku et al., 2004).Therefore, the description of atmospheric aerosols is important to understand global solar heat budgets, water cycle and various dynamics of climate change (Ichuku et al., 2004).Increasing anthropogenic aerosol concentrations lead to form cloud condensation nuclei (CCN) that ranges from smaller effective radii to longer-lived clouds, affecting the hydrological cycle (Ramanathan et al., 2001;Balakrishnaiah et al., 2012).A wide range of studies has shown aerosol and cloud interactions with large uncertainties in the magnitude of radiative forcing (e.g., Twomey et al., 1984;Coakley et al., 1987;Kaufman and Nakajima, 1993;Kaufman and Fraser, 1997a, b;Ackerman et al., 2000;Rosenfeld, 2000;Ramanathan et al., 2001;Rosenfeld et al., 2002;Schwartz et al., 2002;Kim et al., 2003;Andreae et al., 2004;Koren et al., 2004;Penner et al., 2004;Kaufman et al., 2005;Koren et al., 2005;Alam et al., 2010;Balakrishnaiah, 2012;Alam et al., 2014).Alam et al. (2010) and Balakrishnaiah et al. (2012) have examined the spatial and temporal variations in aerosol particles and the impact of these variations over different locations in Pakistan and India.They found highest AOD values during the monsoon period (June-August).Their analysis also showed a strong positive correlation for AOD and cloud parameters in some regions and a negative correlation in other regions.Ramachandran and Jayaraman (2003) investigated AOD values for February-March 2001 over the Bay of Bengal.They found higher AODs on 21 February influenced by the air masses originating either from India or Bangladesh at different heights, indicating the anthropogenic sources.They concluded the difference in AOD values over Chennai that is a coastal urban station of India with that of clean continental land.Reasons may be due to the difference of anthropogenic activities and meteorological conditions, wind patterns and subsequently loading of aerosol particles.Kaufman et al. (2002) observed strong link of anthropogenic aerosols with the climate system and the hydrological cycle.A region with a clean atmosphere (with no pollutant) has relatively large cloud droplets to form clouds for precipitation, although natural aerosols are required to condense upon; a cloud does not form otherwise.In contrast, urban smoke and pollutants affect cloud radiative properties and reduce precipitation at the same time.Thus, anthropogenic aerosols impact the hydrological cycle with a change in cloud cover, cloud property and rainfall patterns in a region.Gupta et al. (2013) examined the quantity and quality of MODIS data over urban areas of Pakistan.They found a lack of availability of MODIS data over Karachi due to the presence of bright surfaces that increases surface reflection.They further added in their findings that high AOD values are shown near the city center and decrease as we move out of that city.The population is more concentrated inside the city with more anthropogenic activities impacts atmospheric aerosols.Biomass burning, industrial activities and transport might also have significant impact on this analysis.Ashraf et al. (2014a) found high AOD values in two seasons of Pakistan i.e., monsoon and pre-monsoon.They associated these higher values of coarse particles with a high frequency of a dust storm in the region as well as fine particles from burning activities in the agricultural fields of Pakistan and India (Indian Gigantic Plain).They found a highly polluted year i.e., 2008 with highest AOD concentrations in the atmosphere.Alam et al. (2014) examined AOD and cloud parameters such as water vapor, cloud top temperature, cloud top pressure and cloud fraction over five cities in Pakistan during the period (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011).The AOD and cloud optical properties correlations changes with changing seasons in Pakistan e.g., AOD showed a positive correlation with CTP and CTT for the spring season and a negative correlation was observed in summer in all investigated regions.
This study focused on three objectives.The first objective was to understand the spatio-temporal distribution of AOD over Sindh, Pakistan.The second objective was to investigate the relationship between AOD and cloud parameters (i.e., Cloud fraction, cloud-top pressure, cloud-top temperature and water vapor).The third objective was to analyze the influence of air mass trajectory on AOD concentrations.Aerosols spatio-temporal distributions, seasonal variations and physical properties may be determined within a short span of time from land, aircraft, or satellite remote sensing (Ichoku et al., 2004).The analysis of air mass trajectories helps to determine natural source of aerosols in southern Pakistan.The months of January (winter), July (summer), April (spring) and October (autumn) were chosen as these are considered as representatives of four seasons in the study region.In particular, AOD-cloud parameter correlation study acquired for thirteen years (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013) suggests new findings in a region.

