UNIVERZA V NOVI GORICI FAKULTETA ZA NARAVOSLOVJE MERITVE IN MODELIRANJE GIBANJA ZRAƒNIH MAS V TROPOSFERI DIPLOMSKO DELO Miha šivec Mentor: prof. dr. S

Transkripcija

1 UNIVERZA V NOVI GORICI FAKULTETA ZA NARAVOSLOVJE MERITVE IN MODELIRANJE GIBANJA ZRAƒNIH MAS V TROPOSFERI DIPLOMSKO DELO Miha šivec Mentor: prof. dr. Samo Stani Nova Gorica, 2016

2

3 UNIVERSITY OF NOVA GORICA SCHOOL OF SCIENCE MEASUREMENTS AND MODELING OF AIR MASS MOTION IN THE TROPOSPHERE DIPLOMA THESIS Miha šivec Mentor: prof. dr. Samo Stani Nova Gorica, 2016

4

5 Zahvala Najprej bi se rad zahvalil mentorju prof. dr. Samu Stani u za vodenje, nasvete in pomo pri izdelavi diplomske naloge. Hkrati bi se rad zahvalil Maru²ki Mole za vso pomo, ideje in lektoriranje. Posebna zahvala gre moji druºini za spodbudo pri tem in vseh vzponih v mojem ºivljenju ter podporo, ko zadeve ne gredo ravno tako, kot je zami²ljeno. i

6 ii

7 Povzetek Skozi zgodovino je bil lovek vedno odvisen od vremena. Ena od prioritet za njegov obstoj je bilo poznavanje vremenskih vzorcev in sposobnost napovedovanja vremena. V zadnjih letih se lahko vreme z uporabo sodobnih merilnikov atmosferskih lastnost, numeri nih modelov in super-ra unalnikov napoveduje natan neje in za dalj²e asovno obdobje kot kadarkoli do zdaj. Hkrati se izbolj²uje tudi razumevanje zikalnega ozadja atmosferskih procesov, ki predstavljajo eno najpomembnej²ih lastnosti na²ega sveta. V diplomskem delu se omejimo na pojave v spodnji plasti atmosfere troposferi, v kateri potekajo vsi vremenski procesi. Pod pojmom vreme razumemo trenutno stanje in dogajanje v ozra ju, ki ga je mogo e opisati s termodinamskimi vremenskimi spremenljivkami in zvezami med njimi. Ena glavnih zna ilnosti troposfere je gibanje zraka v horizontalni in vertikalni smeri. Za zanesljivo modeliranje in napovedovanje vremenskih dogajanj je torej pomembno im bolj²e poznavanje stanja atmosfere, vklju no z zra nimi tokovi, kar zmanj²a napake pri modeliranju. Cilj diplomske naloge je predstavitev novega na ina za merjenje smeri in hitrosti gibanja zra nih mas, ki temelji na kombinaciji aktivnega in pasivnega daljinskega zaznavanja stanja atmosfere. S pomo jo lidarja se dolo i razdalja do objekta, ki sluºi kot sledilec v zra nem toku. V tem primeru so bili sledilci kar oblaki. So asno z lidarskimi meritvami se fotograra del neba, v katerega je usmerjen lidar. Na fotograjah se poi² e izrazite dele oblakov, ki jim je mogo e slediti na ve zaporednih posnetkih. Iz meritev oddaljenosti sledilca ter njegovega premikanja izra unamo hitrost potovanja oblaka in s tem tudi hitrost zra ne mase, ki oblak nosi. Metodo merjenja hitrosti gibanja zra nih mas smo preizkusili na ²tirih testnih primerih v februarju in marcu Meritve so potekale v Ajdov² ini v razli nih vremenskih pogojih. Poleg naprav za daljinsko zaznavanje (lidar in opti ne kamere) smo uporabili tudi prizemne meritve vetra v Ajdov² ini ter podatke vertikalne sondaºe atmosfere v Udinah in v Ljubljani. Rezultati meritev vetra z daljinskim zaznavanjem se v vseh ²tirih opazovanih primerih relativno dobro ujemajo z rezultati sondaºe. Do odstopanj prihaja zaradi krajevnega in asovnega neujemanja sondaº z daljinskim zaznavanjem ter zaradi omejitev metode z daljinskim zaznavanjem pri dolo evanju dejanske smeri vetra. Opaºene omejitve metode je mogo e odpraviti z uporabo vsenebnih kamer ter vertikalnih lidarskih meritev. Klju ne besede: daljinsko zaznavanje, veter, atmosfera PACS: Gv, Gn iii

8 iv

9 Abstract Throughout the history human race depended on weather, so one of the priorities for its survival was to understand weather patterns and to be able to forecast weather. With the development of powerful computers, atmospheric numerical methods and precision instruments for atmospheric monitoring, it is possible to predict weather with greater accuracy and for a longer period of time ahead. At the same time, we are able to gain improved understanding of physical processes that occur in the atmosphere and represent one of most important features in our world. This diploma thesis focuses on the lowest part of the atmosphere troposphere only, as all weather occurs in the troposphere. Weather is a complete collection of momentary thermodynamic states in the atmosphere and is dened with thermodynamic variables and relations between them. The goal of this thesis is development and presentation of a new way to determine the direction and speed of air mass movement, based on the combination of passive and active remote sensing techniques. A lidar is being used to determine the range to an object, in our case a cloud, that can be used as a tracer in the air current. Simultaneously with lidar ranging of clouds that same clouds are being visually monitored in a series of optical photographs. Selecting and following the temporal evolution of distinct cloud features and their range allows us to calculate the speed of clouds. The performance of this method was tested on four cases in Feb. and Mar Measurements were performed in Ajdov² ina in dierent weather conditions. Along with remote sensing (infra-red lidar and optical cameras), ground measurements of wind at Ajdov² ina were performed. Wind speeds and directions obtained from remote sensing were compared to atmospheric sounding data from Ljubljana and Udine at similar heights and performed within as small as possible time window. In all four cases remote sensing results for wind speeds and directions agree relatively well with atmospheric sounding. Deviations are expected to be primarily due to spatial and temporal mismatch between sounding and remote sensing measurements. Another source of uncertainties are the limitations of the present remote sensing method in the determination of the actual direction of the wind, however, theses limitations could be eliminated in the future by using an all-sky camera and vertical lidar conguration. Keywords: remote sensing, wind, atmosphere PACS: Gv, Gn v

