Siddle A 2021 PHD Thesis

Aeronomy of Mars: An Observational Study of the
Dynamic Martian Upper Atmosphere

Alexander Siddle
Department of Physics, Imperial College London
Submitted in partial fulfilment of the requirements for the degree of

Doctor of Philosophy
December 2020
Declarations
I declare that all work presented in this thesis is my own, unless explicitly stated and referenced.
The copyright of this thesis rests with the author. Unless otherwise indicated, its contents are licensed
under a Creative Commons Attribution-Non Commercial 4.0 International Licence (CC BY-NC). Under
this licence, you may copy and redistribute the material in any medium or format. You may also create
and distribute modified versions of the work. This is on the condition that: you credit the author and
do not use it, or any derivative works, for a commercial purpose. When reusing or sharing this work,
ensure you make the licence terms clear to others by naming the licence and linking to the licence text.
Where a work has been adapted, you should indicate that the work has been changed and describe those
changes. Please seek permission from the copyright holder for uses of this work that are not included
in this licence or permitted under UK Copyright Law.
Alex Siddle - December 2020
The work presented in this thesis has contributed to two peer-reviewed published papers:
• Chapter 5 - A. G. Siddle et al. (2019). ‘Global characteristics of gravity waves in the upper
atmosphere of Mars as measured by MAVEN/NGIMS’. in: Icarus 333, pp. 12–21
• Chapter 7 - A. G. Siddle et al. (2020). ‘Density structures in the Martian lower thermosphere as
inferred by Trace Gas Orbiter accelerometer measurements’. In: Icarus, p. 114109
2
Abstract
The Martian upper atmosphere is highly dynamic over both short and long temporal and spatial scales.
Our understanding of this region primarily stems from a wealth of data gathered both in-situ and
remotely from spacecraft orbiting the Red Planet. In 2014 NASA’s Mars Atmosphere and Volatile Evol-
ution (MAVEN) spacecraft began orbiting Mars with the primary objective to probe and characterise
the upper atmosphere using composition data. Data from the Neutral Gas and Ion Mass Spectrometer
(NGIMS) on board MAVEN has been utilised throughout this thesis.
Daily, monthly and seasonal density and temperature variations have been explored. An apparent
day-night asymmetry is observed in temperature with the dayside typically 50-100 K warmer due to
solar-EUV heating. Seasonal analysis has found densities to be significantly enhanced around perihe-
lion compared to other times throughout a Martian year. Perturbations in density and temperature
profiles have been interpreted as vertically propagating gravity waves. Diurnal and seasonal variations
of gravity wave characteristics have been examined with enhanced activity on the nightside. Fully un-
derstanding these features is required if the complexities of the upper atmosphere are to be fully grasped.
The Martian science community was fortunate to have a plethora of spacecraft at Mars during the
June 2018 global dust storm. This thesis has examined the response of the upper atmosphere to such
an important and rare event. A notable expansion of the atmosphere is observed whereby the upper
atmosphere is raised by several kilometres, due to heating in the lower atmosphere. It is hoped that
results can inform efforts made to model and predict the effects of a dust storm. For the first time,
gravity waves at Mars during a global dust storm have been studied. Atmospheric perturbations are
found to be significantly enhanced during the dust storm event.
During 2017/2018, ESA’s Trace Gas Orbiter (TGO) undertook its aerobraking phase to circularise
itself into its science orbit. During this period, density data were able to be retrieved in the lower ther-
mosphere from accelerometer measurements. The retrieval process was not part of this thesis; however,
these data are used for the first time. Standalone analyses are performed with these data but are also
combined with results from MAVEN owing to the two spacecraft sampling similar regions concurrently.
This overlapping period is exploited, and densities are hydrostatically connected to understand the en-
tire thermosphere structure. By combining wave data, it has been inferred that shorter wavelengths are
3
saturated within the thermosphere, whereas larger wavelengths continue to grow with height.
4
Acknowledgements
First and foremost, to Ingo. You taught me so much over the years from trivial things like atmospheric
physics to the more essential skills in life, such as how to think like a scientist, and more importantly
how to have a laugh whilst you’re doing it. I could not have asked for a more friendly, knowledgeable
and humourous supervisor. The pleasure is all mine.
Thanks must go to my original office mates – David, Ewen and Lars - who took me under their
wing and provided much joy and humour throughout the many years we shared an office together. The
arrival of the next cohort – Joe, Emma, Harry, Maks and Ned - saw us kicked out of our office. Truly
unforgivable. Despite that, you’ve all added something unique to the research group, from the revival
of SPAT walks to the introduction of MarsBallTM. A special mention must go to Joe, who followed my
path from Lancaster to Imperial. As such, the patent profession awaits your imminent arrival. Next up
are Earn, Pete and Sadie. You three were always up for a drink, whether it be coffee or alcohol. Both
of which were crucial during the final months of my PhD. And finally my newest office mates Adrian,
Ronan and Tom. I may have only spent half my time in the office, but it’s been a pleasure (and please
keep the office plants alive!)
I want to thank Roger Yelle and Shane Stone for their guidance with everything and anything
MAVEN/NGIMS related. It was a pleasure to meet you, and thank you for your support on my first
paper. Thanks must go to Sean Bruinsma and J-C Marty for undertaking the unenviable task of retriev-
ing data from Trace Gas Orbiter. Such meticulous work often goes underappreciated. This wonderous
dataset was the basis of my second paper, so for that and your support, I am incredibly grateful.
Thank you to Adam and Apostolos for assessing me throughout the years and thereby ensuring I
had a chance of achieving a PhD. Thank you to Adam, Bob and Stephen for taking the time to read
this thesis and asking me all the wrong questions. Joking aside, your pertinent comments have made
this thesis a much stronger piece of work.
Thank you to all my family for the endless support over the past few years. It really does mean
the world. There’s undoubtedly an art to asking about a PhD and then always remembering something
needs doing within seconds of me opening my mouth.
5
Finally, thanks to Southeastern for rarely sticking to their timetable, allowing me more time to work
on the train. Those extra few minutes certainly added up over the years - as did the price hikes.
6
Contents
List of Figures 11
List of Tables 14
Abbreviations 15
1 Introduction 17
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.2 Motivation for Studying Mars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3 Previous Observations of Mars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3.1 Pre-Spacecraft Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3.2 Previous Spacecraft Observations . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.4 Mars Planetary Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.5 Atmospheric Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.5.1 Density Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.5.2 Temperature Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.6 Martian General Circulation Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.6.1 Laboratoire de Météorologie Dynamique Mars Climate Database . . . . . . . . . 33
1.6.2 Other Martian General Circulation Models . . . . . . . . . . . . . . . . . . . . . 34
1.7 Gravity Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.7.1 Gravity Wave Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.7.2 Gravity Wave Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.7.3 Gravity Wave Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.7.4 Modelling Gravity Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.8 Dust Storms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
7
CONTENTS
1.8.1 Dust Storm Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

1.8.2 Dust Storm Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.9 Response of Upper Atmosphere to Dust Storms . . . . . . . . . . . . . . . . . . . . . . 46
1.10 Open Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2 Instrumentation and Data 50

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.2 The Mars Atmosphere and Volatile Evolution Mission . . . . . . . . . . . . . . . . . . . 50
2.2.1 Spatial and Temporal Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.2.2 The Neutral Gas and Ion Mass Spectrometer . . . . . . . . . . . . . . . . . . . 52
2.2.3 Accelerometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.2.4 Comparison Between NGIMS and ACC Data . . . . . . . . . . . . . . . . . . . . 58
2.3 The ExoMars Mission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.3.1 Trace Gas Orbiter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.3.2 Rosalind Franklin Rover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.4 Concurrent MAVEN and ExoMars Data . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3 Data Analysis and Model Comparison 67

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2 Deriving Temperature Profiles From Density Data . . . . . . . . . . . . . . . . . . . . . 67
3.3 Drawbacks of Current Temperature Derivation Technique . . . . . . . . . . . . . . . . . 70
3.4 Comparison Between Derived and Extracted MCD Temperature Profiles . . . . . . . . . 71
3.5 Comparison Between MCD and In-Situ Density Data . . . . . . . . . . . . . . . . . . . 74
3.5.1 Comparison with Viking Landers’ Densities . . . . . . . . . . . . . . . . . . . . 74
3.5.2 Comparison with NGIMS Densities . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.6 Peak Offset in TGO Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4 Background Density and Temperature Analysis 83

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.2 Averaging Temperature Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8
CONTENTS
4.3 Investigating Diurnal Temperature Variations . . . . . . . . . . . . . . . . . . . . . . . 85

4.4 Investigating Diurnal Horizontal Temperature Gradients . . . . . . . . . . . . . . . . . . 90
4.5 Investigating the Effects of the 27-Day Solar Cycle on Atmospheric Density . . . . . . . 92
4.6 Investigating Seasonal Density and Temperature Variations . . . . . . . . . . . . . . . . 97
4.6.1 Identifying Regions Sampled Multiple Times by MAVEN . . . . . . . . . . . . . 97
4.6.2 Density Variability with Season . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.6.3 Temperature Variability with Season . . . . . . . . . . . . . . . . . . . . . . . . 103
4.6.4 Southern Polar Warming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5 Gravity Waves in the Martian Upper Atmosphere 111

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.2 Atmosphere Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.3 Effects of Spacecraft Trajectory on Inferred Gravity Wave Wavelengths . . . . . . . . . . 114
5.4 Comparing Temperature and Density Perturbations . . . . . . . . . . . . . . . . . . . . 117
5.5 Altitudinal Effects on Gravity Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.6 Global Characteristics of Gravity Waves . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.7 Effect of Topography on Gravity Waves . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.8 Comparing Wave Activity Over Successive Orbits . . . . . . . . . . . . . . . . . . . . . 130
5.9 Wave Characteristic Comparison with Other Planets . . . . . . . . . . . . . . . . . . . . 135
5.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
6 Effects of June 2018 Dust Storm on the Martian Upper Atmosphere 140
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6.2 Dust Storm Growth and Atmospheric Expansion . . . . . . . . . . . . . . . . . . . . . . 141
6.3 Dust Storm Decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
6.4 Dust Storm Effects on Gravity Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
6.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
7 Results from Trace Gas Orbiter Aerobraking Campaign 160

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
7.2 Separation of Background and Wave Profiles . . . . . . . . . . . . . . . . . . . . . . . . 161
7.3 Background Density Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
9
CONTENTS
7.3.1 TGO and MAVEN Densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

7.3.2 Comparison with Laboratoire de Météorologie Dynamique Mars Climate Database 167
7.4 TGO Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
7.4.1 Deriving TGO Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
7.4.2 Background TGO Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . 171
7.5 Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
7.6 Combining MAVEN and TGO Data for Continuous Vertical Profiles . . . . . . . . . . . 178
7.6.1 Comparing TGO and MAVEN Densities . . . . . . . . . . . . . . . . . . . . . . 179
7.6.2 Hydrostatically Connecting TGO and MAVEN Density Profiles . . . . . . . . . . 180
7.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
8 Conclusions and Future Work 185

8.1 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Appendices 193
A Visualisation of Regions Sampled Multiple Times by MAVEN 194
B Deriving Lower Atmosphere Heating Caused by a Dust Storm 195
C Software Used 197
D Data Used 198
Bibliography 199
10
List of Figures
1.3.1 Infographic of previous missions to Mars . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1.5.1 Nine density profiles using MAVEN/NGIMS data for orbit 1064 . . . . . . . . . . . . . . 27
1.5.2 Schematic diagram showing monochromatic radiation penetrating a plane and horizont-
ally stratified atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.5.3 Atmospheric temperature profile from entry phase of Mars Pathfinder . . . . . . . . . . 29
1.5.4 Energy balance terms for noon and midnight conditions from 1-D model. . . . . . . . . 31
1.8.1 Early-storm and mid-storm photos taken during the June 2018 dust storm . . . . . . . . 43
2.2.1 MAVEN’s coverage in altitude, local time and latitude . . . . . . . . . . . . . . . . . . 53

2.2.2 Neutral Gas and Ion Mass Spectrometer schematic . . . . . . . . . . . . . . . . . . . . 54
2.2.3 Comparison of NGIMS and ACC density profiles . . . . . . . . . . . . . . . . . . . . . . 59
2.2.4 Boxplots of the ratio between NGIMS and ACC densities during the Deep Dip campaigns 60
2.3.1 TGO to MG05 density ratios for a profile in January 2018 . . . . . . . . . . . . . . . . . 63
2.4.1 TGO and MAVEN’s coverage in altitude, local time and latitude . . . . . . . . . . . . . 65
3.2.1 Diagram showing convergence of temperature profiles after applied offsets . . . . . . . . 69

3.3.1 Binned periapsis temperature difference between inbound and outbound profiles as a
function of solar zenith angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4.1 Comparison between derived and extracted MCD temperatures . . . . . . . . . . . . . . 73
3.5.1 Comparison between MCD and Viking CO2 and N2 densities . . . . . . . . . . . . . . . 75
3.5.2 Ratio of CO2 , Ar and N2 densities between NGIMS and MCD data . . . . . . . . . . . . 76
3.5.3 Ar/CO2 and N2 /CO2 ratios derived from MCD and NGIMS data in solar zenith angle . . 78
3.6.1 TGO density profile examples as function of time from closest approach and altitude . . 80
3.6.2 Time offset as a function of ratio of latitude to longitude traversed during pass . . . . . 81
11
LIST OF FIGURES
4.2.1 First Deep-Dip derived temperature profiles . . . . . . . . . . . . . . . . . . . . . . . . 85

4.3.1 Temperatures interpolated onto the 160 km, 170 km, 180 km, 190 km and 200 km
altitude levels and shown against solar zenith angle . . . . . . . . . . . . . . . . . . . . 87
4.3.2 Average temperature profiles for solar zenith angles less than 30° and greater than 150° . 88
4.4.1 DD1 derived temperature gradient profiles . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.4.2 Horizontal temperature gradients in solar zenith angle binned by solar zenith angle and
altitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.5.1 27-day rolling average density data interpolated at 190 km . . . . . . . . . . . . . . . . 94
4.6.1 Seasonal density variations shown in solar longitude. . . . . . . . . . . . . . . . . . . . 102
4.6.2 Seasonal temperature variations shown in solar longitude. . . . . . . . . . . . . . . . . . 104
4.6.3 Temperature ratios during overlapping sampling periods . . . . . . . . . . . . . . . . . . 107
4.6.4 Average temperatures on the dayside and nightside near perihelion and aphelion inter-
polated at 150 km, 160 km, 170 km and 180 km as a function of latitude. . . . . . . . . 108
5.2.1 Example extraction of perturbations from temperature and density waves . . . . . . . . 115
5.3.1 Estimated wavelengths along spacecraft trajectory shown as a function of angle of tra-
jectory to the normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.4.1 Histograms of wave amplitudes from all studied MAVEN orbits . . . . . . . . . . . . . . 120
5.4.2 Histograms of wave wavelengths from all studied MAVEN orbits . . . . . . . . . . . . . 121
5.5.1 Average gravity wave amplitude as a function of altitude . . . . . . . . . . . . . . . . . 122
5.6.1 Gravity wave amplitudes as a function of solar zenith angle and Mars-Sun distance . . . 126
5.7.1 Averaged gravity wave amplitudes superimposed over Martian topography . . . . . . . . 129
5.8.1 Wave propagation between orbits 4092 and 4093 waves, and orbits 6833 and 6834 . . . 131
5.8.2 Comparison between waves from orbit 3811 and 3812 with the former shifted by 5 km,
1 km and 15 km . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5.8.3 Waves extracted from inbound pass of orbit 6107 and 6108 with corresponding temper-
ature profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.2.1 Binned CO2 , N2 and Ar densities as a function of solar longitude during the onset of the
June 2018 global dust storm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.2.2 Averaged inbound CO2 , Ar and N2 density profiles and ratios for pre-storm and storm
orbits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
12
LIST OF FIGURES
6.2.3 Averaged outbound CO2 , Ar and N2 density profiles and ratios for pre-storm and storm
orbits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.2.4 Post-onset to pre-onset density ratios taken at 170 km as a function of molecular mass . 146
6.2.5 Post-onset altitude as a function of pre-onset altitude for inbound and outbound legs . . 147
6.3.1 Inbound densities throughout June 2018 dust storm . . . . . . . . . . . . . . . . . . . . 151
6.3.2 Outbound densities throughout June 2018 dust storm . . . . . . . . . . . . . . . . . . . 152
6.3.3 Corrected storm decay rates for June 2018 and Noachis dust storms . . . . . . . . . . . 155
6.4.1 Pre- and post-onset density perturbations during the 2018 global dust storm . . . . . . . 157
7.2.1 Example TGO density profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

7.2.2 TGO wave extraction example with wave spectra . . . . . . . . . . . . . . . . . . . . . 163
7.3.1 Interpolated TGO and MAVEN densities at 110 km, 150 km and 190 km . . . . . . . . 165
7.3.2 Comparison between TGO and MCD densities at 110 km in SZA, longitude, LST and
latitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.3.3 Ratio of TGO to MCD densities at 110 km in SZA, longitude, LST and latitude at 105
km, 110 km, 115 km and 120 km . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.4.1 Derived TGO temperatures at periapsis . . . . . . . . . . . . . . . . . . . . . . . . . . 173
7.5.1 Binned TGO and MAVEN wave spectra normalised with scale height . . . . . . . . . . . 175
7.5.2 Example of concurrent MAVEN and TGO gravity waves . . . . . . . . . . . . . . . . . . 178
7.6.1 Ratio between TGO and MAVEN densities at 125 km altitude as a function of solar
zenith angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
7.6.2 Hydrostatically connecting TGO and MAVEN density and temperature profiles . . . . . 183
A.0.1Latitude-local time overlap period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
13
List of Tables
1.4.1 Comparison of Mars and Earth’s planetary properties . . . . . . . . . . . . . . . . . . . 22

1.4.2 Mars’s seasons in Ls and sol range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.8.1 Global-scale dust storms on Mars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.2.1 MAVEN Deep Dip Ephemeris . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.5.1 Extracted periods observed in density due to solar rotation effects . . . . . . . . . . . . 96

4.6.1 Inbound-inbound latitude-local time overlap . . . . . . . . . . . . . . . . . . . . . . . . 98
4.6.2 Outbound-outbound latitude-local time overlap . . . . . . . . . . . . . . . . . . . . . . 99
4.6.3 Inbound-outbound latitude-local time overlap . . . . . . . . . . . . . . . . . . . . . . . 100
5.2.1 Characteristics of extracted example waves . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.3.1 Dust storm decay timescales for Noachis dust storm . . . . . . . . . . . . . . . . . . . . 153
6.3.2 Dust storm decay timescales determined for June 2018 dust storm . . . . . . . . . . . . 154
7.2.1 Periapsis locations for profiles shown in Figures 7.2.1a-c . . . . . . . . . . . . . . . . . . 161
14
Abbreviations
ACC Accelerometer.
amu Atomic Mass Unit = 1.66×10−27 kg.
AU Astronomical Unit = 1.496×108 km.
CME Coronal Mass Ejection.
DD Deep-Dip.
EDM Entry, Descent and Landing Demonstrator Module.
ESA European Space Agency.
EUV Extreme Ultraviolet.
GCM General (Global) Circulation Model.
IMU Inertial Mass Unit.
IUVS Imaging Ultraviolet Spectrograph.
LMD-MCD Laboratoire de Météorologie Dynamique Mars Climate Database.
LST Local Solar Time.
M-GITM Mars Global Ionosphere-Thermosphere Model.
M-S Distance Mars-Sun Distance.
MARSIS Mars Advanced Radar for Subsurface and Ionosphere Sounding.
MAVEN Mars Atmosphere and Volatile Evolution Mission.
15
Abbreviations
MEX Mars Express.
MGS Mars Global Surveyor.
MRAMS Mars Regional Atmospheric Modeling System.
MRO Mars Reconnaissance Orbiter.
MY Martian Year.
NASA National Aeronautics and Space Administration.
NGIMS Neutral Gas and Ion Mass Spectrometer.
ODY Mars Odyssey.
ROSCOSMOS Roscosmos State Corporation for Space Activities.
TGO Trace Gas Orbiter.
UAMS Upper Atmospheric Mass Spectrometer.
VL Viking Lander.
16
Chapter 1
Introduction
1.1 Introduction
Mars’ atmosphere is highly variable over short and long temporal and spatial scales. This thesis aims to
present an exposition on the dynamic nature of the Martian upper atmosphere (above 100 km), which
has been achieved by using observational data from multiple spacecraft currently orbiting Mars. This
chapter introduces the concepts used throughout the study from a history of observations to basic dust
storm theory. Chapter 2 introduces the spacecraft, instrumentation and subsequent datasets used for
analysis throughout this study. Chapter 3 presents data analysis techniques that are utilised throughout
this study, such as an example derivation of temperature from density data. Additionally, comparisons
between observational and model results are presented. Chapter 4 discusses background trends in
temperature and density with particular emphasis placed on daily and seasonal trends achieved by
analysing data taken during periods of near-identical local time and latitude. Chapter 5 examines wave
activity within the upper atmosphere, and their relationships with solar zenith angle and topography are
investigated. Spatially, these are the smallest variations studied. Chapter 6 investigates the effects of
the 2018 global dust storm on the upper atmosphere. Density increases during the event are quantified,
and decay timescales are derived. The impact of the dust storm on waves is also investigated. Results
are compared to previous dust storm investigations. Chapter 7 uses density data from Trace Gas
Orbiter (TGO) to probe deeper in the upper atmosphere. Results from this are compared to those
earlier in the thesis. Moreover, where possible, datasets are ’combined’ with the Mars Atmosphere and
Volatile Evolution (MAVEN) data to construct profiles spanning the upper atmosphere. As evident by
the summary of chapters’ topics above, this thesis covers a diverse range of subjects. As such, the
Introduction endeavours to present the basic knowledge required to understand each chapter more or
17
1.2. MOTIVATION FOR STUDYING MARS
less in the order each topic appears.
1.2 Motivation for Studying Mars
Mars provides an excellent arena to study both geological and atmospheric physics as it shares similarities
with Earth but can also be strikingly different in other aspects. Water is necessary for forming and
sustaining life as we know it, with water being known to have flowed on the Martian surface; so, was
there life on Mars at some part in its past? If not, what other conditions need to be met to sustain life
on the Red Planet? These questions can only be answered by sending spacecraft to Mars, performing
experiments and collecting invaluable data. Mars is untouched and provides a laboratory to study
geological processes by examining surface and below-surface soil. Mars is believed to have had a similar
natural landscape to Earth, given the large quantity of water thought to have been present there based
on geological data. By studying Mars, it may be possible to understand why two potentially once similar
planets have diverged into their current states. Will the natural evolution of Earth tend towards a Mars-
like world? Mars once had a thick CO2 atmosphere which is slowly being lost to space. To what extent
and at which point did the loss of the atmosphere lead to the severely altered landscape that we observe
today? Looking further afield and to the future, newly discovered planets could fall into the Mars-like
category. Therefore, understanding our closet neighbours aid future analyses of distant places. Finally,
an exciting thought is human space travel and the colonisation of Mars. Understanding the extent and
nature of the variability of the Martian atmosphere in addition to its scientific significance is of practical
importance for space mission planning. What effects do dust storms have on orbiting spacecraft? How
do gravity waves impact the expected trajectory of satellites? Although these topics are not explicitly
investigated here, such effects on the atmosphere are assessed for future studies.
1.3 Previous Observations of Mars
1.3.1 Pre-Spacecraft Observations
Observations of Mars are not restricted to modern times. Humans have been fascinated by Mars for
thousands of years. The below information is summarised from https://marsmobile.jpl.nasa.gov/
allaboutmars/mystique/history/. Around 400 BC the Babylonians used celestial bodies to predict
astronomical events and observed Mars, though they saw it as a mysterious red spot in the sky. With
the advancement of telescopes over the next 1000 years, Galileo Galilei was the first person to observe
18
1.3. PREVIOUS OBSERVATIONS OF MARS
Mars using such a device. Our knowledge of Mars strengthens during the 17th century. By using fixed
land features on Mars, a solar day of the Red Planet was refined from 24 hours by Christiaan Huygens to
24 hours 40 minutes by Giovanni Cassini. It is now taken to be 24 hours 39 minutes and 35.244 seconds
(Allison and Schmunk, 2018). In 1877, Schiaparelli produced a map showing what are thought to be
large canals on Mars. By the start of the 20th century, many features of Mars had been discovered.
Yellow clouds had been spotted, which were taken to be dust clouds. The ice caps and their seasonal
expansion and reduction had been noted. The moons of Mars are discovered during this century. Mars’
main features have been identified and contemplated from tens to thousands of years, and it is exciting
to live during a time of great space exploration to understand these distant features further.
1.3.2 Previous Spacecraft Observations
Figure 1.3.1 shows an infographic displaying all successful (and less successful) missions to Mars up
to NASA’s Mars 2020 rover mission (due for launch in July 2020). The first successful flybys of Mars
were in 1964 by Mariner 3 and 4, followed five years later by Mariner 6 and 7. Images from Mariner
4 inferred that a dense atmosphere had not been present since the formation of the cratered surface
(Leighton et al., 1965). Further images from Mariner 6 and 7 agree with this. The Viking 1 and Viking
2 missions in 1975 were the first successful landers and allowed in situ measurements of the atmosphere
to be made for the first time as they descended through the atmosphere. The Upper Atmospheric Mass
Spectrometers (UAMS) on board Viking 1 and 2 gave the first measurements of atmospheric composi-
tion between 100 km and 200 km. CO2 was found to be the dominant species along with N2 , CO, O2 ,
NO and other trace constituents (Nier and McElroy, 1977). Like the Viking spacecraft, Pathfinder used
accelerometer data to determine atmospheric density, pressure and temperature profiles (Seiff and Kirk,
1977; Magalhães et al., 1999). NASA’s Pathfinder consisted of a lander and rover, Sojourner, which
was the first successful rover to land on the Red Planet. Further, Sojourner was the first rover to roam
outside of the Earth-Moon system. Pathfinder was a trailblazer for future rover missions.
Twenty-one years after the successful Viking landings, Mars Global Surveyor (MGS) graced the
atmosphere with its presence. Through its aerobraking campaign, MGS was the first spacecraft to
measure atmospheric density extensively (Lyons et al., 1999). The physics of aerobraking is discussed
in Section 2.2. MGS made significant discoveries including, but not limited to, information about Mars’
10 km layered crust (Malin et al., 1998), ancient lava flows (Hartmann et al., 1999) and evidence of the
presence of sources of liquid water at shallow depths beneath the Martian surface (Malin and Edgett,
19
1.3. PREVIOUS OBSERVATIONS OF MARS
Figure 1.3.1: An infographic from @TheMarsSociety depicting both successful and unsuccessful mis-
sions to Mars (@TheMarsSociety, 2016).
2000). Following on from Pathfinder’s success, NASA’s Mars Exploration Rover mission consisted of
Spirit and Opportunity. Their objective was to study the surface and geology of Mars. Arriving in
January 2004, both rovers explored Mars for over five years, with Opportunity finally succumbing to
the harsh conditions found at Mars. Opportunity discovered signs of water in the first rocks it studied
(Squyres et al., 2004). NASA’s Mars Science Laboratory (MSL) consisted of another Rover, Curiosity.
The objectives of this rover are to study the climate and geology of Mars, leading to conclusions about
the habitability of Mars. As of December 2020, Curiosity is still operational. NASA’s Mars Atmosphere
and Volatile Evolution (MAVEN) arrived at Mars in 2014 and is the first spacecraft to fully characterise
the composition of the thermosphere (above 100 km). Data from this mission are used throughout this
thesis; as such, it is detailed further in Chapter 2. The first of the two-part joint ESA/ROSCOSMOS
20
1.4. MARS PLANETARY PROPERTIES
mission, ExoMars, arrived in 2016. The primary purpose is to gain a better understanding of the trace
gases present in the atmosphere. Acceleration data are used from this spacecraft throughout this thesis;
thus, the mission is discussed further in Chapter 2. The most recent mission to Mars is NASA’s InSight;
a lander tasked with measuring seismic activity within Mars’ interior. InSight successfully landed at
Elysium Planitia in November 2018.
There are many future missions planned for Mars in 2020 alone: UAE’s Hope Mars Mission (cli-
mate and weather - e.g. dust storms), NASA’s Mars 2020 Rover (ancient environment and geological
processes), ESA/ROSCOSMOS’s ExoMars Rosalind Franklin (search for biomolecules or biosignatures
from past life), and China’s Mars Global Remote Sensing Orbiter, Lander, and Small Rover (search for
current and past life). Each dataset will add a new and unique perspective of the Red Planet.
1.4 Mars Planetary Properties
Mars is the furthest terrestrial planet from the Sun and shares similarities with Earth. Table 1.4.1 shows
a comparison between these two planets’ properties. Mars has a radius of 3394 km, which is nearly a
factor of two smaller than Earth’s at 6378 km. Mars’ significantly smaller mass leads to a gravitational
acceleration of ∼3.72 ms−2 at the surface, compared to 9.81 ms−2 at Earth. This smaller acceleration
reduces the velocity needed for particles to escape Mars’ gravitational field, which has implications when
investigating the atmospheric loss of lighter species. The behaviour of species’ densities has a depend-
ence on gravitational acceleration, thus has implications on the density distribution. A Martian year is
equivalent to ∼1.88 Earth years. The length of a solar day (sol) on Mars is slightly longer than a day
on Earth, measuring 24h 39m 35s. This small difference lends itself to the formation of similar tides,
as governed by the rotation period. Waves are produced with harmonics being an integer multiple of
the rotation period. The equilibrium surface temperature is lower on Mars (210 K) than Earth (256 K).
This is primarily due to differences in their respective planet-Sun distance. For Mars, this is 1.38-1.67
AU (1 AU is equal to 1.496×108 km) and 0.98-1.02 AU for Earth. The difference in albedo between
the two planets plays a role too. Earth and Mars have respective values of 0.250 and 0.306, respectively
(McDunn, 2012). The difference in planet-Sun distance causes the solar flux at Mars to be about
35-50% of that at Earth. The planetary inclinations are comparable also; Mars has a tilt of 25° to its
orbital plane and Earth’s tilt is 23.5°.
21
Property Mars Earth
Equatorial radius (km) 3394 6378

Mass (kg) 6.42×1023 5.97×1024
Gravitational acceleration at surface (ms−2 ) 3.72 9.81
Equilibrium temperature (K) 210 256
Mean scale height (km) 10.8 7.5
Mean density (kgm−3 ) 3933 5515
Solar day, sol (h) 24.66 24
Solar year (sol) 668.6 365.24
Planet-Sun distance (AU) 1.38-1.67 0.98-1.02
Axial inclination (°) 25 23.5
Longitude of perihelion (°) 251 281
Longitude of aphelion (°) 71 103
Table 1.4.1: Comparison of Mars and Earth’s planetary properties. Adapted from Zurek et al., 1992
and McDunn, 2012
Assuming solar flux at Mars to be proportional to the Mars-Sun distance squared, then solar flux
varies by ∼38% over a Martian year. The solar longitude, Ls, describe seasons. Solar longitude is the
Mars-Sun angle, measured from the Northern Hemisphere spring equinox where Ls=0°. Ls varies from
0-360° over a Martian year. Table 1.4.2 describes the Martian seasons in detail. The year begins at
Northern spring equinox (Ls=0° ), followed by Northern summer solstice (Ls=90° ), then Northern au-
tumn equinox (Ls=180° ), and finally Northern winter solstice (Ls=270° ). Lewis et al., 1999 introduced
’Martian months’, however, these are not universally used within the community. Each Martian month
covers 30° solar longitude; therefore, the duration of each Martian month varies depending on the season
due to its eccentric orbit. The longest month is between Ls=60 and 90° and has a duration of 66.7 sols.
Aphelion occurs during this month at Ls=71° . The shortest month is between Ls=240 and 270° and
has a duration of 46.1 sols. Perihelion occurs during this month at Ls=251° .
22
Month No. Ls Range (◦ ) Sol range Duration (sol) Details
1 0-30 0.0-61.2 61.2 N. Spring Equinox at Ls=0◦

