Deriving characteristics of thin cirrus clouds from observations with the IRF lidar

(1)

from observations with the IRF lidar

Jennifer Edman

Civilingenjör, Rymdteknik 2019

Luleå tekniska universitet Institutionen för system- och rymdteknik

(2)

Abstract

Cirrus clouds play an important role in radiative transfer, and thus have impact on the energy balance of the atmosphere and the climate of the Earth. Furthermore, they occur often and cover large areas globally at any time. Nevertheless, cirrus clouds are poorly studied, especially in the polar regions. Cirrus clouds are present in a large amount of the 14 years of data produced by the lidar at the Swedish Institude of Space Physics (IRF), but has not been studied to a large extent.

A lidar is an active remote sensing instrument using a laser. This master’s thesis develops and improves programs for analysis of cirrus clouds from this lidar data. It also performs analysis of six case studies chosen from the available data, and statistics of these six cases.

The parameters calculated for each date are the cloud top, base and mean altitude, the geometrical thickness, the depolarisation ratio, the backscatter ratio (BSR), the backscatter coefficient, the extinction coefficient, the optical thickness and the number of cloud layers. No clear correlation between the optical thickness and the cloud top, base or mean altitude was found. There seems to be a weak correlation between increased optical thickness and increased geometrical thickness, which is not unreasonable. The mean cloud layer top altitude was 11.82 km and the mean cloud base was 10.36 km. The mean optical thickness for a cloud layer was 1.46 km, and the average of the cloud layer mean altitude was 11.09 km. It should be noted that the statistical analysis is based on only six cases with a total observation time of no more than 37 hours. A far larger dataset is needed in order to obtain any statistically significant conclusions. The effect of averaging is studied, and it is concluded that averaging over altitude reduced the noise and facilitated the interpolation more than averaging over time did.

Different approaches to obtain the molecular backscatter coefficient are compared, as well as the effect on the simulated molecular signal. Two of these approaches calculate the molecular backscatter coefficient with input of the temperature and pressure either as continuously measured ground vales from the weather station at IRF or as radiosonde profiles for a specific time. In the other two, the molecular backscatter coefficient is obtained from ECMWF data and from the standard atmosphere. Differences in the range 12-35% between the methods are found.

Different approaches to calculate the backscatter ratio (BSR) are also compared. At cirrus altitudes, the decrease in the signal due to the molecular cloudfree part of the atmosphere is still strong, and finding the top and base separately by comparison with the standard deviation of the signal is proven a better method than interpolating between the point where the signal starts to increase and the point where it reaches the same signal value again. Height-normalising the signal provides a more robust method.

For thin cirrus, the signal is not significantly attenuated above the cloud layer, and it is found that a method based on the ratios between the measured signal and the simulated molecular signal at cloud top and base did not produce reliable results for the optical thickness.

In addition to analysing data and data processing methods, new data processing tools in MAT- LAB have been developed and existing functions have been improved. These will be valuable for continued studies with the IRF lidar, for cirrus as well as PSCs and thick and/or low-altitude clouds.

(3)

Acknowledgements

I would like to thank my supervisors and examiner, Peter V¨olger at the Swedish Institute of Space Physics (IRF), Veronika Wolf at Lule˚a University of Technology (LTU) and Thomas Kuhn at LTU, for their support during the thesis work and their feedback on this report. I am grateful for getting the opportunity to to do this master’s thesis at IRF and LTU. I would also like to thank the teachers and lecturers at Lule˚a University of Technology, in particular Mathias Milz who lectured one fourth of the courses I attended. Over the course of these last five years, I’ve learnt a lot. I also wish to thank Maria Winneb¨ack, programme administrator, for making my time at the university easier.

I would like to thank my family for believing in me and supporting me, as well as my friends – old and new –, in particular Lisa and Emma. Lastly, I wish to thank William, for everything.

(4)

Contents

List of figures v

List of tables viii

Acronyms ix

Glossary ix

1 INTRODUCTION 1

1.1 Motivation . . . . 1

1.2 Thesis Aim . . . . 1

1.3 Objectives . . . . 1

1.4 Thesis outline . . . . 2

2 THEORY 3 2.1 Atmospheric concepts . . . . 3

2.2 Cirrus . . . . 6

2.2.1 General description . . . . 6

2.2.2 Classification . . . . 6

2.2.3 Effect on radiative budget . . . . 7

2.2.4 Occurrence . . . . 8

2.2.5 Ice crystal shapes . . . . 9

2.2.6 Formation . . . . 10

2.3 Lidar . . . . 12

2.3.1 General description . . . . 12

2.3.2 System description . . . . 13

2.3.3 Lidar equation . . . . 14

2.3.4 Lidar techniques . . . . 16

2.3.5 Multiple scattering . . . . 17

2.4 Observations of cirrus . . . . 18

2.4.1 Instrument types . . . . 18

2.4.2 Lidar . . . . 18

2.4.3 Previous studies of cirrus . . . . 19

2.4.4 Research relevance of cirrus . . . . 20

2.5 This study . . . . 21

2.5.1 Location and weather . . . . 21

2.5.2 IRF’s lidar . . . . 21

3 METHOD 23 3.1 Data description . . . . 23

3.2 BSR method . . . . 24

3.3 Data processing . . . . 26

3.3.1 show data . . . . 29

3.3.2 beta mol . . . . 33

3.3.3 molecular signal . . . . 35

3.3.4 calculate bsr . . . . 35

(5)