10 vi

11 Contents Zahvala Povzetek Abstract i iii v 1 Introduction 1 2 Troposphere and its properties Clouds Wind HYSPLIT Model Experimental setup and measurement techniques Anemometer Radiosonde Lidar Cameras Wind speed retrieval technique Case studies Case 1 - Pannus cloud Case 2 - Altocumulus Case 3 - Altocumulus Case 4 - Nimbostratus Results Conclusions 33 Acknowledgements 34 Bibliography 35 vii

12 viii

13 1 Introduction The atmosphere is a layer of gases and aerosols that surrounds the Earth. It is not homogeneous; density of the atmosphere is decreasing approximately exponentially with height. About 99 % of atmospheric mass is in its lowest 30 km, which is a very complex environment. Atmospheric pressure is mostly due to hydrostatic pressure as a consequence of the weight of air and is also exponentially decreasing with height. The temperature, that is also one of the main thermodynamic variables, does not manifest such a simple behavior (Fig. 1.1). It changes very dierently in dierent height intervals and is often used to dene atmospheric layers. In this thesis, we focus on the layer that is closest to the ground, called the troposphere. It is further subdivided into the planetary boundary layer (PBL) and free atmosphere. In the PBL the interactions between the atmosphere and the surface take place, such as the exchange of heat and matter, which results in variations of PBL height and structure [1]. Figure 1.1: Average temperature prole for atmospheric layers up to 130 km. All weather phenomena take place in the lowest atmospheric layer, the troposphere. (obtained from [2]). 1

14 As atmospheric properties vary with time, continuous measurements are needed to be able to observe development and propagation of weather phenomena in the troposphere. Except for the lowest part of the PBL, the troposphere is out of reach for most ground based measuring devices. If we want to get data from greater heights, instruments must be sent in the air. This can be done using balloons and airplanes. To get a vertical prole, most commonly balloons with radiosonde are being used. The problem with this kind of measurements is that a balloon stays at a given height for a very short time. As it is very costly to launch a large number of balloons per day in short time intervals, radiosonde measurements have in general poor temporal resolution. Excellent temporal resolution can on the other hand be achieved by remote sensing, which uses electromagnetic or sound waves to detect distant atmospheric features. Data is acquired without direct physical contact between the observer and the observed object. Two types of ground remote sensing instruments are very important in atmospheric research. Weather radar, emitting microwave radio signals, is used to obtain general information about precipitation and precipitating clouds. The distance to the investigated object can be obtained from the time of ight of the transmitted and backscattered signals. Lidar measurement techniques are similar, but instead of microwaves use electromagnetic waves ranging from near UV to near IR light. Shorter wavelength of light makes it possible to detect smaller objects in the air. Air masses are usually invisible to the naked eye. For observation of their movement we need tracers that are carried along by the streams of air, so we can follow their trajectories. Natural tracers for the motion of air masses are clouds. Cloud motion can be recorded with a camera, giving us the possibility to extract its dynamics, if we know how far away they are. The distance of the clouds measured with lidar provides a way to extract the size of their features and length of their paths made in a given time interval. Average speed of the clouds can be obtained from a series of photographs. Under the assumption that the clouds are carried along by the moving air masses, their average velocity and direction are also those of the wind at the cloud height. 2

15 2 Troposphere and its properties Troposphere is the lowest layer of the atmosphere. It is the layer in which all weather happens. Above equator it reaches height of approximately 17 km, in Polar Regions is lower, it reaches 9 km. Thickness in regions in between depends on time of year and is the highest in summer. Troposphere ends with tropopause that has absolute temperature change below 2 C/km and is approximately 2 km thick. In troposphere is approximately 80% of the mass of the atmosphere, which includes most of water vapor of the atmosphere. Temperature is falling while height, except in case of inversion where in rst few kilometers temperature is increasing with height and creates very stable environment. Usual lapse rate is around 6 C/km. The lowest part of troposphere, which is in direct contact with the ground, is called planetary boundary layer (PBL). Its height highly depends on ground conguration, and is changing during the day with highest point being reached in the afternoon. PBL in average reaches 1500 m above the homogeneous ground and 500 m above large bodies of water [3]. Obstacles on the ground create drag on the moving air, which creates mixing and changes in direction of air movement. That is shown as turbulent movement of winds in PBL. Above PBL is free atmosphere that does not depend on the roughness of the terrain. In case of unstable atmosphere, convection starts. Convection breaks PBL and inuence of the ground reaches through the whole troposphere. Convection is stopped by the tropopause at the top of troposphere. On the other hand if temperature inversion appears and creates very stable atmospheric conditions, ground inuence only reaches as high as the top of inversion layer [1]. Weather is a complete collection of momentary thermodynamic states in air [4]. Weather is dened with thermodynamic variables, such as temperature, pressure and moisture of the air. Combining dierent states gives us dierent weather events and phenomena. Events also dier in strength and size. Because troposphere is the rst layer above the ground, creation and evolution of phenomena highly depends on geographic shapes and terrain features such as seas, open planes and high mountain ridges. Extreme landscape features can also create micro climate that diers from climate of surrounding area. Weather is dened by temperature, the amount of precipitation, wind speed, air pressure, etc. 2.1 Clouds A cloud is a visible accumulation of particles of liquid water and/or ice suspended in the free air. It may also include non-aqueous liquid or solid particles such as dust and smoke [5]. Clouds are created in atmosphere as a result of condensation of water vapor. Condensation starts when super-saturation is reached and there are enough cloud condensation nuclei (CCN) in the air. Water vapor starts to condensate on CCN and water droplets are formed. Usual cloud drop has a size between 1 µm and 100 µm, which is smaller than average raindrop. Size of cloud drops is enhanced with condensation, but even though growth by condensation 3

