Human influence on marine low-level clouds

(1)

Moa Sporre Air, Water and Landscape Sciences, 2009

Human influence on marine low-level clouds

(2)

Printed at the Department of Earth Science, Geotryckeriet, Uppsala University, 2009.

(3)

i

Abstract

Human influence on marine low-level clouds

Moa Sporre

A study of air mass origin’s effect on marine stratus and stratocumulus clouds has been performed on clouds north of Scandinavia between 2000 and 2004. The aerosol number size distribution of the air masses has been obtained from measurements in northern Finland. A trajectory model has been used to calculate trajectories to and from the measurement stations.

The back trajectories were calculated using the measurement site as receptor to make sure the air masses had the right origin, and forward trajectories were calculated from receptor stations to assure adequate flow conditions. Satellite data of microphysical parameters of clouds from the Moderate Resolution Imaging Spectrometer (MODIS) has been downloaded where the trajectories indicated that clouds could be studied, and where the satellite images displayed low-level clouds. The 25 % days with the highest number of aerosol with a diameter over 80 nm (N₈₀) and the 35% with the lowest N₈₀ have been used to represent polluted and clean conditions respectively. After screening trajectories and satellite imagery, 22 cases of clouds with northerly trajectories that had low N80 values (i.e. clean) and 25 southerly cases with high N80 values (i.e. polluted) where identified for further analysis.

The average cloud optical thickness (τ) for all polluted pixels was more than twice that of the clean pixels. This can most likely be related to the differences in aerosol concentrations in accordance with the indirect effect, yet some difference in τ caused by different

meteorological situations cannot be ruled out. The mean cloud droplet effective radius (aef) was for the polluted pixels 11.2 µm and for the clean pixels 15.5 µm, which results in a difference of 4.3 µm and clearly demonstrates the effect that increased aerosol numbers has on clouds. A non-linear relationship between a_ef and N₈₀ has been obtained which indicates that changes in lower values of aerosol numbers affect a_ef more than changes in larger aerosol loads. The results from this study also indicate that there is a larger difference in the

microphysical cloud parameters between the polluted and clean cases in spring and autumn than in summer.

Keywords: Low-level clouds, the indirect aerosol effect, satellite retrievals, cloud optical thickness, effective cloud droplet radius

Department of Earth Sciences, Uppsala University, Villavägen 16, SE-752 36 Uppsala

(4)

ii

Referat

Mänsklig inverkan på låga marina moln

Moa Sporre

En studie av luftmassors påverkan på marina stratus och stratocumulus moln har genomförts på moln norr om Skandinavien för åren 2000 till 2004. Luftmassornas

aerosolmängdsdistributioner har erhållits från mätningar i norra Finland. En trajektoriemodell har använts för att beräkna trajektorior till och från mätningsstationerna. Trajektoriorna bakåt i tiden beräknades med mätningsplatsen som slutpunkt för att säkerställa att luftmassorna hade rätt ursprung. Framåt-trajektoriorna beräknades från slutpunkten för att säkerställa lämpliga flödessituationer. Satellitdata av mikrofysikaliska molnegenskaper från MODIS har laddats ner för dagar och områden där trajektoriorna indikerat att molnstudier var

genomförbara och satellitbilderna innehållit låga moln. De 25 % dagarna med högst antal aerosoler med en diameter över 80 nm (N80) och de 35 % dagarna med lägst N80 har valts ut för att representera de förorenade och rena förhållandena. Efter att ha undersökt dessa dagar med trajektorior och konstaterat att det fanns låga moln på rätt platser i satellitdatat återstod 22 rena fall vars trajektorior kom norrifrån och 25 förorenade fall vars trajektorior kom söderifrån.

Medelvärdet av den molnoptiska tjockleken (τ) för alla förorenade pixlar var mer än dubbelt så högt som för de rena pixlarna. Detta kan sannolikt relateras till skillnader i

aerosolkoncentration i enlighet med den indirekta effekten, men vissa skillnader i τ till följd av olika meterologiska situationer kan inte uteslutas. Medelvärdet av den effektiva

molndroppsradien (a_ef) var 11.2 µm för de förorenade pixlarna och 15.5 µm för de rena, vilket ger en skillnad på 4.3 µm som tydligt visar effekten av en ökad aerosolmängd på molnen. Ett ickelinjärt samband har tagits fram mellan N80 och aef vilket innerbär att förändringar i aerosolmängder vid låga halter påverkar aef mer än förändringar i aerosolmängder vid höga halter. Resultat ifrån den här studien indikerar också att det är större skillnader mellan de förorenade och rena fallens mikrofysikaliska parametrar på vår och höst än på sommaren.

Nyckel ord: Låga moln, den indirekta aerosol effekten, satellit observationer, molnoptisk tjocklek, effektiv molndroppsradie

Institutionen för geovetenskap, Uppsala universitet, Villavägen 16, SE-752 36 Uppsala

(5)

iii

1 Introduction

Solid or liquid particles suspended in air are called aerosols. There are natural and

anthropogenic sources of aerosols and globally the mass of natural aerosols are 4 to 5 times larger than the anthropogenic aerosol mass. This ratio can however vary significantly on a local scale (Barry and Chorley, 1992). The main sources of natural aerosols are oceans and deserts and the anthropogenic aerosols are mainly produced by burning of fossil fuel and biomass. Aerosols can become cloud condensation nucleus (CCN) and how likely this is to happen depends on the aerosols size, mixing state, chemical composition and the atmospheric environment (IPCC, 2007a).

The radiation budget of the Earth is affected by green house gases (GHG) but also by aerosols. The anthropogenic aerosols are thought to have a net cooling effect on the climate and might have masked some of the warming caused by anthropogenic GHG that has already occurred (Andreae et al., 2005). The uncertainty of the effect of aerosols on the radiation budget is however much larger than that of the GHG as shown by the error bars in Figure 1.