Site Description
Sindh (23°42′-28º29′N; 67°24′-71°05′E) is the third largest province of Pakistan with a population of approximately 42 million.It is bounded by Balochistan to the west, Punjab to the north, Indian states (Gujarat and Rajasthan) to the east and the Arabian Sea to the south of its location (Fig. 1(b)).The landscape of Sindh is diversified, having Kirthar Mountains to the west, Thar Desert to the east, and Piedmont Plains to the north of Sindh.The climate of Sindh is hot and semi-arid.It has a variation in temperature as inland extreme temperatures to maritime moderate temperatures.For an inland Sindh, the temperature rises above 46°C in May and falls up to 2°C in January.In addition, the coastal region has temperature not more than 40°C and less than 10°C for summer and winter months, respectively.Mainly rainfalls during monsoon season (July-September) that is about 7 inches in a year.Dust Storms hit frequently, mainly during April-May and October-November to the region.
In Sindh, there are ten industrial estates, commonly named as SITE (Sindh Industrial and Trading Estates).Karachi, Hyderabad, Kotri, Sukkur, Tando Adam, Nooriabad, Larkana, Nawabshah and Shikarpur are common stations of Sindh industries.The main industries include cement industries, oil refineries, chemical industries, heavy machinery and pharmaceutical industries, automobiles, shipbuilding, fish factories, sugar and flour mills, fertilizer industries and a steel mill.Their industrial emissions contribute to increase anthropogenic aerosols in the region.Similarly, vehicular emissions from the roads of Sindh, like Super Highway and National Highway and sea salt emissions from Arabian Sea are the major local sources of aerosol in the study area.All the ten sites selected for the study are covering almost whole Sindh with some geographical characteristics.For example, most of the stations chosen for study are cities and industrial centers like Karachi, Hyderabad, Rohri, Larkana, Nawabshah, Jacobabad and Sukkur.Other sites are small cities with some topographical characteristics either mountain or desert like Dadu, Ghotki, Umerkot and Tharparkar.

The Moderate Resolution Imaging Spectroradiometer (MODIS) Instrument
MODIS (Moderate Resolution Imaging Spectroradiometer) acquires data at 36 spectral bands to give atmospheric, oceanic and terrestrial information (King et al., 2003;Ichoku et al., 2004).These bands range from visible to the thermal infrared wavelengths, which are very useful to collect and summarize large statistical and visual impacts of aerosols on climatic conditions (Ichoku et al., 2004).MODIS provides observations at moderate spatial resolution of 250 m (Band 1 & Band 2), 500 m (Bands 3 to Band 7) and 1 km (Band 8 to Band 36) or temporal resolution (1-2 days) over different portions of the electromagnetic spectrum (King et al., 2003;Ichoku et al., 2004;Alam et al., 2010).MODIS data include daily, 8-day or 16-day composite images and products with different time scales.Several climatic parameters also retrieved at a spatial resolution of 5 km and 10 km from MODIS (Alam et al., 2010;Balakrishnaiah et al., 2010;Alam et al., 2011c;Alam et al., 2014).MODIS provides global coverage every 1 to 2 days.This source is also beneficial to represent detailed local, regional and global distributions of several climatic parameters over land as well as over the ocean.These MODIS instruments are equally beneficial to look at atmospheric, terrestrial and ocean phenomenology to monitor various changes and collect statistics.These statistics are especially helpful to analyze aerosol concentrations and the impacts on cloud formation (Alam et al., 2011c(Alam et al., , 2014)).
MODIS aerosol retrievals are expected to be more accurate over the ocean than land (Kaufman et al., 1997;Tanre et al., 1997).Therefore, data accuracy is found to be more over the oceans than it is over land (Remer et al., 2005).All bright surfaces such as deserts and snow covered areas cause problems for the MODIS instrument, as do complex terrains, leading to a large bias in models and needs to have ground-based observations (De Meij et al., 2006;Alam et al., 2010;).There is a difference in the usage of the visible and the thermal bands of MODIS datasets, as the visible bands are used during the day and the thermal bands are used during the day and night both to measure bright surfaces, in particular.MODIS determine the physical properties of cloud as the cloud top pressure and cloud top temperature using infrared bands; optical and microphysical cloud properties using visible and near-infrared bands (Jin and Shepherd, 2008;Alam et al., 2010).
MODIS calculates AOD with an estimated error of ± 0.03 ± 0.05AOD over the ocean and ± 0.05 ± 0.02AOD over the land (it has been learned that 0.03AOD is an offset due to uncertainty in the ocean state and 0.05AOD is another due to uncertainty in aerosol characteristics).Over the land, MODIS measures AOD from the reflection of residual at the top of the atmosphere and uses the 2.1-μm channel for it.Over the ocean, MODIS distinguishes fine particles of pollutants to coarse particles of dust and sea-salt with aerosol spectral signatures of 0.47-2.1 μm range (Kaufman et al., 2002).The MODIS aerosol retrieval algorithm has recently been improved in order to correct systematic biases in the MODIS algorithm used previously (Remer et al. 2005;Levy et al. 2007).
MODIS Level 3 collection 5.1 deep blue (DB) data are available from 2002 to present from Aqua, and 2000 to 2007 from Terra due to the known calibration issues.Therefore, in the present study MODIS DB Terra and Aqua data were used from 2001-2002 and 2003-2013, respectively.This DB algorithm of the MODIS aerosol product provides more reliable and accurate data over bright surfaces such as deserts and snow or ice covered regions.While, the standard MODIS AOD relies on finding dark targets.Deep blue AODs data are more accurate over Sindh due to the advantage of taking dark-surface properties at blue channels (0.412, 0.47 μm) and less absorption of dust at the red channel (0.65 μm).
The monthly MODIS Level 3 data with a 1° × 1° degree spatial resolution was obtained from the NASA DAAC website http://disc.sci.gsfc.nasa.gov/giovanni.AOD was used as a surrogate for aerosol concentration and calculated the correlation coefficient for each set of parameters on a monthly basis, averaged annually, for the period from 2001 to 2013.All data sets were interpolated using software ArcGIS (version 10.1).Spline was found as best technique used to do interpolation of MODIS data.MODIS Level 3 atmosphere products contain no runtime QA flags.Therefore, Level 2 runtime QA flags are used in Level 3 to calculate (aggregate and QA weight) statistics in Level 3. To identify cloud freepixel, the cloud mask algorithm is used.This algorithm classifies each pixel as either confident clear, probably clear, probably cloudy, or cloudy (Hsu et al., 2004).In order to minimize the cloud contamination problem, DB uses AOD spatial variance method to the radiance at 412 nm at every 3 × 3 pixel spatial domain (Hsu et al., 2004).The uncertainty of DB AOD retrievals is expected as ± 0.05 ± 20% × AOD AERONET (Hsu et al., 2006, Huang et al., 2011, Shi et al., 2013).MODIS detailed information on algorithms for the retrieval of aerosol data is available at http://modisatmos.gsfc.nasa.gov.

The Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) Model
In this study, we used the HYSPLIT model to archive data of an air parcel at several localities of Sindh named as 'backward trajectories'.This model computes air mass trajectories with several meteorological parameters like rainfall, relative humidity, ambient temperature and solar radiation flux (Alam et al., 2011c).This meteorological input for the trajectory model was processed as the FNL dataset and reprocessed from NOAA (National Oceanic and Atmospheric Administration) by ARL (Air Resource Laboratory) to understand air masses (Alam et al., 2010).Furthermore, several different pre-processor programs convert NOAA, ECMWF (European Centre for Mediumrange Weather Forecasts), MM5 (The fifth generation NCAR/Penn State Mesoscale Model), NCAR (National Centre for Atmospheric Research) re-analysis, and WRF (weather research and forecasting) model output fields in a format compatible for direct input into the model (Alam et al., 2010).
HYSPLIT has the ability to determine almost 40 forward or backward trajectories at different altitudes of 500, 1000, 1500, 2000, and 2500 (Alam et al., 2010(Alam et al., , 2011c)).All backward trajectories in this study were calculated at 500, 1500 and 2500 m heights above ground level by the model archived from the website (http://ready.arl.noaa.gov/HYSPLIT.php).We analyze the origin of air masses with aerosol such as dust, sea salt, pollutants, etc. at various sites of Sindh.With some limitations of web version all registered PC has no computational restrictions with the user to obtain the necessary meteorological data files in advance.In addition, all unregistered version is similar to the registered version except not work with forecasted meteorological data files (Draxler and Rolph, 2003;Alam et al., 2010).

RESULTS AND DISCUSSIONS
The spatial distribution of annual mean AOD at 550 nm wavelength has been plotted for the period from 2001 to 2013.We also used mean spatial seasonal distribution of AOD for the whole period during 2001-2013.The correlation maps of aerosol and cloud parameters named water vapor (WV), cloud fraction (CF), cloud top temperature (CTT), and cloud top pressure (CTP) have been plotted for the period from 2001 to 2013.The air mass HYSPLIT trajectories demonstrate a source of dust aerosols in the mid-months of all four seasons in Sindh.