16 is fast at the beginning it diminishes with time. At this point growth by collection starts to lead. To get big collector drops, giant CCNs are needed. They act as embryos for collector drops. Bigger droplets start to fall faster than smaller ones. Big collector drops collide with smaller ones and if coalescence occurs (small rain drop does not bounce o), big collector drop becomes even bigger. Under 0 C most of water vapors still condensate as super-cooled water, except in case of CCN with similar crystal structure as ice where formation of ice crystals occurs. At the temperature around -20 C there is almost no liquid water any more. Ice particles can be created by deposition nucleation (already starts as ice particles), homogeneous nucleation from pure droplet, or by freezing at the contact of impurity (heterogeneous ice nucleus) and water droplet [6]. According to the classication by the World meteorological organization (WMO) there are 10 genera of clouds [5]. These genera are then subdivided to species and varieties. They are divided in three etages: low, middle and high, as shown in Fig Limits of the etages depend on latitude, closer to equator, higher are the etages. Limits of etages are not strictly determined and may overlap. The denitions of cloud genera are taken from WMO International cloud atlas [5]. Figure 2.1: Genera of clouds divided in three etages (obtained from [7]). Low clouds reach altitudes up to about 2 km. Those clouds are made of water droplets and are created at temperatures between 0 C and -10 C. The corresponding genera are: Stratocumulus (Sc) is gray or whitish, or both gray and whitish, patch, sheet or layer of cloud which almost always has dark parts, composed of tessellations, rounded masses. rolls, etc., which are non-brous (except for virga) and which may or may not be merged; most of the regularly arranged small elements have an apparent width of more than ve degrees. Stratus (St) is generally gray cloud layer with fairly uniform base, which may give drizzle, ice prisms or snow grains. When the sun is visible through 4

17 the cloud, its outline is clearly discernible. St does not produce halo phenomena except, possibly, at low temperatures. Sometimes St appears in the form of ragged patches. Middle clouds are in temperate regions at altitudes between 2 km and 7 km. With temperatures between -10 C and -35 C, middle clouds are usually mixed clouds, whose major state usually depends on current situation. These are: Altostratus (As, Fig. 2.2) is grayish or bluish cloud sheet or layer of striated, brous or uniform appearance, totally or partly covering the sky, and having parts thin enough to reveal the sun at least vaguely, as through ground glass. As does not show halo phenomena. Figure 2.2: Altostratus, Okre²elj, February 27, Figure 2.3: Cirrus, Cima di Terrarossa, September 27,

18 Altocumulus (Ac) is a white and gray or both white and gray, patch, sheet or layer of clouds, generally with shading, composed of laminae, rounded masses, rolls, etc., which are sometimes partly brous or diuse and which may or may not be merged; most of the regularly arranged small elements usually have an apparent width of between one and ve degrees. Nimbostratus (Ns, Fig. 2.4) is a gray cloud layer, often dark, the appearance of which is rendered diuse by more or less continuously falling rain or snow, which in most cases reaches the ground. It is thick enough throughout to blot out the sun. Low, ragged clouds frequently occur below the layer, with which they may or may not merge. Figure 2.4: Nimbostratus, koje, March 6, High clouds are in temperate regions at altitude between 5 km and 13 km. Because temperatures at these heights are usually below -35 C, the following clouds are ice clouds: Cirrus (Ci, Fig. 2.3) is visible as detached clouds in the form of white, delicate laments or white or mostly white patches or narrow bands. These clouds have a brous (hair-like) appearance, or a silky sheen, or both. Cirrocumulus (Cc, Fig. 2.5) is a thin, white patch, sheet or layer of cloud without shading, composed of very small elements in the form of grains, ripples, etc., merged or separate, and more or less regularly arranged; most of the elements have an apparent width of less than one degree. Cirrostratus (Cs, Fig. 2.6) is a transparent, whitish cloud veil of brous (hair-like) or smooth appearance, totally or partly covering the sky, and generally producing halo phenomena. Clouds with vertical development are not limited to one etage only. They start in one etage and during time expand also to others. Because of the size they can be made of water droplets, ice particles and mixed part all at the same time. These clouds are: 6

19 Figure 2.5: Cirrocumulus, koje, August 18, Figure 2.6: Cirrostratus, Tamar, March 27, Cumulus (Cu, Fig. 2.7) is visible as detached clouds, generally dense and with sharp outlines, developing vertically in the form of rising mounds, domes or towers, of which the bulging upper part often resembles a cauliower. The sunlit parts of these clouds are mostly brilliant white; their base is relatively dark and nearly horizontal. Sometimes Cumulus is ragged. Cumulonimbus (Cb, Fig. 2.8) is a heavy and dense cloud, with a considerable vertical extent, in the form of a mountain or huge towers. At least part of its upper portion is usually smooth, or brous or striated, and nearly always attened; this part often spreads out in the shape of an anvil or vast plume. Under the base of the cloud which is often very dark, there are frequently low ragged clouds either merged with it or not, and precipitation sometimes in the form of virga. 7