2 30-60 61.2-126.6 65.4
3 60-90 126.6-193.3 66.7 Aphelion at Ls=71◦
4 90-120 193.3-257.3 64.5 N. Summer Solstice Ls=90◦
5 120-150 257.3-317.5 59.7
6 150-180 317.5-371.9 54.4
N. Autumn Equinox Ls=180◦
7 180-210 371.9-421.6 49.7
Dust Storm Season begins
8 210-240 421.6-468.5 46.9 Dust Storm Season
Perihelion at Ls=251◦
9 240-270 468.5-514.6 46.1
Dust Storm Season
N. Winter Solstice Ls=270◦
10 270-300 514.6-562.0 47.4
Dust Storm Season
11 300-330 562.0-612.9 50.9 Dust Storm Season
12 330-360 612.9-668.6 55.7 Dust Storm Season ends
Table 1.4.2: Mars’s seasons in Ls and sol range. Notable events are shown in the far right column.
Adapted from http://www-mars.lmd.jussieu.fr/mars/time/solar_longitude.html
23
1.5. ATMOSPHERIC STRUCTURE
1.5 Atmospheric Structure
1.5.1 Density Structure
An atmosphere is defined as the envelope of gases surrounding a planetary body. For a body to possess
a permanent atmosphere the forces acting upon such an atmosphere must be balanced. Assuming the
atmosphere can be approximated by spherical symmetry, the forces acting on a parcel of gas can be
easily calculated. The three forces acting upon a parcel are a downward force caused by the pressure of
the fluid above, an upward force from the fluid pushing upwards and the weight of the parcel causing a
force downwards. For a mathematical explanation, a short cylinder of gas is considered. The cylinder
has a cross-section area, A, and depth, dz, and density, ρ. This gives the cylinder a total volume of
dV and total mass ρdV . The cylinder is located a distance z from the centre of the body. Given all
forces act within the radial direction, vector notation has been omitted. The upward force exerted on
the bottom of the cylinder, Fz , is given by Equation 1.1
Fz = Pz A (1.1)
where Pz is the atmospheric pressure at z. The downward force exerted on the top of the cylinder,
Fz+dz , is given by Equation 1.2
Fz+dz = −Pz+dz A = −(Pz + dP )A (1.2)
where Pr+dz is the atmospheric pressure at z+dz. dP is the change in pressure over dz. The final force
acting on the cylinder is its weight, Fg , given by Equation 1.3.
Fg = −gdm = −gρdV = −gρAdz (1.3)
For the parcel to be in equilibrium, all three forces (Eqs.1.1-1.3) need to sum to zero (Equation 1.4).
Fz + Fz+dz + Fg = 0 (1.4)
By substituting Eqs.1.1-1.3 into Equation 1.4 leads to Equation 1.5
Pz A − (Pz + dP )A − gρAdz = 0 (1.5)
And solving for dP leads to
dP = −gρdz (1.6)
24
As g and ρ are always positive, dP will be negative when dz is positive. Therefore, pressure decreases
with altitude. The pressure gradient in altitude is given by Equation 1.7
dP
= −gρ (1.7)
dz
This is the equation of hydrostatic equilibrium.
The Ideal Gas Law can be used to determine atmospheric pressure as a function of altitude. The
Ideal Gas Law, Equation 1.8, is given by
P = nkT (1.8)
where n is number density, k is Boltzmann constant equal to 1.38×10−23 m2 kgs−2 K−1 , and T is
temperature . As number density is given by
n = ρ/m (1.9)
where m is molecular mass of the gas. Then the Ideal Gas Law can be rewritten as
P = ρkT /m (1.10)
Equation 1.10 can be rearranged for ρ and substituted into Equation 1.7 leading to
dP mgP
=− (1.11)
dz kT
then
dP mgdz
=− (1.12)
P kT
Integrating Equation 1.12 between r0 and r gives
P mg
ln =− (z − z0 ) (1.13)
P0 kT
where P0 and z0 are initial pressure and altitude levels. Then

P mg

= exp − (z − z0 )
P0 kT
(1.14)
(z − z0 )

= exp −
H
where
H = kT /mg (1.15)
25
is the scale height. H is a constant for thermal equilibrium. It represents the distance over which the
number density decreases by a factor of e.
Using the Ideal Gas Law (Equation 1.10) and the description of pressure in an ideal atmosphere
(Equation 1.16), the density distribution in altitude can be derived. This leads to
ρ mg

= exp − (z − z0 )
ρ0 kT
(1.16)
(z − z0 )

= exp −
H
In order to show density behaviour within the upper atmosphere, data from the MAVEN/NGIMS
mission are shown. Figure 1.5.1 shows the variation of nine atomic and molecular species during the
inbound leg of orbit 1064. Shown are the densities of CO2 , Ar, N2 , O, CO, O2 , NO, N and He from 135
km to 350 km. CO2 is dominant below ∼250 km. The lightest elements (e.g., He, O) have the largest
scale heights; therefore, their densities vary more slowly with altitude. The heavier elements such as
CO2 and Ar have smaller scale heights, shown by stronger gradients in Figure 1.5.1, as they are more
affected by gravity. Straight lines signify that data obey the behaviour of an isothermal atmosphere. A
sharp increase in the gradient shows an increase in temperature.
1.5.2 Temperature Structure
In atmospheric physics, heating by radiation is prompted by the absorption and re-emission of electro-
magnetic energy; the main source of such energy is the Sun. With the Sun as a blackbody, it emits
in the shorter wavelengths with a peak in the visible portion of the spectrum at ∼420 nm. Ultraviolet
(UV) and infrared (IR) radiation are also emitted at lower intensities. UV, EUV and X-rays are absorbed
in the upper atmosphere, causing dissociation and ionisation. The Sun’s radiation is absorbed at the
surface and remitted at longer wavelengths. Additionally, solar radiation is absorbed and will lose energy
due to ionisation. Absorption occurs when matter retains radiant energy. The intensity of radiation
within an atmosphere is now discussed.
An infinitely long cylinder filled with a gas with number density n is considered. Each particle has
an absorption cross section denoted by σ. The initial radiation at an initial distance, s0 , is I(s0 ). After
a distance s the photon flux is I(s). Likewise at s + ds, the photon flux is I(s + ds). Now,
I(s0 ) > I(s) > I(s + ds) (1.17)
26
Figure 1.5.1: An example of the variation with altitude of nine atomic and molecular species during
a single deep dip pass on orbit 1064 (Ls=256, LST 11:50 A.M., and latitude 4.5°S at periapsis on this
orbit) is shown. For the trace gas He, gas scattering in the instrument at the lowest altitudes may
distort the profile. N, O2 , O, and NO are derived from open source measurements and the remaining
gases from closed source data. Taken from Mahaffy et al., 2015a.
The change in photon flux is proportional to the length of the considered cylinder, shown by Equation
1.18
dI ∝ Is ds (1.18)
Equation 1.18 can be rewritten using −σn as a constant of proportionality,

dI
= −σnds (1.19)
Is
where dI
Is is the relative change in photon flux over a path length ds. For completeness, this can be
integrated between zero and s to give
Is = I0 exp(−σns) (1.20)
27
This is known as Beer’s Law and relates the reduction in photon flux to the initial photon flux. The
above is for monochromatic radiation only. The form is dependent on wavelength and composition.
This formulation can be extended to derive intensity as a function of altitude (z) and solar zenith angle
(χν ), can be determined using Beer’s Law. The solar zenith angle is the angle between the Sun and
an observer’s zenith. Solar zenith angle is used throughout this study, consequently understanding its
implications on the observed behaviour is important. Figure 1.5.2 shows a graphic of solar zenith angle.
Figure 1.5.2: Schematic diagram showing monochromatic radiation penetrating a plane and horizont-
ally stratified atmosphere. Taken from Schunk and Nagy, 2009
In an atmosphere, the reduction in radiation intensity along an oblique ray path, dzcos χν , is given
by,
dz
dI = σn I (1.21)
cos χν
Equation 1.21 is positive since altitude is increasing. As the angle becomes more oblique, the ray
travels further through the atmosphere to cover the same altitude range. The radiation as a function
of altitude is determined by integrating Equation 1.21 between infinity and z.
Z z Z z
dI σn0
= exp (−z/H) dz (1.22)
∞ I ∞ cos χν
then
σn0 H

I(z) = I∞ exp exp(−z/H) (1.23)
cos χν
28
Radiation, therefore, increases exponentially with height. Locations equidistant from the subsolar point
should receive equal radiation. Radiation decreases further from the subsolar point as solar zenith angle
increases.
Mars’ vertical temperature structure is similar to Earth due to similar physics occurring within the
atmospheres. Figure 1.5.3 shows temperature profiles derived by Viking 1 and 2 and Pathfinder acceler-
ometer data. The figure has been adapted from Magalhães et al., 1999 to show the three central regions
outlined here - the troposphere, mesosphere, and thermosphere. Earth possesses these three regions as
well as a stratosphere located above the troposphere. Mars does not have an ozone layer like Earth or
equivalent. A gas is required that absorbs UV in that region, which leads to heating. The troposphere
and mesosphere are not explored during this study; however, the physics of each region will be briefly
discussed.
Figure 1.5.3: Atmospheric temperature profile from entry phase of Mars Pathfinder. Modified from
Magalhães et al., 1999
29
The region above ∼100 km shown in Figure 1.5.3 is the thermosphere. This is the least well-
understood region of the atmosphere due to a lack of measurements and the key focus of this thesis.
Within this region, temperatures rapidly increase with altitude as absorption of far and extreme UV radi-
ation becomes progressively more dominant than other heating sources. Considering Mars’ eccentricity
(Table 1.4.1), the solar flux present at Mars varies by ∼40% over a Martian year. The effect of changes
in EUV flux has been modelled by general circulation models (Bougher et al., 2015) and derived from
spacecraft measurements (Thiemann et al., 2018). Bougher et al., 2015 found mid-afternoon temper-
atures near 200 km are predicted to vary from 210 to 350 K (equinox) and 190 to 390 k (aphelion
to perihelion). Solar insolation and EUV-forcing heat the thermosphere from below and directly, re-
spectively. (Thiemann et al., 2018). By using MAVEN EUV solar occultations (EUV-SOs), Thiemann
et al., 2018 aimed to decouple these two heating sources. Data were available to allow this; long-term
decreasing solar variability was dominant, leading to a 50% reduction in EUV between the second and
first perihelion. Like in the mesosphere, CO2 cooling acts as an atmospheric thermostat that regulates
lower thermospheric temperatures, near peak EUV absorption altitudes (Bougher et al., 1994). When
considering the overall thermosphere, CO2 cooling is a secondary coolant (Bougher et al., 1999). CO2
cooling arises from the collision of atomic oxygen with CO2 allowing energy exchange between kinetic
and vibrational states of the latter molecules. Hence, cooling rates are proportional to the abundance
of O (Medvedev et al., 2015). There is a depletion in oxygen due to Mars being farther from the Sun,
which reduces CO2 photolysis. (Bougher and Roble, 1991; Bougher et al., 1999). Yelle et al., 2014
constructed a 1-D model of the Martian upper atmosphere; temperature and composition are calculated
by solving the coupled, time-dependent energy balance, diffusion, and continuity equations. The model
domain covers the 100–250 km altitude region at a resolution of 1 km. They used this model to study
the effects of Comet C/2013 A1 (Siding Spring) on Mars’ atmosphere. Stone et al., 2018 repurposed
this model to understand heating and cooling rates in the Martian thermosphere based on results from
MAVEN’s mass spectrometer measurements. They find that near noon, solar UV heating is balanced
primarily by thermal conduction at high altitudes and CO2 cooling at lower altitudes. They attribute a
peak in CO2 cooling to a combination of a decreasing collision rate with altitude and an increasing O
mole fraction. In summary, the temperature in the thermosphere is balanced by the following heating
and cooling terms: absorption of solar EUV and UV radiation primarily by CO2 and atomic O (heating),
molecular conduction (cooling), horizontal advection (heating and cooling), adiabatic motions (heating
and cooling), and 15 µm IR emission by CO2 (cooling). Figure 1.5.4 shows heating and cooling rates
for noon and midnight conditions, taken from Stone et al., 2018.
30
Figure 1.5.4: Energy balance terms for (top) noon and (bottom) midnight conditions from 1-D model.
The legend in the bottom panel refers to both panels. Taken from Stone et al., 2018
The final region, not shown in Figure 1.5.3, is the exosphere. The bottom of the exosphere, the
exobase, is defined as the altitude at which the atmospheric scale height, H, equals the gas molecular
mean free path or, equivalently, the altitude at which an upward-moving atom has a 1/e chance of not
undergoing a collision before escaping to space. Below the exobase, a particle is likely to lose its energy
through collisions and behave thermodynamically, but above, in the exosphere, it will be on a ballistic
31
trajectory. Jakosky et al., 2017 calculated the exobase location using both these criteria. Exobase
altitudes vary from ∼140 km to 200 km and vary due to a combination of variations with latitude, local
solar time, solar zenith angle, and season. It is, therefore, not entirely possible to decouple these factors.
The calculated altitudes are lower than previously quoted. For example, Haberle, 2015 state altitudes
above 220-230 km, with variation caused by solar activity. While the heavier molecules that begin to
escape are typically captured by the planet’s gravitational potential and thus return to the atmosphere
along parabolic trajectories, a proportion of lighter molecules (albeit a tiny proportion, currently) do
indeed escape to space.
Above, the energy budget in each region has been outlined. Additionally, dynamics play an essential
role in the redistribution of energy. Most impactful is convection, whereby winds transport energy from
warmer regions, such as at lower solar zenith angles, to colder regions. This can lead to winter polar
warming which is examined in Chapter 4.
The atmosphere can be ’split’ in another way by considering the dominant diffusion process at dif-
ferent altitudes. In the lower atmosphere (below ∼100 km) mixing by dynamical processes acts to keep
relative chemical compositions constant with pressure. Species’ abundances are controlled by chemistry.
Within this region there a common scale height for all species, given by H = kt/µg, where µ is the
mean molecular weight of species. Turbulence, or eddy diffusion, is dominant and mixes the species
to counteract the distribution caused by gravitational effects. Its dominance is due to a small mean
free path owing to significant pressure and densities. The lower atmosphere is often referred to as the
homosphere. Above this region is the upper atmosphere, also known as the heterosphere. In the het-
erosphere, gases will be distributed vertically according to their scale height due to molecular diffusion.
In Mars’ heterosphere, the more massive species such as CO2 will dominate the lower altitudes above
the homosphere and lighter species such as O will be dominant at higher altitudes (Mueller-Wodarg
et al., 2008). The separation region between the lower and upper atmosphere is called the homopause
or turbopause. Strictly, the homopause is not constant for all species as the diffusion coefficients for
lighter species are larger than those for more massive species, so the altitude of the homopause is lower
for species such as H, H2 and He (Fox, 2004). Jakosky et al., 2017 investigated how the altitude of the
homopause varies using a year’s worth of Neutral Gas and Ion Mass Spectrometer (NGIMS) data. By
extrapolating the ratio of N2 /Ar, they were able to find the altitude at which the ratio is equal to that
found near the surface by Mahaffy et al., 2013. Homopause locations vary from 150 km down to as
32
1.6. MARTIAN GENERAL CIRCULATION MODELS
low as 50 km. Average altitudes are between 80 and 120 km. The rising and falling of the homopause
are likely in response to the behaviour of the lower atmosphere. Seasonal forcing is believed to have a
more substantial effect on the lower atmosphere. In this study, the upper atmosphere refers to altitudes
above ∼100 km and is the main focus of this thesis.
1.6 Martian General Circulation Models
Much of our understanding of Martian atmospheric dynamics comes from the successful implementation
of general circulation models (GCMs). The coupled Navier-Stokes equations of energy, momentum,
and mass continuity are solved on a spherical surface calculate winds, temperatures and composition.
The earliest GCMs were developed for Earth’s atmosphere, and their success lead to the adaption and
employment of these models to other planets. The Laboratoire de Météorologie Dynamique Mars Climate
Database (LMD-MCD) is used in this thesis. It is a database fed by the GCM of González-Galindo et al.,
2015. This model has an online interface (http://www-mars.lmd.jussieu.fr/mcd_python/) which
allows users to produce plots for myriad fields easily.
1.6.1 Laboratoire de Météorologie Dynamique Mars Climate Database
The Mars Climate Database (MCD) v5.3 is a database of atmospheric statistics compiled from state-of-
the-art GCM simulations of the Martian atmosphere (Lewis et al., 1999; Forget et al., 1999; Millour et al.,
2018; González-Galindo et al., 2015). The MCD follows a whole atmosphere approach from the surface
up to ∼300 km in this most recent version. Simulated data are stored on a 5.625°×3.75° longitude-
latitude grid and are interpolated between data points if necessary. Data are available for an entire
Martian year. In essence, for each grid point, the database contains 12 "typical" days, one for each
month. There are many drivers behind variability in Mars’ atmosphere. Two such factors are solar
conditions and suspended dust within the atmosphere. The former is incorporated by simulating solar
minimum, average or maximum conditions. In this present study, climatology scenarios are used; these
are designed to represent a typical Martian year. Solar minimum is used throughout this thesis. The
MCD allows for more infrequent scenarios. There are three dust storm scenarios where opacity, τ , is
set to 5 for solar minimum, average and maximum. ’Warm’ and ’cold’ scenarios allow particularly dust-
heavy periods to be captured outside of dust storms. Lastly, historical conditions from Martian years
(MYs) 24-32 are reproduced using best guesses for actual dust and solar EUV conditions. The MCD
33
is validated using available measurements across a variety of missions. Atmospheric temperatures are
compared to measurements from MGS/TES (Thermal Emission Spectrometer), MRO/MCS (Mars Re-
connaissance Orbiter/Mars Mars Climate Sounder) and MGS and Mars Express (MEX) radio occultation
experiments. Surface pressures and temperatures are compared to TES, Viking, Pathfinder, Phoenix,
and Mars Science Laboratory (MSL) measurements (Millour et al., 2015). Chapter 3 demonstrates the
successes and drawbacks of the MCD in the upper atmosphere.
The MCD interface is open and versatile, allowing users to extract data from 12 climatological
averages and location. For example, the vertical coordinate may be given as an altitude above the
surface or a pressure coordinate. Further positional arguments are local time, latitude, longitude and
solar longitude. Mars’ season can also be determined from entering an Earth date. The MCD can output,
amongst other variables, the following: atmospheric pressure (Pa), density (kg/m3 ), temperature (K),
zonal and meridional components of wind (m/s). Hence, data can be extracted from the MCD along
spacecraft trajectories, leading to comparisons between model and spacecraft data. Number densities
are available from the mass spectrometers on board MAVEN. However, the MCD outputs total density
and volume mixing ratios. Partial densities cannot be derived from these quantities directly. The volume
mixing ratio of the ith species, Xi , is equal to ni /ntotal where ni is the number density of the ith species
and ntotal is the total number density. This is equal to Pi /Ptotal where Pi is the partial pressure of the
ith species and Ptotal is the total pressure. Lastly, the partial pressure is given by the Ideal Gas Law,
Pi = ni kT . Combining the previous three equations leads to Equation 1.24,
Xi Ptotal
ni = (1.24)
kT
This derived number density is now comparable to mass spectrometer data.
1.6.2 Other Martian General Circulation Models
The success of Martian GCMs is owed to their Earth-based ancestors. Modelling work began after the
launch of the early Mars missions, such as the flyby Mariner 4 spacecraft. The first simulation was
performed by Leovy and Mintz, 1969. Within a couple of decades, GCMs had developed further in
their capabilities. The National Center for Atmospheric Research (NCAR) Earth-based thermospheric
general circulation model (TGCM), as described in Dickinson et al., 1984, is one such ancestor. This
34
model was successfully modified for Venus (VTCGM) by Bougher et al., 1988b. Specific processes
such as wave drag and CO2 cooling were introduced. The VTCGM has been validated using Pioneer
data. This was adapted by Bougher and Dickinson, 1988 and later Bougher et al., 1990 for Martian
conditions; this is somewhat easier due to Mars and Venus sharing some thermal and radiative pro-
cesses common to CO2 dominant atmospheres. The MTCGM is a finite difference primitive equation
model that self-consistently solves for time-dependent neutral temperatures, neutral-ion densities, and
three-component neutral winds over the Mars globe. Unlike the MCD, the MTGCM is not a whole
atmosphere model; it stretches from 70-300 km. This model is driven from below by the NASA Ames
Mars MGCM around the 60-70 km level (Haberle et al., 1999). The Mars Global Reference Atmospheric
Model (MarsGRAM) is based on the coupling of MGCM and MTGCM. Bougher et al., 2015 introduced
a new GCM, the Mars Global Ionosphere-Thermosphere Model (M-GITM). This model is born from the
GITM, described in Ridley et al., 2006. The latter is non-hydrostatic. As for previous models, planetary
parameters are adjusted for Mars. M-GITM extends from 0 to 250 km. González-Galindo et al., 2010
examined thermal and wind structures output by two models - LMD-MCD and MTGCM (coupled to
a lower atmosphere model). Common input parameters are used and different dust scenarios are used
at Ls=0°, Ls=90° and Ls=270°. González-Galindo et al., 2010 conclude that both models are in good
overall agreement; however, local features are not necessarily observed in both models, but are globally
similar. Local features are exacerbated during increased dust loading.
Only recently has an empirical model of Mars’ thermosphere begun to be developed. MAVEN’s
vast dataset has allowed this type of model to become a reality. Girazian et al., 2017 outline how, for
any given combination of solar zenith angle, latitude, season and solar activity, the model will provide
vertical profiles of neutral densities and temperature in the range 150-300 km. Functions are created
to ensure that density data are available across the planet. Where needed, data from the M-GITM are
used to bridge any gaps. It is expected that one significant advantage of this class of model outputs
near physically sampled regions is likely to be more accurate than from a numerical model. No literature
is currently available that compares these model densities to observations.
35
1.7. GRAVITY WAVES
1.7 Gravity Waves
1.7.1 Gravity Wave Generation
Gravity waves are generated from a variety of sources including flow over topography, atmospheric
instabilities, and volatile convection; however, any process which perturbs the atmosphere could generate
gravity waves. A basic description of the generation of gravity waves is as follows. If a parcel of gas
in the atmosphere is considered then when it encounters an obstacle, such as a mountain range, it will
be displaced upwards. This is the direction of energy propagation. It may be displaced downwards and
reflected upwards, so this explanation still holds. If the parcel has density ρ0 , volume V and is displaced
from its initial position (z, p) to (z + z 0 , p − p0 ) it will experience a force acting on it. Here, z and
p are altitude and pressure; primed variables are perturbations. As the parcel is displaced upwards, its
density will be greater than the surrounding atmosphere. Gravity acts as the restorative force attempts
to reinstate the parcel to its original location, and it will overshoot. The parcel now has a density
lower than its surroundings and will be restored by buoyancy, again overshooting upon its return. From
Newton’s Second Law, the motion of the parcel is governed by Equation 1.25
d2 z 0
(ρ0 V ) = g(ρV ) − g(ρ0 V ) (1.25)
dt2
By using the Ideal Gas Law, the above equation becomes,
d2 z 0 ρ − ρ0

= g (1.26)
dt2 ρ0
1/T − 1/T 0

=g (1.27)
1/T 0
0
T −T

=g (1.28)
T
If T is the unperturbed background temperature, the temperature of the parcel at z + z 0 is T -Γd z 0

where Γd is the lapse rate equal to -dT /dz. The background temperature at z + z 0 will be T − Γz 0 ,
where Γ is the dry adiabatic lapse rate equal to gcp . cp is heat capacity at constant pressure. Now,
T 0 − T = −(Γd − Γ)z 0 . The motion of the air parcel can be written as
d2 z 0 g
2
= − (Γd − Γ) z 0 (1.29)
dt T
g 1/2
If the oscillation frequency is defined as N = T (Γd − Γ) , then the equation of motion becomes
d2 z 0
+ N 2z0 = 0 (1.30)
dt2
36
1.7. GRAVITY WAVES
The oscillation frequency N is called the Brunt-Väisälä (B-V) frequency and its value is important
in determining the behaviour of gravity waves. There are three regimes the parcel can be found in,
determined by the value of N 2 . For N 2 >0, the oscillation is stable and will continue to oscillate at
its original frequency if no external forces act upon it. For the case N 2 = 0, the weight and buoyancy
terms are equal, resulting in no net force on the parcel, therefore remaining stationary. The last case is
N 2 < 0 where the parcel will continue to move in the direction it was originally displaced and is said
to be unstable. The Brunt-Väisälä frequency is used to quantify the stability of the atmosphere and is
used frequently to calculate gravity wave properties such as potential energy and wind perturbations. If
required, the value is taken from the literature. The described oscillations are rarely found purely in the
vertical direction; waves nearly always have a horizontal component. Hence, periods of gravity waves are
dependent on the ratio between the vertical and horizontal wavelengths. The shortest period waves are
nearly purely vertically propagating and are labelled ’acoustic-gravity’ waves. In contrast, an oscillation
primarily in the horizontal directions takes longer to return to its original position, and as such have
more prolonged periods. These types of waves are ‘inertial-gravity’ (Medvedev and Yiğit, 2019).
1.7.2 Gravity Wave Evolution
Gravity waves couple the lower and upper atmosphere at Mars. Understanding the propagation of
gravity waves is crucial to understanding the dynamics of the atmosphere. In the following section, the
theoretical behaviour of gravity waves is described. Like all waves, gravity waves have speeds associated
with them. The most useful is the intrinsic phase speed (c − u) which is the phase speed (c) measured
relative to the mean flow (u). It is this speed that determines whether waves will be absorbed by the
mean flow. In a conservative atmosphere where dissipation and heating are neglected, the vertical flux
of energy, p0 w0 associated with a gravity wave is related to the vertical flux of net horizontal momentum,
(ρ0 u0 w0 ) by
p0 w0 = (c − ū)ρ0 u0 w0
where p0 , w0 , u0 are the perturbation pressure, vertical and eastward wind speeds and ρ0 is the background
density. This is Eliassen-Palm’s first theorem (Eliassen and Palm, 1961). An upward propagating wave
(p0 w0 > 0) carries eastward momentum if the phase speed is eastward with respect to the mean flow
or c > ū. The ‘noninteraction theorem’ derived by Eliassen and Palm, 1961 states that for steady-state
internal gravity waves (IGWs),
d
(ρ0 u0 w0 ) = 0 for ū 6= c (1.31)
dz
37
1.7. GRAVITY WAVES
u0 and w0 must therefore increase with height as density falls exponentially. If ū = c, the mean flow
absorbs the waves and acts as a barrier to propagation. The wave spends more time in the flow allowing
dissipative processes to act for longer (Vallis, 2006). There is no distinction between the flow and wave,
so the latter is absorbed and can no longer propagate. For cases where the noninteraction theorem is
valid and ū 6= c, waves pass vertically through a sheared environment without extracting or depositing
momentum. Momentum is only extracted from the layers where waves form and deposited where they
are finally absorbed. In a real atmosphere where d 0 0
dz (ρ0 u w ) 6= 0, the condition for ū = c is not ne-
cessary for momentum deposition into the atmosphere. The properties of the medium do not change
in time, nor horizontally, therefore both the absolute frequency and horizontal wavelength are invariant
with altitude. However, the vertical wavelength does vary due to the intrinsic frequency changing. If
the flow speed increases, then the intrinsic frequency decreases. This occurs at the critical layer. At
this layer, the vertical wavelength tends to zero. Waves can grow with altitude until acted on by other
forms of dissipation (eddy and molecular diffusion). The former is due to mixing owing to eddy motion.
The latter is the thermal motion within a mixture. Waves can also break analogously to ocean waves
if their amplitudes grow too large. Understanding how much and where momentum is deposited into
the atmosphere is crucial to understanding the energy and momentum budget and upper atmospheric
dynamics (e.g. Medvedev and Yiğit, 2012). Momentum of the wave can be imparted on the general
background flow, for example with topographically generated gravity waves slowing down the mean
flow. A further threshold for waves to contend with is instability. As a parcel is displaced upwards, it
cools. Whether a parcel continues to oscillate is dependent on the surrounding environment. If the dry
adiabatic lapse rate that is the lapse rate of the parcel is greater than the ambient (or actual) lapse
rate, the parcel is sufficiently cooler than the atmosphere, thus will sink to its original position. For the
opposite scenario, the parcel is warmer than its surroundings and consequently will continue to rise.
Gravity waves carry substantial momentum as they propagate from the lower to the upper atmo-
sphere, so constraining wave energy and momentum deposition is crucial in understanding the dynamics
of the upper atmosphere. As waves dissipate or break, their mechanical energy is irreversibly transferred
to the atmosphere as heat. Secondly, it transpires that as amplitudes decay, the vertical net flux of
sensible heat is no longer zero, as is found for conservatively propagating gravity waves. This flux is
directed downwards, and the divergence of such flux causes heating (cooling) below (above) the region
of dissipation (Walterscheid, 1981). Cooling rates in the Earth’s atmosphere due to gravity waves were
found to be ∼100 K/day. Specific cases can present cooling of 1000 K/day (Yiğit and Medvedev,
38
1.7. GRAVITY WAVES
2009). Gravity waves also dynamically impart their influence on the upper atmosphere. One process by
which this is possible is gravity waves ‘drag’. The divergence of the wave momentum flux can lead to an
acceleration or deceleration of the mean flux. As waves break turbulence and dissipation occur, leading
to a zonal force on the zonal flow. Barnes, 1990 predicted gravity wave drag of ∼1000 ms−1 sol−1 . This
is commensurate to drag found by Fritts et al., 2006 using MGS accelerometer data.
1.7.3 Gravity Wave Observations
Gravity waves have been observed in the Martian thermosphere by previous spacecraft, primarily during
their aerobraking campaigns. MGS density data were initially used for studying larger-scale phenomena,
such as planetary-scale waves (e.g. Keating et al., 1998). Withers, 2006 quantified oscillatory trends in
density profiles and interpreted them as gravity waves based on work by Bougher et al., 1999 and Tolson
et al., 1999. Typical amplitudes were found to be ∼10%. Creasey et al., 2006a followed up on this work
by deriving characteristic along-track wavelengths to be several hundred kilometres. A seasonal trend
was found with significantly larger amplitudes during northern autumn compared to northern spring.
Fritts et al., 2006 further expanded our understanding of wave structures in the upper atmosphere
by utilising MGS and Odyssey (ODY) data. Wave growth with altitude was observed, in some cases
by a factor of five over a 25 km altitude range. Research and interest into upper atmosphere wave
structures were revitalised with the arrival of MAVEN in 2014. Early on, Yiǧit et al., 2015a investigated
CO2 density perturbations in the Martian thermosphere and interpreted them as gravity waves. They
compare observations to outputs from a gravity wave scheme. Similarly, England et al., 2016 extracted
perturbations from CO2 , Ar and N2 density data. The larger dataset allowed monthly means and
seasonal effects to be explored. Terada et al., 2017 similarly investigated Ar density perturbations;
the global distribution of gravity wave amplitudes was examined so potential links with topography
could be found. Wave activity is briefly discussed in Stone et al., 2018 during their derivations of
temperature profiles. Observational studies have been able to estimate the momentum fluxes due to
gravity waves. Barnes, 1990 estimated a drag of ∼1000 ms−1 sol−1 around 50-100 km. By scaling down
momentum fluxes according to the scaling of amplitudes, Fritts et al., 2006 calculated momentum flux
at ∼70-80 km to be ∼1000 ms−1 sol−1 , increasing (decreasing) by a factor of five at 100 km (50 km)
using MGS and ODY accelerometer data. Accelerations caused by such momentum deposition is ∼70
ms−2 . Such momentum fluxes and accelerations are derived from limited data. Much of the following
discussion is taken from the summary of gravity waves presented by Medvedev and Yiğit, 2019 and Yiğit
and Medvedev, 2019. All the above studies and individual results are discussed further in Chapter 5.
39
1.7. GRAVITY WAVES
Additionally, observations of waves in different solar system atmosphere are highlighted.
1.7.4 Modelling Gravity Waves
GCMs such as the LMD-MCD are developing methods in which the effects of gravity waves can be
included in models (Forget et al., 1999). Gravity waves occur at sub-grid scales and as such need to
be parameterised. Yiǧit et al., 2008 outlines a gravity wave scheme which has been implemented into
Martian GCMs. This example scheme accounts for wave dissipation in the upper atmosphere due to
molecular viscosity, thermal conduction, ion friction, and radiative damping in the form of the Newtonian
cooling. Breaking/saturation effects are also included. This is one scheme for which its model results
are discussed. Medvedev et al., 2011 implemented the gravity wave scheme of Yiǧit et al., 2008 into a
GCM described by Hartogh et al., 2005 and Medvedev and Hartogh, 2007. They studied the effects in
the 100-130 km region. Medvedev et al., 2011 found that gravity waves decelerate atmospheric zonal
flows during all seasons. In some cases, zonal flows were reversed. Reversals are primarily driven by
deposition of momentum from gravity waves with intrinsic phase speeds greater than zero. Effects of
gravity waves on atmospheric temperature were briefly studied, and winter polar warming was found.
Medvedev and Yiğit, 2012 studied the thermal effects of gravity waves on the Martian atmosphere. For
perpetual Northern winter (Ls = 270°), they found that heating due to irreversible conversion of gravity
wave mechanical energy into heat is comparable to near-IR CO2 heating. Cooling by wave-induced
downward flux is similar to IR CO2 cooling, in agreement with Parish et al., 2009. Overall, the inclusion
of thermal effects of gravity waves systematically produces a colder thermosphere. Results provided
by the model are in agreement with SPICAM observations (Forget et al., 2009) and ODY aerobraking
temperature retrievals (Bougher et al., 2006). Gravity waves have also been implemented into models
to explain other features within the Martian atmosphere. Yiǧit et al., 2015b looked at CO2 clouds and
the possibility of gravity waves creating pockets in the atmosphere with temperatures below the con-
densation point of CO2 . The temperatures and locations are in agreement with observations; however,
the frequency of cloud production exceeds observations.
For gravity waves to be parameterised and evolved through a model atmosphere, initialisation con-
ditions are needed. By comparing gravity wave characteristics at different altitudes, comparisons to
model results at similar altitudes can be made to fully understand the evolution of gravity waves and
their consequential effects on the upper atmosphere. This is possible with the addition of MAVEN
observations as it is the first mission to study the upper atmosphere; thus a wealth of new data are
40
1.8. DUST STORMS
available to quantify and characterise gravity waves in different regions at different times. This is a key
motivation for this thesis. Models require validation; this is achievable by a comparison of amplitudes
and wavelengths. This is expanded upon in Chapter 5. As shown above, much work has been undertaken
to implement gravity wave parameterisation within GCMs; this, of course, develops our understanding
of physical processes and impact within the upper atmosphere.
Nevertheless, one reason for studying gravity waves is of a practical nature. As future spacecraft
arrives at the Red Planet, an understanding of wave variability is needed for managing orbiters and
ensuring the safety of their on board instruments during aerobraking.
1.8 Dust Storms
Martian dust storms have fascinated scientists for years owing to their unpredictability, complexities, and
rarity. Furthermore, for the first time, the upper atmosphere is probed during such a global event. The
June 2018 global dust storm provides an arena to study all aspects of this episode. The lower atmosphere
houses many spacecraft capable of detecting and measuring elements of the storm, such as the opacity
by Opportunity. Does the upper atmosphere react in a similar way to the lower portion? Moreover,
if not, how does it respond? Many questions like these are answered in Chapter 6. Observations of
previous storms from ground-based and spacecraft instruments are discussed. Next, the physics of dust
storms is briefly outlined.
1.8.1 Dust Storm Observations
Global dust storms are rare occurrences at Mars; Table 1.8.1 lists all known dust storms excluding the
June 2018 dust storm (Shirley, 2015). There was a 15-year wait after the 1956 storm for another then
a succession of 4 storms within ten years developed. MRO was, fortunately, able to study the 2007
dust storm. However, observational data for earlier storms becomes sparse (e.g. Bougher et al., 2001).
Mariner 9 detected the electron density response of the 1971 dust storm. This is discussed and mod-
elled in Wang and Nielsen, 2003. Zurek and Martin, 1993 studied the interannual variability of dust
storms on Mars. By using Earth-based observations of Mars and spacecraft, they attempt to understand
whether the observed rate of dust storm formation is representative of previous decades. Due to Mars’
elliptical orbit, Zurek and Martin, 1993 makes the point that dust storms have occurred without being
detected from Earth evidenced by Viking data. They conclude by saying that the chance of observing
41
1.8. DUST STORMS
Year Mars Year Ls (°) Calendar Dates
1924-1925 -16 310 5 Dec-Jan