Spatial Distribution of Annual Mean Aerosol Optical Depth
The spatial distribution of annual mean AOD at a wavelength of 550 nm has been plotted for the period 2001-2013 (see Fig. 1(a)).The results from Fig. 1 demonstrates that aerosols had a marked impact on ten stations in Sindh, namely Karachi, Hyderabad, Nawabshah, Larkana, Sukkur, Tharparkar, Umerkot, Dadu, Jacobabad and Ghotki.This study includes most of the cities due to their geo-strategic locations, heavy industrial emissions, vehicular emissions and other anthropogenic activities.The spatial gradient of AOD for thirteen years shows high AOD values in central and northern Sindh.This region has Indus River with dense urban populations, residences, commercial and industrial estates that are well connected by the highways of Sindh i.e., Indus Highway and National Highway.High AOD values (> 0.74) are observed over the areas with intense anthropogenic activities.This includes an industrial region of Jacobabad, Rohri (Sukkur), Larkana, Nawabshah, Hyderabad and Karachi.Some other areas with spatial geomorphologic locations such as Dadu and Ghotki has high AOD which may consider a result of wind masses and weather conditions that affects the aerosol loads in each locality.This study also examined a very different case of Karachi.Karachi shows AOD values ranges 0.30 to 0.65 instead to show highest AOD values due to continuous sea breezes from Arabian Sea that keeps AOD in motion.Balakrishnaiah et al. (2012).determined aerosols marked impact on six southern India sites, namely TVM, BLR, ATP, HYD, PUNE and VSK  They observed increased aerosol particles from southern India to northern India up to the Himalayas.Prasad et al. (2004) also examined the spatial gradient of AOD with high AOD values over areas with intense anthropogenic activities.Alam et al. (2010) analyzed AOD distribution for eight cities of Pakistan named Peshawar, Rawalpindi, Zhob, Lahore, Multan, D.G. Khan, Rohri and Karachi.It was found that during summers AOD values were very high in all selected cities, particularly in southern Pakistan, these values were recorded more than 0.9.They also observed an increase in AOD from north to south of Pakistan.Similarly, high AOD in the summer months was reported by Munir and Zareen (2006)

Seasonal Variations in Aerosol Optical Depth (AOD)
Seasonal variations in AOD were demonstrated and calculated for the period 2001-2013 (Fig. 2).Very high AOD values were found in almost whole Sindh during the summer season, particularly in the northern and central parts of Sindh where AODs greater than 1.27 were calculated.In the central parts of Sindh AOD ranges from 1.17 to 1.27.Likewise, in the southern Sindh AODs values greater than 0.98 were reported.Similar increases in AOD during the summer season have previously been reported in other cities of Pakistan (Alam et al., 2010(Alam et al., , 2011a(Alam et al., , b, c, 2012(Alam et al., , 2014(Alam et al., , 2015)).In contrast, very low AOD values were observed in almost whole region during the winter season.During the winter season low AODs values were observed.During the winters, very low AOD values (0.29-0.47) were found in the southeastern and the southwestern parts of Sindh.The reasons for these differences are the arrival of air masses with dust in the region from the southern coast of the Middle East and the monsoon winds during the summer season.The monsoon winds blow from the Indian Ocean after crossing the Indian peninsula, these winds with a very large amount of moisture enters in the northeastern Pakistan.The tail end of these winds enters in the regions of southeastern Pakistan that is specifically Sindh Province.All the northern and the central areas of Sindh are considered as the hotspots of dust aerosols in the summer months.Karachi, which is the coastal city, has a relatively high AOD value (1.17   AOD values have significant seasonal variation, specifically small and high in summer.Tiwari and Singh (2013) studies shown that coarse-mode aerosol particles are dominant in the pre-monsoon, while fine-mode aerosol particles are dominant in the winter season.Other studies have been conducted by Kuniyal et al. (2009) and Gupta et al. (2013), concluded AOD variations were highly dependent on both the travel pathways and the sources (e.g., sea salt, desert dust, smoke from burning biomass, and industrial pollution) of the air masses.

Relationship between AOD and Ångström Exponent
Ångström exponent (Alpha) is the exponent that describes the spectral dependency of aerosol optical depth on wavelength.This spectral dependence of the measured optical depth is determined as <Alpha> that is considered as a good indicator of the size of aerosol particles.High Alpha values indicate a dominance of very fine particles, whereas, low values indicate a dominance of coarse particles.Therefore, <Alpha> is a useful exponent to estimate the size of the aerosol particle.Alpha is determined from the spectral dependence of the measured optical depth as suggested by Angstrom (1963).The spatial correlation plot of AOD at 550 nm versus <Alpha> shows that the <Alpha> increases mostly with AOD over Sindh shown in Fig. 3.Many factors are responsible for high AOD in the region.Sometimes AOD concentrations present near clouds influence weather of the surface and increases size of humidified aerosols in the atmosphere (Balakrishnaiah et al., 2012).
This study analyzed an inverse relationship between AOD and Alpha.Smaller the exponent defines larger the particle size over ten sites of Sindh.Correlations for Alpha vs. AOD ranges from -1.00 to -0.32 over Sindh.This decrease in correlation may be associated with dust loading for most of the Sindh sites except the coastal area, where sea salt aerosols are dominant.In order to investigate the origins of the air masses arriving at the studied sites we performed back trajectory analyses based on the NOAA HYSPLIT (National Oceanic and Atmosphere Administration Hybrid Single Particle Particle Langrangian Integrated Trajectory) model shown in Fig. 4. Most of the air masses reached Sindh from the Dasht (Iran), Hamoun Wetlands (Afghanistan), Cholistan (Pakistan), Thar (India) deserts, and also from the Arabian Sea.Alam et al. (2012) found high <Alpha> values in January and low values in June that indicates the dominance of fine particles in January and coarse particles in June over urban areas of Pakistan.June is one of the months of dust storms in the region.They also found higher AOD values over Lahore than Karachi due to the fact that Karachi is a coastal city of Pakistan.Aerosols keep moving because of constant sea breezes over Karachi.