3.3.5 plot all data . . . . 36

3.3.6 compare averaging . . . . 38

4 RESULTS 39 4.1 Case studies . . . . 39

4.1.1 Selection . . . . 39

4.1.2 2010-01-12 . . . . 40

4.1.3 2011-03-28 . . . . 46

4.1.4 2012-12-21 . . . . 51

4.1.5 2014-01-22 . . . . 57

4.1.6 2017-01-23 . . . . 62

4.1.7 2017-01-27 . . . . 68

4.1.8 Discussion . . . . 74

4.2 Statistics . . . . 78

4.3 Comparison of averaging . . . . 84

4.4 Approaches to molecular signal . . . . 89

4.5 BSR algorithms . . . . 92

4.6 Layer boundaries algorithms . . . . 95

4.7 Approaches to optical thickness . . . . 97

4.8 Saving algorithms . . . . 100

4.9 Discussion . . . . 101

5 CONCLUSIONS 104 5.1 Summary . . . . 104

5.2 Future work . . . . 106

6 References 108

(6)

List of figures

2.1 The US standard atmosphere of 1962. . . . 3

2.2 Example of blackbody curves. . . . 4

2.3 The electromagnetic spectrum. . . . 5

2.4 Examples of the different ice crystal shapes. . . . 10

2.5 Sketch of the set-up of a lidar system. . . . 12

2.6 Comparison of photon-counting and analog channels for the parallel signal. . . . 22

3.1 Example of interpolation of lidar signal. . . . 25

3.2 Flowchart of programs and functions. . . . . 27

3.3 Example how the user is prompted to chose the area of interest. . . . 29

3.4 Example of figures of BSR and depolarisation ratio. . . . . 30

3.5 Example of cloud boundaries. . . . 31

3.6 Example of cirrus backscatter coefficients. . . . 31

3.7 Example of plot of optical thickness. Refer to corresponding figure for each date in section 4.1 and related discussion in the text for more details. . . . 32

3.8 Dialog boxes prompting the user whether to save the parameters, and in which format. 32 3.9 Dialog boxes if some or all of the profiles are already saved. . . . 33

3.10 Dialog box prompting the user to describe the data. . . . 33

3.11 Dialog box prompting the user to choose files. . . . 37

3.12 Statistical plots of geometrical parameters. . . . . 38

3.13 Statistical plots of depolarisation ratio, backscatter coeffficient and optical thickness . 38 4.1 2010-01-12: Time series of height-normalised signal for parallel channel and cloud boundaries. . . . 42

4.2 2010-01-12: Time series of height-normalised signal for perpendicular channel and cloud boundaries. . . . 43

4.3 2010-01-12: Comparison between cloud boundaries from parallel and perpendicular channels. . . . . 43

4.4 2010-01-12: Time series of BSR for parallel and perpendicular channels, and depolarisation. . . . 44

4.5 2010-01-12 22:14: Profiles of height-normalised signal, interpolation and BSR for parallel and perpendicular channels and of depolarisation. . . . 44

4.6 2010-01-12 22:14: Profiles of the measured signal and the simulated molecular signal as well as cloud boundaries. . . . 45

4.8 2010-01-12: Time series of backscatter coefficients for parallel and perpendicular channels. 46 4.9 2010-01-12: Optical thickness against cloud layer top altitude. . . . 46

4.10 2011-03-28: Time series of height-normalised signal for parallel channel and cloud boundaries. . . . 48

(7)

4.14 2011-03-28 22:23: Profiles of height-normalised signal, interpolation and BSR for parallel and perpendicular channels and of depolarisation. . . . 50 4.15 2011-03-28 22:23: Profiles of the measured signal and the simulated molecular signal as

well as cloud boundaries. . . . 50 4.16 2011-03-28: Time series of backscatter coefficients for parallel and perpendicular channels. 51 4.17 2011-03-28: Optical thickness against cloud layer top altitude. . . . 51 4.18 2012-12-21: Time series of height-normalised signal for parallel channel and cloud

boundaries. . . . 53 4.19 2012-12-21: Time series of height-normalised signal for perpendicular channel and cloud

boundaries. . . . 53 4.20 2012-12-21: Comparison between cloud boundaries from parallel and perpendicular

channels. . . . . 54 4.21 2012-12-21: Time series of BSR for parallel and perpendicular channels, and depolari-

sation. . . . 54 4.22 2012-12-21 16:00: Profiles of height-normalised signal, interpolation and BSR for par-

allel and perpendicular channels and of depolarisation. . . . 55 4.23 2012-12-21 16:00: Profiles of the measured signal and the simulated molecular signal as

well as cloud boundaries. . . . 55 4.24 2012-12-21 20:57: Profiles of height-normalised signal, interpolation and BSR for par-

sation. . . . 60 4.32 2014-01-23 00:11: Profiles of height-normalised signal, interpolation and BSR for par-

sation. . . . 65

(8)

4.40 2017-01-24 03:47: Profiles of height-normalised signal, interpolation and BSR for par-

allel and perpendicular channels and of depolarisation. . . . 66

4.46 2017-01-27: Time series of height-normalised signal for parallel channel and cloud boundaries. . . . 70

4.60 Comparison of all six dates analysed. . . . 78

4.61 Distributions of cloud top, base, mean altitude and geometrical thickness for each of the dates. . . . 79

4.62 Cloud top, base and geometrical thickness against optical thickness for each of the dates. 80 4.63 Cloud top, base, mean altitude and geometrical thickness against optical thickness, accumulated for all dates. . . . 81

4.64 Time series of optical thickness with duration of recording for each date. . . . 81

4.65 Distribution of depolarisation ratio with altitude for all dates. . . . 83

4.66 Distribution of backscatter coefficient with altitude for all dates. . . . 84

4.67 Distribution of backscatter coefficient with altitude for all dates; parallel and perpendicular channels. . . . 84

(9)