20 Figure 2.7: Cumulus humilis, Paklenica, April 26, Figure 2.8: Cumulonimbus, koje, July 16, Clouds are moving together with air masses, so they can in principle be used as tracers for air-mass motion. However, not all genera of clouds are useful for this purpose. Tracer clouds have to have separate elements and/or sharp features, that can be marked and followed. Best are small cumulus clouds such as presented in Fig. 2.7, or clouds that accompany other clouds, for example nimbostratus or cumulonimbus. Those clouds are the so called pannus clouds. In the case of pannus clouds, we have a tracer only for a short time, because they move fast and change shape rapidly. In any case, clouds stay at same height for a longer period of time than any radiosonde can. Determining their height and measuring their speed can provide the speed of movement of air-mass motion at that height with good time resolution, as long as useful tracer clouds are available. 8

21 2.2 Wind As any other object with a mass, a mass element of air is exposed to dierent forces when accelerated [1]. Due to Earth's rotation, the reference system for describing the movement of the mass elements of air has to be noninertial. The apparent forces must be introduced to compensate the acceleration of system (in this case centrifugal and Coriolis force). While centrifugal force aects all bodies, Coriolis force depends on their relative velocity and can be described as f d = 2 Ω v, (2.1) where Ω is angular velocity and v velocity in relative system on rotating Earth. Resulting vector from the equation (2.1) is rotated for 90 to the right from v (Northern hemisphere), showing that Coriolis force changes direction of motion of a body. As mentioned before, centrifugal force is experienced by all bodies located in rotting frame. It is strongest when it is perpendicular to the axis of rotation, on Earth this happens at equator. It is written as f c = Ω 2 R, (2.2) where R is a distance from axis of rotation. Centrifugal force opposes gravitational force. Gravitational force appears between bodies with mass and it is shown as attraction between them. Force is proportional to masses of bodies and inversely proportional to the square of distance between them F g = κmm r r 3, (2.3) where κ is gravitational constant, M mass of the Earth, m mass element of air and r distance that separates them. From this we can get gravitational attraction of Earth F g m = g = κm r r 3. (2.4) Combining equations (2.2) and (2.4), we can calculate the sum of centrifugal force and gravitational force, which gives us specic weight from f = g + f c = κm E r r 3 + Ω2 R. (2.5) Pressure is a force per unit area across a surface. Dierence in pressure means dierence in forces on surface. The resulting net force is pressure-gradient force f g = 1 p, (2.6) ρ where p is pressure gradient and ρ density of the air. Pressure gradient force is proportional to the gradient of the pressure and points in the direction of the maximum downward pressure. Pressure gradient is the largest in vertical direction and is roughly four orders of magnitude greater than in horizontal direction. Movement of the air is accompanied with drag. Near the ground in the laminar ow, drag is similar to the drag of solid bodies and depends on velocity and the roughness of the terrain as f t k v. (2.7) 9

22 The description in equation (2.7) is true, when wind is horizontal and homogeneous. But when the ow is turbulent, it is necessary to use description of ow for viscous liquids as f t = K m 2 v. (2.8) K m stands for turbulent diusion. It represents mixing of fast and slow air ows and it is property of ow, not of the liquid. More turbulent is the ow, bigger is K m. Combining all the above mentioned forces together we get the equation of motion for a rotating Earth as d v dt = 1 ρ p 2 Ω v + f + f t. (2.9) Pressure dierence between two places starts the movement, and air starts to ow from higher to the lower pressure. But the airow is deected, because Earth's rotation creates Coriolis eect on moving air mass and so air masses are moving perpendicular to the gradient of the pressure. Dierence in pressure is created due to the dierence in solar irradiation. Pressure gradient force is the one that starts the movement of the air, but it also determines shape and direction of movement of the wind. Here we get two cases of stationary winds, geostrophic and gradient wind. Geostrophic wind happens, if isobars are straight and parallel, so pressure gradient has same direction everywhere. If velocity of the wind doesn't change with time (stationary wind), and friction is negligible (high above the ground), only equilibrium between pressure gradient force and Coriolis force remains. Air mass is moving parallel to the isobars, having (on the northern hemisphere) low pressure on the left side, if we are looking in direction of movement. Gradient wind is still stationary wind, but has bent isobars, giving circular gradient eld. Acceleration (due to a circular motion of air) and both forces (pressure gradient force and Coriolis force) act on air perpendicularly to the direction of velocity. When there is only slight curvature of isobars, there is almost no dierence between gradient and geostrophic wind [1, 6]. Wind elds are computer simulations of wind propagation through time over specic area. Calculation of the wind elds is a part of numerical modeling method. Measurements are used as input data for complex prediction models, that run on supercomputers. Models with higher resolution and ner scale are usually used for weather predictions for smaller area. Resolution and accuracy of the forecast depends on the input data and a type of model, that was used [8]. The strongest wind in Slovenia is Bora wind. It appears in south-west part of Slovenia and along the Croatian coast line. Bora is a gusty wind that blows from the north to northeast and is created when air-masses cross a mountain barrier. When cold air crosses the barrier, it descends down the slopes and accelerates. If there is enough temperature dierence between the cold air above and warm air below, the warming of the descending air is not sucient and the descending air mass is always cooler and thus heavier then the surrounding air. Strong Bora winds are more frequent in the winter, when the temperature dierence between air masses behind the barrier and at its feet is the largest. Average Bora wind speeds are about 10 m/s with much stronger aperiodic wind gusts. In strong Bora events, wind speeds can exceed 50 m/s [9, 10, 11]. 10