1956 1 249 19 Aug-Nov
1971-1972 9 260 22 Sep-Jan
1973 10 300 13 Oct-Dec
1977 12 204 15 Feb-Apr
1977 12 268 27 May-Oct
1982 15 208 Oct
1994 21 254 9 Apr-Jul
2001 25 185 26 Jun-Oct
2007 28 262 22 Jun
Table 1.8.1: Global-scale dust storms on Mars. Adapted from Shirley, 2015
a planetary dust storm in a given year is about one in three and are typically restricted to southern
spring and summer. Further to the study mentioned above, Wang and Richardson, 2015 investigated
the origin, evolution, and trajectory of large dust storms during MY 24-30. Although dust storms were
most frequently observed during the traditional dust storm season, storms can be further seasonally
characterised. Northern hemispheric originating storms occur during northern autumn Ls=180-250° and
winter Ls=305-350°. Southern storms originate earlier in the year around Ls=135-245°.
There are many ways to detect dust storms on Mars. Observations using telescopes are one method
of discovering dust storms onsets. This requires Mars to be the target object. As Zurek and Martin,
1993 noted, dust storms may have been missed due to a lack of images. One advantage of telescopic
images is the ability to observe the global evolution of storms, from initialisation to trajectory to decay;
the spread of the dust storm over the visible face can be very informative. Figure 1.8.1 shows two
images of Mars using Chilescope. The left image is taken a week after Opportunity first detected the
storm on 30th May 2018. Here, south is upwards, so the large India-shaped region is Syrtis Major. The
development of a global dust storm can take several days or weeks to fully mature. The right image
was taken on 7th July 2018 and showed the full extent and severity of the global dust storm. Features
42
1.8. DUST STORMS
of topography are no longer visible as dust is lofted into the atmosphere.
Figure 1.8.1: Early- and mid-storm photos of the same hemisphere of Mars. South is up. Courtesy of
Damian Peach / Chilescope team and Christophe Pellier
A similar method is spacecraft imagery. Very high-resolution images can be attained over a much
more local area. A short orbital period allows the evolution of a dust storm to be studied over a par-
ticular region. This is particularly successful for global dust storms. However, as for telescopic images,
the spacecraft needs to be in a favourable location. Dust storms can be indirectly detected by measur-
ing the increased opacity of the atmosphere using instruments on board rovers such as NASA’s Curiosity.
ESA’s Trace Gas Orbiter (TGO) began its science mission in late April 2018, a few months before the
onset of the 2018 global dust storm event. Solar occultations enabled layers of dust to be identified and
for their impact on atmospheric water vapour to be established. (Vandaele et al., 2019) found layers of
dust were found at altitudes around 25-40 km, in agreement with Gurwell et al., 2005. TGO discovered
an increase in water vapour around these altitudes. One proffered idea stems from the known warming
of the atmosphere, causing more robust circulation. This warming slows or inhibits ice cloud formation,
43
1.8. DUST STORMS
allowing for an increase in water vapour. This result is not unique; Stone et al., 2019 found an injection
of H2 O in the thermosphere by up to a factor of six compared to quiet periods. This highlights the
synthesis between the lower and upper atmosphere.
1.8.2 Dust Storm Theory
Dust storms are not fully understood with much effort being put into modelling the dust storm cycle.
Barnes, 1999 provides a concise exposition of three key dust storm mechanisms: the initial lifting of
dust into the atmosphere, dust heating of the atmosphere, and dust lofting, suspension and horizontal
transport at higher altitudes. One mechanism for lifting dust is saltation (Greeley, 2002). With suffi-
ciently high velocities, forces become strong enough to lift particles from the surface. Upon hitting the
surface again dust can ’splash’, dislodging further dust (Bagnold, 1941). This positive feedback acts as
an injection mechanism. Another mechanism discussed by Barnes, 1999 is the continual suspension and
transport of dust in the atmosphere. In the PBL (planetary boundary layer), turbulent mixing keeps dust
aloft. Convective cells can lift dust high into the atmosphere, penetrating regions otherwise unreachable
by dust. From the television experiment on board Mariner 9, Leovy et al., 1972 determined the vertical
extent of the global dust storm; during the first ∼40 days dust is typically lofted to altitudes of ∼45-60
km (Gurwell et al., 2005).
The next stage in the growth of dust storms is heating of suspended dust. Comparatively, dust is
more efficient at absorbing solar radiation than dominant CO2 in the atmosphere. During a dust storm,
dust particles absorb solar radiation and radiate this back at longer wavelengths, notably infrared. The
increased dust shades the surface as it absorbs incoming solar radiation; however, the atmosphere is
inefficient at absorbing its radiation. Thus approximately only one half of absorbed radiation by dust is
reabsorbed by the surface. The play-off between shading the surface and absorbed re-radiated energy
causes a net cooling effect (Davies, 1979). Although heating occurs in the lower atmosphere, the effects
are profound in the upper atmosphere, as will be shown in Chapter 6. During the 1977 storm, around
the 25 km altitude level temperatures are enhanced by up to 30 K over several days. Heating occurs
within a couple of days from dust injection and is heated throughout the loading process (Martin, 1979,
Jakosky and Martin, 1987). Combining a theoretical approach with observations from Mariner 9, Zurek,
1978 found that heating rates during the 1971 global dust storm may have reached 80 K/day for over-
head sunlight. Further, 20% of direct insolation is absorbed by the dusty atmosphere. As suggested by
Moriyama, 1975 and confirmed by Zurek, 1978, optically thin atmospheres can heat the atmosphere by
44
1.8. DUST STORMS
a few degrees.
The boundary between the warm and cool region becomes unstable and warm air lifts dust into the
atmosphere, typically no higher than 45-60 km (Gurwell et al., 2005). Dust storms are more prevalent
during Southern summer as Mars is near perihelion, where radiative heat sources are most influential.
This creates a more substantial temperature gradient, inducing large storms. Dust storms can last for
days, weeks or months and cover the entire planet. By absorbing solar radiation, the lofted dust places
the surface in a state of shade. In doing so, the storm is being starved of its energy source. This is
coupled with less intense radiation as Mars increases its distance from the Sun. Mars is prone to severe
and long-lasting dust storms due to the inherent arid nature of the planet. The lack of water (or liquid)
means that dust particles do not coalesce as they would on Earth. Another reason global dust storms
are not present at Earth is large bodies of water, such as great lakes or oceans. Dust passing over water
will be removed via spray and necessary moisture within the atmosphere.
Feedback processes due to dust storms such as whether lifted dust results in further radiative forcing
leading to additional dust lifting processes is discussed. Rafkin, 2009 found that the initial response
to dust loading is a decrease in surface pressure as the centre of the dust region. They investigated
’optimal’ conditions for dust storm growth using the Mars Regional Atmospheric Modeling System
(MRAMS) coupled with the Cloud Aerosol and Radiation Models for Atmospheres (CARMA) (Rafkin
et al., 2001, Michaels et al., 2006). They found that location is a crucial factor behind strong dust
storm formation. The optimal location is at subtropical latitudes; here, solar forcing is strong as to
warm the system, and the Coriolis force is sufficient to create circulation. Consequently, higher latitudes
produce a smaller system due to reduced solar input. One ’unknown’ about dust storm initialisations is
the threshold for dust-lifting and subsequent dust-lifting efficiencies; not only is dust needed to be lifted,
but a sufficient amount is needed. Strong feedback occurs for such optimal conditions. Conceivably
counter-intuitive, but excessive dust produces negative feedback leading to the inhibition of the storm.
With focused vorticity, dust can continue to be lifted at night and storms sustained across multiple
sols. And vice versa, a more dispersed storm will dissipate after about one sol. Energy is dissipated
via frictional spin and gravity waves. Gravity waves emanating from localised dust storms were also
observed in Spiga et al., 2013. Another condition that determines the severity of a dust storm is the
initial size of the perturbation; the most considerable perturbations produce the deepest, most intense
circulation patterns.
45
1.9. RESPONSE OF UPPER ATMOSPHERE TO DUST STORMS
Rafkin, 2009 imply that dust storms are tough to predict, as in agreement with previously men-
tioned studies. One factor is the dust distribution. Distribution is meant here as both the size of
dust particulates and geographical location. The former is discussed initially. Studies reviewed shortly
present results outside of dust storm conditions. Using the infrared interferometric spectrometer (IRIS)
on board Mariner 9, Toon et al., 1977 were able to determine dust particle size and composition during
the 1971-1972 dust storm. They found average dust cross-sections between ∼3µm and 6µm, with no
significant change during the storm abatement period. The composition was similar to terrestrial dust,
comprising of SiO2 plus a mixture of other minerals. Pollack et al., 1995 analysed Viking 1 and 2 images
to improve on previous dust size results (Pollack et al., 1977; Pollack et al., 1979). Their latest study
found radii of 1.85±0.3µm at Viking 2 during northern summer when dust loading was low and 1.52±0.3
µm at Viking 1 during the first dust storm. They affirm and reassure that although there is a variety of
dust particle sizes, and thus radiative properties, it does not lead to a substantial change in solar energy
deposition in the atmosphere over the Pollack et al., 1977 and Pollack et al., 1979 estimates. Later
studies have shown dust at Mars is predominantly composed of silicate particles with sizes of the order
1-2µm (Tomasko et al., 1999; Wolff et al., 2006). Rafkin, 2009 took a lognormal distribution with a
mode dust radius of 1µm. Based on observational data, this is a realistic parameter.
1.9 Response of Upper Atmosphere to Dust Storms
During a dust storm, the lower atmosphere is heated, as described above. In turn, the upper atmosphere
expands upwards. Given the unpredictable nature of dust storms, observational data are relatively sparse.
Therefore, dust storms have been incorporated into GCMs in an attempt to reproduce and understanding
atmospheric responses. Hitherto, models and their results have been discussed within the context of the
upper atmosphere. However, dust storms develop in the lower atmosphere. The initialisation of dust
storms in models is considered now. In Bougher et al., 1997 two circulation models are used: a lower
atmosphere model (NASA Ames General Circulation Model - MGCM) where the storm is generated
and an upper atmosphere model where the response is studied (NCAR Mars Thermosphere General
Circulation Model - MTGCM). The MGCM had an upper boundary at ∼100 km, and the MTGCM
had a lower boundary of ∼80 km, thereby providing an overlap of both models leading to coupling
via an exchange of boundary conditions near 70-80 km. Dust storms are simulated by coupling the
46
1.9. RESPONSE OF UPPER ATMOSPHERE TO DUST STORMS
MGCM with an aerosol transport model. Dust is injected into the Southern Subtropics for ten sols and
subsequently evolves self-consistently for another 20 sols. To sequentially couple the MGCM and MT-
GCM codes, the MGCM 1.32 µbar surface heights are extracted and zonally averaged over three sols;
subsequently, these mean heights are specified at the MTGCM lower boundary. At MGS aerobraking
altitudes (∼110 km), densities can increase by a factor of 2 over one day, reaching factors of 5-10 over
several days. Further, compared to a benchmark altitude of 112.5 km for the 1.2-nbar pressure level, the
above-modelled scenario saw an altitude shift of 27.5 km. This is much larger than has been observed
during any recorded dust storm. This result is expanded upon in Chapter 6, and results from previous
observations are discussed.
Medvedev et al., 2013 used a GCM extending from the surface to about 160 km including a spectral
parameterisation of subgrid-scale gravity waves to study the effects of two independent dust storms on
the Martian upper atmosphere. This model is described in Medvedev and Yiğit, 2012. Medvedev et al.,
2013 studied equinoctial and solstitial dust storms from MY25 and MY28. They used dust loading
information from the MGS-TES and MEX-PFS. They ran their model with and without gravity waves.
The equinoctial storm began in the Southern Hemisphere soon after the spring equinox (∼190 Ls),
rapidly increased over a few days extending as far as ∼60°N, and decayed until about 250° Ls. Between
190° and 205° Ls, the amount of airborne aerosol increased more than tenfold over the full range of
latitudes. Similar results are obtained for runs with and without gravity waves included. Direct heating
is expected in regions where aerosols are present; this is observed in the model runs. Temperatures
increase by 30 K in dust regions and decrease by 10-15 K close to the surface. This recent result shows
an agreement with this study. This latter result stems from less solar energy passing through the now
near-opaque lower atmosphere. The middle and upper atmosphere is cooled during this scenario by up
to 30 K. Zonal winds are found to be intensified during dust storms. Haberle et al., 1982 studied the
effects of global dust storms on the atmospheric circulation. Significant changes occur with moderate
injection of dust within the atmosphere where the jet speed and depth of circulation doubled. Further
injection leads to further intensification with polar regions warmed through all altitudes by ∼20-25 K.
Estimated temperature increases in the lower and upper atmosphere.
47
1.10. OPEN QUESTIONS
1.10 Open Questions
As MAVEN is the first dedicated mission to characterise the upper atmosphere using compositional
measurements, much work is required to understand the global overview of the upper atmosphere of
Mars. This thesis aims to answers the following questions.
Models, such as the MCD, allow estimations of atmospheric conditions to be made at a global level.
However, these models need to be validated against observational data to ensure their accuracy. As the
MCD has recently been extended into the thermosphere, this thesis hopes to constrain the model with
recent MAVEN data. Currently, there is no documented validation using thermospheric data. Is there
a systematic difference between NGIMS and MCD densities?
MAVEN’s significant mission length allows a vast dataset to be constructed from which a global
survey of the thermosphere can be undertaken. With this in mind, this allows temporal variations to
be studied in density and temperature. What are typical dayside and nightside temperatures at the
Red Planet and how do they compare with modelling efforts? This has been in local time by Stone
et al., 2018, but this is now generalised in zenith angle. Open questions that can be addressed with this
dataset include how the neutral thermosphere responds to solar-rotation effects; are effects prevalent at
all altitudes studied by MAVEN? This extends previous work by Forbes et al., 2006. Additionally, the
substantial mission length allows seasonal variations to be studied rigorously using a single dataset, for
the first time, by removing other factors, such as latitude and local time. It is expected that densities
and temperatures increase near perihelion, however by how much has not been systematically quantified.
The importance of gravity waves and their effects on the thermosphere, both thermally and dynam-
ically cannot be overstated. As gravity wave schemes become complex, observational data are required
to validate models. Schemes need to ensure they can reproduce observations, leading to accurate mo-
mentum and energy values. This thesis aims to present a global survey of gravity wave characteristics
from which models can be constrained. This is the first study of its kind to fully characterise such waves
in this way. What are typical amplitudes and dominant wavelengths of thermospheric gravity waves?
How do these vary with zenith angle and season? This will hopefully lead to the improve initialisation
conditions and gravity wave inclusion in GCMs should lead to a more accurate representation of the
atmosphere, which is currently lacking. A question born out of the idea of cloud formation is, how
48
1.11. SUMMARY
long can perturbations survive? Is the same wave observable on consecutive orbits and what are the
implications of this?
The 2018 global dust storm gives an opportunity to undertake a case study into how the upper
atmosphere responds and how this compares to previous storms. Although the behaviour, such as at-
mospheric expansion, is expected, we hope to quantify and compare with other instruments and storms.
This is the first observational study to thermospheric gravity waves during a dust storm. The main
open question is, how are gravity waves affected by dust storms? Is there a significant change in the
wave spectrum during a storm? How would this affect the thermal and dynamical structure of the
thermosphere?
The final chapter introduces a brand new density dataset derived from TGO accelerometer measure-
ments, so the questions to answer include what are typical densities and temperatures in this previously
unstudied region, and how do values compare to the MCD? We hope to study wave evolution by utilising
both the TGO and MAVEN datasets. How do amplitudes behave with increasing height? As postulated,
can we observationally show that larger wavelengths dominate with altitude?
1.11 Summary
This chapter has introduced topics that are explored further later in the thesis. The Red Planet’s physical
and orbital properties have been outlined. Furthermore, comparisons have been drawn with Earth.
Typical density and temperature structures have been shown alongside a description of background
physics. The LMD-MCD has been introduced in preparation for validation with observational data in
Chapter 3. The generation, evolution, and implications of gravity waves in the Martian atmosphere
have been summarised. The enormity and rarity of global dust storms have been introduced with the
physics of storm growth and decay highlighted.
49
Chapter 2
Instrumentation and Data
2.1 Introduction
In this chapter, NASA’s Mars Atmosphere and Volatile Evolution mission (MAVEN) and ESA/ROSCOS-
MOS’s ExoMars Trace Gas Orbiter (TGO) missions are introduced. Their main aims are outlined, and
recent results are presented. MAVEN possesses, amongst other instruments, a mass spectrometer and
an accelerometer. TGO possesses just the latter, amongst other instruments. The operations of these
two instruments are described, and data retrieval techniques are outlined. MAVEN has both instruments
and as such densities from these devices are compared. Fortunately for the science community, MAVEN
and TGO sampled the upper atmosphere concurrently and in some instances gathered data in similar
regions. This overlap period is shown, and the related studies performed in this thesis are introduced.
2.2 The Mars Atmosphere and Volatile Evolution Mission
The Mars Atmosphere and Volatile Evolution (MAVEN) mission was designed to explore the planet’s
upper atmosphere, ionosphere, and interactions with the Sun and solar wind. A further aim is to elucid-
ate the loss of Mars’ upper atmosphere over its history and fully characterise its composition (Jakosky
et al., 2015b). These measurements allow the history of Mars’ atmosphere and climate, liquid water,
and planetary habitability to be explored. Launched in November 2013, MAVEN began orbiting Mars
in September 2014. After one year in orbit, NASA announced an extended mission of one year. As
of December 2020, MAVEN is still active. It is the first dedicated spacecraft to gather in-situ data
about the composition and structure of the upper atmosphere and ionosphere of Mars. It is the first
spacecraft to measure atomic oxygen in the upper atmosphere, a crucial measurement for understanding
50
2.2. THE MARS ATMOSPHERE AND VOLATILE EVOLUTION MISSION
CO2 cooling Fox, 2004. Processes that control the atmosphere will be explored, allowing a look back
at Mars under different boundary conditions. Another high-level objective is to measure the loss rate
of the atmosphere to space. And again, the processes that control such loss are to be investigated.
Additionally to the previous goal, with loss rates, it will be possible to step back in time to under-
stand the history and evolution of the atmosphere. The latter points are not directly investigated in
this study, but seminal work on this topic can be found in Jakosky et al., 2015a and Jakosky et al., 2017.
There are two main techniques employed to lower spacecraft into their science orbits, the first is
involves the sole use of thrusters, and the second is aerobraking. The first is quicker, as once a spacecraft
has reached its target body, thrusters are fired to reduce the spacecraft’s speed. The reduction in speed
allows the spacecraft to be captured by the body’s gravity. Further thruster firings will lower the space-
craft’s altitude into its science orbit. This technique is considerably more expensive than aerobraking
owing to the substantial use of propellant. The Mars orbit insertion (MOI) manoeuvre began with six
thruster engines firing briefly to damp out deviations in pointing. Then, the six main engines quickly
ignited and burnt for 33 minutes to slow the craft, allowing it to be captured in an elliptical orbit around
Mars with a period of 35 hours. Six smaller manoeuvres were performed later to bring the highest and
lowest points of the orbit to the altitudes desired for the science orbit. The overall fuel usage, and thus
expenditure, can be dramatically reduced by implementing a technique called aerobraking. Thrusters
are initially used to place the spacecraft in an elliptical orbit. Aerobraking involves flying the spacecraft
through the atmosphere near periapsis. During its pass through the atmosphere, appreciable drag is
exerted on the spacecraft, causing the spacecraft to decelerate and reduces its apoapsis distance given a
loss of energy. At apoapsis, small thruster firings may be used to ensure the spacecraft passes once again
through the atmosphere in the required density corridor. The density corridor is a range of densities that
are a large enough to induce sufficient drag on the spacecraft, but not so large as to cause significant
heating and mechanical damage. Aerobraking is a slower process, but it is made up for in lower cost.
Additionally, data can be derived from its employment. As spacecraft fly through the atmosphere, the
density can be recovered using the accelerations felt by the spacecraft as measured by the on board
accelerometers. This is described later in this chapter. This technique was first used by NASA on the
Magellan mission at Venus in 1993. It was first used as an orbital adjustment technique at Mars for
MGS, leading to many discoveries about the atmosphere of Mars including gravity waves (e.g. Fritts
et al., 2006, Tolson et al., 2007) and longer-term variability (e.g. Keating et al., 1998, Tolson et al.,
2008). Its success is not limited only to MGS, as aerobraking was also employed for Odyssey (ODY)
51
and Mars Reconnaissance Orbiter (MRO).
At periapsis, MAVEN is ∼150 km above Mars’ surface where it can make in situ measurements of
the upper atmosphere and has an apoapsis around 6300 km above the surface allowing for solar wind
conditions to be studied. The elliptical orbit has allowed atmospheric loss mechanisms to be studied; this
has to lead to Jakosky et al., 2015a concluding that the atmosphere is currently losing gas to space via
stripping by the solar wind at a rate of 100 g/s leading to significant loss over long periods. They found
that enhanced solar activity, such as coronal mass ejections, increased the loss rate, thereby confirming
the relationship between the Sun and atmospheric loss. This loss rate can be applied to understand
previous states of the atmosphere. A further study by Jakosky et al., 2017 revealed that Mars has lost
about 66% of its atmosphere into space since its formation.
2.2.1 Spatial and Temporal Coverage
Figures 2.2.1a, b and c show MAVEN’s periapsis location in altitude, local solar time and latitude,
respectively. Panel a shows periapsis altitudes typically above 140 km, with variation throughout the
mission up to 180 km during the latter months of 2017. MAVEN performed Deep-Dip (DD) campaigns
(red) which are visible between 120 km and 140 km. DD campaigns comprise orbits that probe deeper
into the atmosphere, down to ∼120 km. These dives allow data in the lower thermosphere to be
studied, allowing connections between the lower and upper atmosphere to be investigated. Densities are
approximately an order of magnitude larger in this region than at 150 km. Each campaign lasts around
seven days; the first two to three days are occupied with lowering periapsis altitude and the remaining
five days are used to gather density data. DD campaigns are referenced throughout this thesis, and
as such, their locations are detailed in Table 2.2.1. Solar longitude is well sampled. Average latitude,
solar zenith angle (SZA), local solar time (LST) and solar longitude (Ls) values for each DD campaign
are shown. Panel b shows periapsis local solar times which MAVEN cycles through approximately every
six months. Panel c shows periapsis latitudes. MAVEN requires longer to sample all latitudes; hence
MAVEN has sampled each latitude only six times.
2.2.2 The Neutral Gas and Ion Mass Spectrometer
The Neutral Gas and Ion Mass Spectrometer (NGIMS) instrument on board MAVEN is used to study
the composition and structure of the upper atmosphere (Mahaffy et al., 2015b). NGIMS measures both
surface reactive and inert neutral species and ambient ions along the spacecraft track at altitudes below
52
Figure 2.2.1: MAVEN’s periapsis location shown in (a) altitude, (b) local solar time and (c) latitude.
Red points represent Deep-Dip passes. Solar longitude is shown on the bottom axis.
500 km.
There are two operating modes (open- or closed-source) which alternate between odd and even
orbit numbers. Figure 2.2.2 shows a diagram of gas flow through the instrument. The closed source
operation works by opening an aperture permitting a sample of the atmosphere to flow into the ante-
chamber before closing shortly after. The gas will thermalise in the antechamber, thereby becoming
dependent only on mass and charge. Once thermalised, the gas passes through to the ionisation region,
which consists of an electron impact hot filament ion source and becomes ionised and feels the effect
of electric and magnetic fields. Once ionised, the gas flows through the quadrupole deflector and from
this point follows the same path as the open-source gas. A quadrupole deflector consists of a DC quad-
rupole electric field. For the closed source measurements, the rods can be grounded to prevent bending
of the gas. The open-source is used to sample gas constituents that are destroyed or transformed by
collisions in the closed ion source. Constituents such as O and N are sampled via the open-source as
53
Deep Dip () Orbits () Latitude (°) SZA (°) LST (hr) Ls (°)
1 714-747 42.6°N 109.1 18.3 291.1

2 1059–1086 3.8°S 9.3 11.9 328.6
3 1501–1538 62.6°S 110.4 3.5 11.4
4 1802–1838 63.9°S 91.1 16.0 37.5
5 3285–3327 33.2°N 96.5 5.2 166.9
6 3551–3586 2.9°S 166.4 0.7 194.1
7 5574–5620 63.6°N 87.0 20.3 49.4
8 5909–5950 18.9°N 25.0 13.7 76.3
Table 2.2.1: MAVEN Deep Dip Ephemeris. Note. Values are means at periapsis over each Deep Dip
campaign. Adapted from Stone et al., 2018
Figure 2.2.2: Schematic of gas through NGIMS. Taken from Mahaffy et al., 2015b
they are surface reactive and can be absorbed by the ion source surface. The gas is collimated and
passes through a crossed electron beam leading to ionisation. Once ionised, the gas is deflected 90° by
the quadrupole deflector. For the open-source scenario, opposite voltages are applied to opposite rods
54
allowing the path to bend by 90°. After the quadrupole deflector, the paths taken by gases from the
closed and open sources are the same. The gas passes through the ion lens system, which focuses the
beam to ensure as much gas progresses through the instrument as possible. The next stage of NGIMS
is gas flow through the quadrupole mass analyser.
The mass analyser aims to select constituents with certain mass/charge ratios (m/z) for detection
and achieves this as follows. Four hyperbolic quadrupole rods are constructed with opposite voltages
applied to each parallel pair. In order to achieve mass selection an AC current with frequency ω is super-
imposed with a DC current to give final voltages running through the rods of VDC + VAC cos(ωt) and
-VDC +VAC cos(ωt). This results in a 2-D quadrupole field. The ionised particles will gyrate through the
analyser, and only particles with a certain m/z value will successfully pass through for a given voltage
ratio. All other particles will have unstable paths and collide with the rods. In around 4 s the mass
analyser will cycle through all voltage ratios needed to cover the 2-150 amu mass range, allowing the
atmosphere to be sampled with a spatial resolution of ∼100 m. The gas is then passed through another
lens to focus it before entering the detector. The detector consists of electron multipliers that saturate
at several million counts per second with an average gain of ∼5x107 . The electron multipliers allow
the detection of low signals given a background noise of less than one count per minute. The detectors
measure the number of counts of each particle and combining all this information an abundance for each
constituent can be calculated for each sampling period. A reservoir of gas in equal parts of N2 , CO2 ,
Ar, Kr, and Xe is stored in the instrument to ensure the accuracy and sensitivity of the instrument is
satisfactory to mission requirements. At intervals of several weeks, the gas is released slowly in a restric-
ted manner at apoapsis into the closed source. Biases can then be applied should the calibration yield
inaccurate results (Jakosky et al., 2015b). For Ar, the sensitivity is ∼10−3 -10−2 (counts/s)/(part/cc).
Stone et al., 2018 provides an explanation of the reduction techniques used to derive density data
from raw instrument counts. The previous paragraph explained an idealistic situation; however, there
are other signals which can detrimentally contribute to counts. Two such signals are background due to
the desorption of gases from the inner surfaces of the instrument and background signal due to collisions
in the quadrupole mass filter. Desorption is the release of adsorbed particles from a surface, in this case
from the walls of the instrument. The signal due to desorption is greater on the outbound pass; thus,
many studies consider only the inbound pass to reduce uncertainty. Although the background signal
is 3-4 orders of magnitude smaller than periapsis densities, the successful removal of the background
55
signal allows densities at higher altitudes to be retrieved. Like many discrete detectors, NGIMS suffers
from dead time. This is the period after an event (count) that the detector cannot process another
count. The following is taken from Stone et al., 2018. The dead time correction takes the form
m = r exp(rτ ) (2.1)
where m is the measured count rate in units of per second, r the true event rate in units of per second,
and τ the dead time given by
τ = max{A log m + B, 0} (2.2)
The coefficients A and B are determined from the data by comparing signals between molecular
fragments at different counting regimes (Benna and Elrod, 2018). Deadtime coefficients A = 9.49×10−9
s and B = 1.39×10−7 s are used. Correcting for the detector dead time allows us to use count rates up
to 2×107 s−1 . Densities calculated for orbits beyond 748 are divided by a factor of 1.5331 to account
for an observed change in the sensitivity of the instrument following DD1 (Benna and Elrod, 2018).
Corrections are also required to account for an artificial increase in measured density due to an increase
in density caused by interactions between the fast-moving spacecraft (∼4 km/s) and atmosphere. Benna
and Elrod, 2018 explain correction factors to account for spacecraft ram effects.
2.2.3 Accelerometer
Although not part of the science payload, the accelerometers (ACC) on board MAVEN provide a wealth
of data which can be exploited to retrieve atmospheric density, much like observations made by the
accelerometers on board MGS (e.g. Keating et al., 1998; Withers, 2006; Tolson et al., 2007; Tolson
et al., 2008). Acceleration data comes from the inertial measurement units (IMU) which detect any
accelerations, including those due to atmospheric drag. This drag force is related to the density of the
atmosphere and is recoverable. Accelerometers can be constructed using different operating principles.
The basic working principle behind an accelerometer can be thought of as a mass on a spring encased
within a housing. By measuring the extension of the spring, one can derive the acceleration of the
system. Three orthogonal springs can be used to determine accelerations in all directions. Modern
techniques involve the motion of silicon components for a fixed frame that measures the change in
capacitance which is proportional to acceleration. ACC utilises an electromagnetic field which is varied
in time to keep an electronically floating mass stationary relative to the case when the spacecraft senses
56
an acceleration. The current needed to do this is proportional to the total acceleration (Zurek et al.,
2015). The accelerometer cannot implicitly differentiate between accelerations (e.g. between drag and
gravity). Therefore these sources of acceleration need to be removed before recovering density from
drag measurements. Tolson et al., 2008 discuss several of these in-depth. The total acceleration, am ,
felt by the spacecraft is
am = ab + aa + ag + aACS + ω × (ω × r) + ω̇ × r (2.3)
where, from left to right, the terms are the acceleration due to instrument bias, aerodynamic forces,
gravity gradient, attitude control system thruster, and angular motion of the accelerometer about the
centre of mass (last two terms). The aerodynamic forces term arises due to the atmosphere inducing a
drag on the spacecraft. It is this term from which density is recovered once all other accelerations have
been removed. The gravity term arises due to the irregular mass distribution of Mars. The effect of
thruster firings on the spacecraft is understood, so by knowing when these firings occurred, they can be
removed from the data. The acceleration contributions from the angular motion of the accelerometer
are removed using the gyro data, discussed below. Accelerometers are mounted as close as possible
to the centre of mass of spacecraft; this theoretically reduces the angular accelerations felt. Angular
rates for each axis of the accelerometer are calculated, and polynomials are fitted and differentiated
to give the acceleration (Tolson et al., 2008). The effect of solar radiation as a source of acceleration
is important Earth but is assumed to be negligible at Mars (Bruinsma et al., 2004). Once all other
accelerations have been removed from the data, the density of the atmosphere, ρ, can be recovered
from the drag term using
2M