4.68 Comparison of averaging on the parallel signal for 2017-01-23 03:47. . . . . 85

4.69 Comparison of averaging on the perpendicular signal for 2017-01-23 03:47. . . . 85

4.70 Comparison of averaging on the depolarisation ratio for 2017-01-23 03:47. . . . 86

4.71 Comparison of averaging on the parallel signal for 2012-12-21 15:31. . . . . 86

4.72 Comparison of averaging on the perpendicular signal for 2012-12-21 15:31. . . . 87

4.73 Comparison of averaging on the depolarisation ratio for 2012-12-21 15:31. . . . 87

4.74 Example of the effect of averaging on the signal. . . . . 88

4.75 Comparison of different molecular backscatter coefficients. . . . 89

4.76 Comparison of different molecular signals. . . . 90

4.77 Comparison of different molecular backscatter coefficients. . . . 91

4.80 Comparison of results of using the first and the third BSR methods. . . . 93

4.81 Example of using the first BSR method. . . . 94

4.82 Example of using the third BSR method. . . . . 95

4.83 Example of the measured signal being weaker and stronger than the molecular signal. 99 4.84 Example of .txt file with data. . . . . 100

4.85 Example of the effect of correcting negative signal values. . . . 102

List of tables 1 Input and output from function beta mol. . . . 34

2 Input and output from function molecular signal. . . . 35

3 Input and output of function calculate bsr . . . . 36

(10)

Acronyms

ECMWF European Centre for Medium-Range Weather Forecasts.

IRF Swedish Institute of Space Physics (Institutet f¨or Rymdfysik).

LTU Lule˚a University of Technology.

SMHI Swedish Meteorological and Hydrological Institute.

Glossary

R Lidar range (or distance). For ground-based, zenith-looking lidars corresponding to altitude.

α Extinction coefficient, [m⁻¹].

β Backscatter coefficient, [m⁻¹sr⁻¹].

δ Depolarisation ratio, part of radiation changed from parallel to perpendicular polarised.

k Denotes parallel polarised channel.

⊥ Denotes perpendicular polarised channel.

τ Optical thickness or optical depth.

BSR Backscatter ratio, the ratio of actual backscattering and an imagined cloud-free atmosphere.

LR Lidar ratio, the ratio of extinction coefficient to backscatter coefficient.

z Altitude, increasing from ground.

aer As a subscript, denotes aerosol/particulate matter part.

albedo A measure of how much of the incident solar radiation is reflected. From 0 (none) to 1 (all).

cir As a subscript, denotes cirrus part. Here, analogous to aerosols.

cirrus Ice clouds situated in the upper part of the troposphere.

lapse rate The rate at which the temperature (or another atmospheric variable) changes with altitude.

lidar Remote sensing instrument using a laser. ”Light detection and ranging”.

mol As a subscript, denotes molecular part.

SNR Signal-to-noise ratio, the ratio of the power of the signal (meaningful information) to the power of the background noise (unwanted information).

troposphere The lowest part of the atmosphere, up to around 10 km, in which weather takes place and most clouds are found.

(11)

1 INTRODUCTION

1.1 Motivation

Cirrus clouds play an important role in radiative transfer, and thus have impact on the energy balance of the atmosphere and the climate of the Earth. Furthermore, cirrus clouds occur often and cover large areas globally at any time. Nevertheless, cirrus clouds are poorly studied, especially in the polar regions. As they can be very thin, they might be underrepresented in data bases, depending on the lower detection limit of the used measurement method. For example, radar is not sensitive enough to measure the radiation backscattered off these clouds. However, lidar has a higher sensitivity and can detect thin cirrus clouds. The lidar at the Swedish Institute of Space Physics (IRF) is mainly used to measure polar stratospheric clouds. However, in most observations thin cirrus clouds were present. Characterising the optical properties of these clouds and putting them in a broader context of atmospheric conditions in the upper troposphere will facilitate the understanding of thin cirrus.

1.2 Thesis Aim

The aim of this thesis is to characterise the optical properties of thin cirrus clouds in the polar region and study the atmospheric conditions in which they form and exist.

1.3 Objectives

The original objectives for this thesis were the following:

1. Screen all observations from the IRF lidar for cirrus clouds.

2. Derive geometrical height, geometrical thickness and optical properties.

3. Interpret the data by using data on atmospheric conditions from radiosonde soundings and from the European Centre for Medium-Range Weather Forecasts (EMCWF) analyses.

4. Compare with the results from other studies and with data from other lidars.

However, due to unforeseen problems only the first two of these were fulfilled. Rather, another objective was added:

5. Develop and improve programs and functions for analysis of clouds from lidar data.

For the theory part, the following questions were answered:

• What are cirrus clouds? What is the difference between thin, ultra thin and subvisual cirrus?

• How do cirrus clouds form?

• Which properties of cirrus clouds are interesting?

• Why are cirrus clouds interesting and research relevant at all?

• Why are cirrus measurements in high latitudes interesting and research relevant?

• How does a lidar measurement work (in particular for an elastic backscatter lidar)?

(12)

• Which parameters are obtained and how?

• Are there other cirrus measurements in high latitudes? Which methods/instruments are used?

Which parameters are obtained?

• Are there other measurements of thin cirrus? Which methods/instruments are used? Which parameters are obtained?

1.4 Thesis outline

This introduction is followed by a theoretical section. The first part of the theory section is a short introduction of basic concepts in atmospheric science. This is followed by a part on cirrus clouds, and a part on the lidar instrument. The next part combines the previous two by discussing observations of cirrus clouds, by lidar and by other instruments. The last part of the theory section is a brief summary on the details of this specific study. The theory section is followed by the method section.