This figure, from the fourth assessment report by the IPCC, shows the estimated

anthropogenic and natural radiative forcing for the year 2005. Anthropogenic impact on climate includes both the direct and indirect aerosol effects. Scattering of radiation back to space by aerosols in the cloud-free atmosphere is called the direct effect. The indirect effect (or cloud albedo effect) refers to aerosols influence on clouds by changing their reflective and microphysical properties (IPCC, 2007a). The estimation of the indirect effect, shown in Figure 1, is however associated with large uncertainties. The figure also shows that the level of scientific understanding is low for the indirect effect, but that the latter is thought to be larger than the direct effect.

The indirect effect can actually be divided into two parts. The first indirect effect was described by Twomey (1974) who suggested that an increase in CCN due to pollution would lead to clouds with more, but smaller cloud droplets for the same amount of liquid water content (Twomey, 1977; Twomey et al., 1984; Charlson, et al., 1992; Glantz et al., 2000). He also proposed that this would lead to optically thicker clouds that would increase the albedo of the earth. The second indirect effect refers to the phenomena when reduced cloud droplet sizes cause the droplets to collide more seldom which leads to suppressed precipitation and prolonged cloud lifetimes (Twomey, 1991).

The greater uncertainties in the aerosol effect on the radiation budget compared to the GHG have several explanations. Aerosols have shorter lifetimes than the GHG and as they are emitted regionally their distribution and effects on the Earth will vary much more spatially.

The GHG influence the climate continuously while the aerosols are more efficient during day time and summer when there is more incoming solar radiation (Charlson et al., 1992). The fact that the size distribution and chemical composition of aerosols are not well described in the global and regional models is another reason for the large uncertainties shown in Figure 1.

The relationship between the aerosol number concentration and the cloud droplet number concentrations is non-linear and varies between different studies. This means that even if the models can estimate aerosol numbers and composition correctly, the impact on clouds is very

(7)

2

uncertain. All models do however indicate that aerosols cause a net cooling of the Earth (IPCC 2007a).

Figure 1. Global average radiative forcing (RF) estimates and ranges in 2005 for anthropogenic carbon dioxide (CO₂ ), methane (CH₄ ), nitrous oxide (N₂O) and other important agents and mechanisms, together with the typical geographical extent (spatial scale) of the forcing and the assessed level of scientific understanding (LOSU). The net anthropogenic radiative forcing and its range are also shown. These require summing asymmetric uncertainty estimates from the component terms, and cannot be obtained by simple addition.

Additional forcing factors not included here are considered to have a very low LOSU. Volcanic aerosols contribute an additional natural forcing but are not included in this figure due to their episodic nature. The range for linear contrails does not include other possible effects of aviation on cloudiness. (The figure is adopted from the fourth assessment report of the IPCC, Summary for policy makers, IPCC 2007b)

One way to observe the indirect effect has been to study ship induced tracks of thickened cloud in marine stratus and stratocumulus clouds. One such investigation by Durkee et al.

(2000) showed that ships that emitted larger sized particles caused tracks, while ships that emitted smaller particles did not. Based on aircraft measurements they found that the droplet concentrations on average increased by 100% in the ship tracks compared with the

surrounding clouds, and that the droplet radius on average decreased by 3.5 µm. In another study, Twohy et al. (2005) investigated 9 stratocumulus clouds over the northern Pacific Ocean in different pollution regimes. They used an aircraft that sampled the air and also used satellite images to study the clouds. They found that the number concentration of particles above 100nm (N100) measured below the cloud was well correlated with the droplet number concentration of the clouds. An anti-correlation was also found between the cloud droplet effective radius (aef) and N100. No correlation was on the other hand observed between N100

and the optical thickness (τ) of the clouds for the whole data set. The lack of correlation is thought to result from differences in liquid water content and geometrical thickness between the clouds. Nevertheless, when regions of similar liquid water content for different pollution regimes were compared, clouds in air masses with high aerosol numbers were typically associated with higher τ.

A global study of low-level clouds over oceans based on satellite data was done by

Nakajima et al. (2001) using imagery from the Advanced Very High Resolution Radiometer

(8)

3

(AVHRR). They found that there was a positive correlation between the aerosol number concentration and the τ, and a negative correlation with the a_ef. The Ångström exponent

( , where τλ is the cloud optical thickness at the wavelength λ) was used as a measure of aerosol loading which can only be derived in a cloud free atmosphere. The aerosol and cloud data is therefore not simultaneously recorded. The investigation used averages of data from 4 months in 1990 though, so the results should not be significantly affected by this.

Another global study using satellite data was done by Bréon et al. (2002) using the

Polarization and Directionality of the Earth Reflectances (POLDER) instrument onboard the Advanced Earth Observing Satellite (ADEOS). This investigation showed that the aef of the cloud droplets were in the order of 14 µm over remote oceans, while in polluted continental air masses it was only 6 µm. Clouds above ocean areas downwind of continents were also found to have smaller droplets. This investigation did however not focus on low-level clouds but studied clouds at all heights. Parameters for clouds and aerosols were not recorded simultaneously for this study either but seasonal averages were used.

An investigation of low-level clouds during haze events at similar latitudes as the present study has been performed in Barrow Alaska (Garett et al., 2004). Ground based remote sensing instruments were used to study the microstructure of the clouds which was compared to simultaneous aerosol measurements. They found that the aef was lower and the droplet concentrations higher in polluted haze clouds compared to non polluted clouds. Penner et al.

(2004) compared cloud optical properties between a clean site in Alaska and a polluted site in Oklahoma. They used a radiative model to calculate cloud optical properties from measured radiances. Their results lead to the conclusion that the radiative fluxes are considerably affected by the indirect cloud effect.

The Arctic region is important to study as it has turned out to be extra sensitive to the warming of anthropogenic GHG and the atmosphere here shows the largest temperature increase in the world over the past decades. Global climate models further show that these north-most latitudes will experience more warming than the planetary average in the future (IPCC, 2007b).