23 Figure 2.9: Wind chart of Adriatic Sea and its surroundings (obtained from [12]). 2.3 HYSPLIT Model Air mass trajectories anywhere in the world can be modeled reasonably well using established complex numerical modeling tools such as the HYSPLIT model [13, 14], developed by the Air Resources Laboratory of the National Oceanic and Atmospheric Administration, USA. Figure 2.10: HYSPLIT backward trajectories for air masses above Ajdov² ina on Mar. 17, 2016, during a Bora event. Di erent colors denote di erent heights. Red trajectory corresponds to Bora ow from the north-east, while trajectories in higher layers show air mass streaming from the south. 11

24 While based on the same principles as presented in Sect. 2.2, HYSPLIT uses a hybrid calculation method, incorporating the Lagrangian approach with a moving frame and the Eulerian methodology with a xed three-dimensional grid as a frame of reference. HYSPLIT among its other features provides backward trajectory analysis to determine the origin of air masses. In our case, we can use it to estimate wind direction at the cloud base height at the time of the remote sensing measurement. The model was run interactively using the ARL READY web interface ( An example of HYSPLIT backward trajectories is shown in Fig

25 3 Experimental setup and measurement techniques Standard device for wind measurements is an anemometer, which provides a point measurement of wind speed and direction, and is limited to the near ground area. There are two possibilities to measure processes in the atmosphere high above the ground. One is to send instruments, such as radiosondes, to the location where atmospheric parameters need to be known, the other is to use remote sensing techniques. Radiosondes are usually released at pre-determined locations once or twice a day and they only stay at any given height for a very short time, so they have poor spatial and temporal resolution. A way to resolve resolution problems is the use of remote sensing, where you repeatedly measure the eects of the interaction of the emitted signal with the atmosphere along its path. Dierent types of ground remote sensing instruments can be distinguished based on the type of the emitted signal. Instruments commonly used for remote sensing in meteorology are weather radar (RAdio Detection And Ranging), radar wind proler, RASS (Radio Acoustic Sounding System), SODAR (SOund Detection And Ranging) and lidar (LIght Detection And Ranging). Both radars and lidars use electromagnetic signals, but due to the dierence in their wavelengths they are used for dierent purposes. Radars, which generally operate in wavelength range of about 10 cm, are used to monitor precipitation type and intensity, and to track storms. Lidars use much smaller wavelengths (generally between 355 nm and 1064 nm) and can detect much ner atmospheric structures, such as aerosols. SODAR uses sound scattering to detect wind turbulent uctuations. Acoustic waves are emitted in dierent directions and three-dimensional movements of air are calculated from Doppler frequency shifts of the detected backscattered signals. SODAR is generally used as a wind proler. Similarly as SODAR works radar wind proler. It transmits pulses of EM radiation vertically and in at least two nearly vertical directions. Three-dimensional air movement is calculated from detected, backscattered, Doppler shifted signal. Combining radio and sound system gives us RASS. Acoustic sources emit sound into the vertical EM beam produced by radar. The frequency of sound signal varies, and at the points where has EM signal twice the wavelength of the sound, the Bragg scattering occurs. From this speed of sound as a function of altitude, from which virtual temperature can be calculated. 3.1 Anemometer The most basic tool for wind measurements is an anemometer. In an ultrasonic anemometer, wind speed and direction is calculated from the Doppler shifts of sound pulses between the ultrasonic emitters and detectors. Three detectors are usually used for two-dimensional reconstruction of the wind vector and six detectors for three-dimensional reconstruction. For this thesis a twodimensional ultrasonic anemometer Vaisala WMT702 was used. It was mounted on the rooftop 13

26 Figure 3.1: Vaisala WMT702 anemometer, mounted on the rooftop of the University of Nova Gorica building in Ajdov² ina about 10 m above the ground provides horizontal wind speed and direction data (obtained from [15]). Table 3.1: Main operational parameters of the Vaisala WMT702 anemometer used for wind speed and direction monitoring at the University of Nova Gorica (obtained from [16]). Ultrasonic 2D anemometer Wind speed range Speed resolution Speed accuracy Direction resolution Direction accuracy Sampling rate Vaisala WMT m/s 0.01 m/s ±0.1 m/s or 2% of reading, whichever is greater 0.01 ±2 1 s of the University of Nova Gorica building in Ajdov² ina (Fig. 3.1) approximately 10 m above the ground, and operated with a sampling rate of 1 s. 3.2 Radiosonde Package of instruments that is carried in the atmosphere by a weather balloon is called a radiosonde (Fig. 3.2). It is transmitting values of atmospheric variables such as pressure, temperature and relative humidity, and as well the GPS position, from which wind speed and wind direction are calculated (shown in Table 3.2). Radiosondes are launched from same location at the same time every day. Usually there are one ore two launches per day (Ljubljana at 4:30 UTC, Udine at 00:00 UTC and 12:00 UTC). The balloon that is carrying a radiosonde is lled with helium or hydrogen. It starts to rise because of buoyancy, but as it rises, the atmospheric pressure outside of the balloon is reduced, so it expands until burst. Balloon rises with a speed of about 5 m/s and reaches height of approximately 35 km with a drift of up to 300 km from the release point. Radio bands for radiosondes, as set 14

27 by international agreements, range from 400 MHz to 406 MHz and from 1675 MHz to 1700 MHz [17, 18]. Data obtained with the atmospheric sounding is then Figure 3.2: Weather balloon carrying a radiosonde (obtained from [18]). presented in graphs such as Fig Figure 3.3: Radiosonde measurements of the vertical proles of temperature, moist and wind over Ljubljana up to the height of approximately 3 km. Blue line represents temperature of air and red line dew point temperature. Large dierence between blue and red line implies small relative humidity. The lines usually intersect at or very close to a layer of clouds. Wind prole is shown at the right margin of the graph, where its direction and speed is indicated with wind barbs (obtained from [20]). 15