The first part in this section is a description of the data used in this thesis, followed by a description of the method used, the so-called BSR method. The last part of the method section is a description of all programs and functions written and used, and a description of how the data was processed. The following section is the results section. The first part of this section is case studies of the six selected and analysed dates. This is followed by statistics based on these dates. After this, there are several parts concerning the solutions on different encountered problems, comparisons of different algorithms and the like. The last part of the results section is a general summarising discussion. The last section is the conclusion section, consisting of a summary and an outlook on future work.

(13)

2 THEORY

2.1 Atmospheric concepts

The atmosphere The atmosphere can be divided into four layers according to its thermal structure;

the troposphere, the stratosphere, the mesosphere and the thermosphere (see e.g. Salby, 2012). These layers can be seen in fig. 2.1 of the US standard atmosphere. The US standard atmosphere is a static mean model of how e.g. the temperature and pressure change with altitude. The atmosphere consists partly of molecular gases, such as nitrogen and oxygen, and partly of aerosols (or particulate matter), such as water droplets, ice crystals and soot particles.

Figure 2.1: The US standard atmosphere of 1962. Figure from Cmglee (2015).

The troposphere is the lowest layer of the atmosphere. Salby (2012) explains that the troposphere extends from the ground up to about 10 km altitude, or rather from a pressure of about 1000 hPa at ground up to 100 hPa. However, the top of the troposphere can be as high as 20 km in the equator, and as low as 7 km above the poles in winter (University Corporation for Atmospheric Research, 2019). Geerts and Linacre (1997) describe that these differences in tropopause altitude are due to the difference in temperature of the air mass in the troposphere. A higher temperature in the troposphere can lead to increased deep convection (i.e. thunderstorms), which pushes the tropopause upwards and increase the range of the troposphere. On the other hand, colder regions have a lower tropopause due to the negative radiation balance at the surface, which limits convective overturning, and the tropopause is not pushed upwards. Salby (2012) explains that the lapse rate (the rate at which the temperature decreases with altitude) is – in absence of temperature inversions – nearly constant throughout the troposphere with a global mean of 6.5 K/km. The troposphere is characterised by

(14)

convective overturning, which is why its name means ”the turning sphere”, and is where weather takes place. Near the ground there can be instances where the temperature increases with altitude rather than decreases (negative lapse rate); this is called a temperature inversion (University Corporation for Atmospheric Research, 2019).

Salby (2012) further explains that the upper boundary of the troposphere is known as the tropopause.

The tropopause is characterised by a temperature minimum and a lapse rate close to zero. The official definition by World Meteorological Organization (1992) is stricter; in that the tropopause is defined as ”the lowest level at which the lapse rate decreases to 2^◦C/km or less, provided that the average lapse rate between this level and all higher levels within 2 km does not exceed 2^◦C/km”. Thomas and Stamnes (1999) further describe that above the tropopause is the stratosphere, which means

”the layered sphere”. It is an apt name; there is only very little vertical mixing. Going upwards, the temperature increases again, and the lapse rate is thus negative. This increase in temperature is due to absorption of solar UV (ultra violet) light by ozone molecules. In the stratosphere, so-called Polar Stratospheric Clouds (PSCs) are present during the polar winters. PSCs typically only occur at temperatures below 195 K (−78^◦C), mostly at altitudes between 15 and 25 km. They can consist either of water ice, of frozen nitric acid trihydrate (NAT) or of supercooled ternary solution (STS) of sulfuric acid, nitric acid and water. As a point of reference, commercial aircraft typically fly at altitudes corresponding to the tropopause or lower stratosphere, as the turbulence is much decreased there compared to in the troposphere.

Figure 2.2: Example of blackbody curves and comparison with classical theory. Figure from Darth Kule (2010).

The electromagnetic spectrum All objects with a temperature which is above 0 K (−273.15^◦C) emits energy in form of electromagnetic radiation. A blackbody is an idealisation, an object which absorbs all of the radiation incident on it. The spectral distribution of a blackbody’s radiated thermal energy depends on the its temperature (see e.g. Carroll and Ostlie, 2007). Some examples of blackbody curves can be seen in fig. 2.2. The blackbody curve of the Sun peaks at about 0.6 µm, while the one of the Earth peaks at about 10 µm.

The main part of the radiation from the Sun and the Earth corresponds to the parts of the electromagnetic spectrum known as the visible and the infrared (IR). As seen in fig. 2.3, the visible part of

(15)

Figure 2.3: The electromagnetic spectrum. Figure from Philip Ronan, Gringer (2013).

the spectrum covers the range of wavelengths between 380 and 750 nm. This is the range where the spectral flux or irradiance of the Sun is the greatest (see e.g. Thomas and Stamnes, 1999). Radiation at these wavelengths are referred to as solar, visible or shortwave. On the other hand, the Earth’s emission in form of heat corresponds to the wavelengths of λ > 3.5 µm. This type of radiation can be referred to as terrestrial, infrared or longwave.

Scattering There are different descriptions for how light scatters of a particle, depending on the relative size/wavelength of the particle and the radiation.

Wandinger (2005) describes that when the particle is very small compared to the wavelength of the incident radiation, the scattering can be described by the Rayleigh approximation of scattering (or Rayleigh scattering). For lidar applications, molecular scattering is well described by the Rayleigh model. As oxygen and nitrogen accounts for 99% of the molecular part of the atmosphere, they can be considered to cause all Rayleigh scattering in the atmosphere. The intensity of Rayleigh scattered radiation is proportional to λ⁻⁴, where λ is the wavelength. It dominates the backscattered lidar signal at short wavelengths.