The aim of this study was to investigate the effect of pollution on marine stratocumulus and stratus clouds over the oceans around the northern parts of Scandinavia. The two parameters τ and aef has been compared between polluted and clean clouds for selected days from the years 2000 to 2004. The origin of the air has been calculated using a trajectory model. Aerosol number distributions at an aerosol measuring station in Värriö, in northern Finland have been used to estimate the aerosol loading of the air. τ and aef has been determined using satellite imagery from the MODIS instrument onboard the Terra and Aqua satellites. The scientific objectives of this study were to:

Investigate the anthropogenic aerosols impact on the microphysical and optical properties of low-level clouds over a remote ocean region.

Examine if there are any seasonal differences in the effects aerosols have on low- level clouds.

Relate air-mass origins (marine, continental, fires) to aerosol number size

distributions and cloud microphysical properties in an attempt to better understand

(9)

4

the influences anthropogenic aerosols as well as meteorology have on low-level marine clouds.

Low-level clouds over the ocean were chosen as the objects of study, as marine

stratocumulus and stratus clouds are more sensitive to anthropogenic aerosols than other clouds types, as they normally have a low droplet number (Twomey, 1991). These clouds have a large affect on the planetary albedo since they cover a surface that without clouds has a low albedo. 25 % of the Earth’s surface is covered by marine stratus clouds, so changes in their reflectivity would noticeably affect the radiation budget of the Earth.

In Section 2, some background theory for this study is explained. Section 3 describes the method used in this study. The results from the investigation are presented in Section 4, and a discussion of the results is given in Section 5. Section 6 provides some conclusions to the results in this study.

2 Theory

2.1 Aerosol characteristics and processes

Aerosols exist in various shapes and sizes that span 5 orders of magnitude. To handle the large size range, a lognormal distribution is often used when plotting aerosol distributions.

The aerosol size distribution can be divided into sections based on the aerosols particle diameter (Dp). The coarse mode includes aerosols with a Dp larger than 1 µm and the fine mode contains particles with a Dp less than this size. The fine mode can then be divided into the nucleation mode (0.01< Dp<0.03µm), Aitken mode (0.03< Dp< 0.1µm) and accumulation mode (0.1< D_p< 1µm) (Targino, 2005). New particles are often formed during nucleation events which mean that the amount of very small particles increase considerably during a period of time. Nucleation mode particles are exposed to coagulation and condensation that cause them to grow into Aitken mode particles or dry deposition that remove them from the atmosphere. Changes in aerosol numbers due to coagulation and dry deposition are slower in the Aitken mode size range. The accumulation mode particles are mainly affected by in-cloud processes and wet deposition. The time scale at which a particle remains in the troposphere as a single particle can vary from seconds to weeks (Tunved, 2004).

If air is cooled so that it reaches supersaturation with respect to water, water vapour condenses on the aerosols. Aerosols of compounds that are soluble are more likely to act as CCN than non-soluble aerosols (Targino, 2005). The size of the particles also affects their ability to act as CCN, and for the Scandinavian region it has been shown that particles around 80 nm are potentially good CCN’s (Komppula et al., 2005). Most of the formed cloud

droplets are dried out again and if a droplet has collected more aerosol particles due to coalescence, or if in-cloud oxidation processes have transformed SO2 to sulphate, the size of the resulting dried out aerosols will be larger than the original size. This is called in-cloud processing. There is a minimum formed in the aerosol distribution at around 100nm by in- cloud processing and a large amount of particles larger than this is often formed (Hoppel et al., 1994).

(10)

5 2.1 The investigated area and air masses

The climate of the Arctic is characterised by a high albedo from snow and ice surfaces, and a winter that is both lengthy and dark. The parts of the Norwegian Sea and Barents Sea, that are the focus for this study are however different from the rest of the Arctic, in the sense that less sea ice is formed here due to the Gulf Stream, which brings warmer water to this area. This results in a low albedo of the open ocean areas.

An example of pollution in the Arctic is haze events that were discovered in the 1950s by pilots flying in the Arctic region (Mitchell, 1957). The haze events are areas of lowered visibility in the atmosphere around 1000 km wide and are thought to originate from strong pressure gradients in an east west direction. These pressure gradients cause episodic bursts of polluted air from Europe and Siberia that then reach the Arctic (Barrie, 1986). Since the precipitation frequency and intensity is small and mixing slow during the winter in the Arctic, these pollutants are accumulated in the atmosphere with the highest concentrations in the bottom 2 km. When the Arctic Front is weakened during spring, the pollutants are dispersed (Radke et al., 1989). This causes the haze events to peak during early spring (Engvall, 2004).

The situation in the area investigated in this study is slightly different since the Arctic front is not placed south of the area, which means that aerosols are not accumulated in the same way here. The region is however, still affected by pollution bursts from the European continent.

Information on an air mass’ past can be deduced from its size distribution of aerosols. In a study by Tunved et al. (2003) it was found that continental air masses arriving at high latitudes in Scandinavia usually have larger amounts of accumulation mode particles.

Maritime air masses on the other hand have low aerosol numbers or larger amounts of Aitken mode particles. It was also found that as air moves northward over Scandinavia the amount of aerosols decreases, and as air moves southward over Scandinavia the amount increases, especially Aitken mode sized particles. When the air masses are moving southward, they usually have low aerosol amounts to start with, and are responsive to nucleation events that cause the aerosol number load to increase. The air masses moving northward have high amounts of aerosols already, which make them less responsive to nucleation events. This means that one can expect aerosol numbers at stations in southern Scandinavia to be larger than those at stations further north, as was also noticed in this study. In another study by Tunved et al. (2005), it is concluded that the thermodynamic properties of an air mass are more conservative than the aerosols. This indicates that the aerosol distributions can vary a lot between stations even though they are in the same air mass. One can expect that the air that arrives at the focus area of this investigation from the south would have high aerosol

concentrations, especially in the accumulation mode part of the size distribution. The air that reaches this area from the north or west should have a lower aerosol load and especially accumulation mode loading. Since larger particles are more likely to act as CCN, there will be more CCN available in the southerly air masses. This is expected to cause more, smaller droplets in the clouds in the continental air masses.