ρ= az (2.4)
V 2 Cz A
where M is the mass of the spacecraft, V is the velocity of the spacecraft, Cz is the aerodynamic drag
coefficient along the z-axis (pointing along trajectory), A is the area of the spacecraft perpendicular
to its velocity vector, and az is the acceleration of the spacecraft centre of mass due to aerodynamic
forces along the z-axis. Cz is a dimensionless quantity that accounts for the shape of the spacecraft
and friction between the atmosphere and spacecraft. Cz is a function of density; therefore, the previous
equation is solved recursively. M decrease as fuel is used for propulsive manoeuvres.
At 160 km, density is recoverable to ∼0.1kg/km3 for MAVEN which is equivalent to ∼5% noise
level (Zurek et al., 2015). The IMU has limited resolution for measuring accelerations, which introduces
57
uncertainties. Error is potentially introduced when fluid models are used to determine the drag coeffi-
cient. The model includes the interaction of the gas flow with the spacecraft due to small-scale surface
roughness, surface temperature and composition, gas composition and perhaps chemistry. All these
errors equate to ∼10% of the drag coefficient (Zurek et al., 2015). Within each IMU is a three-axis
ring laser gyroscope (RLG); each RLG measures rotation in one axis. For ease of explanation, it is
assumed that beams travel in a circular motion around the rotation axis. In practice, mirrors are used to
redirect beams. An initial input laser beam is split, causing rays to travel in a clock- and anticlockwise
direction. A detector is located equidistant from the beam source. Without rotation, both beams are
detected at the same time. With a clockwise rotation, the clockwise (anticlockwise) beam takes a longer
(shorter) time to reach the detector unit. By calculating the time difference, the phase difference can
be determined. The output voltage is proportional to the rotation; therefore RLGs measures rotation
rate about its sensitive axis (Zurek et al., 2015).
Like all instruments, and as explained for NGIMS, calibration and biasing need to be performed to
convert raw instrument readings into physical quantities. Instrument bias may be caused by heating of
the IMU throughout an aerobraking pass and is seen as a systematic increase or decrease in acceler-
ometer readings about their actual values. Thus, there are several occasions during the mission that
the accelerometer bias is determined. These calibrations are generally performed before any propulsive
manoeuvre where the accelerometers are used to determine the manoeuvre. Additionally, the bias is
re-determined by fitting a line to the data before and after periapsis when the spacecraft is little affected
by atmospheric drag. This linear fit can be removed from the data, reducing the bias (Zurek et al., 2015).
2.2.4 Comparison Between NGIMS and ACC Data
Along-track densities are retrieved by both NGIMS and ACC along all orbits, and therefore densities
can be directly compared for each orbit. However, only DD orbits are used here. As NGIMS is capable
of measuring the densities of multiple species, the total density is calculated using densities for CO2 ,
Ar and N2 . At altitudes studied here, CO2 the dominant species, thus adding other more species to
the total density makes a scant difference. Figures 2.2.3a-d show density profiles from four MAVEN
orbits: 1061, 1062, 1063 and 1065, respectively. These orbits occurred during DD2 and occurred near
noon, around 5°S and Ls=256°. NGIMS and ACC profiles are shown by blue and red lines, respectively.
Inbound and outbound profiles are shown by solid and dashed lines, respectively.
58
Figure 2.2.3: NGIMS (blue) and ACC (red) density profiles for orbit (a) 1061, (b) 1062, (c) 1063 and
(d) 1065. Inbound and outbound legs are shown by solid and dashed lines, respectively.
Figure 2.2.3a exhibits apparent wave-like behaviour in the NGIMS data for both inbound and out-
bound legs, especially below 150 km. However, this behaviour is not repeated in ACC data, as similar
behaviour would be expected for both instruments. Panel b shows excellent agreement not only between
the instruments but also the passes; this implies very weak or no horizontal density gradients. Some
wave activity is still visible below 150 km in the inbound NGIMS profile. Panel c displays different
profiles. Again, perturbations are observed in only the NGIMS data. Lastly, profiles in panel d show
remarkable agreement, particularly on the outbound leg. Two main inferences can be made: densities
retrieved from both instruments are in general agreement with each other, and wave activity is more
prominent in the NGIMS dataset, implying the resolution of ACC data is not sufficient for wave ana-
lysis. Fitted background profiles would invariably improve agreement between the datasets. Third-order
polynomials are fit to the logarithm of the densities, which act as background profiles by removing wave
activity. Agreement between the datasets can be quantified by taking the ratio between NGIMS and
59
ACC densities at given altitudes, denoted as ρN GIM S /ρACC . Altitudes of 120, 125, 130 and 135 km
are used. Therefore, for each orbit, there are eight ratios - an inbound and outbound ratio for each of
the four altitudes. Figures 2.2.4a-d show boxplots of ratios for seven DD campaigns at the four fixed
altitudes, respectively. Shown also is a dashed horizontal line that shows the expected ratio of one.
DD3 (1501–1538) has been omitted based on a substantially broader range of ratios (0.1-3) than found
for other campaigns. This exclusion allows other campaigns to be compared more easily.
Figure 2.2.4: Boxplots of ratios using interpolated density data from NGIMS and ACC at (a) 120 km,
(b) 125 km, (c) 130 km and (d) 135 km for DD1-8, excluding DD3. A ratio of one is shown by a dashed
horizontal line.
Figure 2.2.4a shows boxplots displaying ratios at 120 km. Only DD5 and DD6 dipped down to 120
km. For each campaign, median NGIMS and ACC densities are within 10% of each other. Furthermore,
50% of densities are within 20% of each other. Panel b shows ratios at 125 km. DD4 and DD8 probed
down to this altitude. DD5 and DD6 stray away from their original behaviour at 120 km; the former has
a median of around 0.8 while the latter ratio improves to near one. The range of values also diminishes,
60
2.3. THE EXOMARS MISSION
suggestive that ACC and NGIMS densities are comparable. Both DD4 and DD6 show NGIMS and ACC
densities are within 10% of each other; this is in excellent agreement. All DD campaigns probed down to
at least 130 km, as visible in panel c. Here, all median ratios are within 25% of one with ranges stretching
from ∼0.4 to 1.4. A general shift to lower ratios is evident for all DD campaigns indicating ACC densities
are increasing more rapidly than NGIMS densities. Panel d shows ratios at 135 km and repeats the
findings found at 130 km (panel c). Ratios continue to decrease, again suggestive of ACC densities
becoming larger than NGIMS. One finding that has been implied in the above discussion is that ratios
are distinctly different; there is no systematic difference between NGIMS and ACC densities, as may have
been expected, shown by the large range of ratios. This suggests that a more complicated explanation
that goes beyond a simple and predictable offset is required. It is known that both instruments perform
most effectively at lower altitudes (higher densities); thus, uncertainties become fractionally smaller for
higher densities. This is largely evident here. It is worth noting that, henceforth, ACC data are not used
in this study for MAVEN. NGIMS are data preferred owing to the nature of studies undertaken; the first
of two main reasons is the ability to differentiate between species, allowing to study the composition
change, e.g. during dust storms. Secondly, ACC data exhibits significant noise (where uncertainties are
greater than 5% of measured values) above ∼160 km, whereas NGIMS CO2 data remain noise-free well
above 200 km. This allows a greater range of altitude variations to be studied.
2.3 The ExoMars Mission
ExoMars is a joint ESA-ROSCOSMOS two-part mission to search for signs of past life, variations in
water, and investigate atmospheric trace gases and their sources (Vago et al., 2017). The first part of
this mission, launched in 2016, was Trace Gas Orbiter (TGO). TGO is tasked with identifying sources of
methane and other trace gases that may evince biological or geological processes. Given its short lifetime
of the order a few hundred years, when exposed to solar radiation any detection of methane implies it
has been released recently. Thus far, TGO has detected no methane (Korablev et al., 2019). TGO’s
science orbit fortuitously began in April 2018, a few months before the June 2018 global dust storm
event. Layers of dust were found around 25-40 km altitude (Vandaele et al., 2019). TGO will operate
as a communications relay for the second part of the ExoMars mission, the Rosalind Franklin rover.
Rosalind Franklin is scheduled to launch in July 2020 with the primary goal to search for subsurface
chemical and morphological life signatures. Both phases are outlined further below.
61
2.3.1 Trace Gas Orbiter
The first phase of the mission consisted of Trace Gas Orbiter (TGO) and a test lander, Schiaparelli.
TGO has a suite of four instruments that are used to achieve the mission goal of finding methane and
other atmospheric trace gases at Mars. TGO performed aerobraking between March 2017 and February
2018 to circularise its initial orbit to its final 400 km science orbit. Aerobraking was paused in summer
2017 owing to Mars’ conjunction and resumed until February 2018. Routine commands and updates
were not able to be reliably sent due to fear of deterioration (Scuka, 2017). Launched with TGO was
an Entry, Descent and Landing Demonstrator Module (EDM), Schiaparelli. The EDM’s mission was to
investigate local electricity and meteorology. Ultimately, it provided the opportunity to test technologies
for future planetary landers. Unfortunately, the landing of Schiaparelli was unsuccessful. The IMU
became saturated, and navigation data placed Schiaparelli below the surface. The combination of the
parachute being released too early and thrusters firing for 10% of the anticipated time lead to systems
being deployed at 3.7 km as if Schiaparelli had landed, but had in fact crashed.
TGO contains two IMUs, each containing three ring laser gyroscopes and three accelerometers. The
operations of accelerometers and ring laser gyroscopes were outlined above in the context of MAVEN.
The same working principles apply to TGO. Densities have been retrieved and are used in this study. The
uncertainty in the derived density is the sum of a systematic part due to the uncertainty in CD , estimated
to be 5% at these very low altitudes, and a noise and bias part due to the accelerometer. The systematic
error has no impact on analyses of relative variations, e.g. wave perturbations. The bias and noise of the
accelerometer were estimated through analysis of data outside the sensitivity range, i.e. above 130 km.
The biases were estimated by computing averages before and after perihelion and were negligible (of the
order 10−5 m/s2 ). The formal error cannot be computed based on the accelerometer specifications (as
was done, in principle, for identical instruments on Venus Express in Mueller-Wodarg et al., 2016) as raw
measurements are not available, but rather pre-processed accelerations for TGO. The formal noise of the
Venus Express accelerometers, after processing of the velocity increments one second apart, was 0.001
m/s2 (1σ). Using that cut-off value leads to valid TGO densities to a maximum altitude of 115 km.
However, after careful inspection of the acceleration and the resulting density profiles, the noise on the
TGO accelerations appears to be on average twice lower, 0.0005 m/s2 assuming that lower value at the
cut-off threshold leads to a maximum valid altitude of 120±4 km. Another way of determining the valid
altitude range of the TGO densities is by using a comparison with a model, Mars-GRAM 2005 (MG05).
62
The TGO-to-MG05 density ratios were computed for all profiles. The density ratios start to present a
typical noise behaviour, minimal density ratios followed (preceded) by fast and large oscillations for the
outbound (inbound) leg, for altitudes above 120-130 km. This is illustrated in Figure 2.3.1.
Figure 2.3.1: The TGO-to-MG05 density ratios for a profile in January 2018 (blue line). x-axis is CNES
Julian date. Altitude is shown on secondary y-axis (grey line).
TGO carries a suite of instruments capable of detecting trace gases within Mars’ atmosphere. The
Nadir and Occultation for Mars Discovery (NOMAD) instrument consists of three spectrometers; these
can detect methane and hydrocarbons (Vandaele et al., 2011). The Atmospheric Chemistry Suite
(ACS) complements NOMAD by using infrared instruments to examine the chemistry and structure of
the atmosphere (Korablev et al., 2018). The Color and Stereo Surface Imaging System (CaSSIS) is
assigned to investigating the surface of Mars by capturing high-resolution images. This enables sources
and sinks of trace gases to be identified (Thomas et al., 2017). Finally, the Fine Resolution Epithermal
Neutron Detector (FREND) figuratively digs into the surface. The detector is tasked with identifying
reserves of subsurface hydrogen up to one metre below the surface (Mitrofanov et al., 2018).
63
2.4. CONCURRENT MAVEN AND EXOMARS DATA
2.3.2 Rosalind Franklin Rover
The second stage of the ExoMars mission is to land the Rosalind Franklin rover on the surface of Mars
with the end goal of searching for the existence of past life on Mars. The launch date is set for 25th
July 2020 (as of December 2020). Densities may be recoverable from accelerometers during the rover’s
descent from which a vertical temperature profile can be derived and compared to model outputs. This
has been done previously with Pathfinder and ExoMars Schiaparelli (Magalhães et al., 1999; Aboudan
et al., 2018). Beyond this, the Rosalind Franklin rover will not explicitly complement the work presented
in this study, but will without a doubt improve our understanding of the Martian surface.
2.4 Concurrent MAVEN and ExoMars Data
During TGO’s aerobraking phase, MAVEN continued its science mission. For the first time, Mars’
upper atmosphere has been simultaneously sampled by two spacecraft. This exciting prospect allows
the decoupling of long-term spatial and temporal variations in the upper atmosphere to begin. A more
global picture of the density and temperature structure of the atmosphere can be formed. It should be
noted that Mars Express is still in orbit and performs occultations to determine vertical neutral density
profiles up to ∼50 km (e.g. Tellmann et al., 2006). The effects of significant events at Mars, such
as coronal mass ejection (CMEs) or dust storms, can be studied nearly concurrently at two separate
locations. With two spacecraft, there is a higher probability of observing any noticeable effects of such
events. Thiemann et al., 2015 examined the dayside neutral density response in the upper atmosphere to
solar flares and found significant heating caused by the flares, responding and recovering rapidly. With
two spacecraft the effect of these flares could have been studied at two separate locations concurrently.
Circulation may cause heating on the nightside and different latitudes may respond differently. In
Chapter 7, these datasets are combined to investigate density and wave structures. Figure 2.4.1 shows
TGO and MAVEN’s periapsis coverage in (a) altitude, (b) local time and (c) latitude during TGO’s
aerobraking phase. In panel a, the lowering of TGO is visible in April 2017 as it descends from ∼120 km
to its characteristic periapsis altitude of ∼105 km. The effects of conjunction are clear in summer 2017.
MAVEN’s periapsis altitude remained relatively invariant at ∼150-170 km, with two dips to ∼130 km. No
nominal DD campaigns were undertaken during this period. Panel b shows local solar time coverage. All
times were sampled by TGO and nearly twice by MAVEN. In February 2018 both spacecraft sampled the
midnight-dawn sector simultaneously. Panel c shows latitudinal coverage. TGO predominantly sampled
the southern hemisphere, proceeding closer towards to pole as the aerobraking campaign advanced.
64
2.4. CONCURRENT MAVEN AND EXOMARS DATA
During this period, MAVEN principally remained in the northern hemisphere, however, moved in the
southern hemisphere during the latter three months of the aerobraking campaign. Furthermore, sampled
similar latitudes concurrently with similar local times in February 2018.
Figure 2.4.1: TGO (green) and MAVEN’s (blue) periapsis (a) altitude, (b) local time and (c) latitude
during TGO’s aerobraking phase.
TGO probes deep into the atmosphere, typically down to ∼105 km. This is ∼15 km lower than
MAVEN’s DD orbits and larger than Mars’ average scale height; therefore, it is expected to see potentially
very different behaviour. As discussed in Jakosky et al., 2017, the homopause location is reasoned to
hover around 100 km, but cannot be resolved by accelerometer measurements alone. Although ACS
is capable of deriving vertical CO2 density (and temperature) profiles, these are only retrievable at the
morning and evening terminators in the 10-80 km altitude range. The homopause is unlikely to be
found in this range, but rather ∼75-125 km (Jakosky et al., 2017). Towards the latter end of TGO’s
aerobraking phase, MAVEN and TGO sampled similar regions concurrently. This overlap is exploited in
65
2.5. SUMMARY
Chapter 7 to garner complete thermospheric profiles from the two distinct altitude ranges. Both density
and temperature profiles can be derived by hydrostatically connecting the profiles. These profiles are
useful for understanding the unsampled structure but also for comparison with models as a method of
validation. Complete profiles are of importance for modelling diffusion.
2.5 Summary
In this chapter, the MAVEN and ExoMars missions have been outlined, including a summary of their
objectives. The instrumentation used to gather in situ atmospheric data has been discussed. A summary
is shown below.
• MAVEN is the first spacecraft to primarily probe the upper atmosphere of Mars. Its on board
suite of instruments allows the thermosphere to be characterised by composition. Its coverage in
altitude, local time, and latitude has been presented.
• NGIMS is the main source of data throughout this study. NGIMS measures the species with
masses 2-150 amu, which covers the major constituents in Mars’ upper atmosphere, with CO2 , Ar
and N2 density data used throughout this study. The accelerometer on board MAVEN has been
discussed.
• NGIMS data are compared to accelerometer data by interpolating onto a fixed altitude grid and
taking the ratio at each altitude. Densities are more alike at low altitudes (large densities) but
differ significantly at high altitudes. Along-track densities are typically within ∼20% of each other.
• The first part of the ExoMars mission, TGO, is introduced. TGO undertook an aerobraking cam-
paign to circularise its orbit from which densities have been derived from accelerometer measure-
ments. During the latter part of TGO’s aerobraking campaign, it concurrently measured a similar
region to MAVEN; the potential science associated with this fortunate occurrence is highlighted.
66
Chapter 3
Data Analysis and Model Comparison
3.1 Introduction
First, as temperature is not a directly measured quantity by either spacecraft, a technique is introduced
to derive temperature profiles from density data. This technique is verified using density and temperature
data extracted from the Mars Climate Database (MCD). Potential drawbacks and uncertainties with this
technique are discussed. The MCD is used further to compare with NGIMS observations of individual
species, such as CO2 , Ar and N2 , as well as Ar/CO2 , N2 /CO2 ratios, to establish if there any systematic
difference between the MCD and observations. Finally, an interesting property within TGO density
profiles is observed; the maximum density is consistently located away from the closest approach. This
effect is quantified, and potential explanations are proffered.
3.2 Deriving Temperature Profiles From Density Data
Temperature is not a parameter directly measured by MAVEN. It therefore needs to be derived from
measured densities. The hydrostatic equation (Equation 1.7, Chapter 1) relates vertical changes in
pressure to density. Here, p, z, ρ and g are pressure, altitude, density, and gravitational acceleration,
respectively. Integrating the hydrostatic equation with altitude downwards allows one to determine a
pressure profile; from this, a temperature profile can be determined using the Ideal Gas Law. Given
a column of atmosphere, the pressure at any given altitude, z, is derived by adding the weight of the
atmosphere above, expressed by
Z ztop
P (z) = P (ztop ) + ρ(z 0 )g(z 0 )dz 0 (3.1)
z
67
3.2. DERIVING TEMPERATURE PROFILES FROM DENSITY DATA
where P (ztop ) is the pressure at the top of the column, ρ(z 0 ) is the mean density in height element dz0
and g(z 0 ) is gravitational acceleration at altitude z 0 .
During its flyby through Mars’ atmosphere, MAVEN travels both vertically and horizontally. Near
periapsis, MAVEN is travelling quasi-horizontally, so any changes in density are likely to be due to
horizontal structures, rather than vertical. Mueller-Wodarg et al., 2006 and Stone et al., 2018 discussed
the effect of correcting along-track densities to account for horizontal structures. Stone et al., 2018
found a shift of ∼10 K after applying corrections, as such profiles are uncorrected and should be treated
with caution. Further, for wave studies, a systematic shift will not significantly affect the results. Another
issue to address in temperature derivation is the initial upper boundary condition, P (ztop ). As Equation
3.1 is integrated downwards, and pressure is equal to the weight of the atmosphere above, an initial
pressure is needed and shown later to be non-negligible. An initial temperature is required to determine
P (ztop ) and is estimated from the top 10 km of the region studied, assuming to be isothermal. The
scale height, H, is determined for this region by constructing a plot of log(ρ) in altitude and taking the
gradient in this region using the method of least squares. The temperature can be calculated from the
scale height using Equation 3.2 and assuming a mean molecular mass of m (equal to 43.03 amu).
mg
T =H (3.2)
k
where terms have been previously introduced. This derived temperature can be used in the Ideal
Gas Law to retrieve the corresponding pressure. P (ztop ) values are typically of the order 10−12 Pa at
∼240 km. Several scale heights below ztop , the initial choice of P (ztop ) becomes unimportant and
profiles from varying upper boundary conditions converge, as shown in Mueller-Wodarg et al., 2006 and
Snowden et al., 2013. The derived P (ztop ) value is used in Eq. 3.1 to derive the pressure profile using
a vertical step of 0.1 km. The Ideal Gas Law is then implemented to derive the temperature profile.
Figures 3.2.1a and b show how temperature profiles converge when the upper boundary temperature
condition is varied by ±25 K and ±50 K. From these offset temperatures, new pressure profiles are
derived. In this case, temperatures converge within 40 km. This is likely to vary depending on the initial
upper altitude. This has important implications for TGO temperature derivations where densities are
only available over a 15 km range.
The error associated with temperature has been found numerically rather than analytically, fol-
lowing the method of Mueller-Wodarg et al., 2006. For each leg, 1000 random density profiles were
68
3.2. DERIVING TEMPERATURE PROFILES FROM DENSITY DATA
Figure 3.2.1: Temperature profiles from orbit 1060 derived from (a) inbound and (b) outbound Ar
densities. The black line signifies the temperature profile derived using the upper boundary condition
described. Profiles are shown for upper boundary conditions shifted by ±25 K and ±50 K. Temperature
profile derived by Stone et al., 2018 is shown by the green dashed line.
created as follows. 1000 density profiles were produced assuming ρ(z) = ρobs (z) + ∆ρ(z) where ρobs (z)
is the observed nominal density and ∆ρ(z) a perturbation chosen randomly to lie within the range
-σρ (z) ≤ ∆ρ(z) ≤ σρ (z). σρ (z) is the combination of error in density measurement and random un-
certainty associated with the measurement. From these new profiles, temperature profiles were derived
using the method described by Equation 3.1. Temperature errors are assumed to equal one standard
deviation at each altitude level, with error bars being small due to small associated uncertainties. σ
values are altitude dependent, with values of less than 10 K near periapsis and ∼25 K around 200 km.
Inbound and outbound temperature profiles inferred for orbit 1060 are shown in Figures 3.2.1a and
b. Ar density data are used due to its inert nature. Clear wave-like behaviour can be seen, and the
general trend is in agreement with what is to be expected (e.g. Bougher et al., 2015; Bougher et
al., 2017). Near-isothermal behaviour at higher altitudes is observed in Figure 3.2.1b as conduction
becomes dominant in this region. These are typical profiles with temperature values between 100 and
300 K dependent on location, local time and season. Uncertainties at higher altitudes are around 15
69
3.3. DRAWBACKS OF CURRENT TEMPERATURE DERIVATION TECHNIQUE
K and reduce to around 5 K near periapsis. The temperature profile presented in Figure 6 in Stone
et al., 2018 from orbit 1060 is identical to the profile presented here; both studies have used different
techniques to determine P (ztop ). They determine upper boundary temperature and pressure values by
fitting the density at high altitudes, assuming isothermality, to an equation of the form,
GM mi 1 1

Pi = Po,i · exp − (3.3)
kTi r ro
where Pi is the pressure of the ith species, Pi is the pressure of the ith species at the lower boundary
of the fitted region. G, M , mi , k are the gravitational constant, Mars mass, the mass of species i and
Boltzmann constant, respectively. Ti is the temperature of the ith species. r and ro are the distances
above Mars’ surface and centre to the lower boundary of the fitted region, respectively. This temperat-
ure profile derived by Stone et al., 2018 is shown by the green dashed line in Figure 3.2.1a. Their upper
boundary temperature is ∼50 K cooler than found in this study; this is likely due to the number of data
points considered for the boundary condition. However, as expected, temperature profiles converge after
several scale heights. The agreement between the green profile (Stone et al., 2018) and our converging
profiles validates our technique.
3.3 Drawbacks of Current Temperature Derivation Technique
The above technique Equation 3.1 only considers vertical variations in density when deriving temperature
profiles; however, MAVEN’s orbit trajectory becomes quasi-horizontal with respect to Mars’ surface as
it approaches periapsis. The variations in the apparent vertical density profile are caused by vertical and
horizontal variations and therefore may not be in hydrostatic equilibrium. Mueller-Wodarg et al., 2006
studied the effect of horizontal density variations on derived temperature profiles for Titan’s atmosphere.
An artificial horizontal density gradient was introduced into an isothermal model, and densities were
extracted along an imaginary spacecraft trajectory, and temperature profiles were derived in the manner
explained above (Equation 3.1. Even with an ideal isothermal atmosphere, the extracted temperature
profiles diverged away from the isothermal temperature by over 5 K. Even without knowledge of hori-
zontal density variations in the atmosphere their effects can be seen by comparing an orbit’s inbound
temperature profile to its outbound temperature profile. Physically, periapsis temperatures should be
identical for inbound and outbound profiles during each orbit; the spacecraft is sampling the same region
at the same time so profiles should converge to a single temperature, regardless of integrating inbound or
70
3.4. COMPARISON BETWEEN DERIVED AND EXTRACTED MCD TEMPERATURE PROFILES
outbound data. This effect is exacerbated in a real atmosphere with non-uniform density gradients and
waves. The extent of this effect can be quantified by determining the difference between inbound and
outbound periapsis temperatures for each pass. These differences are denoted ∆T . Figure 3.3.1 shows
∆T as a function of periapsis solar zenith angle. On the dayside, ∆T is typically less than 50 K which
for an atmosphere with along-track temperatures of ∼300 K gives a difference of ∼20%. Potentially
earlier than expected, ∆T begins to increase fairly rapidly around 40°. By 80°, ∆T values are 80-90 K.
This is significant if atmospheric temperatures are 300 K. Beyond 100° ∆T values gradually decrease to
60-70 K. Additionally, the trajectory length in solar zenith compounds this as passes spanning a large
solar zenith range result in larger periapsis temperatures. Temperature gradients are investigated in
more detail in Chapter 4.
Figure 3.3.1: Binned periapsis temperature difference between inbound and outbound profiles as a
function of solar zenith angle
3.4 Comparison Between Derived and Extracted MCD Temperature