On the other hand, Mie scattering is more complex, and describes the scattering of radiation with arbitrary wavelength on a dielectric sphere of arbitrary size. Thus it include the solution for Rayleigh scattering, but is generally used to describe cases where the size of the scattering particle is comparable to the wavelength of the scattered radiation, or larger. When the former is the case, the intensity of the scattered radiation depends strongly on the wavelength, whereas for very large particles the scattering intensity is wavelength-independent. Thus, for aerosol particles in the size range of about 50 nm to a few micrometers one can scatter light with different wavelengths off them to acquire information on size and other parameters.

As long as the wavelength of the radiation is large compared to the size of the particle, the particle’s shape does not matter much. However, many aerosols are both large and non-spherical, such as ice crystals, soot agglomerate and mineral dust. For these, Mie scattering does not work, but one rather has to apply intricate non-spherical scattering theories. For particles such as ice crystals, the geometrical optics approximation can be used, in which one traces the light ray as it enters the particle and is affected by refraction, and as it reflects off the particle’s facets. Large and non-spherical particles

(16)

are easily detected by a lidar. If scattering linearly polarised light at 180^◦(backscattering it), spherical particles do not change the polarisation of the light. However, non-spherical particles do change the polarisation (depolarises it).

Optical thickness One important parameter concerning clouds is their optical thickness (or optical depth), denoted by τ . Thomas and Stamnes (1999) explain that the optical thickness describe how far light can travel through a material (i.e. the atmosphere or a cloud), before it is attenuated to a certain degree. The optical thickness is defined as

τ = lnΦⁱ_e

Φ^t_e = − ln T. (1)

Here, Φⁱ_e is the incident radiant flux, Φ^t_e is the radiant flux transmitted through the material, and T is the material’s transmittance.

2.2 Cirrus

2.2.1 General description

Cirrus are clouds consisting of ice particles which are situated in the upper troposphere, at altitudes between 5 and 13 km in Sweden (Swedish Meteorological and Hydrological Institute, 2013). Due to the cold temperature at such high altitudes, the water vapour saturation pressure is low, leading to a low ice water content. Thus, cirrus are often optically thin and may have a stringy, feather-like appearance.

Lynch (1993) further describe cirrus as detached and without self-shadowing, appearing as white, fibrous tufts. They often have delicate filaments. Very thin cirrus might appear blue, as they do not scatter much light. While cirrus are ice crystal clouds, not all clouds consisting of ice crystals are cirrus. Most stratiform (hazy, featureless) clouds during the polar winter consist either fully of ice crystals, or a mixture between ice crystals and droplets (so-called mixed-phase). Such clouds are in the mid-altitudes of the troposphere, lower than cirrus. Another example is ice fog – or diamond dust – which consists of tiny ice crystals and exists near the ground.

2.2.2 Classification

Sassen and Cho (1992) describe the most used classification of cirrus into subvisual, optically thin and optically thick, depending on their optical thickness:

• Subvisual - τ <0.03,

• (Treshold visual - 0.01<τ <0.1),

• Optically thin - 0.03<τ <0.3,

• Optically thick - τ >0.3.

Sassen and Cho (1992) point out that lidar-derived τ typically saturates at approximately 2.0-3.0 in dense cirrostratus.

(17)

Plana-Fattori et al. (2009) studied the characteristics of high clouds at Lannion and Palaiseau in France, and found that optically thick high clouds (τ >0.3) tended to have large geometric thicknesses and high cloud top altitudes, while optically thin high clouds (τ <0.3) tended to have large altitudes (mean height) and small geometric thickness.

However, Keckhut et al. (2006) point out that it has not been shown that the classifications based on optical thickness correspond to cirrus types with significantly different formation processes, morpholo- gies or similar. Rather, the classification by Sassen and Cho (1992) is based on human vision and the difference in observations between lidar and human vision (e.g. subvisible clouds are not visible to a human, but are detected by a lidar). Furthermore, Sassen and Cho (1992) describes that classification of a cloud by a human observer on the ground depends on the solar angle and the viewing angle. As the scattering in the forward direction is strong near the solar disc, the visibility of clouds is increased overhead around local noon. Thus, they observed threshold-visible clouds with τ as small as 0.01 although the limit for subvisual clouds are 0.03.

Keckhut et al. (2006) further point out that for liquid clouds a number of different classes was developed early, based on visual information of morphological characteristics such as geometrical form, spatial extent, colour, contrast and texture, for different altitudes. These classes have been proven useful in determining the formation processes of clouds and parametrization of them in models of general circulation. Cirrus are on the other hand generally described as a homogeneous class of ice clouds, without regard to their altitude. Using data from lidars in France and applying principal component analysis, cluster methods, and linear discriminant analysis, Keckhut et al. (2006) found four distinct classes of cirrus. Almost all observations belonged to one of three classes; geometrically thin clouds above or at the local tropopause (in France, 11.5 km), geometrically thin clouds at a lower altitude (8.6 km), and geometrically thick clouds (thickness of 3.2 km), below the tropopause at an altitude range inbetween the first two classes (9.8 km). Their fourth class, episodic highly scattering cirrus, consisted of only two cases, and further investigations on larger datasets are needed to determine whether this is a distinct class. Their data did not include the depolarisation ratio, which they mean would have been a good way of discriminating between different classes, as it contains information of the ice particle shapes.

One problem with identifying cirrus (in the sense of high-altitude clouds consisting of ice crystals) is that different authors and instruments use different criteria to identify ice clouds. Plana-Fattori et al.

(2009) describe that one method is to only include clouds with temperatures below -38^◦C (235 K), as no liquid water is present below this temperature, while other criteria can regard the base altitude, the scattering ratio, the depolarisation ratio, the visual appearance or different combinations of these parameters.