2.2 Cloud optical and geometrical thickness

Cloud optical thickness at a wavelength λ and droplet radius r is defined as:

(1)

(11)

6

Where h is the cloud depth, k_E is the extinction coefficient, z is the height above ground, Q_E(r/λ) is the extinction efficiency from Mie theory, n(r,z) is the amount of droplets

(Twomey, 1977). The τ of stratus and stratocumulus clouds is strongly affected by the liquid water content and the droplet size distribution (Durkee et al., 2000). Geometrical thickness of the clouds and the droplet number concentration also affects the τ and the former parameter is the more dominant of the two (Twohy et al., 2005).

Low-level clouds are defined as clouds with a base below 2000 m and include stratus, stratocumulus and nimbostratus clouds. Stratus clouds usually have a cloud base at around 400 m and the other two cloud types at about 1000 m. This changes with latitude and are approximately 1000 m higher at the equator. The average cloud depth of stratus and

stratocumulus clouds is 600 m and 1000 m respectively. At mid-latitudes the two clouds types usually have a cloud top height of 1000 m, which increases in summer (Kokhanovsky, 2006).

In this investigation a cloud top height of 1200 m was chosen as a maximum for all seasons.

2.3 Effective radius and liquid water path

The effective radius is by definition the third moment of the size distribution of the cloud droplets divided by the second moment of their size distribution (Platnick et al., 2002). This results in the equation:

(2) Where

(3)

and f(r) is the global distribution function of the cloud droplets and n is here either 2 or 3. The average volume of the droplets divided by the average surface of the droplets is proportional to the aef (Kokhanovsky, 2006). The aef is often the parameter used to describe the distribution of droplets in calculations of radiative transfer (Platnick et al., 2002).

The liquid water path (LWP) is defined as:

(4)

Where z1 is cloud bottom height, and z2 is the cloud top height and Cw is the liquid water content of the cloud. Satellite measurements have shown that typical values of the liquid water content are between 50 and 150 gm^-2 (Kokhanovsky, 2006). Several studies and large eddy simulations of stratocumulus clouds indicate that the LWP does not change with

changing aerosol load, while other studies show that the LWP decrease in polluted conditions (IPCC, 2007a). LWP will change with changing liquid water content and geometrical

thickness of the clouds.

3 Method

3.1 Aerosol data and station descriptions

Aerosol distributions from Värriö in northern Finland for the years 2000 to 2004 have been used in this study as a measure of pollution level of the air masses moving out over or in from the oceans north of Scandinavia. Aerosol data from two other stations in Finland (Pallas and Hyytiälä) has been used to control the results in Värriö. It was ensured that the Värriö number

(12)

7

size distributions did not display results significantly different from the other stations to avoid local pollution episodes and to detect possible problems with the instrumentation in Värriö.

The aerosol station in Värriö (67º46’ N and 29º35’E) is located 400 m above sea level on a hill top surrounded by pine forest. The instrument used to observe the aerosols is a

Differential Mobility Particle Sizer (DMPS) and the measurements provide an estimate of the number size distribution between 8 and 906 nm every 10 minutes. The station is located far away from any local sources of anthropogenic aerosols but when the air is arriving from the Kola Peninsula, the station usually measures high levels of accumulation mode particles (Tunved et al., 2003).

The Pallas station (68º00’N and 24º14’E) is also surrounded by pine forests and aerosol number size distributions between 7 and 490 nm is observed by a similar DMPS as in Värriö.

The station in Hyytiälä (61º51’N and 24º17’ E) is located in boreal forests and the two differential mobility analysis systems at the station produces new aerosol distributions every 10 minutes. These stations are not affected by local sources of anthropogenic aerosols either (Tunved et al., 2003). More detailed information about the stations can be found in Tunved et al. (2003).

In this study, the number of aerosols with a diameter over 80 nm (N₈₀) was used as the criteria for defining clean and polluted air. It has been shown that the typical activation diameter is around 80 nm at high latitudes in Scandinavia (Pallas measurement station, Komppula et al, 2005). The mean activation diameter indicates the size at which the aerosols start to act as CCN. A daily average of the aerosol number distribution was calculated and then N₈₀ was calculated from the daily average distributions. The 25% days with the highest N₈₀ were chosen to represent the polluted cases for further screening with trajectory data. The 25 % days with the lowest N80 were initially used to select the clean cases. However, since the amount of cases fulfilling this criterion was low, the limit was changed to 35%. Data from the N₈₀35-50% percentile was also used to find clean cases with southerly trajectories.

3.2 Trajectory model

Air mass transports were estimated using the trajectory model HYSPLIT4 (Hybrid Single- Particle Lagrangian Integrated Trajectory 4). The model uses gridded meteorological data as input and can be run online or downloaded and run on a computer. More information about the model can be found in Draxler and Hess (1997). The downloaded version has been used in this study. The trajectories were calculated for 5 stations, the three described above but also for Birkenes (58º30’N and 8º25’E) in Norway and Ny Ålesund (79º00’N and 12º00’E) on the island of Svalbard. The locations of the five stations are displayed in Figure 2. New

trajectories were calculated every hour and the trajectories calculated 10 days backward in time. Meteorological data from the National Centre of Environmental Predictions (NCEP) Global Data Assimilation System (GDAS) Final data set is put into the model every 6 hours.

The trajectories start at a height of 100 m above ground and were calculated for the years 2000 to 2006. The boundary layer height (BLH) in the model is defined as the height at which the potential temperature exceeds the ground level value by 2K. The temperature difference is sought for from above to avoid shallow stratifications near the ground. The minimum BLH for all times is however 250 m (Draxler and Hess 1997).

(13)

8

The trajectories for the days from the 25% highest respective 35% and 50% lowest N₈₀ were studied and sorted according to direction manually. For the polluted cases, days with

trajectories from west- and central Europe and south-western Russia were chosen for further studies. When selecting the clean cases, trajectories from both the south and north were considered. When the trajectories were arriving from the north, it was made sure that they had passed the investigation area within two days before arriving in Värriö. The cases with

approved directions were studied further by calculating trajectories for all five stations and also by comparing the aerosol distribution for all three stations (Värriö, Pallas, Hyytiälä). The studies of all trajectories were done to make sure the air movement to Värriö was part of a large scale air movement out over or in from the oceans north of Scandinavia.