Analysis of Influence of Air Mass Trajectory on AOD Concentration
In Figs.4(a)-(h) back trajectories are representatives of four seasons, i.e., summer, winter, autumn and spring.These were drawn for the 500, 1500, and 2500 m altitude for the measurements to study complex behavior of aerosol near the earth's surface.These were conducted on typical days of four seasons, i.e., 15 th January (winter), 15 th April (spring), 15 th July (summer) and 15 th October (autumn).These trajectories are the part of GDAS (Global Data Assimilation) dataset reprocessed by National Centers for Environmental Prediction (NCEP) of the Air Resource Laboratory.The NOAA HYSPLIT model is helpful to understand the origins of the air masses arriving in the studied region (Draxler and Rolph, 2003).We found similar trajectories for other time periods, but this study includes only selective back trajectories as these are the representatives of the entire season.
The cruise locations on these days of months were  could be observed over this region.Sindh is considered as lower Indus Plain that is influenced by four seasons: summer with monsoon time period (June-August), winter under the influence of western depressions (December-February), spring (March-May) and autumn (September-November).
Dust storms and thunderstorms are characteristics of April, May, October and November (Prasad et al., 2004(Prasad et al., , 2005(Prasad et al., , 2007)).In addition, local pollution in cities and industrial estates, like Karachi, Hyderabad, Nawabshah and Rohri also seen to have contributed to an increase in the aerosol concentrations in the studied region (Gupta et al., 2006;Alam et al., 2011c).
On 15 January the air parcels at 2500 m originated from UAE and Oman arid regions and before reaching Karachi, Hyderabad, Nawabshah and Tharparkar traveled through the Arabian Sea.For Rohri and Jacobabad 2500 m trajectory originated from the Dasht Desert of Iran.A short distance has been traveled from Thar to Nawabshah, Makran for Rohri, Coast of Iran for Hyderabad, and Persian Gulf for Karachi and Tharparkar at 1500 m.The 500 m air trajectory originates from the Arabian Sea and Thar Desert of Pakistan and India.On 15 July the 2500 m air trajectories had north, northeast, and east origin of Pakistan.The 500 m air trajectories originated from the Arabian Sea and spent about 12 hours in the sea before reaching Karachi, Hyderabad and Tharparkar.The 1500 m air parcels originating within Sindh had little journeyed with zigzag air mass pattern.It is therefore analyzed that air masses in January traveled long distances mostly from south-west before reaching Sindh whereas, in June traveled shorter distances mostly from the northeast of Sindh.The air trajectories spent more time over sea in January; in contrast, in June these trajectories spent more time over land that is the reason why AOD values are recorded highest in summer months and lowest in winter season.Most of the locations from where air masses originated are arid or semi-arid in all seasons thus there is a great chance of dust particles blown with air masses to increase aerosols concentrations in Sindh.A similar source was found by Ramachandran and Jayaraman (2003) that most of the air masses with the altitude of 500, 1000 and 5000 m originated from North Atlantic Ocean passes through Africa having world's largest Sahara Desert and the Arabian peninsula with the states like Yemen, Oman and Saudi Arabia traveled through Afghanistan and southern Pakistan before reaching India.The possibility of the air masses getting mixed with the anthropogenic particles during travel cannot be eliminated.From air mass trajectories it has been clear that aerosol concentrations are influenced by arid and semi-arid regions of middle-east, the Arabian Sea, the Indian Ocean and industrial estates in Sindh (Ramachandran and Jayaraman, 2003;Alam et al., 2011c).Most of the air masses reached to various cities of Pakistan from Dasht (Iran), Thar (India), Arabian Sea, Afghanistan, and Sahara desert (Alam et al., 2010(Alam et al., , 2011c(Alam et al., , 2015)).Therefore, both local sources and long-range transported aerosol lead to high AOD in the region.