2.2.3 Effect on radiative budget

Salby (2012) explains how clouds affect the radiative budget of the Earth. In general, clouds act cooling in the shortwave (SW) range, as they are highly reflective and reflect incident sunlight, thus reducing the amount of solar radiation reaching the surface. On the other hand, water and ice absorb strongly in the infrared (IR) part of the spectrum. Clouds thus absorb and reemit the outgoing terrestrial radiation, and act warming in the longwave (LW) part of the spectrum. Thus, clouds tend to counteract the greenhouse effect in SW, but reinforce it in LW. The net effect depends on the cloud’s temperature and therefore, as seen in fig. 2.1, on its altitude. Keckhut et al. (2006) clarify that a cirrus cloud at high altitude – which therefore is cold – influences the infrared radiation budget more

(18)

strongly than the same cloud on a lower altitude would do. This is because the LW emission from a cloud is described by blackbody emission according to the temperature in its top layer (P. V¨olger, personal communication, 2019-07-09). On the other hand, for a low-altitude cirrus which is warmer, the effect of longwave absorption is weaker, and the effect of reflecting solar radiation may dominate.

Thus, high clouds tend to have an overall cooling effect, while low clouds are overall warming.

In Kiruna, at a high latitude above the polar circle and with cold air temperatures, all clouds, especially high ones, have a net warming effect (V. Wolf, personal communication, 2019-04-08). According to C´ordoba-Jabonero et al. (2016), tropical cirrus situated at high altitudes tend to have a net warming effect, while low-altitude cirrus over the polar region rather have a cooling effect due to the increased albedo. Thus, these sources contradict each other.

C´ordoba-Jabonero et al. (2016) point out that apart from their altitude, the balance between solar albedo cooling and infrared greenhouse warming also depends on the microphysical properties of the ice crystals. Furthermore, there are feedback mechanisms. Milz (2018) explains that as the mean state of the atmosphere changes due to climate change, the clouds will likely change their characteristics.

However, this will change the state of the atmosphere, and so on. Feedbacks can either enhance or counteract the initial change. As an example, if the amount of atmospheric CO2 increases and causes increased atmospheric temperatures, more water might evaporate from the oceans and cause increased water vapour content in the atmosphere, which in turn could cause thicker and more reflective clouds.

This would reduce the surface heating, which would lead to lower atmospheric temperatures. However, according to current models clouds have in general an enhancing effect on the warming. There is however a large range of predicted feedbacks for clouds. Thus, clouds are the largest contributor to the uncertainty of climate sensitivity.

C´ordoba-Jabonero et al. (2016) point out that the effect of anthropogenic climate change, such as anthropogenic greenhouse effect and contamination in the upper troposphere from aircraft, on cirrus cloud formation and occurrence has not been studied. For example, cirrus contrails due to aircraft could increase the albedo of the upper troposphere, which in turn could modify the warming effect from greenhouse gases, as less sunlight would reach the lower atmosphere.

McFarquhar et al. (1999) specify that in addition to the non-negligible effect of cirrus on the radiative budget, their other effects should not be ignored. These include increasing the amount of water vapour in the lower stratosphere and the vertical motions in the upper troposphere.

2.2.4 Occurrence

Cirrus cover a large part of the globe at any time; using lidar in the central Pacific Tropics, McFarquhar et al. (1999) observed spatially thin cirrus with base above 15 km altitude 29 % of the time. Keckhut et al. (2006) point out that in the first height-resolved cirrus climatology from mid-latitude data, the occurance frequency of cirrus was 50 %, regardless of the time of the year. According to Salby (2012), the general frequency of occurrence of cirrus clouds are 30 %. Whatever the exact value of the frequency of occurrence is, cirrus are present a large amount of time, and their effect on e.g. the radiative budget is not fully studied and understood.

Plana-Fattori et al. (2009) report that in the data from the TOVS Path-B satellite instrument, the global frequency of high clouds (defined in the dataset as above the 440 hPa pressure level) is 30%, while the frequency over the northern hemisphere mid-latitudes is 28%. The annual cycle of high cloud frequency is stronger over land than over ocean; in spring and summer the frequency of occurrence

(19)

is more than 35%, while in autumn and winter it is less than 25%. However, only 40% of the high clouds in this statistic have an optical thickness of τ < 1.4. Furthermore, about 2/3 of the high cloud occurrences involve multi-layered systems. However, this is a passive remote sensing instrument, a spaceborne infrared vertical sounder. Its spatial resolution is 1^◦ (longitude and latitude), and its limit for cloud optical thickness is τ ∼ 0.1. The limit of 1.4 for the optical thickness does not correspond to the limits for optically thin or thick clouds, but is rather due to the instrument used.

According to McFarquhar et al. (1999), subvisible cirrus almost only occur at tropical latitudes (20^◦S- 20^◦N). Plana-Fattori et al. (2009) find that high, thin clouds with optical thicknesses of τ < 0.1 are more relatively frequent at their tropical sites than at their mid-latitude sites, and point out that this is in accordance with earlier space-borne lidar observations. Thus, it would seem that while subvisible clouds are relatively more frequent at tropical sites, they do appear outside of the Tropics, according to more recent studies.

However, Plana-Fattori et al. (2009) point out that it is not unexpected that different regions with different atmospheric conditions yield different distributions of optical thickness. There can be a strong linkage between prevailing weather at a site and the occurrence and properties of cirrus clouds at the same site, as weather have a large influence on values and variations of temperature, pressure and water vapour content in the atmosphere.

The lidar ratio (LR) is an important parameter, defined as the ratio of the extinction coefficient to the backscatter coefficient, which is described in more detail in section 2.3.3. C´ordoba-Jabonero et al.