Profiles
The Mars Climate Database (MCD) was introduced in Chapter 1, and its output is now compared with
NGIMS densities. This is undertaken as a method of validation which is currently minimal in the upper
71
atmosphere. Solar minimum conditions are used within the MCD. Seasonal change is accounted for by
the use of solar longitude within the model, which also accounts for Mars-Sun distance. Throughout the
MAVEN mission, F10.7 varied between ∼20 and 80 sfu, according to Mars Initial Reference Ionosphere
(MIRI) model (Mendillo et al., 2018). At Earth, during the MAVEN mission, F10.7 did not exceed 140
sfu; this is the value used as the MCD solar average condition, thus, as the majority of the mission
possessed F10.7 values less than this, solar minimum conditions are used. Equivalent values are approx-
imately 80 sfu at Earth. Furthermore, solar cycle 24 was quieter in terms of activity than previous cycles;
therefore solar average was commensurate with typical solar minimum conditions. The MCD outputs,
amongst other variables, the total pressure, temperature, density and volume mixing ratios; from these,
the individual species’ densities can be determined. The MCD allows validation of techniques within a
known environment. In this section, densities are extracted from the MCD using MAVEN spacecraft
trajectories and then used to derive temperature profiles, as shown in this chapter. These are then
compared to extracted temperature profiles along the passes. Figures 3.4.1a and b show derived and
extracted temperatures, respectively. Temperatures are averaged in solar zenith angle and altitude with
bin sizes of 5°×5 km. Figure 3.4.1a shows derived temperature profiles from along-track MCD densities.
As expected, a clear trend in solar zenith angle is observed with the warmest temperatures (>220 K)
at the subsolar point and the coolest on the nightside (∼150 K). The cause of the anomalously warm
temperatures around 180 km at 40-60° is unknown. Figure 3.4.1b shows the extracted along-track
temperatures from the MCD. These are consistently cooler compared to derived temperatures. This
is in agreement with Stone et al., 2018 who found derived temperatures to be warmer than corrected
temperatures, of which the extracted MCD temperatures can be taken as analogous. Figure 3.4.1c
shows a more informative plot where the difference between the derived and extracted temperature has
been calculated. At solar zenith angles below ∼30°, temperature differences are 20-40 K which are
larger than expected for this region. The expected behaviour is that temperature differences should
be largest across the terminator, where density gradients are expected to be most substantial. Locally,
a slight increase in the difference is observed across the terminator, by up to 40 K. A more obvious
increase at all altitudes would be expected. Although this finding does not support the expected beha-
viour, neither does it contradict the behaviour as the MCD density gradients are weaker than found in
observations. With weakened gradients, temperature differences are reduced as observed in the above
figures. Quantitatively, at 150 km, MAVEN densities decrease by just over an order of magnitude across
the terminator, whereas the MCD estimates a more modest decrease of around a factor of two. Thus,
density gradients are significantly less within the model. Overall, derived temperatures are typically
72
within ±20 K of ‘actual’ temperatures. This gives confidence in our method of deriving temperatures.
However, caution should still be taken when analysing observational temperatures; they are statistically
likely to be warmer than actual temperatures given the results in Figure ?? in agreement with Stone
et al., 2018.
Figure 3.4.1: (a) Binned derived temperatures using extracted along-track MCD densities from MAVEN
trajectories. (b) Binned extracted along-track MCD temperatures. (c) Difference between derived and
extracted temperatures. A bin size of 5°×5 km is used in all panels.
73
3.5. COMPARISON BETWEEN MCD AND IN-SITU DENSITY DATA
3.5 Comparison Between MCD and In-Situ Density Data
3.5.1 Comparison with Viking Landers’ Densities
CO2 , Ar and N2 densities are extracted from the MCD along MAVEN trajectories taking into account
altitude, local time, latitude, longitude and solar longitude. Both Viking 1 and Viking 2 possessed
on board upper atmospheric mass spectrometers (UAMS) which allowed neutral densities within the
atmosphere to be sampled during each lander’s descent. Viking 1 landed on 20 July 1976 and Viking
2 on 3 September 1976. Nier and McElroy, 1977 present these neutral density profiles for CO2 , N2 ,
CO, O2 and NO between 100-200 km, along with early estimates for diffusion coefficients. Number
densities shown have a lower limit of ∼106 cm−3 ; thus, densities for less dominant species (NO and
O2 ) are shown only up to ∼135 km and 160 km, respectively. CO2 is measurable up to ∼200 km.
For each species, data are available at increments typically ∼10 km. Nier and McElroy, 1977 includes
drawn lines that are the least-squares fit of a straight line to the data, and it is this line to which MCD
comparisons are made; therefore, any wave activity is not shown. This best fit line acts as a background
profile, therefore removes perturbations caused by wave activity. It must be noted that Viking 1 and 2
are landers, therefore produce one vertical profile each, as such MCD densities are extracted vertically
upwards from the landers’ touchdown location (Viking 1: 22.48° N, -49.97° E, 16 hr LT and Ls=97° .
Viking 2: 47.97° N, -225.74° E, 10 hr LT and Ls=117.6° ). Figures 3.5.1a and b show CO2 (green) and
N2 (magenta) data gathered from the Viking 1 and Viking 2 landers, respectively. Solid (dashed) lines
show Viking (MCD) data. Densities between Viking 1 (Figure 3.5.1a) and extracted MCD densities
show a remarkable resemblance at altitudes 140-200 km for both studied species. Typical densities fall
in the range 1013 -1015 m−3 . From both Viking and MCD data, the atmosphere appears isothermal as
shown by a constant scale height. Below 140 km, Viking and MCD densities diverge, suggesting a
cooling in the MCD. The opposite behaviour is seen from Viking 2 results. Viking 2 still behaviours
isothermally, but is cooler than predicted by the model above 140 km. At 200 km, MCD densities are
over an order of magnitude larger than gathered by Viking 2. Densities become more comparable with
decreasing altitude. Similar density structures are observed below 140 km where densities differ by less
than a factor of five. Similar temperatures are observed.
74
Figure 3.5.1: CO2 and N2 densities as a function of altitude taken from the MCD and (a) Viking 1
and (b) Viking 2. CO2 and N2 densities are shown by green and blue lines, respectively. Viking (MCD)
data are shown by solid (dashed) lines, respectively.
3.5.2 Comparison with NGIMS Densities
In this section, NGIMS to MCD ratios are determined for all available orbits presented such that a global
comparison can be made. This acts as further validation of the MCD. This section aims to answer the
following question: Is there a systematic difference between NGIMS and MCD densities?
Data are binned in solar zenith angle and altitude with a bin size of 5°×5 km. This bin size allows
orbit to orbit variability to be removed while retaining any larger trends. Figures 3.5.2a-c show NGIMS
to MCD density ratios for CO2 , Ar and N2 , respectively, as a function of solar zenith angle and altitude.
In each plot, the colour scale is centred around one, the value taken if NGIMS densities equal those
output by the MCD. Ratios above one (red) indicate the MCD is underestimating densities. A discussion
of each species’ ratio follows.
Figure 3.5.2a shows the ratio of MCD to NGIMS CO2 densities. There is clear day/night asymmetry
in the ratios, as on the dayside ratios are typically larger than one with values increasing towards lower
solar zenith angles up to a factor of five at the subsolar point. On the nightside densities typically differ
75
Figure 3.5.2: Ratio of (a) CO2 , (b) Ar, and (c) N2 densities between NGIMS and MCD data. Ratios
are centered about unity. A bin size of 5°×5 km is used.
by a factor of two; this is fairly invariant in zenith angle. The strong contrast in densities is potentially
due to heating and circulation issues, with the MCD typically producing cooler temperatures. Figure
3.5.2b shows a comparison of Ar densities. Whereas for CO2 there was a clear day/night asymmetry, no
such trend is observed. Instead, it appears that overall Ar is consistently overestimated across nearly all
zenith angles by typically over a factor of two than those found from NGIMS. Again, this overestimation
increases towards the nightside. Underestimations are seen below 30° solar zenith angle. This region is
one of the least sampled; therefore, the uncertainty is larger than around the terminator. MCD densities
are approximately a factor of two larger. Figure 3.5.2c shows N2 ratios and similar behaviour to that
presented for Ar. N2 ratios are overestimated for solar zenith angles bar those below ∼30°. The only
appreciable difference between Ar and N2 ratios is more N2 bins have values closer to unity. A subtle
altitude trend can be seen with ratios minimally diverging away from unity with increasing altitude. The
day/night asymmetry in the CO2 ratios suggest the daily cycle is not fully understood and captured
76
by the MCD. Ar and N2 discrepancies most likely manifest from this. For all three species, the largest
discrepancies occur at the sub-solar point. Thus, it is most likely the local time dependency is not being
successfully captured.
Ar/CO2 and N2 /CO2 ratios using MCD and NGIMS data are compared next. The aim of this
exercise to understand how MCD relative abundances compare to observational data, shown identically
to Figures 3.5.2a-c. Figures 3.5.3a and b show Ar/CO2 ratios taken from MCD and NGIMS data,
respectively. Note the colour scale difference from the previous figures. Ar densities are consistently
less than CO2 densities; therefore, a scale centred around unity is not used. MCD Ar/CO2 ratios reveal
a trend in solar zenith angle with dayside ratios of ∼0.075, increasing to over 0.1 on the nightside. A
similar trend is observed in the NGIMS data; however, ratios are about a factor of two smaller than
values output from the MCD. Day and nightside ratios are ∼0.025 and 0.05, respectively. As well as
day/night asymmetry there appears to be an altitudinal dependence, especially on the nightside where
ratios are maximised to values around 0.1. The clear impact of scale height is evident here as the heavier
CO2 densities drop off more rapidly than Ar, thus increasing the ratio observed. However, the ratios are
not in agreement.
Figures 3.5.3c and d show N2 /CO2 ratios taken from MCD and NGIMS data, respectively. Unlike
for Ar, N2 densities are comparable to CO2 ; thus, the colour scale is centred on one. MCD results are
discussed first. The observed structure is more complex than found for Ar/CO2 . On the dayside CO2
is dominant by up to a factor of ten, shown by the swathe of blue coloured bins. On the nightside CO2
is dominant up to ∼180 km. Above this N2 quickly dominates by over a factor of two. Again, the
influence of scale height is visible. Results from NGIMS data are in very good agreement with the MCD
ratios. Unlike as predicted by the MCD, CO2 is the dominant species on the dayside up to 170 km,
suggesting the MCD is predicting cooler temperatures than observed. Evans et al., 2015 used dayglow
observations by IUVS to derive CO2 and N2 density profiles from 18 October 2014 on 18 November
2014. By averaging in latitude and longitude, N2 /CO2 ratios at 170 km range from ∼0.02 to 0.05. This
is in good agreement with the averaged values found in this study.
The above two sections have compared in-situ spacecraft data with MCD data taken from the same
trajectories. For the Viking comparison, the MCD does not successfully capture the in-situ densities for all
altitudes; this is unanticipated given the MCD is validated against Viking landing data. More noticeable
differences are observed in MCD/NGIMS comparisons. The MCD does not successfully capture trends
77
Figure 3.5.3: Ratio of Ar/CO2 derived from (a) MCD and (b) NGIMS data. Ratio of N2 /CO2 derived
from (c) MCD and (d) NGIMS data. A bin size of 5°×5 km is used.
in solar zenith angle with an underestimation (overestimation) on the dayside (nightside), which is likely
a consequence of the MCD unable to recreate the local time dependency. The relative abundance of
Ar to CO2 is most alarming. The deficiency to capture these trends can be somewhat justified by the
validation undertaken to constrain MCD densities (and temperatures), as the model top of the MCD
has been extended upwards on numerous occasions from ∼90 km to ∼300 km from its inception to
recent years (Forget et al., 1999; González-Galindo et al., 2015; Millour et al., 2018). As has been
done in the lower atmosphere with landers and rovers, upper atmosphere densities need constraining
by observational data. Currently, MAVEN provides the only substantial dataset that characterises the
composition of the upper atmosphere, thus with only one version since MAVEN’s arrival (as of December
2020), the task of validating high altitudes has most likely just begun. Hopefully, our work highlights
the current deficiencies with the model such that the respective team can address them. Overall, it is
not surprising there is currently a systematic difference between model and observational densities, as
78
3.6. PEAK OFFSET IN TGO PROFILES
this is the first documented validation exercise. With time, these comparisons will allow the community
to improve their models with newly added physics as we learn more about the atmosphere.
3.6 Peak Offset in TGO Profiles
During the derivation of along-track densities from accelerometer data, it was observed that the loca-
tion of maximum density did not always coincide with the closest approach. Gravity waves introduce
horizontal density variations that may exceed vertical density changes, causing this apparent behaviour
along the spacecraft trajectory. This offset has been explored and quantified using the difference in time
between these two events. For each orbit, a background density profile is fit. Here, a third-order poly-
nomial is fit to the logarithm of the density profile. Figures 3.6.1a-c show three density profiles derived
during TGO’s aerobraking period. Periapses occurred at 2017-12-22 20:57:50, 2018-01-25 15:39:55 and
2018-02-05 15:54:44, respectively. Dashed vertical lines on each show two events. These are the point
of closest approach and the maximum modelled density. Each profile displays a different offset charac-
teristic. Figures 3.6.1a, b, and c show an offset of 25 s, 0 s and -15 s between closest approach and
maximum density, respectively. Here, the positive (negative) offsets occur when the closest approach
is prior (post-) point of maximum density. Figure 3.6.1d shows each orbit’s trajectory in local solar
time (LST) and latitude. Colours correspond to respective density profiles in Figures 3.6.1a-c. Circles,
squares, and crosses signify the start, closest approach, and end of the trajectory, respectively. These
are shown on the density profiles, likewise. In light of discussion Chapter 2, data are restricted to below
120 km.
These offsets are non-trivial and persistent through TGO’s aerobraking phase and are therefore in-
vestigated further. As this effect is thought to be potentially caused by horizontal density gradients,
the distances travelled by TGO in latitude (Slat ) and longitude (Slon ) along each pass are considered.
A ratio of Slat /Slon is computed for each pass. Hence, a ratio below one signifies TGO has travelled
more distance in longitude than latitude, and vice versa for ratios larger than one. Figure 3.6.2 shows
the absolute time offset as a function of this ratio. The ratio is shown in log scale.
The majority of passes traversed farther in latitude than longitude, apparent by a substantial portion
data above a ratio of one. In this regime, offsets can range from 0 s to around 60 s with a peak in
offsets in the range 20-40 s. This equates to an offset distance of approximately 100 km. In the cases
79
Figure 3.6.1: (a) TGO density profile shown in time from closest approach and altitude (solid red
line). Pass occurred at 2017-12-22 20:57:50. Circle, square and cross signify the start, point of closest
approach and end of pass. Fitted density profile is shown by dashed red line. ∆T is time between
closest approach and maximum density. (c) and (d) follow the same format as (a), but for passes at
2018-01-25 15:39:55 and 2018-02-05 15:54:44. (d) Latitude and local time locations for passes shown
in (a)-(c), coloured accordingly.
where TGO travelled further in longitude than latitude (ratio below one), different behaviour is observed.
Close to one, offsets are commensurate to those already discussed. However, as longitude becomes the
dominant direction of travel, time offsets lessen. The majority of offsets under this regime are less than
20 s, with smaller offsets typically associated with lower ratios.
From the above result, it can be inferred that the conjectured horizontal background densities are
more prevalent and dominant in the latitudinal direction. One possible reason speculated is the different
timescales that densities vary in longitude and latitude. In the former direction, planetary rotation en-
sures strong contrasts in density do not arise between the dayside and nightside (Bougher et al., 1988c).
80
Figure 3.6.2: Time offset as a function of ratio of latitude to longitude traversed during pass
This is also evident later in Section 7.3.1. Moreover, for the latter case, a latitudinal gradient is likely
induced in response to seasonal behaviour and remains established for longer than gradients in longitude.
A short study was performed to understand how the geometry of Mars affects the observed densities.
A simple hydrostatic model was created for two planetary formations - spherical and ellipsoidal. The
ellipsoidal model was based approximately on Mars with an equatorial (polar) radius of 3390 km (3370
km). With the model atmosphere above these surfaces, density data were collected along TGO orbits.
Owing to an ideal atmosphere the spherical planet produces expected profiles; the only important factor
dictating measured density is the altitude. Because of this, the closest approach is coincident with
maximum density for all orbits. A different outcome is recognised when an ellipsoidal atmosphere is
used. In this case, there is frequently a discernible offset between closest approach, as observed in the
data. Similar behaviour was seen for ascending and descending behaviour, as above. However, for the
radii stated above, offsets were significantly shorter than found observational; the majority were less
than 5 s. For model offsets to be commensurate with observational values, the difference between the
radii needs to be several tens of kilometres, which is not realistic for Mars, even including the rough
topography. The planetary geometry does alter the offsets, however to a much lesser degree than
experienced. There are other factors beyond this which most likely exacerbate this effect. The impact
81
3.7. SUMMARY
of specific topography has also been explored. No correlation was found between offsets and pronounced
topography, such as mountainous regions or vast ridges.
3.7 Summary
In this chapter, the data analysis techniques used throughout this study have been presented. NGIMS
data are compared to the Mars Climate Database. An offset is observed between closest approach and
maximum density in TGO profiles. The main results are outlined below.
• MAVEN does not directly measure temperatures so have been derived from NGIMS density data.
A method has been presented, which integrates density downwards producing a vertical pressure
profile. Subsequently, the Ideal Gas Law is used to derive a temperature profile. The advantages
and drawbacks of this technique have been discussed. Using densities extracted from the MCD
along spacecraft trajectories, temperatures have been obtained using the integration method.
Overall, derived temperatures are typically within ±20 K of extracted temperatures, and gives
confidence in our implemented technique.
• Comparisons between NGIMS and MCD densities are made. It appears that CO2 is overabundant,
whilst Ar and N2 are both under calculated. Possible drawbacks of the model have been discussed.
MCD Ar/CO2 and N2 /CO2 ratios have been compared to identical orbits’ ratios from NGIMS.
The local time dependence on density appears not to be captured successfully. Subsolar MCD
densities are up to a factor a five smaller than observed by NGIMS. This behaviour is reversed on
the nightside.
• Examination of TGO density profiles has revealed significant offsets between the maximum ob-
served density and closest approach density. In an ideal atmosphere, there is no offset; maximum
density is found at closest approach. It has been speculated strong latitudinal background density
gradients preferentially cause this finding. Potential other causes have been discussed.
82
Chapter 4
Background Density and Temperature

Analysis
4.1 Introduction
This chapter investigates background density and temperature structures and trends using data from
NGIMS. A simple solution to solving the disparate periapsis temperatures scenario is proposed by making
use of MAVEN’s short orbital period, as this allows average temperature profiles to constructed from
a set of consecutive orbits given sampling location changes relatively slowly. This chapter examines
temporal trends on three distinct timescales: diurnal, monthly and seasonal. The former is possible due
to MAVEN sampling the full range of solar zenith angles multiple times. The second due to MAVEN’s
short orbital period, allowing the atmosphere to be sampled continually (∼5 orbits per day) at similar
orbit-to-orbit locations. The latter is feasible due to MAVEN’s lengthy nominal and extended missions.
By this virtue, occasions arise when latitudes and local times are sampled numerous times, allowing these
factors to be decoupled from the overall trends in density and temperature. By presenting atmospheric
variability over different timescales, it offers a platform from which models can be constrained. For
example, what effect does the 27-day solar rotation have on the thermosphere? Is this visible in the
model data? What are the implications of seasonal behaviour in the upper atmosphere? These are
crucial for understanding the underlying physics controlling atmospheric structure and are explored in
this chapter.
83
4.2. AVERAGING TEMPERATURE PROFILES
4.2 Averaging Temperature Profiles
In Chapter 3, it was shown that inferred inbound and outbound periapsis temperatures are rarely identical
when physically they should be. By considering campaigns of consecutive orbits, average temperature
profiles can be determined with more agreeable periapsis temperatures, which are more physically mean-
ingful. This is demonstrated by using orbits undertaken during DD1 (orbits 714-747 - details in Table
2.2.1). Figure 4.2.1a shows MAVEN’s trajectory through solar zenith angle and altitude for orbits during
DD1. Red (blue) signifies inbound (outbound) orbits. Periapsis solar zenith angles are typically between
105° and 115° with MAVEN spanning 30° . MAVEN sampled a local time and latitudinal range of 17.5-
19.7 hr and 20.7-61.7°. Figure 4.2.1b shows derived temperature profiles from DD1, using the method
presented in Chapter 3, as faded lines. Colour coding is conserved. Orbit-to-orbit variability is seen
with temperatures ranging between 100 K for outbound passes to over 300 K for some inbound passes.
Similarly, each profile exhibits large variations likely caused by pervasive wave activity in the atmosphere.
These perturbations are explored further in Chapter 5. Considering just a single temperature profile may
give an unrealistic profile for a given zenith angle range. To get a clearer idea of typical background
structures, temperature profiles from DD1 are averaged to create single inbound and outbound profiles,
shown by the dashed lines in Figure 4.2.1b. As expected, temperatures close to periapsis are in agree-
ment. This technique is therefore useful for determining an average temperature profile from a range
of consecutive orbits where wave activity and strong horizontal density gradients are present. Inbound
and outbound behaviour over a short duration can be compared owing to minimal local time, latitude
and seasonal variations. By considering consecutive orbits, realistic average temperature structures can
be derived. Here, for example, temperatures across the temperatures range from 150-300 K, whereas
nearer the nightside the structure isothermal around 150 K. These conclusions could not unambiguously
be drawn by considering only individual profiles.
A slightly different approach in deriving average temperature profile is now outlined. From the
same range of orbits, average inbound and outbound Ar density profiles are created using a bin size
of 2 km, as used above. From these average density profiles, temperature profiles are derived using
the method described in Chapter 3. These newly obtained temperature profiles are shown by dotted
lines in Figure 4.2.1b. These are in good agreement with the averaged temperature profiles. For an
isothermal atmosphere, this is expected. However, by averaging out divergences between true and
derived temperature profiles caused by horizontal density variations, it is equivalent to initially averaging
84
4.3. INVESTIGATING DIURNAL TEMPERATURE VARIATIONS
Figure 4.2.1: (a) Trajectories of DD1 orbits (714-744) in solar zenith angle and altitude. Red (blue)
lines show inbound (outbound) trajectories. (b) Derived inbound and outbound temperature profiles
from DD1. Average profiles are shown by dashed lines. Temperature profiles derived from average
density profiles are shown by dotted lines.
out horizontal density variations and then deriving a single profile. This result demonstrates that either
technique can be applied, with the same structures derived. Either way, it is better to characterise
temperatures by averaging, not considering only individual profiles
4.3 Investigating Diurnal Temperature Variations
Any planetary body which is illuminated by the Sun can be divided into two regions - a lit dayside and
a dark nightside. The terminator is defined as the locus of points on a planet or moon where the line
through its parent star is tangent. For an observer on the surface of a planet, the Sun disappears behind
the solid body at a solar zenith angle of 90°. This assumes a perfectly spherical shape and treats the
Sun as a point source. The terminator zenith angle increases with altitude. Beyond the terminator, the
atmosphere is no longer directly solar heated.
The diurnal nature of temperature variations is investigated and will address the following open ques-
tion: What are typical dayside and nightside temperatures at the Red Planet, and how do they compare
85
with modelling efforts? One method to understand such structures is by considering fixed altitudes and
investigating the temperature variation in solar zenith angle. Inbound and outbound background tem-
peratures are interpolated at 160-200 km in 10 km increments. Background temperatures are derived by
fitting a hyperbolic tangent function to individual temperature profiles; this removes wave activity Hedin
et al., 1983. The reasons for studying these altitudes stem from both practical and scientific interest.
Practically, nearly all orbits have a periapsis altitude of 160 km, or lower and therefore more data are
available for analysis, at the selected altitudes, MAVEN is sampling vertical variations, more so, than
near periapsis, and finally, these altitudes are several scale heights below the considered upper boundary,
so this eliminates uncertainty in the upper boundary condition. Scientifically, this is the least-studied
region within the atmosphere, thus will extend our understanding. With this in mind, Figure 4.3.1 shows
interpolated background temperatures as a function of solar zenith angle. Thousands of data points are
presented, thus for clarity, a hyperbolic tangent function has been fitted to each altitude level, and are
shown in corresponding colours.
The dayside is discussed first. Temperatures range from ∼150 K at 160 km to over 300 K at 200
km. As background temperatures have been used, variability due to wave activity is minimal. As a
consequence of this, orbit-to-orbit variability is most visible alongside seasonal behaviour. The latter is
explored later in this section. Average temperatures, shown by the hyperbolic fits, span a smaller range
from 200 K at 160 km to 230 K at 200 km. It is clear that average temperatures become more similar
with increasing altitude; this is clear evidence of the atmosphere becoming isothermal. The reasoning
for this was highlighted in Chapter 1; as the density in the thermosphere is tenuous, the mean free
path of particles becomes comparable to the scale height towards the exobase; thus particles travel vast
distances before colliding, so a large region of the atmosphere is isothermal due to conduction (Schunk
and Nagy, 2009). Further, vertical conduction is large within the thermosphere, and this isothermal
behaviour is shown below. Solar radiation also decreases with altitude; thus, less heating occurs. Tem-
peratures are relatively invariant on the dayside up until ∼75 °. The behaviour at the terminator is
explored in more detail in the next section. Temperatures decrease by ∼40-50 K on the nightside,
with the coolest temperatures seen at the lowest altitudes. The range of average temperatures is more
compact, spanning only 20 K. However, similar variability as observed on the dayside is apparent. One
remark is the removal of wave activity from temperature profiles insignificantly changes average profiles,
and similar orbit-to-orbit variability is still observed.
86
Figure 4.3.1: Temperatures interpolated onto the 160 km, 170 km, 180 km, 190 km and 200 km
altitude levels and shown against solar zenith angle. A hyperbolic tangent function is fit to each altitude
level and coloured accordingly.
The behaviour exhibited by the thermosphere is one that is primarily solar driven. If the atmosphere
were heated only by local solar absorption, temperatures on the night side would be significantly cooler,
as found for Venus (Bougher et al., 1990). Modelling studies, such as Bougher et al., 1990, have invest-
igated the general circulation at Mars and the subsequent effect on temperature. Bougher et al., 1990
saw diurnal behaviour with temperature peaks near 1500 LT (195-305 K) and troughs around at 0500
LT (145-177 K). These values are in good agreement with the results presented in this study. Bougher
et al., 1990. carried out a numerical experiment to highlight the importance of planetary rotation on
temperature dynamics by neglecting the effects of Mars rotation, and the results resembled behaviour
modelled for Venus by Bougher et al., 1988a. Near-symmetric day to night wind and temperature
distributions were found for this scenario, which is not observed in the data. More enhanced day-night
density and temperature contrasts are found when rotation is neglected. Therefore, rotation buffers the
87
day-night temperature contrast. Understanding day/night temperatures allow atmospheric dynamics

to be understood fully. Temperature variation across the terminator may lead to useful insight about
the dynamics in this region. During the discussion above, it was noted that temperature structures
vary between the dayside and nightside, and this is briefly explored now. Temperatures derived for
solar zenith angles below 30° and above 150° are taken as typical dayside and nightside temperatures,
respectively. The altitude range has been extended down to 120 km to include the DD orbits. The mean
and standard deviation are calculated for each altitude level. Figure 4.3.2 shows the average dayside
and nightside temperature profiles.
Figure 4.3.2: Temperatures binned in altitude for solar zenith angles less 30° (red) and greater than
150° (blue). Shaded regions show one standard deviation.
The dayside is warmer than the nightside as outlined earlier. The vertical structure in both regions
is more apparent in this format. Near isothermal behaviour is observed above 170 km with the aver-
age temperatures increasing moderately with altitude from 220 K to 240 K. Below this, temperatures
88
decrease more rapidly down to ∼150 K at 130 K. This hyperbolic-like behaviour has been assumed for
fitting purposes on, for example, Venus (Hedin et al., 1983). This behaviour has been explained in
Chapter 1. The standard deviation also decreases with decreasing altitude; this is likely a manifestation
of both sampling and environmental reasons. The former is due to fewer orbits at these lower altitudes,
thus little opportunity for significant variability. The latter could be borne from the effective thermostat
in the thermosphere, which restricts substantial temperature changes. The nightside exhibits a similar
trend to the dayside, but with less rapid cooling with decreasing altitude. Temperatures range from
100 K to 160 K. Bougher et al., 1990 noted that dayside temperatures appear to vary much more than
global mean values, implying a small response by Mars nightside temperatures to changing solar fluxes.
Less variation is seen for dayside values here. Modelling efforts have shown that the thermosphere is
a capable thermostat via cooling mechanisms (Bougher et al., 1990). There are two similar reasons
behind this. The first is the general temperature trends. The upper thermosphere is heated at different
rates throughout a sol, as demonstrated by a difference of 70 K between day and night at 200 km.
A descent through the thermosphere leads to a convergence of day and nightside temperatures. The
dayside profile only reaches 130 km; however, a simple extrapolation would lead to further convergence
between the profiles. The convergence to common temperatures suggests an effective cooling mechan-
ism in and below this region on the dayside. The optical depths for EUV and UV wavelengths approach
one, so no more dayside heating occur are low altitudes. CO2 cooling is dominant at Venus, however,
due to Mars’ 10-fold smaller O abundance on Mars than Venus, CO2 cooling is rendered less effective
for moderating solar flux changes on Mars (Bougher, 1995). It is believed that winds and conduction are
more effective than the cooling mentioned above (Bougher and Roble, 1991). This effective thermostat
restricts severe day/night temperature contrasts. Further evidence is the uncertainty in profiles. At the
top of the thermosphere, one standard deviation on the day- (night-)side is 60 K (115 K). This is in part
due to the choice of the upper boundary condition and highly variable solar fluxes. The amount of flux
varies with season, heliocentric distance, local time and solar cycles. The range of temperatures rapidly
decreases with descending altitude. By 130 km, the day and night side standard deviations are 20 K
and 30 K, respectively. This narrow range of temperatures shows how efficient cooling mechanisms are
to keep lower thermospheric temperatures within ∼30 K. These results are in very good agreement with
those found by Bougher, 1995, as they found highly varying exobase temperatures (170-305 K) and less
varying temperatures around 120-130 km of 160-190 K over the solar cycle.
Along-track density profiles have been extracted from the MCD, and temperature profiles derived
89
4.4. INVESTIGATING DIURNAL HORIZONTAL TEMPERATURE GRADIENTS
from these. From Chapter 3, it was shown derived temperature profile differ from extracted temperat-
ures, thus for a fair comparison with NGIMS results, derived profiles are used. Data taken at solar zenith
angles below 30° or above 150° are binned in altitude, as undertaken for NGIMS data. The averaged
temperature profiles are shown by dashed lines in Figure 4.3.2. Near the subsolar point (red), the MCD
successfully captures the observed behaviour below ∼150 km with rapid cooling of 50 K over 20 km.
Above 150 km, observations continue to show a dependence on altitude for temperature, as further
warming of 50 K is seen up to 190 km. The MCD implies near-isothermal behaviour of 200 K above
150 km. The MCD clearly is too cool on the dayside by ∼20 K. On the nightside (blue), the MCD
consistently predicts warmer temperatures by up to ∼25 K at 130 km, reducing to ∼10 K at 190 km.
Derived temperatures from NGIMS densities are as expected, based upon previous data as described
above and with comparison to models. As alluded to in Chapter 3, and a more surprising result, is
that there is no clear systematic difference in the model and observed temperatures; the MCD is cooler
(warmer) on the dayside (nightside) than observed in the data.
4.4 Investigating Diurnal Horizontal Temperature Gradients
Average temperature profiles, as shown above, are used here to quantify horizontal temperature gradi-
ents in solar zenith angle, with a particular focus across the terminator. As before, orbits from DD1 are
used to aid with the explanation. Their trajectories through solar zenith angle and altitude are shown
in Figure 4.4.1a. Red (blue) lines are inbound (outbound) legs. Two initial solar zenith angle ranges
of width 2° are used to filter temperature data, as shown by the black dashed boxes in Figure 4.4.1a,
for example. The corresponding inbound and outbound temperature data and averaged temperature
profiles are shown in Figure 4.4.1b. The temperature gradient is determined at each altitude level by
dividing the difference in average inbound and outbound temperature by the change in solar zenith angle
(°). Figure 4.4.1c shows a vertical profile of temperature gradients based on the average inbound and
outbound temperatures in Figure 4.4.1b. All gradients are negative, which suggests temperatures are
decreasing towards the nightside, as expected. Gradients range from 0 K/° to -9 K/° with an apparent
increase in variation with altitude. This is repeated for all combinations of solar zenith angle ranges; the
inbound range is kept constant whilst the outbound range is shifted by 1° . Once all outbound ranges
have proceeded, the inbound range is shifted by 1°. Each gradient also has a zenith angle associated
with it, given by the average zenith angle of the ranges. By cycling through all zenith angles, many
90
4.4. INVESTIGATING DIURNAL HORIZONTAL TEMPERATURE GRADIENTS
gradients can be determined from just a small range of orbits.
Figure 4.4.1: Left pane: Trajectories of orbits 714-744 through solar zenith angle and altitude. Red
(blue) show inbound (outbound) trajectories. Dashed black boxes show the solar zenith angle ranges
used to determine temperature gradients. Middle panel: Temperature data (dots) from shown solar
zenith ranges. Dashed lines show average temperature profiles. Right panel: Horizontal temperature
gradient in K per solar zenith angle degree.
Figure 4.4.2 shows as global overview of gradient data binned in solar zenith angle and altitude.
Temperatures decrease most rapidly near the terminator as solar insolation is significantly reduced. Up
to ∼75° temperature gradients rarely exceed ± 5K/°with no clear trend in zenith angle or altitude.
Closer to the terminator, gradients increase by a factor of at least two to a maximum of -10 to -15
K/° with a noticeable decrease in altitude. Beyond the terminator, nightside gradients are similar to
dayside values.
91
4.5. INVESTIGATING THE EFFECTS OF THE 27-DAY SOLAR CYCLE ON ATMOSPHERIC
DENSITY
Figure 4.4.2: Horizontal temperature gradients in solar zenith angle binned by solar zenith angle and
altitude. Bins which contain data are bordered.
4.5 Investigating the Effects of the 27-Day Solar Cycle on Atmospheric

Density
The following timescale considered is related to the rotation of the Sun. This section aims to answer
the question as to whether solar rotation effects are observable in the thermosphere. The Sun goes
through both decadal and monthly cycles of change, as well as unpredictable events such as coronal
mass ejections (CMEs) and flares. During the 11.6-year solar cycle, the level of solar radiation varies
between solar maximum and minimum (Clette et al., 2014). The appearance of the Sun varies during
the 11.6-year cycle, with sunspot numbers significantly increased during solar maxima. Coronal mass
ejections are large clouds of plasma which may be released from the Sun, and as plasma leaves the Sun
it drags the magnetic field with it. The number and severity of CMEs increase during solar maxima and
is maximised during the declining phase. Currently, it is not possible to study this cycle in its entirety
with MAVEN data alone as data cover less than half the solar cycle, and few places have been sampled
more than once, as will be shown in the next section. A more examinable period is the 27-day solar cycle
caused by differential rotation of the Sun. The equatorial region rotates faster than the polar regions
with periods of 24 days and 30 days, respectively. Magnetic field lines are forced to twist due to dif-
92
DENSITY
ferential rotation which creates active regions that release enhanced solar energy, including periodicities
in extreme ultraviolet (EUV) radiation responsible for heating the thermosphere (Hall and Hinteregger,
1970; Forbes et al., 2006). Hall and Hinteregger, 1970 studied the 27-day rotation effects using the
Orbiting Solar Observatory; this orbiting satellite is located at 550 km above the Earth’s surface and
has a wavelength range 1310 to 270 Å. In the 27-day variation reported in their study, an increase
of about 40% in the solar EUV is identified while the 10.7-cm flux increased from 111 to 201. EUV
flux is typically scaled by the Mars-Sun distance; however, the periodic nature is retained. As noted
in Forbes et al., 2006, relatively little is known about the solar rotation effects on the thermosphere.
However, many studies have examined these effects on the Martian ionosphere. For example, Withers
and Mendillo, 2005 studied the response of the electron density peak to solar rotation and found a
periodicity a ∼26 days.
The approach outlined in Forbes et al., 2006 is used here to identify the solar rotation effects on
the Martian thermosphere. They compare the response of Earth (at 420 km) and Mars’ (at 320 km)
thermospheres to solar rotation effects. Earth’s thermospheric data is derived from the Challenging
Minisatellite Payload (CHAMP) accelerometer measurements. Mars data is inferred from the precise
orbit determination of MGS. Concurrent measurements were taken between 01 January to 31 December
2003 and compared; these results are discussed later. Densities are interpolated to respective altitudes.
Short-term variations caused by pervasive wave activity, amongst others, are averaged out by calculating
a 27-day rolling mean. The window is centred about the date and is shifted by one day. Any potential
solar rotation effects are isolated by subtracting the interpolated data from the rolling mean. In order to
derive relative perturbations, the residual densities are divided further by the mean densities. To remove
large variability, as in Forbes et al., 2006, a 5-day rolling average is determined from the residual. Figure
4.5.1a shows interpolated Ar densities at 190 km for all orbits taken during 2015. The 27-day rolling
average is shown by blue crosses. The large variation seen between DOY 150 and 250 is primarily due
to sampling the nightside. Predictably, the 27-day rolling mean captures the general trend successfully.
In the following, Earth days are used as found in NGIMS data files and periods are converted to
Martian sols. Figure 4.5.1b shows the mean residual density after the subtraction of the 27-day rolling
mean and application of the further 5-day rolling mean. There is a clear periodic motion between DOY
260-340 in the 5-day mean values. This covers nearly three solar rotations. A sinusoidal function is fit
using the method of least squares to the averaged residual data; this is a simple method to determ-
93
DENSITY
Figure 4.5.1: (a) Interpolated Ar density data at 190 km (red circles) for 2015 and 27-day rolling mean
(blue crosses). (b) The 5-day rolling mean of residual densities (red circles) and a sinusoidal fit between
DOY 240 and 340 (blue line)
94
DENSITY
ine amplitudes and periods. An example fit is shown by the blue line in Figure 4.5.1b. For this fit,
the amplitude is 11.8% with a period of 26.6 days. The same procedure is performed with data from
2016, 2017 and 2018 using data at 160 km, 170 km, 180 km, and 190 km for all years. Intervals that
show clear periodic nature, as shown for 2015 data, are selected, and sinusoidal functions are fit as
before. Table 4.5.1 shows five intervals that have been manually identified that exhibit clear wave-like
behaviour along with their respective fitted parameters. Amplitudes vary greatly over the four sampled
years. Amplitudes varied greatly with year from under 4% in 2016 to nearly 30% in 2018. The average
is 14.3±8.4%. For each chosen interval amplitudes generally increase with increasing altitude. The rate
of increase is not constant across the years, nor for each interval. For example, in 2015, the amplitude
decreased by 2.9% over 30 km, however, in 2018, the amplitudes decrease by 13.2% over the same
range. Forbes et al., 2006 only presents results at 390 km; therefore, direct comparisons cannot be
made. Nor can results found here be extrapolated upwards due to inconsistent trends in altitude. Seem-
ingly larger amplitudes with higher altitudes imply that these regions are more susceptible to solar-driven
effects. The mean period is 25.5±1.9 days. Withers and Mendillo, 2005 note a 27-day period at Earth
corresponds to a 26-day period at Mars due to a difference in their orbital periods. This may explain
the reduced average observed period. By identifying intervals where there is clear periodic behaviour,
Forbes et al., 2006 perform fits to data; however, they do not explicitly extract nor state amplitudes and
periods. They find relationships between the F10.7 index and density changes instead. Nonetheless, by
inspection, amplitudes are ∼10-20% but can extend to over 40% within these intervals. Periods are ∼30
days; however, this is restricted by the manual subdivision of the DOY scale. The overall agreement
between the two datasets suggests that strong solar rotation effects are present within the thermosphere.
There are several occasions where parameters could not be fit. One factor is the lack of data.
During 2016, periapsis altitude was raised to above 180 km; therefore, data are not available below this
altitude. Where fits are available, data are taken from the dayside where the atmosphere is directly
heated by the Sun, thus limiting the opportunity to study this effect. Even with missing data, much
can be inferred from the limited results. This response to the solar-rotation impacts is not exclusive
to Mars, as alluded to with quoted studies. Guo et al., 2007 used retrieved densities from CHAMP
accelerometer measurements. Explicit periodic behaviour is observed within the Earth’s atmosphere at
410 km. Maximum amplitudes of oscillations are less than 50% but typically more significant than those
found at Mars. This is anticipated from the decrease in solar flux with distance from the Sun. Although
not investigated at Mars, Guo et al., 2007 used several solar activity proxies such as soft X-ray, XUV,
95
DENSITY
Year DOY Altitude (km) Amplitude (%) Period (Days) Period (Sols)
160 8.9 27.9 27.1