28 Table 3.2: Example of a data set obtained from radiosonde measurements in Ljubljana on Feb. 19, at 5:30 LT. PRES is the pressure, HGHT is height above sea level, TEMP is temperature, DWPT is dew point temperature, RELH is relative humidity, MIXR is mixing ratio between mass of water vapor and mass of dry air, DRCT is the wind direction (azimuth), SKNT is the wind speed (obtained from [19]). For the purpose of this thesis only wind speed and direction were used. PRES HGHT TEMP DWPT RELH MIXR DRCT SKNT hpa m C C % g/kg knot Lidar The distance from the observer to the cloud was measured using lidar, a remote sensing device that measures backscattering of emitted laser pulses on aerosols and molecules along the emitted direction. This detected lidar signal is described with a singlescattering lidar equation as P (r) = P 0 k cτ 0 2 Aβ(r) r 2 T 2 (r), (3.1) where P (r) is a received backscattered power from distance r, P 0 is a power transmitted by laser, k is system eciency, c speed of light, τ 0 pulse duration, A eective area of the detector, T (r) describes how much light gets lost while traveling to distance r and back, β(r) is a backscattering coecient. β(r) is used to determine the strength of the signal. A basic lidar consists of emitter (pulsed laser) and a receiver (a telescope) connected to a light detector, amplier, digitizer and a computer that collects data (Fig. 3.4). The main advantage of a lidar is good temporal and spatial resolution of its measurements. In this study we used the mobile Mie scattering lidar, constructed at the University of Nova Gorica (Fig. 3.5). Lidar uses Nd:YAG pulsed laser with a wavelength of 1064 nm and a pulse repetition rate of 10 Hz. Receiver of backscattered light is Newtonian telescope with diameter of 300 mm. Detector is an avalanche photodiode. Received measurements are an average over 10 16

29 pulses, yielding spatial resolution of 3.75 m and temporal resolution of 1 s. In clear weather conditions, the maximum detectable range of this device is approximately 10 km [15, 21]. Figure 3.4: Schematic diagram of the infra-red mobile lidar used to measure the cloud range (obtained from[21]). It consists of a receiver (telescope), bi-axial transmitter (laser) and spectroscopic lter (Zoom-in A), used to separate dierent types of the backscatter signal. Figure 3.5: Mobile Mie scattering lidar of the University of Nova Gorica with accompanying data acquisition system as used in the eld (obtained from [15]). 3.4 Cameras Information on the dynamics of cloud movement was retrieved from a series of optical photographs obtained from two dierent cameras. Cameras were oriented 17

30 Table 3.3: A summary of main parameters of the cameras, used for wind speed measurements (obtained from [22, 23]). Camera type Canon EOS 1000D MAGINON IPC-20C Resolution Horizontal aperture angle (β) Vertical aperture angle (α) Horizontal angular resolution /pixel /pixel Vertical angular resolution /pixel /pixel Temporal resolution 15 s 60 s in the same direction as lidar as shown in Fig. 3.6, so that consequent photos cover the same area. The picture rate must be high enough to get multiple pictures of the same cloud structure moving across the eld of view of the camera. To trace cloud motion, a tracer is manually positioned at a distinctive cloud feature in each of the photos in the sequence (Fig 3.7). As the marker position is set manually, the positioning error in pixels was estimated for each investigated case (each series of cloud photos) separately. The error is aected by both the repeatability of manual marker positioning as well as temporal persistence of the cloud shape. We estimated the statistical marker positioning error as the variance of a series of independent marker placements on the same cloud feature on the same photo. The obtained variance of dn pixels is further used in the wind retrieval procedure. In our measurements, two dierent types of cameras were used. Their properties are summarized in Tab Wind speed retrieval technique Wind speed at the cloud base height was extracted using lidar and camera data, under the assumption that clouds move together with the air masses and that during the measurement, the distance R from the experimental setup to the cloud did not change considerably. In the time series of photographs taken with the cameras, a prominent cloud feature was selected and marked with a tracer (Fig. 3.7). The distance from the experimental setup to the tracer point was determined from lidar measurements. In the photos, the reference frame (x, y) was selected so that its initial point was in the upper left corner of a photo, x-axis pointing to the right and y-axis pointing down. Relative position of a marker in the picture frame was calculated as x(t) = 2 n x(t) N x R tan β/2 y(t) = 2 n y(t) N y R tan α/2, (3.2) where N x N y is the resolution and α, β are horizontal and vertical aperture angles of the camera, R is the distance from lidar to the cloud and n x and n y are the pixel coordinates of the marker (Fig. 3.6). Positioning errors can be obtained 18

31 Figure 3.6: Side view of the study area with lidar laser beam shown as thick red θ is the lidar elevation angle, h the β and α are horizontal and vertical apertures of the line and camera eld of view shown in yellow. height of the observed object, camera, and W is the horizontal distance at range R. Yellow arrow denotes the North and φ is the azimuth angle of the lidar beam. as the total derivative of the position, 2 2 2nx 2R 2 tan β/2 δnx + tan β/2 δr2, δx = Nx Nx 2 2 2R 2ny 2 2 δy = tan α/2 δny + tan α/2 δr2. Ny Ny 2 Statistical marker positioning errors δnx and δny (3.3) were estimated for each series of photographs separately on a sample of ten independent positioning trials. The range uncertainty δr was determined from the variance of the cloud range during the observed period. As the marker in general moved both in x and in y direction, the marker position and the corresponding uncertainty were taken as s = p x2 + y 2, δs = p δx2 + δy 2. (3.4) In the reference frame of the photograph, the 2D wind speed and direction can thus be described as s v = dx dt γ = arctan 2 + dy dt 2 vy. vx = q ds vx2 + vy2 =, dt (3.5) 19