(2016) found that polar cirrus clouds had higher LR values (from 21 sr for subvisual cirrus and 28-29 sr for threshold visual to 42 sr for thin and 32 for thick) than subtropical cirrus clouds (corresponding values was 10 sr, 17-29 sr, 18-32 sr and 25-27 sr). It should also be noted that the LR of subvisual cirrus is lower than the LR for thin and thick cirrus, regardless of the latitude. C´ordoba-Jabonero et al. (2016) explain that this result likely reflect a difference in internal state between the different classes of cirrus clouds. Thus, there might be a variation in microphysical properties of different cirrus classes and over different regions.

C´ordoba-Jabonero et al. (2016) further found that at subtropical latitudes, optically thick cirrus are located at lower altitudes (cloud top altitude at 1.2 km below tropopause level) than subvisual cirrus (0.5 km below). They did however find that in the polar regions, all cirrus categories are found at similar altitudes (about 9.5 km altitude), and explain this with the polar atmosphere being more stable.

2.2.5 Ice crystal shapes

Cheng et al. (2009) describe that cirrus consists of non-spherical ice crystals. The ice crystals have a large variability in size, shape and density, depending on the ambient temperature and humidity.

Wolf et al. (2018) explain that cloud ice crystals can be classified into five different groups:

• Compact particles, which has no pronounced shape and include spherical particles,

• Rosettes, ice particles with two or more arms,

• Plates, hexagonal particles where hexagonal base facet are longer than the prism facets,

• Columnar particles, hexagonal particles where the prism facets are longer than the hexagonal base facet,

(20)

• Irregular particles, ice particles which don’t fit any of the above groups.

Figure 2.4: Examples of the different ice crystal shapes. Figure with courtesy from Wolf et al. (2018).

Examples from these categories of different ice crystal shapes can be seen in 2.4. Olofson et al. (2009) describe that from field studies it seems like columns and polycrystals are more common in the upper, colder part of a cirrus cloud, while plates dominate the lower, warmer part of the cloud. Kahnert et al. (2008) have shown that to predict the radiative forcing of cirrus, it is not enough to know the ice particle mass or number density. Knowledge about ice particle shapes and their adequate description in models are important in order to accurately determine the radiative impact of Arctic cirrus.

2.2.6 Formation

As mentioned in section 2.2.5, the ambient temperature and humidity distributions, as well as vertical velocity in the upper troposphere, are important parameters for the formation of ice particles. Eix- mann et al. (2010) explain that cirrus ice particles form when air parcels rise adiabatically until the supersaturation required for ice formation is reached. The nucleation of ice particles depends not only on the thermodynamic structure of the upper troposphere, but also on the presence of aerosol. This because there are two major forms of nucleation; heterogeneous nucleation and homogeneous freezing of aqueous solution droplets. These are explained in more detail below.

Salby (2012) explains that ice can only form in the atmosphere under the certain circumstances providing homogeneous and heterogeneous freezing, which is why nearly 50% of clouds warmer than -10^◦C and nearly all clouds warmer than -4^◦C contain no ice at all. This does however depend significantly on the location. Such clouds instead consist of droplets of water in the metastable supercooled state. However, as the temperature decrease to -20^◦C, 90% of clouds consist of at least some ice.

For homogeneous freezing, an ice embryo formed by chance collection of water molecules inside a droplet is needed. However, this process is not favoured at temperatures above -36^◦C (237 K). To

(21)

understand why, the process of homogeneous nucleation of a water droplet is described below. Homo- geneous freezing occur in a analogous manner.

The Gibbs free energy is a thermodynamic potential, and is a measure of how much non-expansion work can be performed by a thermodynamically closed system. The change in Gibbs free energy ∆G depends on the saturation ratio S = _e^e

0 as

∆G ∼ ln e

e₀, (2)

in which e is the vapour pressure for a certain amount of the gas, and e₀ is the saturation vapour pressure. For subsaturation (_e^e

0 < 1) and saturation (_e^e

0 = 1), Gibbs free energy increase monotonically with droplet radius. As equilibrium in a system is achieved by reducing its energy, droplet growth is unfavoured when _e^e

0 ≤ 1. Thus, any droplet with such small radius formed by chance collision of water molecules will evaporate. However, under supersaturated conditions (_e^e

0 > 1), Gibbs free energy exhibits a maximum for a certain droplet radius. Beyond this radius, the free energy decreases. Thus, any droplet formed with a radius larger than this critical radius will grow spontaneously through condensation.

At temperatures around −36^◦C, the free energy of formation of ice increases for ice embryos smaller than a certain diameter. Any ice embryo, formed by chance collision of water molecules, with a radius smaller than this diameter will disperse. When the temperature increase above −36^◦C, the critical size to be exceeded is too large, and homogeneous freezing does not occur. Thus, homogeneous freezing only take place in high and cold clouds.

Salby (2012) further explains that heterogeneous freezing occurs when water molecules accumulate on a so-called freezing nucleus, an existing particle with a molecular structure similar to ice. The ice embryo formed in this way is larger than one formed by homogeneous nucleation, and can thus grow at warmer temperatures. Once the ice particles have formed, they can grow in different ways. One way is by deposition of water vapour on the ice particle. As the saturation vapour pressure with respect to ice is lower than the saturation vapour pressure with respect to water, a cloud which is nearly saturated with respect to water may be supersaturated with respect to ice. Thus, the growth of a particle is accelerated if it freezes. Ice particles can also grow by colliding with supercooled droplets, so-called riming. As more and more droplets collide with the ice crystal and freeze, it assumes an irregular shape. Finally, ice particles can grow through coagulation, a process favoured at temperatures warmer than −5^◦C. As mentioned earlier, clouds with a temperature above -4^◦C rarely contain ice. When they do, coagulation is thus the dominant growth mechanism. These different freezing and growth mechanisms leads to the variety in shapes and sizes discussed in section 2.2.5.