Figure 2. Map showing location of the stations used in the study.

The HYSPLIT4 model was then also used to calculate forward trajectories for the polluted cases and the clean cases arriving from the south. This was only done for the days that had passed the screening with respect to the back trajectories. These trajectories were calculated at 00:00, 06:00, 12:00 and 18:00 for the chosen day and run four days forward. The plots of these trajectories displayed if the air moved out over the ocean north of Scandinavia, and if it did so, also where and at what time. The days that the air did not move out over the ocean were dismissed.

For the clean cases with trajectories arriving from the north, shorter back trajectories were plotted to visualize how long ago the air had been over the north Norwegian Sea or Barents Sea. Both this information and the forward trajectories were then used to determine how long the time gap between aerosol measurements and the satellite data should be.

3.3 Satellite data

Only satellite data for the period March to October was used in due to lack of sunlight and hence satellite data of τ and aef during the winter months. A brief description of the satellites, radiometers and processing will be presented here but they are more thoroughly described in Platnick et al. (2002).

The satellite data is from the MODIS sensor onboard the Terra and Aqua Satellites, which were launched by NASA in1999 and 2002 respectively. MODIS has 36 channels spread between 0.415 and 14.235 µm and is a whiskbroom scanning radiometer. Both Terra and

(14)

9

Aqua are sun synchronous, polar orbiting satellites with an altitude of 705 km and each of them has a global coverage every 2 days (Platnick et al., 2002). At higher latitudes where this investigation is performed the coverage is better though, with several satellite passages per day. Terra has a descending orbit and crosses the equator at 10.30 local solar time, whilst Aqua has an ascending orbit, crossing the equator at 13.30 local solar time (Platnick et al., 2002).

The satellite products used in this study are the cloud top temperature (day), τ, a_ef and LWP.

The cloud top temperature has a horizontal resolution of 5 km and the remaining parameters have a horizontal resolution of 1 km (Platnick et al., 2002). These parameters are a part of the secondary products from MODIS called Cloud Products that are available for download without any cost from http://ladsweb.nascom.nasa.gov/data/search.html. The cloud products are the results of processing from the original data from various bands received by MODIS.

The cloud top temperature was used to remove pixels containing high level clouds in the satellite scenes, since the focus of this study is to investigate low-level clouds. Bands in the 11 µm region are used by the algorithms calculating the cloud top temperature for low-level clouds. The cloud top pressure is then derived from the cloud top temperature using the NCEP GDAS data. For clouds at higher altitudes the cloud top pressure is derived from satellite data and the database used to derive the cloud top temperatures (Platnick et al., 2002). Since the cloud top pressure is inferred from the cloud top temperature for low-level clouds, it was here decided to use the cloud top temperature as a limit for determining low-level clouds.

Described next is some of the processing procedures of the MODIS cloud products τ, a_ef and LWP. All the microphysical properties of the clouds are calculated by comparing radiation signals from visible and infrared channels and iteratively line these against libraries of pre- calculated values. These values have been calculated for homogeneous clouds that are plane- parallel with a black surface underneath (Nauss et al., 2005, Platnick et al., 2002). Separate libraries exist for water and ice clouds. Whether the pixel is overcast and the phase of the cloud in the pixel first has to be known before its τ, a_ef or LWP can be determined. A logical decision tree with several tests is used to determine the cloudiness and phase. For the overcast part, the satellite data is classified using a cloud mask with 4 different confidence levels indicating if the pixel contains clouds. Only the pixels with the two highest confidence levels for clouds are classified for τ, aef and LWP (Platnick et al., 2002). Determining the cloudiness level is important because broken cloud can affect the retrieved results of the τ and the a_ef. If the surface underneath is darker than the clouds, broken clouds will cause an increase in the aef and a decrease in the τ. If the surface is lighter than the clouds, the opposite will occur (Nauss et al., 2005). The decision tree also decides the phase of the clouds and each pixel can only be classified as water or ice even though the clouds can be a mixture between the two (Platnick et al., 2002). In this study, the only cases with a low cloud level are included in the analysis low so minimal problems should occur with mixed phase clouds.

The albedo of the area beneath the clouds can affect the retrieval of the cloud parameters.

To correct for this, the algorithm uses a variable surface albedo that has been obtained for the entire Earth during clear conditions. This varies with vegetation during the seasons, and snow and ice cover is taken into account by using a snow and ice mask provided by the National Snow and Ice Centre’s Near real-time Ice and Snow cover. The satellite data is also corrected

(15)

10

for the effects the atmosphere overlying the clouds, might have on the radiation received by the satellite (Platnick et al., 2002).

The aef, τ and LWP data analysed has been derived using the shortwave infrared band 7 (2.1 µm) in combination with one of the visible bands 1, 2 or 5 (0.65, 0.86, 1.2 µm). The visible band is chosen by the algorithm and the choice depends on the underlying surface with band 1 used for land, band 2 for oceans and band 5 for ice or snow (Platnick et al., 2002). Since this study is taking place over ocean areas that are ice free most of the year, one can assume that mainly band 2 has been used in the pixels studied. The channels in the visible range of the spectrum are more sensitive to τ and channels in the infrared region more sensitive to the aef

of the cloud droplets (Kokhanovsky et al., 2005).

Areas, dates and satellite products were defined on the internet page presented above and up to 14 sets of cloud products could be produced by both satellites per day. All of these did however not display the area of interest. For spring and autumn cases, the number of images was less due to fewer hours of sunlight. The satellite images were investigated and one chosen per case. Which image that was chosen, depended on how well they displayed the area and how well they matched the progress of the trajectories in time. It was also ensured that the images contained low-level clouds where the trajectories indicated that clouds could be studied. This meant that the images of cloud top temperatures were first studied, and if low- level clouds were present the images of τ were examined to make sure that the clouds were not convective. Convection could be seen as a spotty pattern in the τ images. For the clean and polluted cases before July 2002 only images from Terra were used. For the clean cases this was done to speed up the screening procedure and because Terra provided sufficient images.