32 Figure 3.7: Sequence of photos used to determine wind speed at cloud base height, taken with Canon EOS 1000D camera on Feb. 19, In the top left corner of each photo is time and in the bottom right corner a scale that represents 5. Distance to the cloud was 1900 m. Red markers at a prominent cloud feature were manually added to the photos. Our data is N tracer coordinates with t long time intervals between them. Path traveled and its uncertainty can be calculated from data as s i = (x i x 1 ) 2 + (y i y 1 ) 2, δs i = 2 2 s i (xi x 1 ) 2 δx 2 + (y i y 1 ) 2 δy 2, (3.6) where x 1 and y 1 are positions of rst marker. From data, we can calculate average speed for every time interval between two photos as v xi = x i+1 x i t v yi = y i+1 y i t (3.7) If we assume δx i δx and δy i δy within the same series of photos, the error for the speed is written as ( ) 2 ( ) 2 ( ) 2 δxi+1 δxi δx δv xi = + = 2, t t t ( ) 2 ( ) 2 ( ) 2 δyi+1 δyi δy δv yi = + = 2, (3.8) t t t 20

33 giving us, δv xi = 2 δx t and δ yi = 2 δy. (3.9) t Average speed in the plane of photo and its error are calculated as v i = v xi2 + v yi2, δ v i = 2 2 vxi2 δx v i t 2 + v yi2 δy 2. (3.10) In order to be able to compare these results with other measurements, the wind speed and direction need to be transformed in the ground reference frame. Our ground reference frame was chosen so that its x g axis was pointing east (E), y g axis was pointing north (N) and z g axis was pointing up, its origin being collocated with our experimental setup. The obtained wind speed vector was transformed from the picture frame with two consecutive rotations around x g and z g axis as v xg cos ϕ sin ϕ v x v yg = sin ϕ cos ϕ 0 0 cos θ sin θ v y, (3.11) v zg sin θ cos θ 0 where ϕ is the azimuth angle and θ = 90 θ is the zenith angle of the lidar beam. Wind direction with respect to the ground can be obtained as a scalar product of the transformed wind speed vector and the direction of the north as Λ = arccos v yg v g. (3.12) After the transformation between reference frames, wind azimuth angle can be expressed in terms of the velocities in picture frame as Λ = arccos v x sin ϕ + v y cos ϕ cos θ v 2 x + v 2 y cos 2 θ. (3.13) Lidar data was processed using a custom data acquisition software developed by the Center for atmospheric research. It is based on C++ and CERN Root package. Marker position data obtained from the photos was processed with an original code in Python. 21

34 22

35 4 Case studies To verify the performance of the remote sensing wind speed measurements, four test case studies were performed. All the measurements were performed in Ajdov² ina (45 53' 7 N, 13 54' 47 E) using mobile Mie scattering lidar for cloud ranging and an optical camera, covering the same part of the sky, for the extraction of the dynamics of cloud motion. Measurements were performed at dierent times of day, under dierent meteorological conditions. Radiosonde data from two closest launching points, Ljubljana and Udine were also retrieved [19] for the launches as close in time to the remote sensing measurements as available and are summarized in Tab No radiosonde data for the exact times of the lidar and camera measurements exist. Table 4.1: Radiosonde wind data from Ljubljana and Udine for the four considered test cases in Feb. and Mar No radiosonde data for the exact times of the lidar and camera measurements exist. Date Ljubljana Udine Time Height Speed Direct. Time Height Speed Direct. LT m m/s LT m m/s Feb. 19 5: : Mar. 16 5: : Mar. 17 5: : Mar. 30 6: : Table 4.2: Beginning and duration of measurement for the four considered test cases in Feb. and Mar. 2016, along with the azimuth of lidar beam and its elevation. Date Time Duration Azimuth Elevation s Feb : Mar : Mar. 17 7: Mar : The details on the experiment conguration, including its duration and the direction of the lidar, are summarized in Tab Temporal and angular resolution of the measurement depended on the type and resolution of the camera used. Canon EOS 1000D with resolution acquired pictures in 15 s intervals while the acquisition rate of MAGINON IPC-20C with resolution ) was once 23

36 every 60 s. The details of cloud range and height for each investigated case, as well as the positioning and ranging uncertainties originating from lidar and camera measurements are summarized in Tab Table 4.3: Cloud height and range and the uncertainties of lidar measurements and marker positioning for the four considered test cases in Feb. and Mar Range and range uncertainty were obtained from lidar data while the path uncertainty is dominated by the marker positioning error. The rst case is exception, where domination of range uncertainty is seen as change of error in time. Date Time Height Range δr Resolution δn δx δy δs LT m m m m m m Feb : Mar : Mar. 17 7: Mar : Case 1 - Pannus cloud The measurements were performed on Feb. 19, 2016, between 12:49 and 13:19 local time (LT). The sky was completely covered with clouds (Fig. 4.1), where the main cloud base was at height of 2250 m and the monitored pannus clouds below were at height of 1100 m. Within the observation window of 180 s the pannus cloud range varied for about 14% (Fig. 4.2) as the clouds changed rapidly. Relative marker position with respect to the rst measured location and average wind speed between successive measurements are shown in Fig The cloud traveled about 700 m in 165 s, giving us the average speed of approximately 3.9 m/s. Figure 4.1: Cloud coverage above Ajdov² ina on Feb. 19, 2016 at 12:55 LT. A number of pannus clouds were present at about 1100 m, below the main cloud layer at 2250 m. 24