170 11.2 26.8 26.0
2015 240-340
180 11.7 27.5 26.7
190 11.8 26.6 25.8
160 † † †
170 † † †
2016 260-320
180 * * *
190 3.7 27.7 26.9
160 5.7 19.8 19.2
170 5.3 22.8 22.1
2017 250-300
180 8.4 23.7 23.0
190 11.7 23.4 22.7
160 16.1 23.3 22.6
170 18.7 24.3 23.6
2018 50-90
180 27.6 23.2 22.5
190 29.3 23.0 22.3
160 * * *
170 15.1 17.7 17.2
2018 240-290
180 * * *
190 14.9 26.6 25.8
Table 4.5.1: Thermospheric response to solar rotation effects. Table outlines selected intervals for
which periodic activity is identified. Fitted wave amplitudes and periods are shown in percent and days,
respectively. † - Not enough data. * - Data not periodic
96
4.6. INVESTIGATING SEASONAL DENSITY AND TEMPERATURE VARIATIONS
and F10.7 and found thermospheric densities correlate well with these solar irradiances variations. The
observation of periodic behaviour in density data at timescales of solar rotation is not unexpected, given
behaviour observed in electron data (Withers and Mendillo, 2005) and neutral data at higher altitudes
(Forbes et al., 2006). However, this is the first study of its kind in this region, and shows that the
effects are less profound at lower altitudes.
4.6 Investigating Seasonal Density and Temperature Variations
The longer-term trend with season is now examined, and the seasonal effects are quantified - how
much do densities and temperatures vary season? Are the effects negligible? If not, how much do
they differ throughout a Martian year? In order to study the variation in the background atmosphere
due to solar longitude, the effects caused by latitude and local time need to be removed. The general
trend of temperature in latitude and local time for each season or part thereof could be found and
systematically removed from temperature profiles, then averaged temperature profiles from each season
could be compared to examine the seasonal effect. This technique relies heavily on the statistics of the
thermosphere. Decoupling all these factors and deducing profiles which are purely dependent on the
season is difficult. An alternative method is proposed below.
4.6.1 Identifying Regions Sampled Multiple Times by MAVEN
Orbits have been identified that sample similar latitude and local times, but different solar longitudes.
This process is simplified further by removing the effect of altitude by considering one fixed altitude
along each orbit. In the following study, all data are interpolated to 200 km. On the requirement
that each orbit has a periapsis below 200 km, each orbit will contribute two data points for analysis
at two separate latitudes and local times - one for inbound and one for outbound. The next step
identifies groupings of orbits that have similar inbound and/or outbound locations. An initial range
of orbits is considered over which latitude and local time at 200 km do not vary dramatically, for this
a range of 10 orbits is chosen which equates to around two days worth of data. From this group of
orbits, the maximum and minimum latitudes and local times are found. The second range of orbits
is scanned through located at least 200 orbits later. Again, the maximum and minimum latitude and
local time are computed and are compared to those from the first range. Orbit ranges are classed as
overlapping if the maximum latitudes are within 2° of each other, and likewise, minimum latitudes must
be within 2° of each other. Maximum local times must be within 0.5 hr. Similarly for minimum local
97
times. Ranges are manually inspected to see whether sets can be combined. As each orbit contributes
two data points, there are three possible occasions where orbits may overlap. These are by comparing
inbound to inbound, outbound to outbound and inbound to outbound legs. All three are considered
here and shown in Tables 4.6.1-4.6.3, respectively, alongside the date, orbit, Ls, latitude and local time
ranges of each overlapping period. A visualisation of overlapping occasions in local time and latitude
is provided in Appendix A. These ranges are not exclusive to solely studying density and temperature.
They can be utilised to study the seasonal trend in any measured quantity. It should be reiterated that
these overlap regions apply to 200 km and will inevitably vary when considering other altitudes.
Date Orbit (#) Ls (°) Latitude (◦ ) Local Time (hr)
30 Nov 2014 - 06 Dec 2014 333-365 243.6-247.5 72.4 N-74.3 N 3.7-5.7

24 Jun 2017 - 01 Jul 2017 5300-5340 24.0-27.5 72.3 N-74.4 N 3.8-6.0
23 Dec 2014 - 28 Dec 2014 455-481 258.4-261.5 60.4 N-63.6 N 23.9-0.0
06 Feb 2016 - 12 Feb 2016 2635-2670 105.4-108.4 60.2 N-64.0 N 23.9-0.0
26 Feb 2015 - 06 Mar 2015 795-838 298.5-303.4 13.2 N-18.3 N 15.8-16.6
26 Mar 2017 - 03 Apr 2017 4825-4870 339.3-343.8 13.3 N-18.7 N 15.7-16.5
17 Apr 2015 - 29 Apr 2015 1059-1125 327.2-334.0 25.0 S-19.8 S 10.7-11.8
08 Oct 2015 - 21 Oct 2015 1995-2060 52.2-57.6 26.6 S-18.0 S 10.7-11.8
12 Dec 2015 - 21 Dec 2015 2335-2385 80.4-84.5 20.1 N-26.3 N 5.3-6.1
29 May 2016 - 06 Jun 2016 3240-3280 160.4-164.4 20.1 N-26.2 N 5.3-6.0
16 Apr 2016 - 22 Apr 2016 3010-3040 138.4-141.2 55.7 N-59.9 N 10.0-10.7
25 May 2017 - 30 May 2017 5141-5168 9.8-12.2 55.5 N-60.1 N 10.0-10.7
10 Aug 2016 - 17 Aug 2016 3635-3674 201.3-205.6 33.5 S-29.2 S 22.4-23.1
15 Jan 2017 - 23 Jan 2017 4460-4500 299.8-304.3 34.6 S-28.2 S 22.2-23.0
Table 4.6.1: Seven pairs of occasions where MAVEN sampled the same region of the atmosphere in
latitude and local time during inbound legs at 200 km.
98
08 Dec 2014 - 13 Dec 2014 376-401 248.8-251.6 62.4 N-65.3 N 8.4-9.2

14 May 2016 - 20 May 2016 3156-3192 152.3-155.7 62.8 N-66.5 N 8.4-9.3
03 Aug 2015 - 09 Aug 2015 1648-1676 22.4-24.8 66.5 S-63.3 S 0.9-1.7
06 Jan 2017 - 11 Jan 2017 4413-4436 294.4-297.0 67.4 S-65.2 S 1.1-1.8
27 Sep 2015 - 02 Oct 2015 1934-1964 47.1-49.6 65.2 S-61.8 S 13.8-14.6
04 Nov 2016 - 10 Nov 2016 4082-4114 254.7-258.6 64.0 S-60.3 S 13.7-14.6
27 Jan 2016 - 04 Feb 2016 2584-2622 101.2-104.3 19.5 N-25.0 N 1.9-2.5
14 Jul 2016 - 19 July 2016 3486-3515 185.4-188.5 21.1 N-25.9 N 1.9-2.4
23 Sep 2016 - 28 Sep 2016 3866-3895 228.4-231.9 33.2 S-29.5 S 18.9-19.5
26 Feb 2017 - 04 Mar 2017 4678-4713 323.9-327.6 34.8 S-29.5 S 19.0-19.6
24 May 2016 - 30 May 2016 3210-3243 148.8-160.7 57.2 N-61.0 N 7.2-7.9
04 Jul 2017 - 09 Jul 2017 5356-5382 28.9-31.2 56.5 N-59.8 N 7.1-7.7
Table 4.6.2: Six pairs of occasions where MAVEN sampled the same region of the atmosphere in
latitude and local time during outbound legs at 200 km.
99
16 Nov 2014 - 17 Nov 2014 260-266 234.8-235.5 70.0 N-70.5 N 9.7-9.9

07 May 2016 - 09 May 2016 3123-3131 149.0-149.8 68.9 N-59.5 N 10.1-10.4
16 Mar 2015 - 23 Mar 2015 891-929 309.2-313.4 3.1 N-9.0 N 14.2-1439
20 Apr 2017 - 27 Apr 2017 4960-4993 352.7-355.8 3.7 N-7.3 N 14.2-14.7
07 May 2015 - 14 May 2015 1167-1208 338.2-342.2 36.3 S-30.4 S 9.3-10.0
06 Nov 2015 - 15 Nov 2015 2146-2195 64.8-68.8 37.6 S-31.0 S 9.2-10.0
02 Jul 2015 - 03 Jul 2015 1469-1476 6.8-7.4 70.6 S-70.0 S 2.8-3.1
29 Dec 2016 - 02 Jan 2017 4368-4389 289.1-291.6 71.5 S-69.7 S 2.7-3.5
23 Aug 2015 - 30 Aug 2015 1753-1789 31.5-34.7 57.1 S-52.6 S 15.5-16.3
21 Oct 2016 - 28 Oct 2016 4013-4046 246.2-250.3 55.8 S-51.8 S 15.5-16.3
28 Oct 2015 - 05 Nov 2015 2100-2141 61.0-64.3 12.6 S-7.4 S 9.4-10.0
12 May 2015 - 19 May 2015 1198-1233 341.2-344.7 11.0 S-6.5 S 9.5-10.1
29 Dec 2015 - 06 Jan 2016 2430-2468 88.2-91.4 65.1 N-39.5 N 3.7-4.4
24 June 2016 - 30 June 2016 3379-3411 174.3-177.6 36.2 N-41.2 N 3.8-4.4
23 Feb 2016 - 25 Feb 2016 2728-2738 113.3-114.2 69.7 N-70.5 N 21.2-21.6
24 Aug 2017 - 25 Aug 2017 5623-5632 51.5-52.2 71.6 N-72.2 N 21.4-21.7
16 Jun 2016 - 22 Jun 2016 3333-3371 169.7-173.5 5.9 N-11.6 N 3.7-4.4
07 Jan 2016 - 15 Jan 2016 2477-2518 92.2-95.5 4.4 N-10.2 N 3.7-4.4
27 Aug 2016 - 05 Sep 2016 3727-3772 211.8-217.1 45.2 S-40.1 S 20.5-21.4
09 Feb 2017 - 16 Feb 2017 4589-4625 314.3-318.2 46.2 S-41.4 S 20.6-21.4
14 Oct 2016 - 19 Oct 2016 3973-4001 241.4-244.8 68.1 S-65.2 S 14.9-15.9
19 Sep 2015 - 25 Sep 2015 1891-1925 43.5-46.3 69.4 S-66.1 S 14.8-15.9
10 Dec 2016 - 17 Dec 2016 4272-4306 277.8-281.8 60.8 S-56.8 S 2.5-3.3
24 Jul 2015 - 28 Jul 2015 1591-1611 17.5-19.2 58.9 S-56.3 S 2.6-3.1
01 Feb 2017 - 09 Feb 2017 4546-4590 309.5-314.4 22.3 S-16.9 S 20.6-21.3
04 Sep 2016 - 11 Sep 2016 3768-3804 216.6-220.9 21.5 S-15.7 S 20.7-21.4
12 Apr 2017 - 18 Apr 2017 4916-4947 348.4-351.4 25.6 N-29.3 N 14.4-14.9
19 Mar 2015 - 26 Mar 2015 906-945 310.9-315.1 26.0 N-30.6 N 14.4-15.2
Table 4.6.3: Fourteen pairs of occasions where MAVEN sampled the same region of the atmosphere
in latitude and local time during inbound (upper rows) and outbound (lower rows) legs at 200 km.
100
4.6.2 Density Variability with Season
Now that latitudinal and local time variations have been removed, the seasonal effect is investigated.
For each range in Tables 4.6.1-4.6.3, the mean density and solar longitude are calculated. There-
fore, for each overlap there are ρ1 , ρ2 , Ls1 and Ls2 values. A new quantity denoted ρs is given by
|ρ1 − ρ2 |/(ρ1 + ρ2 ). ρs produces a value between 0 (signifying densities within an overlapping pair are
identical) and 1 (signifying densities are very different).
Figure 4.6.1 shows density variation, ρs , in solar longitude. Mars’ orbit is shown by the black circle.
The arrow shows the direction of Mars’ orbit. Grey dashed lines mark out four seasonal reference points
(Ls=0°, 90°, 180°, and 270°). For each overlapping pair shown in Tables 4.6.1-4.6.3, a line joins the two
different solar longitudes from which ρs was calculated; therefore there are 27 lines from the 27 pairs.
Each line is coloured by ρs . For example, the near-vertical green line to the left of the Sun is the pair-
ing 23 December 2014-28 Dec 2014 and 06 February 2016 - 12 February 2016 (second row, Table 4.6.1).
The expectation is that overlapping ranges with similar solar longitudes (shorter lines) should yield
low ρs values as Mars has not precessed far through its year and vice versa, different solar longitude
(longer lines) should have larger ρs values associated with them. There are several occasions where sim-
ilar locations were sampled between Ls=90° and Ls=180°, as shown in the upper-left portion of Figure
4.6.1. ρs values are expectedly low, barely exceeding ∼0.3. During these times, the Mars-Sun distance
is maximal; thus, influences are likely to be minimal. A grouping of pairs is observed with similar Ls
differences between northern spring and northern summer. ρs values during these seasons are typically
no less than 0.6. Thus, it can be inferred that densities vary to a more considerable degree in this region.
The last primary result this plot reveals is the potential effect of the ’dust storm season’ on densities.
As discussed in Chapter 1, the dust storm season straddles perihelion (Ls∼271°). Lines emanating from
this region take ρs values greater than 0.8; this is true for those extending to Ls∼45° which signals
a definite seasonal variation, but also those that stretch from Ls∼280-290° to only Ls∼0-20°. These
latter solar longitudes differences are similar to those shown in the upper left. ρs values are substantially
different, suggesting the dust storm seasons plays a significant role in driving density changes.
Studies so far have not rigorously studied seasonal variation in such a manner. Liu et al., 2017 studied
the seasonal variation of longitudinal structures using NGIMS data. Although seasonal density variations
101
Figure 4.6.1: Density variations throughout a Martian year. Each line represents an overlapping period
shown in Tables 4.6.1-4.6.3. One line is an occasion where MAVEN sampled the same region of the
atmosphere in latitude and local time. Lines have feet at sampled solar longitudes. Lines are coloured
by how different densities are between sampled periods. 0 (1) is no (large) change.
were not explicitly explored in their study, seasonal variations can be inferred from their work. Similar
latitudes and local times were sampled in May 2015 and November 2015; these cover substantially more
orbits. Solar longitudes for the two months were Ls∼340° in May, close to northern spring equinox,
and Ls∼70° in November, close to northern summer solstice and aphelion. Using densities presented in
their study, ρs values at 200 km for these months are ∼0.75. This is in line with expectant values from
Figure 4.6.1. The orbital ranges used in this study are sufficiently large to produce comparable results
to a study using a month’s worth of data.
The increase in densities around perihelion owing to an expansion in the atmosphere caused by
lofted dust. To achieve ρs values upwards of 0.8, densities must vary by a factor of 10. Furthermore,
102
to achieve values above 0.98, which is an upper limit in our study, the discrepancy in densities must be
a factor of 100. Such large seasonal changes have not been quantified observationally
4.6.3 Temperature Variability with Season
An identical examination to that above is now performed using temperature data. Temperatures are in-
terpolated onto the 200 km altitude level. Figure 4.6.2 shows an identical plot to Figure 4.6.1, however,
seasonal variability in temperature is quantified as |T1 − T2 |, where T1 and T2 are average temperat-
ure values for the sampled regions. Whereas for density variations, differences vary by over an order
of magnitude, average temperatures typically vary by no more than ∼100 K, therefore only absolute
temperature differences are taken forward.
The main contrast between seasonal density and temperature variations is the relative variability
throughout a Martian year. This is further evidence that the thermosphere is an effective thermostat,
restricting temperatures to within a fixed range. Explicit behaviour is not apparent as it was in the
previous section with variation during the dust storm season not evident. Consistent values are observed
between northern summer and autumn, with typical differences of 40-60 K. The above results have been
achievable due to MAVEN’s extensive mission length; more pairings will occur if additional, currently
unused, data from 2019 onwards are utilised. Hence, model comparisons are necessary. Bougher et al.,
2015 studied seasonal variations using the Mars Global Ionosphere-Thermosphere Model (M-GITM).
While M-GITM has the capabilities to discern effects caused by the 11-year solar cycle, the consequence
of such an external factor cannot be yet be identified in the data. Bougher et al., 2015 present results
from solar minimum, moderate and maximum conditions. MAVEN arrived during the declining solar
phase (moderate to minimum); as such, it is these conditions from Bougher et al., 2015 that are used
for comparison (Lee et al., 2017). Exospheric dayside (lower zenith angles) are considered first in
Bougher et al., 2015. For solar moderate (minimum) conditions, temperatures lie in the range 260-300
K (190-220 K). As this is for an idealised situation, one would expect to see more substantial variation
due to wave activity, e.g. gravity and thermal waves. The most considerable variations, as seen in
observations, are between perihelion and aphelion (Ls=251° and 71°, respectively). Many of the data
emerging from around perihelion are taken on the dayside, thus, on the whole, are comparable with
Bougher et al., 2015. Bougher et al., 2017 used NGIMS and Imaging Ultraviolet Spectrograph (IUVS)
data to examine seasonal and solar activity trends in scale heights and temperatures. One key finding
in agreement with results here is significant heating between aphelion and perihelion of several tens of
103
Figure 4.6.2: Temperature variations throughout a Martian year. Each line represents an overlapping
period shown in Tables 4.6.1-4.6.3. One line is an occasion where MAVEN sampled the same region
of the atmosphere in latitude and local time. Lines have feet at sampled solar longitudes. Lines are
coloured by how different temperatures are between sampled periods.
104
K. One of the more substantial differences is observed between Ls∼260° and Ls∼105°. These occurred
in December 2014 (end of solar maximum) and February 2016 (start of solar minimum), respectively
(Table 4.6.1 - second row). The M-GITM predicts a temperature difference above 200 K; part of this
substantial difference can be attributed to temperatures modelled during peak solar conditions, rather
than declining phases. Cross-spacecraft comparisons can be made as initially presented in Bougher et al.,
2000. The most pertinent spacecraft for this particular comparison are Viking 1 and 2 during minimum
conditions at Ls∼95-120° and Mariner 9 nominal mission during solar maximum conditions at Ls∼300°.
Respective temperatures are 186 K, 145 K and 325±40 K; these are per expected differences found
in this current study. The next two largest differences cannot be attributed to different solar phases.
Apart from the highlighted cases above, it can be assumed that a large portion of the pairings occur
during solar minimum and moderate conditions. According to the M-GITM, temperature differences
should scarcely exceed 70 K. Further, spacecraft temperature profiles during solar minimum vary by
a similar amount, thereby showing agreement with MAVEN results presented here. The results found
here indicate further that the M-GITM is a robust model which can successfully account for seasonal
variations. Martian seasons are affected by both planetary obliqueness and distance from the Sun. If
the former is temporarily neglected, the effect of the latter can be explored and is undertaken below.
Given the clear seasonal dependence on density, as shown in the previous section, the lack of trend in
temperature is unexpected. This highlights how effective the atmosphere is at controlling temperature
via its thermostatic features, e.g. thermal conduction and CO2 cooling. This lack of trend is worth
studying further.
Further Investigation Into Seasonal Temperature Variations
At any given distance from the Sun, the solar flux (F ) is inversely proportional to the Mars-Sun distance
(R) squared. If it assumed that thermospheric temperature (T ) is proportional to F then T ∝ R2 , then
the constants of proportionality need not be known if comparisons between the two scenarios are made.
The expectation is then T1 /T2 =(R2 /R1 )2 , For each pair shown in Tables 4.6.3-4.6.1, the first range is
subscript one, the latter range two. This relationship is tested now using the pairs in Tables 4.6.1-4.6.3.
For each overlap, a value T1 /T2 is determined using temperatures from earlier. To get the uncertainty in
this value a random temperature from range one is divided by a random temperature in range two and
repeated 1000 times from which the standard deviation is calculated and taken to be the uncertainty.
This is repeated for all overlap occasions. Figure 4.6.3a shows T1 /T2 against (R2 /R1 )2 . The expected
result is T1 /T2 =(R2 /R1 )2 which is shown by the black dashed line. This is the expected behaviour for
105
a purely, constantly heated atmosphere. Overall, the behaviour exhibited by the data is in agreement
with the expected trend to within the tolerance of the uncertainty.
A measure of the differences between the actual and expected T1 /T2 values are determined for each
pair. The expected value is given by (R2 /R1 )2 , as governed by the above law, and the difference is
therefore |T1 /T2 -(R2 /R1 )2 |. Figures 4.6.3b and c show these differences as a function of local time and
latitude, respectively. Solar zenith angle is not used as this varies within each pair owing to planetary
obliqueness and M-S distance. From the basic theory explained early in this section, it may be expected
that only the dayside follows the behaviour as only this portion of Mars is directly heated by solar
radiation. There is a slight trend with local time with a grouping of points located around noon with
low |T1 /T2 -(R2 /R1 )2 | values (<0.25). As expected, this implies more direct heating on the dayside.
Slightly larger values are seen in the dawn and dusk sectors. Figure 4.6.3c plots the differences in
latitude. Near the equator, differences are typically ∼0.1. Differences increase towards the northern
polar region as differences can exceed 0.3. Differences in the southern hemisphere are of similar value
to near the equator. Near the poles, solar radiation passes through more atmosphere to reach a given
altitude compared to at the equator, as shown in Chapter 1. By sampling the same season, local time
and latitude should result in small differences. Overall, the behaviour exhibited is what is to be expected
by a solar-driven atmosphere.
4.6.4 Southern Polar Warming
In Section 4.6.3, results have shown that there are certainly seasonal variations in density and temper-
ature. The unanticipated small differences between expected and actual T1 /T2 values in the southern
hemisphere lead to an interest in potential polar warming. This is inspired by a study by Bougher et al.,
2006. Initially, only the orbits shown in Tables 4.6.1-4.6.3 were used, however this severely limited
confidence in results. Thus, a new approach was taken, whereby all data could be included. Four alti-
tude levels are selected - 160 km, 170 km, 180 km, and 190 km and data are interpolated accordingly.
Although passes probe deeper than these altitudes, good coverage is available for all latitudes. Data are
then categorised into one of four bins. Orbits can either be on the dayside (06 - 18 LST) or nightside
(18-06 LST) and either near perihelion (1.38-1.52 AU) or near aphelion (1.52-1.66 AU). Increasing the
number of bins reduces the number of orbits per bin; thus, four bins are used for brevity and clarity.
Figures 4.6.4a-d show temperature as a function of latitude for the four considered altitudes. Here, red
(blue) lines symbolise orbits near perihelion (aphelion) and solid (dashed) lines signify dayside (nightside)
106
Figure 4.6.3: (a) T1 /T2 against (R2 /R1 )2 at 200 km. Crosses are average T1 /T2 . Error bars are
one standard deviation. The dashed line shows the expected fit of T1 /T2 =(R2 /R1 )2 . (b) Difference
between average T1 /T2 (or (R2 /R1 )2 ) and expected value as a function of local time. Error bars are one
standard deviation. The dashed line is a best fit. (c) Difference between average T1 /T2 (or (R2 /R1 )2 )
and expected value as a function of latitude. Error bars are one standard deviation.
orbits. A rolling mean is employed with a 5° window.
Figure 4.6.4a shows temperature at 160 km. The expected behaviour with the warmest temperat-
ures found near perihelion on the dayside is seen. Substantial warming occurs across the equator with
temperatures exceeding 240 K. Coolest temperatures are observed on the nightside around perihelion,
extending down to 160 K. Aphelion temperatures are fairly invariant throughout the day, with temper-
atures between 160 K and 200 K. Figure 4.6.4b shows temperatures at 170 km. Temperatures across
all cases are more alike at this altitude and begin to exhibit behaviour observed at higher altitudes. The
dayside perihelion temperatures remain the warmest. A peak of 20 K in the dayside aphelion temperat-
ure is apparent around 0-40°N. This is most likely due to northern summer occurring near aphelion, thus
being directly heated. Northwards of ∼40°S, nightside perihelion and aphelion are near identical with
temperatures generally 180-200 K. Southwards of ∼40°S sees an interesting trend. Significant warming
of ∼30 K is discernible towards the south pole. In the opposite, and more expected manner, nightside
107
Figure 4.6.4: Average temperatures on the dayside (solid lines) and nightside (dashed lines) near
perihelion (red) and aphelion (blue) interpolated at (a) 160 km, (b) 170 km, (c) 180 km and (d) 190
km as a function of latitude.
108
4.7. SUMMARY
aphelion temperatures cool by ∼30 K towards the south pole. Dayside warming/cooler is observed but
occurs nearer the south pole. Figures 4.6.4c and d show very similar behaviour to that observed at 170
km (panel b). The cooling around aphelion agrees with findings from MGS during the southern winter
solstice (Ls=70-90°); temperatures near 120 km increase slightly from the equator (130–140 K) toward
mid-latitudes (up to 160 K), while dropping to 100 K near the South pole (Keating et al., 2003). ODY
data revealed strong northern polar warming within the 100-130 km altitude range during its aerobraking
campaign (Ls=265°–310°). At 110 km, warming increased from ∼100 K to 170-200 K, maximising on
the nightside. At 120 km, warming was still observed but to a lesser degree; temperatures reached
160-170 K. Obvious northern warming is not seen for any of the four cases presented here. Bougher
et al., 2006 suggested a strong inter-hemispheric circulation during northern winter to be responsible
for this warming. This is driven by the stronger insolation and dust heating near perihelion. This cir-
culation most likely explains the strong polar warming, not previously observed at these altitudes. Bell
et al., 2007 modelled this further using the M-GITM to understand vertical dust transport. Bougher
et al., 2006 did not simulate southern warming in light of aerobraking findings; however, this should be
considered in the future.
4.7 Summary
In this chapter variations in background densities and temperatures are investigated primarily on three
timescales: diurnal, 27-day and seasonal.
• Clear diurnal behaviour is observed with dayside temperatures in the region 150-300 K, increasing
with altitude. Nightside temperatures are found to be typically between 100 and 250 K. Evidence of
an effective thermostat is presented with dayside and nightside temperatures appearing to converge
in the lower thermosphere. Temperature gradients are most significant across the equator taking
values of ∼10-15 K/°. This is expected within a solar-driven thermosphere.
• The effect of the 27-day solar rotation has been explored. The removal of a 27-day rolling mean
reveals a strong periodic trend, with densities varying by up to 10%. This is in agreement with
results at higher altitudes. This trend is not always visible within the data, inferring some spatial
and temporal dependence. The most apparent periodicities occur on the dayside.
• The seasonal variation in density has been investigated using the overlapping orbits. Commensur-
ate with small differences in sampled solar longitudes, density differences are comparatively small.
109
4.7. SUMMARY
This is true during Northern summer and autumn. The most considerable density differences occur
when one of the two ranges in a pair is sampled during the dust storm season (around perihelion).
This is true even for solar longitudes that are separated by 90°.
• Seasonal temperature variations have also been studied. The general trend is ambiguous. However,
a general agreement with the sparse previous measurements during solar minimum and the Mars
Global Ionosphere-Thermosphere model (M-GITM) is observed. In line with the results from M-
GITM, temperatures typically vary by no more than ∼60 K across solar minimum and moderate.
As MAVEN approached during the declining phase, it is unknown how much temperatures would
vary by.
• Evidence of southern polar warming on the nightside near perihelion is presented. Temperatures
increase by up to 30 K. This has not been observed in data before. However, it has been explained
using dust-induced circulation as proposed by Bougher et al., 2006.
Whilst this chapter has investigated density and temperature variations of substantial time scales (one
day to several months), the next chapter probes variations along each orbit. These variations are
interpreted as waves and are characterised to understand them fully.
110
Chapter 5
Gravity Waves in the Martian Upper

Atmosphere
5.1 Introduction
In this chapter, gravity waves within the Martian atmosphere are introduced, and their role described
therein. This chapter aims to answer the following questions: what are typical amplitudes and dom-
inant wavelengths of thermospheric gravity waves? How do these vary with zenith angle and season?
This will hopefully lead to the improve initialisation conditions and gravity wave inclusion in GCMs
should lead to a more accurate representation of the atmosphere. Gravity waves are extracted from
NGIMS data. Waves in the Martian atmosphere are time-varying structures with spatial scales, as
measured via wavelengths, of tens to hundreds of kilometres in the vertical and horizontal direction.
These wavelengths and amplitudes are extracted from the data, and their diurnal and seasonal variations
are explored. Waves are expected to grow with altitude, as described in Chapter 1; this is tested by
examining how amplitudes vary with altitude. As topography is a known mechanism of wave generation
at the surface, possible correlations between amplitude and topography are explored. Finally, waves
from consecutive orbits are shown to demonstrate wave propagation within the atmosphere.
Gravity waves are imperative to study in the context of the Martian atmosphere, especially within
the upper atmosphere. As described in Chapter 1, wave amplitudes grow with altitude to conserve
momentum. It is within the upper atmosphere that these waves begin to dissipate and break, thus
depositing substantial momentum and energy within this region. This can lead to the slowing of flows
111
5.2. ATMOSPHERE PERTURBATIONS
and warming/cooling of the atmosphere. Therefore, gravity waves as a source of momentum and energy
are crucial to our understanding of the upper atmosphere. On a more practical level, more spacecraft
will unquestionably arrive at Mars in the not-too-distant future. Waves are related to variability in the
atmosphere; understanding this variability is essential for predicting atmospheric conditions encountered
by future spacecraft to Mars. Where do we see the largest gravity waves? What are their typical amp-
litudes? These questions are answered below.
5.2 Atmosphere Perturbations
To extract perturbations from the temperature and density profiles, unperturbed background profiles
need to be determined. Only Ar profiles are used in this analysis. England et al., 2016 showed that
near-identical wave structures are observed in at least CO2 , Ar and N2 profiles. Inbound and outbound
legs are treated separately and therefore have different background profiles. There are many options
which have been considered and used previously, including running averages (Creasey et al., 2006b;
Fritts et al., 2006). Temperature is highly variable in the thermosphere, so there is not a single function
which can be used universally for all temperature profiles. The general behaviour shows an increase
in temperature with altitude, which, at high altitudes, becomes isothermal (Bougher et al., 2017). A
hyperbolic tangent function is often used for the background temperature in the thermosphere, given
by Equation 5.1,
z − zmin