The satellite images selected were downloaded and the program Transform© used to

transform the hdf data files into text files readable by Matlab®. For the polluted cases the area of the pixels to be studied was chosen in the transform program by comparing the program’s cloud top temperature image to the satellite image online and the trajectories to make sure the right area was chosen. The comparison had to be done because no geographical information was available when transform displayed the data. This was very time consuming and for the clean cases an area between 71 and 75º N and 15 º of longitude between 0 and 45 º E was chosen for each case, depending on where the trajectories came from. This could be done for the cold cases since the areas of low-level clouds often were smaller than those of the polluted cases, and because the trajectories arrived from the north, they were always within certain latitudes. The polluted data was more spread around the coasts at different latitudes. For the polluted cases, it also had to be checked thoroughly that the clouds were where the polluted air moved out over the ocean.

3.4 ECMWF data

Data from the European Centre for Medium-Range Weather Forecasts (ECMWF) has been analysed to determine the mean temperatures at 1200 m above ground level for April, July and September. Monthly average temperatures from the 21 lowest pressure levels for the three months from each year were analysed. The data had a spatial resolution of 17 km which is less than the satellite data. The spatial average temperature over a region between 71-75º N and 5-45º E was then calculated for each pressure level. This area was chosen since most

(16)

11

satellite scenes were situated within this region. The altitude to the pressure levels was then calculated using the hydrostatic equation. Average temperatures and pressures between the levels were used to calculate the air density. Through interpolation the temperature at 1200 m was then obtained for different months. An average temperature at 1200 m for each month has been estimated according to the years 2000 to 2004. From this a function for cloud top

temperatures depending on the month of the year was created using linear interpolation. The values for this function are shown in Table 1.

Table 1. Maximum cloud top temperatures estimated for the different months.

The ECMWF forecasted BLH for the three months above of the years 2000 to 2004 has also been analysed in the present study. However, for the year 2000, no data is available for April.

An average BLH for each month was also calculated for the data from the different years.

4 Results

4.1 Aerosol data, trajectories and satellite data

Figure 3 shows the daily average N80 for the years 2000 to 2005, obtained at the Värriö measurement station. The yellow areas mark the summer months. The highest N₈₀ values occur in summer or late spring most years. In 2001 the values appear low in summer, however the data in Pallas from the same time period does not present such low values (not shown).

The winter 2000/2001 has higher N80 than the other winters.

Figure 3. Daily average of N₈₀ at Värriö for the years 2000-2004. Yellow areas mark the months July to August.

The total number of cases for the N80 25 percentile was 331, of which 50% were observed during the darker months of the year when no satellite data could be used. For the polluted cases only14 % was observed during the winter period. Periods with high N80 are much more frequent during the summer months, which also is evident from Figure 3. Using the 35

Month March April May June July August September October

T_top max (K) 261.5 265.7 269.9 274.1 278.3 275.3 272.3 269.3

(17)

12

percentile as criteria for low particle concentrations, 48% of the days was found to occur during winter.

Based on investigations of the back trajectories, 79 days from the polluted data set were determined to originate from a suitable direction. A significant amount of the 25% most polluted days had trajectories which did not originate from the south, but passed over the Kola Peninsula shortly before arriving in Värriö. These pollution episodes were considered to be local and were therefore not further investigated. After studying the forward trajectories of the polluted cases, 48 days were considered suitable for further investigation using satellite data.

During the study of the forward trajectories, cases were dismissed because the trajectories did not reach the ocean or because there were model problems and no trajectories could be calculated. For the clean 35 percentile, 126 cases had suitable trajectories from the north and an aerosol number distribution not too different from Pallas. For the 16 southerly clean cases in the 35 percentile, forward trajectories were calculated. Unfortunately, none of these cases could be included in the study either because the trajectories did not reach the ocean or the no suitable clouds were present in the satellite imagery.

One extra case from 2006 has been included in the study since it was known that the investigated air mass originated from severe agricultural fires in south-western Russia. The case is interesting since agricultural fires generated extremely high amounts of accumulation mode particles into the air (Stohl et al., 2007). No aerosol data from Värriö has been available for the year 2006 for this study, so data from Pallas has been used as a proxy of N80 for this case. The value of N₈₀ from this case is well above the 75% limit of N₈₀ observed at Värriö.

After the study of the satellite images, 25 polluted (including the case from 2006) and 22 clean cases remained. Some days were rejected because the satellite images displayed only higher clouds, clear areas, only small areas of low clouds between high clouds and holes, or convective clouds. Some days there was no satellite data available due to problems with the satellite. Furthermore, there seemed to be a greater amount of high level clouds in the clean satellite imagery. When there were low-level clouds present they were often convective due to cold air from the north moving out over a warmer ocean. Thus, even though there were more than twice as many clean cases compared to polluted ones investigated in satellite imagery, there were fewer clean cases approved. 3 semi-clean cases from the 35 to 50% interval with southerly trajectories also had suitable satellite data and are used in some parts of this study.

The frequent presence of high level and convective clouds during the clean days also resulted in less amounts of pixels available in the analysis of the clean cases compared to the polluted ones. When determining the τ for all polluted cases a total of 1150880 pixels were used while for the clean cases 347539 pixels were used. There are more than 3 times as many pixels used for the polluted cases. The number of pixels used for each case is shown in Tables 2 and 3. The amount of pixels varies a lot from case to case. The case with the lowest amount of pixels covers an area of 519 km². When studying the numbers of pixels in the tables, one can also see that for most cases the values of τ are higher than the values of aef. The reason for this is not known, but may be connected to the processing of the satellite data and the

matching towards the pre-calculated libraries. For example, all water phase pixels are classified as broken cloud when the a_ef is larger than 30 µm (Platnick et al., 2002). Some pixels might have been classified as broken clouds in the a_ef processing but cloudy in the τ processing.