Kr¨amer et al. (2016) explain that there are two different formation mechanisms, or origins, of cirrus.

The first is in situ origin, where the ice crystals are formed directly from water vapour. The second is liquid origin, where liquid water drops are uplifted, and subsequently freezes as the temperature is low enough. Wolf et al. (2018) studied Arctic cirrus clouds above Kiruna, and found that the cloud’s origin (liquid or in situ) had effect on the cirrus particle sizes, shapes and number concentrations.

Cirrus with in situ origin had smaller particle sizes, consisted mainly of compact and irregular ice particles and a large range of number concentrations. On the other hand, cirrus with liquid origin had larger particles and a wider range of sizes, the dominating particle shapes were irregular, rosettes and columns, and the number concentrations were low.

Rossby waves – or planetary waves – are linked to the Earth’s rotation and the variation of the Coriolis force with latitude (Salby, 2012). They are associated with the jet stream and with pressure

(22)

systems. Eixmann et al. (2010) explain that while the understanding and modelling of cirrus clouds’

microphysical properties and radiative impact have advanced, the dynamical linkage between cirrus clouds, both of natural and anthropogenic (contrail) origin, and different weather systems is less explored. They further report that the most striking features of cirrus in the Northern Hemisphere midlatitudes are produced by tropopause disturbances associated with Rossby waves. In Rossby wave breaking events, especially in anticyclonic, poleward and downstream ones, the tropopause altitude can increase significantly, up to 12 km (in midlatitudes) during winter. Furthermore, Eixmann et al.

(2010) found that subtropical water vapour was transported northeastward and upward during such an event. They found that in the transition between moist air and cold air, cirrus was formed.

2.3 Lidar

2.3.1 General description

The lidar (Light detection and ranging) is a remote sensing instrument, using pulsed laser light in analogy to the radar which uses radio waves. Wandinger (2005) describes that lidar, along with radar, is one of the most important instruments in atmospheric science, and in particular for profiling the atmosphere. In fig. 2.5, a schematic sketch of the set-up of a lidar system is shown.

The lidar is an active remote sensing instrument. This means that it gathers information on the object of interest from a distance by measuring e.g. electro-magnetic waves emitted or reflected by the object. As an active instrument it supplies the source of energy that the object is probed with, in contrast to passive instruments which uses energy from e.g. the sun.

Figure 2.5: Sketch of the set-up of a lidar system. Figure from Athina.Argyrouli (2015).

(23)

2.3.2 System description

The source of energy, the transmitter, is for the lidar a pulsed laser. The light pulses have in general a length of a few to several hundred nanoseconds. A beam expander is often included in the transmitter in order to reduce the divergence of the beam. The transmitted laser beam is directed to the atmosphere, where a small part of the transmitted energy is backscattered by atmospheric particles and molecules. This backscattered light is detected by the receiver unit, where a telescope collects the backscattered photons. After the telescope, the light is separated according to its wavelengths and/or polarisation states by an optical analysing system. The selected light is directed onto detectors, which convert it to an electric signal. This electric signal is recorded electronically by a computer.

The laser wavelength used depends on the application; wavelengths in the range 250 nm to 11µm are common. The source of the laser light can for example be a noble gas and a reactive gas (in the case of high-power excimer lasers), or Nd:YAG crystals, in which frequency tripling and doubling with nonlinear crystals can be used to convert the primary radiation to the more suitable wavelengths 532 and 355 nm. Both Nd:YAG lasers and excitable lasers can be used not only as direct laser emitters, but also to pump secondary laser sources. However, Wandinger (2005) explains that these two types nowadays tend to be replaced by tunable, solid-state lasers based on crystals, doped crystalline lattices or similar.

The laser beam is by its nature already highly collimated, but a beam expander can be used to increase its collimation even more, to values as small as 100 µrad. Then, the field of view of the receiver telescope can be almost as small. This is beneficial as the background light from the atmosphere is significantly reduced, and the receiver detects fewer of the photons which underwent multiple scattering.

The diameter of the primary telescope optics is typically in the range 0.1 to a few meters, depending on the application. Mirror telescopes are used by most lidars; only the ones with small apertures can use lenses. At the local plane of the receiver optics, a field stop is placed and determines the field of view.

The emitter and receiver optics can be geometrically arranged relative to each other in two different ways. These are biaxial and coaxial systems. In coaxial systems, the laser beam is emitted along the receiver telescope’s optical axis. On the other hand, in biaxial systems the emitter is located at least a telescope radius away from the receiver, so that their optical axes are separated and the backscattered light enters the receiver telescope slightly from the side. Due to this and other geometric factors, the signal from the range close to the lidar is compressed to a certain degree, as the laser beam cannot be fully imaged onto the receiver. Thus, only a part of the lidar return is actually measured. This varies with beam diameter, and apart from the type of geometrical arrangement also depends on receiver field of view, the laser beam diameter, shape and divergence as well as the telescope’s imaging properties, such as focal-length-to-diameter-ratio. This part varies from zero at the location of the lidar to unity where the receiver field of view and the transmitted laser beam fully overlap, typically at a few kilometers distance, and is called the overlap function.

In front of the detector, optical elements such as interference filters, polarizers, spectrometers and interferometers can be applied. The detector typically consists of either photomultiplier tubes or photodiodes. Photomultiplier tubes and avalanche photodiodes can be operated in Geiger mode, in which each photon is counted individually.

By measuring the time interval between the laser pulse leaving the transmitter and the backscattered part of it being detected by the receiver and knowing the speed of light, one can determine the distance to the particles that backscattered the light. Furthermore, the temporal resolution ∆t correspond to