Relationship between Aerosol Optical Depth and Cloud Parameters
The MODIS provides an enormous valuable data to understand how aerosols relate cloud parameters (Myhre et al., 2007;Alam et al., 2010;Alam et al., 2014).This section deals with the relationships between AOD and cloud parameters for the ten sites of Sindh through the spatial correlation map.We used the correlation maps between AOD and cloud parameters for the time period from 2001 to 2013.For the entire study area correlation maps used for a real comparison between AOD and WV, AOD and CF, AOD and CTT, and AOD and CTP.It has been noticed that higher aerosol loading favors the distributions of higher clouds and larger cloud fractions, either the MODIS AOD data is calculated from all-data or the stringently less-cloudy data sets.Most of the data are noisier when recorded high clouds datasets with a significant deficiency in the sample size (Balakrishnaiah et al., 2012).These relationships are discussed in the following sub-sections.

Relationship between Aerosol Optical Depth and Water Vapor
In order to understand the aerosol impact and the process of the hydrological cycle, we investigated the change in column water vapor in relation to aerosols (Platnick et al., 2003;Myhre et al., 2007;Balakrishnaiah et al., 2012).The spatial correlation between AOD and WV (Fig. 5) reveals, that AOD and WV have a stronger positive correlation over western and central Sindh than on eastern Sindh.The highest positive correlations (correlation coefficient greater than 0.67) were found for most of the eastern Sindh sites including Karachi, Hyderabad, Nawabshah, Larkana, Sukkur, Ghotki and Dadu.In addition relatively lower correlations (greater than 0.62) was found for other site, such as Umerkot.The lowest levels of correlation coefficient (greater than 0.22) were found on eastern most strip of Sindh that's the part of Tharparkar.Similar results were found by Alam et al. (2010Alam et al. ( , 2014) that shows a strong positive correlation between AOD and WV at higher latitudes than at lower latitudes of Pakistan.They analyzed highest positive correlation greater than 0.8 at Peshawar and Rawalpindi and lower levels of correlations greater than 0.7 over other studied sites of relatively lower latitudes of Pakistan.In addition, Alam et al. (2014) found no significant relationship between AOD and WV during the winter season due to absence of dust aerosols consequently lowers the WV in several cities of Pakistan.They related these results with the fact that aerosols might be removed by monsoon rains of summer and the AOD values reduced by colder ground conditions.An elevated WV values was founded along the coast like Karachi where industrial and vehicular emissions is also very high.
The water absorbing ability of aerosols may vary for different types of aerosol.Shi et al. (2008) and Alam et al. (2011) found during their studies that coarse fractions of aerosols i.e., dust particles are mostly insoluble but may become hygroscopic when coated with sulphate or other soluble inorganic aerosols during transport.In addition, Li and Shao (2009) found that when these dust particles well coated with a nitrate may become hygrophobic.Indeed, the concentration of aerosols with their types and nature changes the climate with meteorological parameters such as precipitation, temperature, wind speed and wind direction (Aloysius et al., 2009;Alam et al., 2011).During the summer season, most of the air masses come from Arabian Sea shown in Fig. 4, that is the source of water vapor and higher values of AOD in the region.IPCC (2013) describes the role of water vapor on the scattering behavior of aerosols.All hydrophilic aerosols are important to form clouds with the availability of water vapors to change in both direct and indirect radiative forcing on climate.Therefore, the correlation of aerosol and water vapor is an important topic for modeling of aerosol and for remote sensing to understand.