(18)

13

More information about for example the overpass times of the satellite scenes, time period between the in-situ aerosol data and satellite images and which satellite sensor that has been used can be found in Tables 2 and 3

Table 2. Properties of the polluted cases. Dates are from the day of the aerosol measurements.

Table 3. Properties of the clean and semi-clean cases. Dates are from the day of the aerosol measurements.

Year Month Day N80

(cm^-3) Mean latitude (°)

Mean longitude

(°)

Time of satellite image (UTC)

Satellite

Days between aerosol measurements and stallite data

Mean τ Mean a_ef (µm)

No of pixels for

τ

a_ef

Mean LWP (gm^-2)

2000 4 17 390 72 37 11:05 Terra 1 14.13 9.61 73173 72555 87.38

2000 4 18 498 72 37 10:10 Terra 1 16.84 8.66 87233 86374 92.28

2000 9 22 867 71 38 11:15 Terra 1 27.20 11.82 119359 116118 217.77

2000 10 18 582 70 44 09:20 Terra 2 19.68 15.31 55532 49439 266.69

2001 4 24 720 73 46 14:30 Terra 2 16.00 10.09 26045 29438 187.39

2001 4 25 636 71 42 09:40 Terra 1 25.89 10.13 133153 137668 188.67

2001 6 7 1261 74 29 15:50 Terra 1 10.98 12.47 9977 9558 87.59

2001 6 8 1567 72 16 17:15 Terra 2 10.31 11.31 14820 13750 75.92

2001 7 14 290 73 47 08:40 Terra 2 9.55 13.23 122021 120860 82.00

2002 5 2 1201 72 45 15:35 Terra 1 8.13 10.43 23867 18279 62.85

2002 7 1 624 72 15 18:20 Terra 2 13.66 10.44 29818 29124 90.69

2002 7 2 617 72 33 17:25 Terra 2 10.77 10.74 14780 14529 76.52

2002 7 3 757 70 58 07:00 Modis 2 9.42 8.56 519 687 67.12

2002 7 7 863 68 44 16:05 Terra 2 5.25 12.10 984 770 46.91

2002 7 26 839 69 56 08:30 Terra 2 13.69 11.49 6422 6345 103.88

2003 7 24 499 69 47 16:15 Terra 2 10.39 13.60 24686 23309 88.85

2003 7 26 966 73 27 06:35 Modis 2 14.53 12.55 19484 18736 119.93

2003 9 12 1008 69 45 08:10 Modis 2 8.80 14.37 17856 15004 101.06

2003 9 23 415 75 43 11:50 Terra 1 8.86 14.20 18086 14817 85.39

2004 7 2 712 72 21 12:15 Terra 2 13.43 11.15 29903 29726 96.00

2004 7 4 708 77 34 16:20 Terra 1 9.39 13.64 7034 7029 82.28

2004 7 8 857 69 42 07:45 Modis 3 9.89 11.60 14689 14649 74.70

2004 7 13 515 72 20 19:20 Terra 1 12.52 11.50 104938 93954 96.06

2004 9 25 336 73 18 10:55 Terra 2 13.06 11.71 102950 103140 95.52

2006 5 2 3173 74 25 09:20 Terra 1 12.22 8.34 93551 90124 57.33

Year Month Day N₈₀ (cm^-3)

Min longtitude

(°)

Max longitude

(°)

Time of satellite image (UTC)

Satellite

Days between aerosol measurements and stallite data

Mean τ Mean a_ef (µm)

τ

a_ef

Mean LWP (gm^-2)

2000 7 1 62 25 40 10:15 Terra 2 10.33 10.39 69458 68174 60.67

2001 9 29 50 15 30 10:55 Terra 1 5.60 18.43 23860 13400 109.87

2002 4 8 65 20 35 12:40 Terra 2 10.10 14.76 14740 12933 88.96

2002 8 2 101 0 15 10:40 Terra 2 6.18 13.62 4842 4667 32.81

2002 8 3 64 5 20 11:20 Terra 2 7.66 11.44 42259 41551 46.33

2002 8 5 76 15 30 11:10 Terra 2 4.27 16.35 13950 12712 42.79

2002 9 19 47 25 40 09:00 Terra 2 7.53 21.02 8275 5847 112.94

2002 10 5 69 30 45 11:20 Terra 1 7.12 18.89 1183 786 115.58

2002 10 8 24 15 30 09:30 Terra 2 4.37 14.90 22177 17742 42.32

2003 8 24 42 15 30 14:20 Terra 2 12.18 14.64 40404 38447 113.19

2003 8 25 37 15 30 10:10 Terra 2 9.79 17.40 75339 69684 112.76

2003 8 26 68 25 40 10:55 Terra 2 7.52 17.26 13107 12049 82.02

2003 8 27 24 30 45 10:40 Terra 1 3.14 20.40 15218 10560 47.97

2003 9 2 77 20 35 08:25 Terra 1 5.01 21.74 6282 5458 44.64

2003 9 20 65 15 30 12:25 Terra 2 5.20 20.51 30235 19484 92.85

2004 5 13 77 15 30 11:15 Terra 2 3.14 20.47 8521 3659 57.17

2004 5 25 90 30 45 11.35 Terra 2 6.27 10.92 10479 8362 42.65

2004 6 8 78 0 15 18:20 Terra 2 2.13 17.30 4048 3044 21.11

2004 8 13 53 15 30 16:30 Terra 2 4.17 16.83 3050 1798 56.58

2004 8 24 36 10 25 12:55 Terra 2 4.61 21.38 16856 10249 78.82

2004 8 26 48 35 50 11:05 Terra 2 3.53 15.68 7314 6685 19.84

2004 9 11 62 0 15 11:50 Terra 1 5.71 15.32 19892 16976 62.20

semi- clean cases

2000 3 16 150 25 40 13:25 Terra 2 4.14 16.44 16416 7577 81.54

2003 3 4 117 30 45 10:30 Terra 1 10.34 14.57 40759 37086 95.52

2004 9 12 138 30 45 12:20 Terra 1 4.16 21.05 4867 2440 85.94

(19)

14 4.2 Cloud Optical thickness

Results of the satellite retrieved τ, obtained for the clean and polluted cases of the investigated period, are shown in Figure 4. The normalized distribution functions shown in the figure include all pixels from the 25% most polluted and 35% cleanest cases. The mean τ of all the polluted pixels is 15.9 and for the clean pixels 7.22 as can be seen in Table 4. Thus the mean τ is more than twice as high for the polluted conditions compared to the clean conditions. In Figure 4 it is visible that the distribution of τ for the polluted pixels peak at higher values than the clean distribution which has its highest percentages at the lowest bin of τ. The polluted distribution also has higher percentages at higher τ compared to the clean distribution.