T (z) = Tmin + (Tmax − Tmin ) tanh (5.1)
∆z
where Tmin and Tmax change depending on location, local time and season, z is the altitude above the
surface, zmin is periapsis altitude and ∆z is the altitude range where the temperature is not isothermal.
A hyperbolic tangent function is used as this allows the isothermal behaviour at high altitudes in the
thermosphere and exosphere to be captured. This function will also successfully capture the rapid de-
crease in temperature in the 120-140 km region. This function contains no oscillations and reduces the
risk of introducing artificial waves to the fit. The background temperature is fit using the method of
least squares. The isothermal temperature calculated earlier (Equation 3.2) is used as an initial value for
Tmax and the temperature at periapsis as Tmin . ∆z is initially set as 100 km. All three are optimised
to give the best-fit background profile using the method of least-squares. A representative set of orbits
were examined, and the profiles were captured successfully by the fit. Figures 5.2.1a and d show the
112
5.2. ATMOSPHERE PERTURBATIONS
inbound and outbound temperature profiles for orbit 1060. Temperature increases with altitude as heat
is conducted downward. The isothermal behaviour can be seen in panel d with a value of 260 K. Panel a
does not display such clear behaviour, suggesting the upper atmosphere continues to be heated at higher
altitudes. The profiles have clear waves present. To extract perturbations from the density profiles, an
exponential function could be fit, but this assumes the atmosphere to be isothermal, which has been
shown not be the case. A third-order polynomial is chosen and fit to the logarithm of the density.
This low order polynomial is chosen to reduce the possibility and impact of edge effects. Inbound and
outbound density profiles are shown for orbit 1060 in Figures 5.2.1b and e.
Normalised perturbations are determined by subtracting the data from the background profile and
normalising by dividing this residual by the background profile. As seen in Figures 5.2.1c and f, gravity
waves are composed of smaller waves with varying wavelengths. The main properties considered here
are amplitude and wavelength. A fast Fourier transform (FFT) is performed on the extracted normalised
perturbations allowing the wave properties to be identified. A single FFT is performed over the entire
altitude range for each leg. Due to wave growth/decay, amplitudes do not remain constant for a given
wavelength. Extracted amplitudes are averages over the sampled range since a straightforward Fourier
transform does not allow for accounting of wave decay. Using the amplitudes and wavelengths from the
four most dominant waves, a new wave can be constructed using the wavetrain described by Equation
5.2.
3
2π

(5.2)
X
An sin z + φn
n=0
λn
where An , λn and φn are the amplitude, wavelength and phase of the nth wave. φn is initially set to 0◦
for all n. The phase of each composite wave is unknown and therefore needs to be fit for comparisons
with the observed perturbations. This is determined by least-squares fitting. The fits were automated
given the large number of orbits being analysed, with manual checks performed on a representative
sample set. In some cases, two waves could capture the gravity wave successfully, in others more were
needed. It was found that, overall, gravity waves were generally captured sufficiently well with four
waves. By considering four waves, the varying wavelengths can be captured. By fixing the number
of waves, comparisons between properties can be made, and the fitting process is done automatically
without human checks. As discussed, four waves are used to reconstruct gravity waves. The initial
wavetrain is used as an estimate for the best fit wavetrain. The wavetrain is refined further using the
method of least squares by adjusting amplitudes, wavelengths and phases of each wavetrain during the fit.
113
5.3. EFFECTS OF SPACECRAFT TRAJECTORY ON INFERRED GRAVITY WAVE
WAVELENGTHS
Figures 5.2.1c and f show the perturbations (faded lines) and their wave fits (dark lines) for both
the inbound and outbound leg of orbit 1060. This orbit has been shown as it highlights ’typical’ gravity
waves found throughout this study. The inbound wave profiles (Figure 5.2.1c) clearly exhibit behaviour
described earlier; temperature (red) and density (blue) waves are in near perfect antiphase as described
by Equation 5.3, discussed in Section 5.4. Both structures in the temperature and density fields are
captured well by the wave fits. The perturbations are centred around zero, which shows waves have been
successfully extracted from the background profiles. Wave amplitudes are around 10%, and wavelengths
are 20-30 km. As shown later, this wave is typical. The amplitudes are larger than uncertainties which
suggest physical waves are present. The density amplitude around 160 km is underestimated by about
5%. There are shorter wavelengths of around 5 km which have not been captured. These are visible up
to around 180 km. Given their relatively small amplitudes of a couple of percent, we do not consider
them in this analysis. Pressure perturbations are shown by the grey dashed line. The amplitude of such
waves is typically an order of magnitude lower than density and temperature amplitudes. The outbound
wave profiles (Figure 5.2.1f) show an example of a long-wavelength gravity wave. Temperature (red)
and density (blue) are still in anti-phase like the inbound leg but to a lesser degree. The general trend
of the wave is captured well. Amplitudes are about 10%, and vertical wavelengths have values between
60 km and 80 km; as discussed later in this chapter, these wavelengths are longer than typical values.
Smaller-scale perturbations with visible wavelengths of a few km are visible but not included in the
analysis. Their amplitudes are at least an order of magnitude smaller than the more dominant waves
and therefore not considered here. The similarity between fitted temperature and density perturbations
in respective inbound and outbound legs are shown by similar amplitudes and wavelengths in Table.
5.2.1. This gives confidence in the methods used.
5.3 Effects of Spacecraft Trajectory on Inferred Gravity Wave Wavelengths
In this short section, the effects of trajectory on measured gravity wave wavelengths are investigated.
This is aided by the use of the MCD which initialises gravity waves in a random direction. Although
this scenario is highly idealised, the results highlight potential problems with current techniques used
to derive wavelengths. Five scenarios are considered where wavelengths of 5 km, 10 km, 15 km, 20
km and 25 km are used. As shown above, these are typical wavelengths. As gravity waves are dir-
114
WAVELENGTHS
Figure 5.2.1: Panels (a) and (d) show temperature profiles (red solid line) for inbound and outbound
legs, respectively. Red dashed lines are fitted background profiles. Panels (b) and (e) are Ar abundance
profiles (blue solid line) for inbound and outbound legs, respectively. Blue dashed lines are fitted
background profiles. Panels (c) and (f) show the inbound and outbound extracted waves. Red are
temperature structures and blue are density structures. Faded lines show the actual waves. Dark lines
show the fitted wavetrains. The grey dashed lines show pressure perturbations. Data is taken from
orbit 1060. Periapsis location: -2.1◦ latitude, 149.9◦ longitude, 12.1 LST, 327.3◦ Ls . Error bars are one
standard deviation of the random temperature profiles.
ected randomly in the MCD, the only possible way to guarantee accurate wavelength measurements
is to fly vertically through the wave. An orbiting spacecraft does not this. At periapsis, MAVEN is
moving quasi-horizontally, therefore would not measure a vertical wavelength. This section attempts
to measure the vertical wavelength along a spacecraft trajectory. In Section 5.2, a robust technique
used to extract gravity wave characteristics from data is described. This technique determines the most
dominant wavelengths for an entire leg. As the effect of spacecraft trajectory on the measurement of
wavelength is sought, ideally wavelength as a function of a trajectory parameter is wanted. One such
115
WAVELENGTHS
A1 A2 A3 A4 λ1 λ2 λ3 λ4
T 11.3 6.2 5.6 5.5 23.6 32.6 16.3 20.9

Inbound
ρ 10 7.6 7.1 3.9 25.3 16.8 33.7 19.1
T 8.3 3.3 1.7 1.7 73.4 47.9 37.1 17.8
Outbound
ρ 11 4.2 2.8 2.3 56.1 37.2 24.7 122
Table 5.2.1: Amplitude and wavelengths of constituent waves that form the wave extracted from orbit
1060. Inbound and outbound leg values are shown. A are amplitudes (%) and λ are wavelengths (km)
parameter that is used now is the angle to the planetary normal. This is 90° at periapsis and increases
away from closest approach. As gravity waves in the MCD are periodic at the initial wavelength, a
simple analysis can reveal wavelength as a function of angle to the normal. Peaks (maximum) in density
are identified, and the difference in altitude between adjacent peaks is determined, giving an estimate
of the wavelength. The average angle to the normal between adjacent peaks is used as a measure
of the angle. Thus, several values for the wavelength can be measured along the trajectory. Figure
5.3.1a shows apparent wavelength as a function of angle to normal for 5 km, 10 km, 15 km, 20 km
and 25 km model wavelengths. Data points darken with increasing wavelength length. In all cases,
the measured wavelength near periapsis (90°) is ∼0 km, as expected. A horizontally moving space-
craft cannot detect vertical variations. Below angles of ∼92°, all measured wavelengths are identical,
regardless of the model wavelength. Therefore, using these data to determine vertical wavelengths is
futile. The measured wavelengths increase away from periapsis as the spacecraft begins to travel with a
vertical component with respect to the Martian surface. The measured wavelengths appear to plateau
to values less than the model wavelengths. With longer orbital data, wavelengths at higher altitudes
could be measured. To understand the general behaviour of how successfully wavelengths are captured,
measured wavelengths are normalised by the known model wavelength. Figure 5.3.1b shows normalised
wavelengths, λ̃. Similar behaviour is seen between all wavelengths. λ̃ values are ∼0 near periapsis, as
expected. These increase with increasing trajectory angle. λ̃ plateaus around 0.8 at around 96°. This
is equivalent to an altitude of ∼240 km. Therefore, over an entire trajectory, wavelengths are almost
certainly always underestimated if they are purely vertical.
116
5.4. COMPARING TEMPERATURE AND DENSITY PERTURBATIONS
Figure 5.3.1: (a) Estimated wavelengths as a function of angle of trajectory to Mars’ normal for
simulated 5 km, 10 km, 15 km, 20 km and 25 km vertical wavelengths. (b) Estimated wavelengths
normalised by the original known simulated wavelength as a function of angle of trajectory to Mars’
normal.
As explained in Chapter 8, this work could be explored further with the introduction of MCD v6.0,
where wave can propagate horizontally.
5.4 Comparing Temperature and Density Perturbations
Using the Ideal Gas Law, the perturbations of a gas with pressure p, density ρ, temperature T and mean
molecular mass m may be expressed as
δρ δp δm δT
= + − (5.3)
ρ p m T
As only Ar is considered here, δm is zero. As seen from Figures 5.2.1c and f, density and temperature
perturbations appear primarily in antiphase for all altitudes. Using Equation 5.3, we may calculate δp/p
and obtain typical values of up to 1%, whereas δρ/ρ reach 10% and δT /T reach slightly smaller, but
similar values as Eq. 5.3 needs to balance. δp/p values are calculated using the same method to extract
density perturbations. This shows that pressure perturbations are overall a factor of about 10 less than
117
density and temperature perturbations.
Histograms were constructed to get an overview of gravity wave properties in the upper atmosphere.
Figures 5.4.1a and b show the amplitudes of wave structures in the temperature and density fields,
respectively, without spatial or temporal distinction. Each bin has a width of 2.5%. The histograms
are in agreement with each other. The median temperature and density amplitudes are 10% and 8%,
respectively. Standard deviations are 11% and 10%, respectively. A broad peak in amplitude is seen
around 10%. To understand the distribution of waves further, the histogram is broken down into the
four waves which comprise each fitted wavetrain. Less dominant waves are shown by faded lines. The
most dominant constituent wave has the broadest distribution in both density and temperature. This
highlights the variability in characteristics and is discussed further later. Density amplitudes are system-
atically smaller than temperature amplitudes by about 2-4%.
Figures 5.4.2a and b show the distribution of gravity wave wavelengths. They have been broken
down identically to Figures 5.4.1a and b. As found for the case of amplitudes, the results for tem-
perature and density wavelengths are in very good agreement with each other. From Equation 5.3,
perturbations in density and temperature fields are expected to be in anti-phase. Due to this, density
and temperature wavelengths must match to ensure waves remain in anti-phase. The histograms take
into account all waves are nearly centred on the median values of 27 km and 24 km for temperature and
density, respectively. Standard deviations are 26 km and 24 km, respectively. A clear second peak is
observed in Figure 5.4.2a around 80 km. The largest apparent perturbations may be due to the chosen
temperature function not capturing temperature profile. The subsequent residuals may be interpreted as
a large wave by the Fourier transform. Much can be learned about the lower thermosphere by analysing
the DD campaigns. For wavelengths to be captured with confidence, they need to be shorter than the
altitude range examined. During the DD campaigns, the spacecraft sampled a more extended altitude
range, down to ∼120 km, hence allowing us to extract longer wavelengths potentially. We analysed the
DD orbits using the techniques described in Section 5.2. However, we did not detect longer wavelengths
from the DD data, instead, finding characteristics similar to those from higher periapsis orbits. As
shorter wavelengths are thought to be filtered out with altitude, we may have expected to see shorter
wavelengths at lower altitudes. This is found not to be the case, implying they may have already been
filtered out.
118
Gravity waves have vertical and horizontal components; however, these cannot be independently
determined due to the unknown and varying angle between the wave phase velocity and spacecraft
velocity. If we interpret the structures to be caused by vertical variations, then typical wavelengths are
found to be 10-30 km. By interpreting structures as horizontal waves, wavelengths are around 100-300
km. Henceforth, we focus on vertical wavelengths. A discussion and comparison with previous results
now follows. Creasey et al., 2006a found vertical wavelengths to be bimodally distributed with peaks
between at 8-10 km and 13-15 km, which are around a factor of two less than here. As wavelengths scale
with scale height, normalised wavelengths need to be considered. Similar scale heights are observed near
the surface (Lewis, 2003), and in the upper atmosphere, thus different wavelengths are being observed.
They used occultation data from MGS and retrieved temperature profiles below 30 km altitude. In their
study, waves appear periodic in the vertical dimension. We, therefore, interpret variations as vertical
waves. Previous studies have shown shorter wavelengths to be filtered by the mean flow, which may be
occurring here. Perturbations with shorter wavelengths induce local temperature gradients which may
exceed the environmental lapse rate and cause waves to break at lower altitudes. Larger wavelengths
pass through the flow and are seen in the thermosphere. Creasey et al., 2006b reveal dominant hori-
zontal wavelengths of 100-300 km along the orbital path. There has not been a comprehensive study of
vertical wavelengths in the thermosphere, so comparisons between techniques cannot be made. Another
possibility is secondary wave production. Near the surface, the topography is likely a prolific source of
gravity wave generation; however, waves may break, causing the generation of secondary waves. This
has been discussed in the context of Earth (Moudden and Forbes, 2011). In Gardner and Schunk,
2011 a large-scale gravity wave is studied in a global thermosphere-ionosphere model to determine the
3-D characteristics of the wave as it propagates upward through the Earth’s thermosphere. Waves are
initiated around 100 km. They found that as the gravity wave breaks, it deposits energy, and a second
wave is generated from the original gravity wave. The newly generated gravity wave has a wavelength
which exceeds the original. On Mars, if waves are breaking around 100 km, secondary wave generation
may explain the longer wavelengths observed in the upper atmosphere. Waves appear saturated on the
dayside so wave breaking may be occurring below 120 km. There is no definitive evidence for this,
but the results from Gardner and Schunk, 2011 may explain what is happening at high altitudes in
Mars’ atmosphere. On the nightside waves continue to grow with breaking occurring around 180-200
km; secondary waves be produced in this region. Equally, short wavelengths (<20 km) are theoretically
dissipated below 100-120 km with longer vertical wavelengths (50-100 km) propagating up to 150-200
km (Imamura et al., 2016). The most dominant wavelengths here are around 30-50 km. The waves
119
observed near the surface are significantly shorter than noted here (Ando et al., 2012; Creasey et al.,
2006a). Results presented provide evidence for either (or both) wave dissipation and secondary wave
production.
Figure 5.4.1: Panel (a) shows a histogram of gravity wave amplitudes extracted from temperature
structures. Grey shows the distribution of all gravity wave amplitudes extracted. Red shows the break-
down of the total into the four most dominant waves. Less dominant waves are shown by lighter red.
Grey is the sum of all reds. Panel (b) same as (a) but showing amplitudes extracted from density
structures. T 0 and ρ0 are temperature and density amplitudes, respectively.
Amplitudes observed here are of the order of those found in previous studies which have used MGS,
Odyssey (ODY) and MAVEN datasets (Creasey et al., 2006b; Creasey et al., 2006a; Fritts et al., 2006;
Yiǧit et al., 2015a). Fritts et al., 2006 used MGS and ODY data and found gravity wave amplitudes of
5-50%. From Figure 5.4.1a and b we observe that amplitudes rarely exceed 30%, but do reach 50%.
This discrepancy may be explained by the different regions of the atmosphere, which were sampled by
MAVEN and MGS. MGS typically sampled a lower region of the atmosphere with periapsis altitude
around 100-120 km. Terada et al., 2017 applied a smoothing technique to NGIMS Ar density data
to extract perturbations. They found root-mean-square amplitudes of 6-30% with an average value of
around 15% for vertical waves. These are in agreement with the results found in this study, especially
the most dominant waves. The very good agreement between our results and those found by Terada
120
5.5. ALTITUDINAL EFFECTS ON GRAVITY WAVES
Figure 5.4.2: Panel (a) shows a histogram of gravity wave wavelengths extracted from temperature
structures. Grey shows the distribution of all gravity wave wavelengths extracted. Blue shows the
breakdown of the total into the four most dominant waves. Less dominant waves are shown by lighter
red. Panel (b): as (a), but showing wavelengths extracted from density structures.
et al., 2017 give us confidence in the technique used. We find that structures in the temperature and
density fields are typically in anti-phase; however, in the cooler regions, pressure perturbations become
large enough to see significant effects. As the mean molecular mass is constant (44 amu), pressure
perturbations were calculated by the addition of density and temperature perturbations. Pressure per-
turbations are typically no more than 50% of the respective density and temperature perturbations.
Unlike on Titan, where pressure perturbations are taken to be negligible, pressure perturbations in Mars’
atmosphere are typically several percent (Mueller-Wodarg et al., 2006).
5.5 Altitudinal Effects on Gravity Waves
We investigate how gravity waves evolve vertically within the atmosphere. For each orbit, we split the
orbit into 30 km regions, each shifted by 10 km allowing Fourier transforms to be applied in different
regions as a method of understanding growth and/or decay. For example, we may study the 140-170
km, 150-180 km, 160-190 km regions of a leg. For each section, we identically extract perturbations,
121
as described in Section 5.2. We determine dominant amplitudes and wavelengths for each section. The
average solar zenith angles for each section are determined.
Figure 5.5.1: Gravity wave amplitude as a function of altitude. Amplitudes are separated in solar
zenith angle with each panel covering 30◦ SZA. Amplitudes are extracted in 30 km section as explained
in Section 5.5.
We consider six solar zenith angle regions; each region spans 30◦ in SZA. The amplitudes determined
above are initially binned into one of the six solar zenith angle regions. Amplitudes are then binned
in altitude and the mean calculated. Figures 5.5.1a-f shows the result. We observe amplitudes on the
dayside to remain fairly constant in altitude with typical values around 10%. As discussed further later in
this chapter, it is thought that processes which act to grow amplitudes are well balanced with dissipative
processes. On the nightside, we observe that amplitudes grow from around 10-15% at 120 km to over
20% at 180 km. For altitudes above 180 km on the nightside, we find amplitudes to begin to decrease
by several percent.
122
By considering the conservation of momentum in a conservative atmosphere, amplitudes are expec-

ted to grow exponentially with altitude as density decreases. Our results suggest that the growth process
is well balanced by dissipative processes in the thermosphere, such as an increased viscosity. As temper-
ature increases, decay processes are enhanced, balancing growth. On the nightside, we observe growth
in amplitudes up to around 180 km. From this result, it appears the effect of dissipative processes are
reduced, allowing growth processes to dominate there. The amplitudes do not increase exponentially;
therefore wave decay is still occurring. Beyond 180 km on the nightside, amplitudes begin to decrease
by 2-5%. This behaviour is consistent on the nightside and is potential evidence for wave breaking.
The induced temperatures gradients may exceed the adiabatic lapse rate, causing wave motion to be no
longer sustained. The breaking of gravity waves will deposit momentum into the atmosphere, poten-
tially causing changes in the flow. The effects of such breaking will form part of a future study. Using
the gravity wave scheme employed by Yiǧit et al., 2008; Yiǧit et al., 2015a investigated the amplitude
growth/decay in the Martian atmosphere. They found amplitudes grow rapidly from the surface up to
around 100 km and amplitudes are modelled to be on average about 50% at this altitude. Upwards
of 100 km, the average amplitude decreases rapidly, returning to values below 10% at 140 km. We
do not observe such decreases in amplitude at these altitudes. Gravity wave propagation is strongly
affected by background winds (Yiǧit et al., 2015a). This is a factor which has not been investigated in
this study but may account for some of the behaviour observed. Winds in the upper atmosphere are
substantially changed by the presence of gravity waves. This is caused by momentum deposition above
the mesopause by gravity waves. Both the easterly and westerly jets are impacted by gravity waves with
the easterlies are weakened (Medvedev et al., 2013). Yiğit and Medvedev, 2017 modelled the effect of
gravity waves on Earth’s thermosphere. Tidal amplitudes increase or decrease depending on latitude
due to interactions with gravity waves.
Fritts et al., 2006 used the assumption of saturated gravity waves to estimate net horizontal mo-
mentum flux, from which momentum deposition can be derived. This can be presumed to occur on the
dayside (Figure 5.5.1a-c). As the B-V frequency has not been explicitly calculated in this study, it is
taken to be 2×10−2 s−1 (from Medvedev and Yiğit, 2019). Momentum fluxes calculated in the ther-
mosphere are a few hundred m2 s−2 for small amplitudes (∼0.1), reaching 1000 m2 s−2 at amplitudes
close to 0.5. Fritts et al., 2006 find values of ∼2000 m2 s−2 . As similar amplitudes were used in Fritts
et al., 2006 (∼0.2) as used here, the likely candidate for the discrepancy is the ratio of horizontal to
123
5.6. GLOBAL CHARACTERISTICS OF GRAVITY WAVES
vertical wavelengths. In this study, vertical (horizontal) wavelengths are taken to be 10-30 km (200-300
km, taken from England et al., 2016), whereas Fritts et al., 2006 took horizontal wavelengths of ∼100
km. This may explain subtle differences, but overall values are in good agreement.
5.6 Global Characteristics of Gravity Waves
The previous section looked at the general properties of gravity waves; this section will look at how
location, local time and season affect gravity wave characteristics. As DD campaigns make up a small
portion of the total number of orbits and have similar amplitudes and wavelengths to non-DD orbits,
they have been included in the following analysis. Solar zenith angle has been used to describe the loc-
ation of MAVEN. The temperature of the atmosphere is expected to decrease with increasing distance
from the Sun. For example, mid-afternoon equatorial temperatures near 200 km vary from 190-390
K between aphelion and perihelion throughout the solar cycle (Bougher et al., 2015). Terada et al.,
2017 have highlighted the effect temperature has on the growth of gravity waves in the thermosphere.
Warmer regions of the atmosphere restrict amplitude growth as dissipative processes are enhanced with
warmer temperatures. The combination effects of solar zenith angle and M-S distance may, therefore,
affect wave properties, which we will examine in the following section.
The data were 2-D binned in solar zenith angle and M-S distance, and the median amplitudes and
wavelengths were calculated for each bin. Bin sizes of 11.25◦ ×0.025AU were used, giving a total of 192
bins. 80 out of 192 bins contain data. The coverage is good in M-S distance. Figures 5.6.1a and b show
polar plots of gravity wave amplitude and wavelength, respectively, as a function of solar zenith angle
and M-S distance. In each polar plot, the upper semicircle presents results from temperature structures
and lower semicircles from density structures. The azimuthal component is the solar zenith angle, and
the radial is M-S distance. Temperature is then decreasing both radially outwards and towards larger
zenith angles.
Temperature amplitudes are considered first (upper semicircle, Figure 5.6.1a). Below solar zenith
angles of about 45◦ coverage of M-S distance is limited; values range between 1.45 and 1.6 AU, so
perihelion and aphelion are not sampled. In this region, amplitudes are found to be less than 5% with
little variation in either solar zenith angle or M-S distance. The most sampled quadrant is between 45◦
124
and 90◦ with only coverage missing from the 1.45-1.475 AU bin. There is greater variation in this region
compared to the previously discussed region. Near periapsis, amplitudes are similar to low solar zenith
angle values (<5%). The effect of M-S distance is more evident in this quadrant. Moving towards
aphelion leads to amplitudes increasing by a factor of around 2 to about 10%. The next quadrant
(90◦ -135◦ ) shows a large increase in amplitudes with values typically greater than 15%. The effect of
M-S distance seems to be weakened in this region. Amplitudes are fairly constant. The final quadrant
(135◦ -180◦ ) is the least sampled and as such little can be concluded from it. M-S distance is only
sampled at distances less than 1.5 AU and amplitudes are greater than 15%. Considering the contours
of constant solar zenith angle and M-S distance reveals their relative importance in affecting gravity
wave growth. As shown in Chapter 4, temperature is well correlated with solar zenith angle. There-
fore, there is a relationship between gravity wave characteristics and temperature, which is discussed
later. There is a larger variation in amplitude over the solar zenith angle range than over M-S distance.
The most well-sampled distance is around perihelion of Mars, with only low solar zenith angle data
missing. Here, amplitudes range from their minimum values (<2% at 45◦ ) to their maximum (>15%
around 170◦ ). This range of values is much larger than over M-S distance range. This large and rapid
change in amplitude suggests that gravity waves exhibit highly variable behaviour. Amplitudes vary
greatly by several factors throughout a sol. A near-identical trend is found in the density structures
(lower semicircle, Figure 5.6.1a). Density amplitudes are typically larger than their temperature equi-
valents, as also shown in Table 5.2.1. This is most obviously seen around solar zenith angles around 45◦ .
Figure 5.6.1b shows 2-D binned vertical wavelengths as a function of solar zenith angle and M-S
distance. The same bins are used as discussed for amplitude analyses. Temperature wavelengths (upper
semicircle, Figure 5.6.1b) are considered first. Compared to presented amplitude results, there is not a
clear trend with either variable. There is a slight pre-terminator increase in wavelength towards aphelion.
Given that this trend is not seen elsewhere, it is likely due to factors other than temperature. Median
wavelengths are fairly constant with values between 20 km and 30 km. Possible reasons are discussed
later.
The effect of gravity waves on the atmospheric temperature has been explored in great detail using
models (e.g. Medvedev et al., 2011; Medvedev and Yiğit, 2012). Gravity waves affect the mean flow in
the upper atmosphere by closing jets and, in some cases, reversing the flow. Dependent on location and
gravity wave dynamics, gravity waves can warm or cool the atmosphere via dissipation. We have found
125
90 90
a b
135 45 135 45
T 1.4 1.5 1.6

T 1.4 1.5 1.6
180 0 180 0
135 45 135 45
90 90
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0 5 10 15 20 25 30