Relationship between Aerosol Optical Depth and Cloud Fraction
The spatial-temporal correlation between AOD and CF has been plotted for various locations over Sindh during the period 2001-2013 as shown in Fig. 6.Overall the correlation for AOD and CF was found higher at lower Sindh and lower/negative at upper Sindh.For example, the correlation between AOD and CF was greater than 0.23 in Karachi whereas, in Dadu, Larkana, and Sukkur it was less than Karachi and greater than 0.08 and in some cases less than zero or negative.Such areas are biomass and dust prone areas that affect the correlation between two parameters.Kaufman et al. (2005a, b) examined the influence of AOD on meteorological parameters such as cloud cover and air pressure and temperature differences.They found that various cloud properties and aerosol concentrations are being affected by changing in atmospheric circulations for instance, a low pressure region may tend to accumulate aerosol particles and water vapors that lead generating favorable atmospheric conditions to increase cloud cover.
The increased aerosol exposure to solar radiation will result in stronger aerosol absorption effect on clouds.Physically, aerosol absorption of solar radiation can evaporate or thin cloud optically (Hoeve et al., 2011), and that the decrease of CF with AOD.Thus absorption effects suppress the formation and growth of clouds.Some desert clouds are dust formed clouds made with dust and tiny droplets in a case of presence of humidity in surroundings as studied by Rosenfeld et al. (2001).Similar results were found by Balakrishnaiah et al. (2012) that the correlation between AOD and CF was found to be higher than that of coastal sites compared to continental.Nakajima et al. (2001) found that most of the global regions adjacent to the coasts of the continents, have shown positive correlation between AOD and CF.In addition, most of the tropical continental regions have shown a negative correlation with small CF and large AOD value due to the total absence or a very few low clouds interacting with aerosols or may be due to the presence of dust particles that are not very effective as CCN.
For most of the southern Sindh sites AOD and cloud fraction are positively correlated such as in Karachi, Hyderabad, Nawabshah, Thatta, Badin, Tharparkar, and Umerkot.As we move towards the north of Sindh correlation starts to reduce for the rest of the Sindh sites like for Dadu, Larkana, Khairpur, Jacobabad and Ghotki especially the negative correlation has been noticed in eastern and western Sindh region.In these areas negative correlation was found in the range from -0.56 to -0.60.Similarly, Alam et al. (2010) found the correlation between CF and AOD higher at lower latitudes and lower at higher latitudes of Pakistan.They mentioned all those regions of Pakistan dominated by biomass and dust aerosols have the marked increase in the correlation.All meteorological factors influenced that relationship of AOD and CF as well.Alam et al. (2014) analyzed the correlation between AOD and CF, and found lower in urban areas and higher in arid and coastal stations of Pakistan.During the summer months, it was found that AOD and CF were positively correlated as both values were in increasing trend.In contrast, the correlation was found negative during winter months.

Relationship between Aerosol Optical Depth and Cloud Top Temperature
In order to retrieve many properties of the atmosphere closest to surface that is important to know the atmospheric temperature at the level of the cloud top (Balakrishnaiah et al., 2012).In Fig. 7, the spatial correlation of AOD and CTT has shown over the period 2001 to 2013.This demonstrates that CTT increased as AOD increased in almost all northern Sindh sites, whereas, started to decrease and shown a negative correlation towards southern Sindh.Similarly, Alam et al. (2010) found a positive correlation between AOD and CTT at higher latitudes like Peshawar, Zhob, and Rawalpindi and a negative correlation at lower latitudes like Karachi.This is the opposite case as we discussed earlier for AOD and CF correlation map in which CF was observed positively correlated with AOD over southern Sindh, whereas, negatively correlated over northern Sindh.There was a positive correlation between AOD and CTT at Ghotki, Jacobabad, Dadu, Sukkur, Larkana and Nawabshah indicate a consistent positive correlation at higher latitudes of Sindh (High values are up to 0.17).In addition, Karachi, Hyderabad, Umerkot and Tharparkar following a negative correlation at lower latitudes of Sindh (Low values are not less than -0.58).Therefore, mainly a positive correlation between AOD and CTT has been observed for most of the stations of Sindh may be due to large-scale meteorological situations in MODIS data.This correlation might be involved in a very complex interplay among convection, layer of boundary and largescale cloud parameterizations in the region (Quass et al.,  2009; Alam et al., 2010).Several atmospheric and surface properties may retrieve by accurate information on cloud top temperature.It plays an essential role in the net earth radiation budget studies (Balakrishnaiah et al., 2012).Furthermore, Alam et al. (2014) found a negative correlation between AOD and CTT in the winter season, except for Lahore and D G Khan.On the other hand, a consistent negative correlation was observed at coastal sites of lower latitudes like Karachi due to the presence of fine mode aerosol particles that may couple within the water column to show decreased CTT.In addition, anthropogenic aerosols may change the cloud-top temperature by their induced secondary circulations in the atmosphere (Balakrishnaiah et al., 2012).
It has been found that the aerosol concentration influence CCN for most of Sindh areas as aerosol acts on clouds to change the size and composition of cloud particles with a change in nearby temperature conditions in the atmosphere.Some southern most sites are insensitive to changes in CTT with changing aerosol concentrations.The anthropogenic aerosols increase CTT by the induced secondary circulation, and thus change the cloud top temperature (Santer et al., 1995;Sekiguchi et al., 2009;Balakrishnaiah et al., 2012).