37 Figure 4.2: Time series of logarithm of range corrected lidar return signal for Ajdov² ina Feb. 19, Range is measured from the lidar site. Time interval with the availability of camera data is marked with vertical red lines. Figure 4.3: Temporal dependence of marker position and velocity for measurement starting at 12:55:30 LT on Feb. 19, Height of the clouds was 1220 m above sea level. Position is shown relative to the location of the rst marker. For comparison, radiosonde data from Ljubljana (1441 m above sea level, taken at 5:30 LT) is added in green and radiosonde data from Udine (1178 m above sea level, taken at 13:00 LT) is added in blue. 4.2 Case 2 - Altocumulus The measurements were performed on Mar. 16, 2016, between 13:00 and 15:00 LT, during the rst day of a two day Bora wind episode. The base of the observed clouds was at 3555 m, where the cloud range within the observation window varied for about 0.1% (Fig. 4.6). Relative marker position with respect to the rst measured location and average wind speed between successive measurements are 25

38 shown in Fig The cloud traveled about 5000 m in 480 s, giving us the average speed of approximately 10.1 m/s. Figure 4.4: Cloud coverage above Ajdov² ina on Mar. 16, 2016 at 14:10 LT, during the rst day of a two day Bora episode. Figure 4.5: Temporal dependence of marker position and velocity for measurement starting at 14:10:19 LT on Mar. 16, Height of the clouds was 3675 m above sea level. Position is shown relative to the location of the rst marker. For comparison, radiosonde data from Ljubljana (2974 m above sea level, taken at 5:30 LT) is added in green and radiosonde data from Udine (3759 m above sea level, taken at 13:00 LT) is added in blue. 26

39 Figure 4.6: Time series of logarithm of range corrected lidar return signal for Ajdov² ina Mar. 16, Range is measured from the lidar site. Time interval with the availability of camera data is marked with vertical red lines. 4.3 Case 3 - Altocumulus The measurements were performed on Mar. 17, 2016, between 00:00 and 08:00 LT, during the second day of a two day Bora wind episode. The base of the observed clouds was at 2231 m, where the cloud range within the observation window varied for about 0.7% (Fig. 4.8). Relative marker position with respect to the rst measured location and average wind speed between successive measurements are shown in Fig The cloud traveled about 4100 m in 600 s, giving us the average speed of approximately 6.9 m/s. Figure 4.7: Cloud coverage above Ajdov² ina on Mar. 17, 2016 at 7:00 LT, during the second day of a two day Bora episode. 27

40 Figure 4.8: Time series of logarithm of range corrected lidar return signal for Ajdov² ina Mar. 17, Range is measured from the lidar site. Start and end of the used data is marked with vertical red lines. White gap implies there is no lidar data. Figure 4.9: Temporal dependence of marker position and velocity for measurement starting at 7:00:21 LT on Mar. 17, Height of the clouds was 3370 m above sea level. Position is shown relative to the location of the rst marker. For comparison, radiosonde data from Ljubljana (3065 m above sea level, taken at 5:30 LT) is added in green and radiosonde data from Udine (2308 m above sea level, taken at 13:00 LT) is added in blue. 28

41 4.4 Case 4 - Nimbostratus The measurements were performed on Mar. 30, 2016, between 09:30 and 17:30 LT. The base of the observed clouds was at 1806 m, where the cloud range within the observation window varied for about 0.3% (Fig. 4.11). Relative marker position with respect to the rst measured location and average wind speed between successive measurements are shown in Fig The cloud traveled about 4000 m in 360 s with an average speed of approximately 11.3 m/s. Figure 4.10: Cloud coverage above Ajdov² ina on Mar. 30, 2016 at 10:00 LT. Figure 4.11: Time series of logarithm of range corrected lidar return signal for Ajdov² ina Mar. 30, Range is measured from the lidar site. Start and end of the used data is marked with vertical red lines. White gap implies there is no lidar data. 29

42 Figure 4.12: Temporal dependence of marker position and velocity for measurement starting at 10:00:06 LT on Mar. 30, Height of the clouds was 1926 m above sea level. Position is shown relative to the location of the rst marker. For comparison, radiosonde data from Ljubljana (1494 m above sea level, taken at 6:30 LT) is added in green and radiosonde data from Udine (1761 m above sea level, taken at 14:00 LT) is added in blue. 30

43 4.5 Results Wind speed and direction obtained in the four investigated cases are summarized in Tab They are in relatively good agreement with radiosonde data from Ljubljana and Udine, where the known reasons for the disagreement include spatial and temporal dislocation of remote sensing experiment and radiosonde launching grounds as well as poor spatial and temporal resolution of radiosondes. Table 4.4: Comparison of the results obtained with remote sensing technique and data received from radiosonde sounding in Ljubljana and Udine. Date Remote sensing Ljubljana Udine Speed Direct. Speed Direct. Speed Direct. m/s m/s m/s Feb ± Mar ± Mar ± Mar ± Comparing wind properties obtained from anemometer measurements 10 m above the ground to wind speed and direction in higher layers (Tab. 4.5), we can see that both direction and speed change change substantially with height. The dierence in direction is especially notable during the Bora episode (Mar. 16 and Mar. 17, 2016). In both layers we observed winds with comparable speeds, but the observed wind directions were almost perpendicular to each other. The observation is also in agreement with the HYSPLIT backward trajectory prediction for Ajdov² ina for the corresponding days (Fig. 4.13). During the Bora episode HYSPLIT trajectories at low altitudes arrive to Ajdov² ina from the north north-east, while at higher altitudes they arrive from the south-east. Another conrmation is the observation of Kelvin-Helmholtz clouds above Ajdov² ina on Mar. 16, 2016 [24], which occurred due to the shear between atmospheric layers with dierent velocities, directions and densities. Table 4.5: Comparison of remote sensing results with anemometer data from Ajdov² ina for corresponding time intervals. Date Measurements Anemometer Time Height Speed Direct. Speed Direct. LT m m/s m/s Feb : ± Mar : ± Mar. 17 7: ± Mar : ±