Figure 4. Normalized distribution function of τ for polluted and clean pixels.

Table 4. Means for polluted and clean distributions. Values within brackets are medians.

Since the cloud optical property is both affected by pollution and dynamical and

thermodynamically processes, it is not possible to determine how much of the differences in the distribution of τ shown in Figure 4 that are caused by the aerosol load. In that perspective, the study by Twohy et al (2005) found no significant correlation between τ, τ is calculated from the LWP and aef, and aerosol number concentrations.

Unfortunately, it has not been possible to determine the geometrical thickness of the clouds in this study, although it is well known that this quantity significantly affects the τ. The meteorological situation for the polluted and clean days was probably very different. For the clean cases the air arrives from the north. This usually involves advection of cold and dry air over a warmer ocean. This increases the temperature and water content in the marine

boundary layer and causes unstable stratification. Open cells which indicate convection were often observed in the satellite imagery of the northerly trajectories. These cases were rejected in the present study. During most of the year the polluted air arriving from the south is relatively warm and moist. As this air moves out over a cold ocean a possible convective boundary layer would collapse and become stable. Formation of advection fog at times would also be favorable when warmer air moves out over a colder ocean. Some of the clouds in the study are connected to frontal systems which also affect the depth of the clouds. Thus, the

Mean τ Mean a_ef (µm) Mean LWP (gm^-2) Polluted cases 15.9 (10.7) 11.2 (10.0) 123.1 (77.4)

Clean cases 7.2 (3.8) 15.5 (13.8) 78.5 (45.1)

(20)

15

meteorological situation varied with origin of the air but also was also very different from case to case.

The satellite retrieved LWP, which also affects τ, is higher for the polluted days than the clean days as is shown in Figure 5. The mean LWP is 123 gm^-2 for the polluted distribution and 78.5 gm^-2 for the clean distribution. The higher values for the former cases associated with southerly transport are expected as these air masses are warmer and contain more moisture than the cold air arriving from the north. The LWP also depends on the geometrical thickness of the clouds. The clouds from the south most likely have both a higher geometrical thickness and water content. Figure 6 shows a relatively strong correlation between mean τ and mean LWP for the clean and polluted cases.

Figure 5. Normalized distribution function of LWP for polluted and clean cases.

Figure 6. Mean LWP for each case versus mean τ for each case.

The τ is also affected by a change in the cloud droplet distribution presented in the next section.

4.3 Effective radius

Results of the satellite retrieved aef for polluted and clean conditions are displayed in Figure 7.

The mean aef of the polluted pixels is 11.2 µm and the mean of the clean pixel is 15.5 µm which results in a difference of 4.3 µm. In Figure 7, one can see that the peak of the polluted distribution is placed at a slightly shorter radius than the peak of the clean distribution. The

(21)

16

distribution for the clean conditions has a larger spread compared to the polluted distribution, with higher percentages in the 20-30µm range for the former cases. The fact that retrievals of aef for water clouds is restricted to droplet sizes less than 30µm explains the sharp decrease for these droplet radiuses, particularly for the clean pixels. Figure 7 suggests that there are few pixels that have been classified as ice. However, minor amounts of ice pixels can have an aef

less than 30µm.

Figure 7. Normalized distribution function of a_ef for polluted and clean pixels.

The satellite retrieved aef is a better indicator of the effect of pollution on the clouds than the τ, since it is only to a minor degree affected by the cloud thickness (Twohy et al., 2005), especially as no convective clouds are included in this study. The a_ef will also be affected by the amount of available moisture in the cloud, which the LWP parameter is an indication of.

Even so, the probable higher moisture contents for the polluted cases (Figure 5), supported by the southerly transports, does not results in larger mean aef compared to the clean condition (Figure 7). This further strengthens indication that higher N80 values instead had a

significantly larger influence on the microphysical properties of the clouds in the sense that more droplets with smaller sizes were formed. The significant difference in droplet radius (4.3 µm) between the polluted and clean conditions, clearly shows that aerosols carried northward in air masses from Europe influenced the clouds present over the oceans around the northern parts of Scandinavia.

In the study by Durkee et al (2000) they found a similar difference in a_ef (3.5 µm) between the surrounding clouds and the ship tracks. In a study by Garrett and Zhao (2006), the retrieved aef was 12.9 µm for their clean cases and 9.9 µm for their polluted cases, thus the difference was smaller and the values lower than the present results. The levels of N80 are not known though, so differences in aerosol number loads could play a role in the discrepancy.

The lower values of a_ef are probably caused by lower air moisture content. The LWP in their study were on average 33.5 gm^-2 and 31.1 gm^-2 for the polluted and clean cases, respectively.

The present mean aef for the polluted cases is significantly higher than the one found by Bréon et al. (2002), whom have estimated a minimum of 6 µm based on satellite data. This was however obtained over land where the aerosol load was probably significantly higher than in this study. Their largest value of a_ef retrieved over remote oceans was 14 µm. Furthermore in a study by Breon and Doutriaux-Boucher (2005), cloud microphysical parameters from

Human influence on marine low-level clouds

Moa Sporre Air, Water and Landscape Sciences, 2009

Human influence on marine low-level clouds

Abstract

Human influence on marine low-level clouds

Referat

Mänsklig inverkan på låga marina moln

Table of Contents

1 Introduction

2 Theory

3 Method

4 Results