Relationship between Aerosol Optical Depth and Cloud Top Pressure
The spatial correlation between AOD and CTP over the period 2001 to 2013 indicates overall a negative correlation in Sindh (see Fig. 8).It was observed that as CTP decreased with an increased in AOD especially at lower latitudes, whereas in relatively mid-latitude this decrease has been observed moderate.Overall, at coastal sites such as Karachi there was a significant decrease in CTP with AOD whereas at continental sites such as Hyderabad, Nawabshah, Larkana, Sukkur, Dadu, Jacobabad and Ghotki this decrease was only moderate.The correlation was found lower in Karachi (-0.45),Tharparkar (-0.35),Hyderabad (-0.25),Nawabshah (-0.16) and more than -0.16 at Ghotki.In addition, Alam et al. (2014) found a consistent positive correlation between AOD and CTP at Peshawar (higher latitudes) and a negative correlation at Karachi (lower latitudes).Furthermore, they have investigated a positive correlation of CTP with AOD for the entire region of Pakistan in the spring season.These correlation facts may be caused by the large-scale meteorological variations.
A similar study has been conducted over different cities of Pakistan by Alam (2010) in which they showed a correlation of AOD and CTT or CTP.Overall, they found a negative correlation of CTP with AOD in the southern region of Pakistan as there was a significant decrease in CTP relative to AOD, moderately decreases at mid latitudes and increase at high-latitudes.Kaufman (2005) reported that CTP decreased with an increased of AOD due to the suppression by increasing the cloud lifetime thus affecting the cloud albedo and CTP.Balakrishnaiah (2012) and Myhre (2007) investigated the same correlation in order to understand atmospheric characteristics that changes with time.Tripathi et al. (2007) analyzed reduced updrafts of vertical winds in the winter seasons and very strong drafts in the monsoon seasons of Indo-Gangetic Plain.They suggest that all these vertical winds of the region are strongly correlated with changing CTP and CF.Koren et al. (2005) also found the vertical wind velocity as a very important meteorological parameter to influence cloud properties.They investigated the Atlantic convection clouds are restructured and invigorated by vertical winds.These vertical updrafts at 500 mb are correlated to the CTP and CF larger than 0.9 for almost the entire studied region, 0.6 for some regions and -0.7 for other regions.All these variations suggest associations between meteorology and aerosol in these regions.

CONCLUSION
All aerosol characteristics based on satellite data were not deeply studied over Sindh.To the best of our knowledge this is the first attempt to analyze and assist aerosol characteristics over this region.Thus, this paper fills the scientific gap in understanding of aerosols and its distributions, concentrations and variations over the region.AOD spatial and temporal distributions were analyzed through the use of the MODIS satellite data.The various sources of aerosols arriving to Sindh were identified through the HYSPLIT model.The Angstrom exponent was determined through the MODIS dataset that may change if the concentration of small and large aerosol particles changes.A correlation of AOD products with Cloud parameters such as CF, WV, CTT, and CTP is necessary for climatological studies and to improve the accuracy and the coverage achievable with MODIS sensor.Cloud parameters were investigated in order to represent MODIS information that was rarely studied for the understanding of atmospheric conditions over this region of Pakistan.
In this study of aerosol spatio-temporal distribution, maximum AOD was found in the summer season and minimum AOD was recorded in the winter season.The highest AOD values were observed in cities of Sindh where industrial and vehicular emissions with various anthropogenic activities increases the number of concentration and distribution over Sindh.Similarly, AOD concentration is also influenced by water cycle of the southern humid regions that converts particles in hygroscopic aerosols.The desert of Sindh has high seasonal concentrations mostly from the Arabian Sea during the monsoon windy season.It has also been found that the water vapor increased as AOD increased over most of Sindh except the desert, CF increased with AOD, where as CTP and CTT have shown a positive correlation in the northern Sindh, a negative in the central Sindh and a highly negative in the southern Sindh.The observed relationship between AOD and CF, the co-variation of AOD with CTP, CTT and WV may be attributed to the large-scale meteorological variation.
between 24°-25°N and 66°-71°E location of Pakistan.Alam et al. (2014) found high AOD values specifically over Karachi, Lahore and D.G. Khan.They attributed these variations to the reason that these are urban, industrial and coastal areas.The dust from Cholistan and Arabian Sea also contributed persistence of dust aerosols along with local vehicular emission.
) in the summer and low AOD value (0.38) in the winter season.During the autumn season, Dadu had specifically the highest AOD value (0.72) of the season while other areas of Sindh had AOD values range from 0.54 to 0.66.In addition, the spring months are the period of dust storms in the region that's why after winters again AOD had shown increased values (0.71-0.86).The southeastern and southwestern Sindh had low AOD values, which are greater than 0.49 during the autumn season.

Fig. 2 .
Fig. 2. Monthly mean AOD retrieved from MODIS for the period 2001-2013 over ten sites in Sindh, Pakistan.

Fig. 4 .
Fig. 4. Back trajectories of different heights in meters ending at 2300 UTC for six stations in Sindh, Pakistan dated January 15, July 15, April 15, and May 15, 2013.