X0 (km)
Figure 5.6.1: Panel (a) shows a polar plot of the gravity wave amplitudes from temperature (top
semicircle) and density (bottom hemisphere) data as a function of solar zenith angle and M-S distance.
Panel (b) is same as (a) but showing gravity wave wavelengths. The four most dominant waves are
considered from each orbit.
that gravity wave amplitudes show clear trends with solar zenith angle and M-S distance. Processes
which dissipate waves are enhanced in the warmest regions; therefore amplitudes are smaller here. Dis-
sipative processes that decay gravity waves include eddy and molecular viscosity, thermal conduction,
radiative cooling, and ion drag (Yiǧit and Medvedev, 2010). A trend suggests that regions, where waves
have been damped, have been warmed. An explicit relationship between amplitude and temperature
has not been computed in this study. Terada et al., 2017 derived temperatures for each orbit and found
amplitudes increase during cooler orbits. They suggest the anti-correlation between amplitude and tem-
perature is primarily caused by breaking/saturation of gravity waves due to convective instability. They
derive a threshold of convective instability for the perturbation number density. In the short-wave limit,
assuming a temperature of around 250 K and vertical wavelength 25 km, the threshold for amplitude
is about 11% which increases to 18% for a temperature of 150 K. The threshold condition is propor-
tional to wavelength, so the larger amplitudes are likely associated with larger wavelengths. In the
long-wavelength limit, where amplitudes are typically larger, the threshold is about 86%. This value is
not exceeded in this study.
126
Given the results presented here, SZA/LST plays a more dynamic role in controlling amplitude
than M-S distance given the more rapid, larger temperature changes. Dissipative processes such as
molecular diffusion, whose coefficient increases with EUV flux, are enhanced with temperature (Yiǧit
and Medvedev, 2010). Therefore, cooler regions in the atmosphere can support most gravity waves,
whereas gravity waves will dissipate in the warmer regions. Using the LMD MCD, we find viscosity
to be maximum where the temperature is maximum. The increased viscosity may explain the damped
amplitudes in the warmer regions. Creasey et al., 2006b extracted gravity waves from MGS accelerometer
data. They looked at seasonal effects by investigating data at similar latitudes and local times. They
found larger amplitudes in northern autumn compared to northern spring. They suggest aerobraking
during spring/summer as waves are most damped, which is in agreement with findings here; aerobraking
should occur in the warmest regions. Density is also highest in these regions. Yiǧit et al., 2015a studied
NGIMS CO2 perturbations in the upper atmosphere. They were far more restricted in the number of
orbits available; however, they found amplitudes to be inversely proportional to scale height which agrees
with the results presented previously. Overall, from this study and others, temperature appears to be
moderately important in controlling gravity wave amplitudes in the Martian atmosphere shown by the
strong correlation with solar zenith angle. It should be noted that horizontal background winds are more
important in controlling the upward propagation of gravity waves. If the background horizontal wind
speed equals the gravity wave phase speed, the wave can no longer propagate (Fritts and Alexander,
2003; Lindzen, 1981). This may explain the variation in observed amplitudes.
This section has answered the posed open questions with wave amplitude and wavelength distribu-
tions shown. As discussed in this chapter, derived amplitudes are comparable to previous studies, so
not surprising. However, the wealth of data increases our confidence in determined amplitudes. This
is the first substantial study to derive vertical wavelengths, and as such adequately answers the open
question. Vertical wavelengths are a factor of 10 shorter than horizontal wavelengths, as theoretically
predicted. The open question of diurnal and seasonal variation of wave characteristics is surprisingly
clear, with obvious day/night asymmetries in amplitudes, which has not been shown before. As waves
are filtered as they propagate upwards by background flows, it is not too unexpected that there is not
a similar day/night trend.
127
5.7. EFFECT OF TOPOGRAPHY ON GRAVITY WAVES
5.7 Effect of Topography on Gravity Waves
One source of gravity waves is flow over topography, so we investigate the vertical evolution and search
for potential links with topographical features. We use the data produced in Section 5.5. Figures 5.7.1a
and b show gravity wave amplitudes in latitude and longitude for both the lower (120-150 km) and upper
(150-180 km) regions, respectively. We have superposed a topography map to compare with gravity
wave amplitudes. Figure 5.7.1a shows amplitudes extracted in the lower region. We see the coverage
is mainly between 15◦ S and 40◦ N with good coverage in longitude. Data are also seen below 40◦ S.
The Tharsis region is centred near the equator in the western hemisphere. This region contains Arsia
Mons, Pavonis Mons, and Ascraeus Mons located at 8.35◦ S 120.09◦ W, 1.48◦ N 112.96◦ W and 11.92◦ N
104.08◦ W. These volcanoes are visible on the superposed map. Olympus Mons is located at 18.65◦ N
133.8◦ W.
Figure 5.7.1b shows amplitudes in the upper region. The coverage is much better with all latitudes
being sampled. We observe similar results as in the lower region. In similar regions, amplitudes are
comparable between the two altitude regions suggesting that waves remain constant. There appears to
be a latitudinal dependence on amplitude seen in both altitude ranges. This dependence is clearer in
5.7.1b due to more coverage at these altitudes. Terada et al., 2017 determined wave amplitudes using
a different method, but found a similar dependence. They removed the temperature dependence to
produce corrected amplitudes. These corrected amplitudes still show a visible, but weaker dependence
on latitude. The increase in amplitudes in the Northern hemisphere is therefore potentially caused by
a sampling bias; the atmosphere was likely sampled during cooler periods in the Northern hemisphere.
Even with the removal of this bias, the dependence is still visible. The Southern hemisphere has more
topography than the Northern hemisphere. As the topography is one of the main drivers behind gravity
waves, the latitudinal dependence is contradictory to what is expected. The slight longitudinal depend-
ence in the Northern hemisphere agrees with our understanding of gravity wave generation. Amplitudes
over the mountainous Tharsis region appear enhanced, especially at higher altitudes. This is in agree-
ment with Terada et al., 2017, even after their corrections.
Data and modelling studies have tried to elucidate the connection between gravity wave activity and
topography (e.g. Creasey et al., 2006a: Miyoshi et al., 2011). Creasey et al., 2006a looked at waves in
the troposphere and found gravity wave activity to be higher in some areas of elevated topography such
128
5.7. EFFECT OF TOPOGRAPHY ON GRAVITY WAVES
Figure 5.7.1: (a) Averaged amplitudes determined from binning in latitude and longitude using amp-
litudes extracted between 120 and 150 km. (b) Same as (a), except amplitudes extracted in the 150-180
km region are used.
as the Tharsis region, however in other relatively similar areas of topography the activity is comparatively
lower. Seasonally averaged latitude-longitude plots of amplitude and wavelength were produced, but
no direct link with topography was found in either case. As latitude and longitude are well sampled by
129
5.8. COMPARING WAVE ACTIVITY OVER SUCCESSIVE ORBITS
MAVEN, the distribution of amplitudes in Figures 5.6.1a and b would be expected to be more random
if there was a link in this region. The strong trend with solar zenith angle and M-S distance shows the
possible importance of temperature in controlling gravity waves; once a gravity wave has been generated
it is the conditions within the atmosphere which control its evolution. Wave amplitudes are significantly
larger in the warmer regions of the atmosphere. Waves deposit their momentum directed mainly against
the local wind, therefore providing wave drag on the mean flow (Kuroda et al., 2015). The dissipation
of energy into flow may erase any potential evidence of its origin.
There is inherent bias while analysing potential topography dependencies on gravity wave properties.
Although density and temperature variations have been interpreted as vertical gravity waves, MAVEN
spans tens of degrees in latitude during each pass due to its horizontal component. For each orbit, we
have sliced the wave, as in Section 5.5, and associated each with the average longitude and latitude for
that section. By slicing up the wave, each bin will contain wave data from multiple waves. This allows
local time effects to be averaged out, and in theory, leaves the topography dependence. This should show
stronger relationships with topography than considering the entire leg and using periapsis latitude and
longitude. Bias due to spacecraft trajectory on gravity wave amplitudes is less than for wavelengths.
Close to periapsis MAVEN is travelling quasi-horizontally, so although we have interpreted waves as
vertical, the variations are due to horizontal variations. This will affect the measured wavelengths.
We have not accounted for this trajectory bias. This problem is lessened away from periapsis; the
higher altitude wavelengths are likely more accurate. We find no correlation between wavelength and
topography.
5.8 Comparing Wave Activity Over Successive Orbits
Hitherto, each extracted wave has been treated independently. In this section waves from consecutive
orbits are compared to each other. This process allows potential wave evolution to be studied within a
relatively short time period of MAVEN’s orbit (∼4.5 hr). This is of particular importance when invest-
igating the lifetime of waves - do they persist for several hours? Or are they too short to sample twice?
This may have implications for the formation of thermospheric CO2 clouds, for example.
This is undertaken in the following way. Two waves are considered; the first is denoted as Υ1 which
is the density wave of the ith orbit. The second is Υ2 taken from the i + 1th orbit. The likeness
130
Figure 5.8.1: Examples of waves from two consecutive orbits exhibiting similar characteristics. (a)
Waves extracted along inbound passes for orbit 4092 (red) and 4093 (blue). (b) Waves extracted along
inbound passes for orbit 6833 (red) and 6834 (blue)
of the waves is quantified by computing the correlation coefficient between Υ1 and Υ2 . And this is
repeated for all orbits. A normal distribution is observed with a median around zero. Thus, indicating
no consistent correlation between consecutive waves. There are some cases where consecutive waves
have large correlation coefficients. This suggests waves can remain fairly stationary between consecutive
orbits. Figures 5.8.1a and b show two examples where consecutive waves appear to be near-identical,
suggesting the wave is not propagating. This is possible to see in the global structure of the waves
where similar dominant wavelengths and amplitudes are seen in phase. Amplitudes in both cases are
∼20%, and vertical wavelengths are ∼30 km. Smaller-scale structures vary between orbits. Identical
waves, like those shown, account for only ∼0.5% of all cases.
131
Some basic conclusions based on the above can be made. There are no cases where three con-
secutive orbits exhibit near-identical wave characteristics; hence it can be inferred that such structures
can survive up to ∼9 hours. There are no occasions where successive inbound-inbound waves occurred
on the same orbit as consecutive outbound-outbound passes. Each subsequent orbit is sampled at a
different longitude than the previous; however, local times are very similar between consecutive orbits.
Tidal modes may be weakened, therefore enhancing the persistence between successive orbits.
A more likely scenario is that after 4.5 hours, waves have propagated upwards and this may be
captured on the next orbit. This is attempted in this section, whereby a comparison between two or-
bits is as follows. Υ1 and Υ2 are interpolated onto a fixed altitude grid with spacing 0.1 km which
is the approximate resolution of NGIMS data. With Υ2 kept stationary as the ’reference’ wave, Υ1 is
shifted upwards by 0.1 km as if it were propagating in an ideal manner. A new correlation coefficient
is calculated between the waves. An example is shown in Figure s5.8.2a-d. Figure 5.8.2a shows the
extracted inbound density perturbations from orbit 3811 (faded red line) and 3812 (faded blue line).
Similarities are observed between the two waves; amplitudes are ∼20%, and vertical wavelengths are
∼30 km. Figures 5.8.2b-d show the wave extracted from orbit 3811 (Υ1 ) shifted by 5 km, 10 km, and
15 km, respectively. Faded lines show the original waves. Bright lines show the shifted waves. The first,
non-shifted waves are shown by faded lines. After a 5 km shift, the two waves are strongly in phase. It
could be inferred that MAVEN has sampled the same wave on consecutive orbits, but shifted upwards.
After a further 10 km shift, the waves are in anti-phase with each other. This is repeated for all waves,
and the correlation coefficient is calculated for all shifts. By restricting required coefficients to values
over 0.9, an emergent trend is seen. Correlation coefficients are maximised after a 30-40 km shift. If an
orbital period of ∼4.5 hrs is used, then wave propagation occurs at a speed of ∼6.5-9 km/hr. This is
the vertical group velocity of the wave. .
The implications of consecutive waves being similar are now addressed. Temperature profiles de-
rived using NGIMS data exhibit temperatures below the CO2 condensation point (R Yelle 2019, personal
communication). This may be the first evidence of CO2 clouds in the thermosphere. Such clouds have
been identified in the mesosphere by SPICAM (Forget et al., 2009). Temperature is an important factor
in cloud formation; however, there are more subtle aspects that need to be considered. One is density;
there need to be sufficient CO2 particles for nucleation to take place. From Section 5.1 it is seen that
temperature and density are in strong anti-phase; by their very nature, the lowest temperature regions
132
Figure 5.8.2: (a) shows the original waves from orbit 3811 (red) and 3812 (blue). (b), (c) and (d)
show the wave from orbit 3811 shifted upwards by 5 km, 10 km and 15 km, respectively. The original
waves are shown by faded lines. Bright lines show shifted waves.
allow for the best chance of cloud formation due to the highest densities. Another consideration is
the lifetime of waves which the above work has attempted to address. More specifically, how long do
these pockets of low temperature last? Naturally, a longer lifetime leads to an increased chance in
cloud formation. Two aspects are considered - similarity between consecutive orbits’ waves and low
temperatures. In doing so, the first criterion that is to be met is that perturbations are similar over
consecutive orbits. From Figure 1.5.3 the CO2 saturation threshold temperature is shown. At 100 km
the required temperature for saturation is ∼100 K. A basic extrapolation up to altitudes considered here
results in a saturation temperature of ∼50-80 K. An assumption that a minimum temperature of 100
K is needed to achieve cloud formation is used. CO2 profiles have been constructed using the method
outlined in Chapter 3.
133
Several occasions have been identified which meet the above criteria that consecutive waves need a
correlation coefficient greater than 0.75, and absolute temperatures need to be less than 100 K. Figure
5.8.3a shows extracted inbound waves from orbit 6107 (red) and 6108 (blue). Periapsis locations are
5.6°S, 10.8 LST and Ls=90°, and 5.9°S, 10.7 LST and Ls=91°for orbit 6107 and 6108, respectively.
A large perturbation is seen around 190 km in both waves. Average amplitudes are ∼50-75%, and a
long wavelength of ∼50 km is observed. Figure 5.8.3b shows the individual temperature profiles for the
orbits mentioned above, coloured accordingly. On the first pass, temperatures at minimised to ∼70 K
at 198 km and ∼80 K at 196 km on the following pass. The consistency between consecutive wave
and consecutive temperature profiles is indicative of persistent behaviour lasting several hours. These
profiles provide evidence for potential cloud formation, having met two criteria.
Figure 5.8.3: (a) Extracted inbound density waves from orbit 6107 (red) and 6108 (blue). Periapsis
locations are 5.6°S, 10.8 LST and Ls=90°, and 5.9°S, 10.7 LST and Ls=91°for orbit 6107 and 6108,
respectively. (b) Inbound temperature profiles for orbit 6107 (red) and 6108 (blue).
134
5.9. WAVE CHARACTERISTIC COMPARISON WITH OTHER PLANETS
The results found in this section are conceivably the most surprising so far in this thesis, given
consecutive waves have not been studied before. The implications of studying waves in such a way are
profound. Firstly, vertical group velocities can be calculated more accurately allowing wave dynamics to
be observationally explored. Secondly, potential cloud formation can be studied further.
5.9 Wave Characteristic Comparison with Other Planets
As briefly mentioned earlier in this chapter, gravity waves are not exclusive to the Red Planet. This
short section aims to review studies at Venus, Earth, Jupiter, Saturn and Titan and draw comparisons
between them.
As one of our closest neighbours, Venus has been well sampled. Radio occultation profiles from
Magellan revealed perturbations around the cloud tops (∼65 km) with amplitudes and vertical wavelengths
of 4 K and 2.5 km, respectively. This amplitude is equivalent to 1-2% assuming a background temper-
ature of 200-250 K, as suggested in Hinson and Jenkins, 1995. Although interesting, 65 km marks the
approximate bottom of the mesosphere and thus studies within the upper atmosphere need to be con-
sidered for comparison (Pätzold et al., 2007). Garcia et al., 2009 used the Visible and Infrared Thermal
Imaging Spectrometer-Mapper (VIRTIS-M) on-board Venus Express to explore gravity waves within
the lower thermosphere (110-140 km). Amplitudes were found to be ∼0.5%. Horizontal wavelengths
possess an extensive range from 90 to 400 km. More recently, Venus Express (VEX) undertook a two-
week aerobraking campaign in Venus’ lower thermosphere (130–140 km). Densities were recovered from
accelerometer data, and analysis similar to that found in this chapter was performed. Mueller-Wodarg
et al., 2016 found amplitudes and horizontal wavelengths of ∼10% and ∼100-200 km, respectively.
Earth is the most well-studied planet for gravity waves, and as such, myriad articles have detailed
experimental results. A select few publications are discussed. Space shuttle re-entry accelerometer data
implied amplitudes in the mesosphere/thermosphere of typically less than 5% (Fritts, 1989, Fritts, 1989).
Due to the trajectory of these shuttles, only horizontal wavelengths were inferred; typical lengths are
10-1000 km in the 60-140 km region. Garcia et al., 2009 used accelerometer data from Gravity Field and
Steady-State Ocean Circulation Explorer (GOCE) to retrieve thermospheric densities and subsequent
relative density perturbations. During solar minimum conditions, maximum amplitudes of 6% around
∼270 km were found. However, more typical amplitudes are less than half this, which is in agreement
135
with Park et al., 2014 who used a similar technique using data from CHAMP located around 300-400
km. This study found amplitudes of <2%, suggesting attenuation with altitude. Horizontal wavelengths
are found to be ∼150-600 km.
Galileo is only one of two spacecraft to orbit Jupiter, arriving in 1995 and finally succumbing to
Jupiter’s harsh radiation environment in 2003 when it was purposefully flown in the Gas Giant. Young
et al., 1997 used directly-measured temperatures profiles from the Atmospheric Structure Instrument
(ASI) on board the Galileo probe to infer gravity waves in Jupiter’s thermosphere. Temperature perturb-
ations are ∼50-80 K superimposed on a 600-800 K background temperature, consequently amplitudes
are ∼5-10%. Vertical wavelengths are ∼90-150 km. By modelling these waves, Young et al., 1997
required amplitudes of ∼2-7% in the troposphere to explain amplitudes in the thermosphere. This
agrees with wave growth seen for Mars. A further study by Young et al., 2005 used ASI data in the
stratosphere and found temperature fluctuations of ∼5%.
Titan is Saturn’s largest moon and the second largest in the solar system behind Ganymede. It
is the only known moon with a dense atmosphere. Titan’s atmosphere has been sampled in-situ by
both flybys and landers. Hinson and Tyler, 1983 first inferred gravity wave signatures in Titan’s lower
atmosphere in occultation data from Voyager 1. Mueller-Wodarg et al., 2006 used in-situ data from the
Ion and Neutral Mass Spectrometer (INMS) on board Cassini from two Titan flybys (24 October 2004
and 16 April 2005). Closest approaches occurred at 1176 km and 1025 km with altitudes up to 1600
km. Strong perturbations in N2 and CH4 densities were found and interpreted as vertically propagating
waves. Spectral curves were fit to fluctuations with typical vertical wavelengths ranging from 170 to 360
km and density and pressure amplitudes reaching 4–12%. Temperature perturbations were 5-10 K. A
second opportunity to study waves in was provided by the descent of the Huygens probe through Titan’s
atmosphere. Temperature and density profiles were recovered from the Huygens Atmospheric Structure
Instrument (HASI) data. Above 500 km, temperature perturbations of 10–20 K about a mean of about
170 K were observed (Fulchignoni et al., 2005). These equate to amplitudes of ∼6-11% which are very
similar to values found by Mueller-Wodarg et al., 2006. Likewise, Lorenz et al., 2014 used Huygens
temperature descent data and retrieved constant perturbations of ∼2 K between 60 km and 140 km.
Wavelengths are 3-8 km. These are significantly different from those found by (Mueller-Wodarg et al.,
2006). One reason is likely the sampling range; the range is not large enough to deduce wavelengths
found by Mueller-Wodarg et al., 2006. Additionally, it is expected for shorter wavelengths to be damped
136
with altitudes, leaving the longer wavelengths to remain at higher altitudes from the original ensemble.
During its nominal mission, Cassini did not graze Saturn’s atmosphere as done during Titan’s flyby.
After 13 years, Cassini was running out of fuel and instruments exceeded their expected lifetime; thus,
a decision was taken to dispose of Cassini in Saturn’s atmosphere purposefully. Cassini continued to
gather data until it burnt up in the atmosphere. For the first time, Mueller-Wodarg et al., 2019 dis-
covered atmospheric waves in Saturn’s thermosphere using INMS. A spectral fit to perturbations reveal
the presence of waves with amplitudes reaching 5–10%. If interpreted as vertical (horizontal) waves,
wavelengths would range from 100–300 km (1550–3500 km). No other studies are available for com-
parison.
Finally, Hinson and Magalhães, 1993 discovered inertio-gravity waves in Neptune’s atmosphere with
the aid of Voyager 2 occultation data. Amplitudes are minimal, taking values of 0.1-1 K which equate
to a few percent in an atmosphere with temperatures typically less than 80 K.
This section is not an exhaustive review of planetary gravity waves, however, does provide an over-
view of typical wave characteristics. Data are severely sparse for some bodies; however, notwithstanding
this, conclusions can be made. Wavelengths across the solar system exhibit enormous variation com-
pared to altitude, from a few km to several hundred km. Planetary size appears indicative of the
expectant wavelengths. The vast vertical extent the Gas Giants’ atmosphere allow such large waves to
be generated and sustained. Terrestrial planets are much smaller and as such have vertical wavelengths
an order of magnitude lower. Mars’ are 10% that of Jupiter’s. Such large waves on Saturn would be
planetary-size on Mars. On the contrary, amplitudes are not comparable with body size. The most
obvious comparison is with the largest body, Jupiter, and Mars; Jupiter is 20 times larger (by radius)
than Mars. However, amplitudes are comparable. It is not unusual for Mars’ nightside amplitudes to
be double those found on Jupiter. Earth is twice as large (by radius) than Mars’ but typically possesses
amplitudes much less than the Red Planet. These simple conclusions reveal the relative importance
of wave generation mechanisms. Mars has significant topographical features compared to Earth, thus
may explain such large amplitudes. The implication of such large amplitudes includes the importance
of waves in atmosphere dynamics. From Fritts et al., 2006, a simple relationship is proffered with
amplitude being proportional to the Brunt-Väisälä frequency and inversely proportional to gravitational
acceleration and vertical wavenumber. Given Mars’ gravity is nearly a factor of three smaller than all
137
5.10. SUMMARY
the planets’ discussed above, this may explain the disparity between planets’ wave amplitudes assuming
similar Brunt-Väisälä frequencies.
Observations of perturbations are not limited to neutral density profiles. Mayyasi et al., 2019 showed
evidence of ion-neutral coupling within Mars’ thermosphere as indicated by correlated ion and neutral
profiles from NGIMS. These correlations imply that neutral perturbations drive oscillations in the plasma.
This phenomenon is not evident in all profiles, however. Interactions with magnetic fields and the solar
wind induce additional structures which are not visible in neutral profiles.
5.10 Summary
In this chapter, perturbations from profiles of densities and temperatures have been extracted and
interpreted as vertical gravity waves. The main findings are highlighted below.
• Perturbations have been extracted from density and temperature profiles and interpreted these as
vertically propagating gravity waves. The amplitudes and vertical wavelengths of structures in the
density and temperature fields have been examined. They have been found to average amplitudes
and vertical wavelengths to be around 10% and 25 km respectively, with slightly larger values in
density perturbations.
• It has been shown how solar zenith angle and Mars-Sun distance affect the thermospheric gravity
wave amplitude. A strong correlation between solar zenith angle and gravity wave amplitudes
has been found. With an increase (decrease) in solar zenith angle (temperature), we observe an
increase in amplitude. The enhancement of dissipative processes such as diffusion and viscosity
occurs with increasing temperature. This is one proposed explanation.
• The evolution of gravity waves in altitude has been examined by extracting amplitudes and
wavelengths in 30 km sections along each leg. At solar zenith angles below around 90◦ amplitudes
appear invariant with altitude. From the conservation of momentum, amplitudes are expected to
grow with altitude. Therefore, decay processes appear to balance growth. Amplitude growth has
been observed with increasing altitude at solar zenith angles beyond 90◦ . Amplitudes begin to
decrease above 180 km, which may present evidence of breaking.
• As gravity waves play an essential role in the Martian atmosphere, the predicted evolution of
gravity waves in different regions within models could be compared to the results presented in this
138
5.10. SUMMARY
study. This would potentially help to improve gravity wave schemes.
• Waves extracted from consecutive orbits have been compared in an attempt to understand wave
lifetimes. In ∼1% cases waves on successive orbits exhibited structures too similar to be a coin-
cidence. One possible explanation is that other wave modes (tides) are weakened. In a smaller
subset of cases, there is evidence of cold pockets (<100 K) of air surviving several hours. This
aids the potential for CO2 cloud formation.
139
Chapter 6
Effects of June 2018 Dust Storm on the

Martian Upper Atmosphere
6.1 Introduction
Global dust storms are rare, global events at Mars, with the last but one dust storm occurring in 2007.
Fortuitously, MAVEN was operational during the June 2018 dust storm. NGIMS continued to collect
data throughout the dust storm, creating a new dataset to study. This chapter exploits data from this
grand event, including studying individual species’ responses to the dust storm. Previous studies have
shown evidence of upper atmospheric expansion due to heating in the lower atmosphere, such as during
dust storms. Results are compared to those found from other instruments on board MAVEN, such
as the Imaging Ultraviolet Spectrometer (IUVS), and to previous studies, such as regional dust events
sampled during aerobraking periods (Bougher et al., 1997). The rate at which dust storms effects decay
is calculated and compared to previous studies. Given the rarity of such events, modelling efforts of
storms are discussed and compared. It is crucial to understand dust storms in their entirety. Pertinently,
Opportunity, unfortunately, perished as a consequence of the June 2018 dust storm preventing the rover
garnering enough power via its solar panels owing to increased dust opacity. By understanding how
long dust storms preside on Mars, it will be possible to test spacecraft to such extremes without the
necessary power. Lastly, continuing the study of gravity waves, waves extracted during the dust storm
are compared to those before the storm onset. Little is known regarding the potential alteration of
gravity waves during a dust storm. Work speculated on Earth-based waves is utilised to understand any
effects on waves at Mars. The open questions this chapter hopes to answer are how are gravity waves
140
6.2. DUST STORM GROWTH AND ATMOSPHERIC EXPANSION
affected by dust storms? Is there a significant change in the wave spectrum during a storm? How would
this affect the thermal and dynamical structure of the thermosphere?
6.2 Dust Storm Growth and Atmospheric Expansion
On 30 May 2018, NASA’s Mars Reconnaissance Orbiter (MRO) detected a dust storm which progressed
towards NASA’s Opportunity rover located at Perseverance Valley on the plains of Meridiani Planum
(centred at 0.2°N 357.5°E). Opportunity was put in safe mode to conserve power as the dust storm
blocked out the sunlight necessary to power its solar arrays and keep instruments warm. The dust storm
had ballooned with such intensity that by 13 June 2018 it had reached NASA’s Curiosity rover at Gale
Crater (5.4°S 137.8°E). Dust storms can be identified by rovers and landers on the ground with optical
instruments by measuring the opacity level, or τ , of the atmosphere. Outside of the dust storm season,
τ is typically below one. As the dust storm season approaches, τ increases as dust is lofted up into
the atmosphere, as discussed in Chapter 1, with values rarely exceeding two. During the 2007 global
dust storm, τ peaked around five, and during the 2018 dust storm, τ exceeded ten. By measuring the
opacity, conditions within the atmosphere can quickly be assessed by science teams and appropriate
action taken. During the global dust storm MAVEN continued to collect data. Post-event analysis of
density data is possible, and evidence for the dust storm in the upper atmosphere is presented. As a
dust storm develops, the atmosphere expands due to heating. Dust is lofted into the atmosphere up
to several tens of kilometres and efficiently absorbs solar radiation. Dust storms present themselves as
sudden and long-lived increases in densities compared to the background atmosphere.
Regional dust storms are more common than global dust storms and thus have a higher probability
of being observed with an orbiting spacecraft. MGS was fortunate to gather in situ density data during
its aerobraking phase throughout the Noachis dust storm (Keating et al., 1998). MGS Thermal Emis-
sion Spectrometer (TES) detected the regional dust storm on 25 Nov 1997 in the southern hemisphere
around the Noachis Terra region (40°S 20°E). Density increases were observed in the lower (∼60 km)
and upper atmosphere (∼126 km); these coincided with warming and hydrostatic expansion. Relative
pressure increases were a factor of 2 (3) in the lower (upper) regions. Therefore, the lower atmosphere
accounts for 2/3 of the atmospheric inflation. Additional heating above 60 km is required for further
expansion (Bougher et al., 1999). Bougher et al., 1999 used and expanded on the study by Keating
et al., 1998 by attempting to reproduce observations using two models (MGCM and MTGCM) out-
141
lined in Section 1.9 (Bougher et al., 1997). Seasonal-orbital and latitudinal variations were modelled
Bougher et al., 1997 and compared to MGS accelerometer data. The general behaviour of the storm was
captured relatively successfully. The underlying and unpredictable large variability caused by pervasive
wave activity is not reproduced within the models. As described earlier, dust is injected into the lower
atmosphere, and the system is allowed to evolve. At 130 km MGS densities increased by 200%, with the
model predicting a modest increase of ∼70%. Similarly, the F1-ionospheric peak is underestimated; the
model increase is 3 km compared to an observational rise of 8 km. This modest expansion corresponds
to a lessened temperature increase. TES observed warming of 10-15 K, whereas modelled zonally aver-
aged temperatures are 7-10 K. Bougher et al., 1997 cite "(a) missing aerosol heating of the atmosphere
below 30 km during the Noachis storm, (b) missing dynamical heating at Northern mid-latitudes above
60 km during the storm, and (c) missing wave effects (including longitudinally fixed planetary waves
and gravity waves)" as potential shortcomings of their model.
This section aims to examine how individual species are impacted by the global dust storm. Inbound
and outbound CO2 , N2 and Ar densities have been interpolated at 180 km, with similar behaviour ob-
served at other thermospheric altitudes. Background densities have been binned using a width of 0.5° Ls.
This creates a smooth time series by removing orbit-to-orbit variability. Figures 6.2.1a and b show these
interpolated data from Ls=165° to Ls=210°. A large period is shown to highlight variation caused by
sampling location. For this study, the first detection of the storm is taken at Ls=189° (vertical dashed
line), in agreement with Elrod et al., 2019. There are several reasons for this choice. Firstly, there is
a gap in data at this solar longitude; MAVEN was put into safe-mode by the science team as the dust
storm spread. Secondly, although Opportunity first detected the dust storm on 30 May 2018 (Ls∼185° ),
it took over two weeks for the storm to develop and circumnavigate Mars fully. Therefore, the chosen
solar longitude is a compromise between these two significant dates. The dotted line signifies an appar-
ent peak in the storm around Ls=193°. The following effect is discussed in more detail in later sections.
Densities increase to a maximum just before Ls=180°. This is caused by sampling move towards the
dayside, and a reduction in densities is seen beyond this caused by MAVEN proceeding towards the
nightside. This decreasing trend is interrupted by the dust storm; a sudden increase is observed, which
is not expected to arise purely due to benign seasonal or geographic changes in sampling locations. This
is indicative that the effects of a dust storm are being observed in the upper atmosphere. At the begin-
ning of the dust storm, NGIMS undertook two ten orbit campaigns to determine wind measurements
by swinging the boresight direction by ±8°(Benna et al., 2019). Due to this, regular neutral densities
142
cannot be produced. Hence, there are missing data. Although all species presented here have increases
associated with them, an unexpected decrease in O densities was by Elrod et al., 2019. One explanation
proffered is an oxygen sink in the form of water ice clouds.
Figure 6.2.1: Binned CO2 (blue), N2 (blue) and Ar (red) densities at 180 km as a function of solar
longitude for (a) inbound and (b) outbound passes. The onset is shown by the grey dashed lines around
Ls=189°. The initial growth peak is shown at Ls=193° by the grey dotted lines
Two ranges of orbits are considered to isolate the effects of the dust storm. The first is a pre-onset
range from 1-3 June 2018 (Ls=185-189°). This period is used when pre-storm conditions are required
and contains around 20 orbits. The second is a post-onset range from 12-15 June 2018 (Ls=192-194°).
This range straddles Ls=193°which has been taken as the peak of the storm in the upper atmosphere.
Now, the effect of the dust storm is studied for all altitudes by interpolating background densities onto
a fixed altitude grid with spacing 2 km, from which the mean density and standard deviation at each
altitude level are calculated for inbound and outbound legs. All orbits within the pre-onset and post-
onset range are utilised. Variability in profiles during the dust storm is studied later. These average
profiles are shown in Figures 6.2.2a (inbound) and 6.2.3a (outbound). Ar, CO2 and N2 are coloured red,
green and blue respectively, with pre-onset densities and standard deviations shown by solid lines and
shaded boxes, respectively and post-onset densities and standard deviations shown by dashed lines and
143
hatched boxes, respectively. Colours are kept consistent. In line with Figures 6.2.1a and b, densities are
generally larger during the dust storm for a given altitude, typically taking values nearly one standard
deviation away from pre-onset values, suggesting a significant increase. The largest increase is seen for
CO2 . The ratio is taken between post-onset and pre-onset densities at each altitude level to quantify
the increase in densities for each species. Figures 6.2.2b (inbound) and 6.2.3b (outbound) show these
ratios as function of altitude. N2 varies the least with post-onset density increases of up to 20%. Ar
varies by ∼30-50%, followed by CO2 , which varies by ∼50-70%.
Figure 6.2.2: Averaged inbound CO2 , Ar and N2 density profiles for pre-storm (solid) and storm
(dashed) conditions. Shaded (hatched) boxes show one standard deviation for pre-storm (storm) con-
ditions. Right: Ratio of storm to pre-storm densities as a function of height for CO2 , Ar and N2
In both inbound and outbound profiles, there is an apparent mass dependence on the density in-
crease. It was initially believed to be due to the expansion of the entire atmosphere, such that there was
scale height dependence as heavier constituents increase more rapidly. This was assessed by dividing
the post-onset to pre-storm density ratios by the respective mass, referred to as the relative ratio, and
revealed an interesting result. For both inbound and outbound profiles, Ar and CO2 relative ratios are
near-identical taking values in the range 0.034-0.036 amu−1 . N2 exhibits significantly different beha-
viour; for similar relative ratios to be observed in N2 , a slight decrease in density is required during
144
Figure 6.2.3: Averaged outbound CO2 , Ar and N2 density profiles for pre-storm (solid) and storm
(dashed) conditions. Shaded (hatched) boxes show one standard deviation for pre-storm (storm) con-
ditions. Right: Ratio of storm to pre-storm densities as a function of height for CO2 , Ar and N2
the global dust storm. The original claim that the atmosphere expands relative to each species scale
height has been refuted. To further understand the potential mass dependence, Figure 6.2.4 shows the
post-onset to pre-onset density ratios plotted as a function of molecular mass at 170 km. This dataset
is extended by a result found by Elrod et al., 2019 who found O decreased by ∼20% during the global
dust storm. A clear trend is observed with the lightest species behaving differently during the storm for
a given altitude. Densities of species with masses less than ∼23 amu are predicted, and found for O, to
be lower during the storm period. Elrod et al., 2019 suggests the idea that an oxygen sink is induced
or enhanced during the storm, such as water ice clouds. With the addition of data in this study, it is
more likely that dynamics drive the distribution, rather than chemistry. The fairly negligible increase in
N2 is unanticipated, even with the consideration of scaling by mass. This is an interesting, outstanding
puzzle to solve with suggested work in the final chapter.
In Chapter 1, it was discussed that the atmosphere expands during a dust event owing to heating
in the lower atmosphere. Implications of this include physical processes that manifest at certain alti-
tudes before the storm occurs at higher altitudes during storms, such as the electron density peak, as
145
Figure 6.2.4: Post-onset to pre-onset density ratios taken at 170 km from Figures 6.2.2b and 6.2.3b
as a function of molecular mass (blue circles). O has been added from Elrod et al., 2019 (red plus).
Black dashed line shows line of best fit.
shown by Wang and Nielsen, 2003. The vertical extent of atmospheric expansion can be quantified by
considering at which altitudes the same densities are observed pre- and post-onset. For example, Ar
densities of ∼1013 m−3 are located at ∼170 km before storm onset. After the onset, the expansion of
the atmosphere causes this density to be sampled at ∼175 km. This calculation has been performed for
all sampled species and their densities. Figures 6.2.5a and b show the post-onset altitude as a function
of pre-onset altitude for CO2 (green), Ar (red) and N2 (blue). Shown for reference is line showing post-
onset altitudes equal to pre-onset altitudes (y=x, black dashed line). The vertical distance between
each species’ line and y=x line gives the altitude shift. For CO2 the altitude shift is ∼6-8 km. For Ar,
the altitude shift is ∼5-7 km. N2 has the smallest shift of no more than 4 km. The mass dependence
described above has been applied to altitude shifts. Relative to each species’ scale height, the altitude
shifts are found not to be equal.
Chaufray et al., 2019 performed a similar investigation using MAVEN/IUVS data obtained during
the 2018 global dust storm event. IUVS measures Lyman-α emissions from interplanetary and Martian
hydrogen at the limb and through the extended corona of Mars. From these brightness data, the CO2
146
Figure 6.2.5: Post-onset altitude as a function of pre-onset altitude for (a) inbound and (b) outbound
legs using results from Figures 6.2.2a and 6.2.3a. CO2 , Ar and N2 altitude shifts are shown by green,
red and blue lines, respectively. The no-shift scenario is shown by the black dashed line.
absorption optical thickness can be derived. Before the dust storm, the altitude at which the optical
thickness was equal to one was relatively constant with a value of 114.5±2 km, where the uncertainty is
one standard deviation. At the onset of the storm, this reference altitude increases up to 119±2 km. As
expected, this is in very good agreement with the CO2 altitude shift found in this thesis. Chaufray et al.,
2019 associate this with heating in the lower atmosphere. They estimate the global heating required
using Equation 6.1. Constant heating is assumed within a deep region, ∆Z < 50km. This is derived in
Appendix B.
hHi2 1
β= = hHi1
(6.1)
hHi1 1− ln n2
∆Z n1
where hHi1 is the average scale height prior to the onset within the heated region . Likewise, hHi2
is the average scale height after the storm onset. n2 /n1 is the fractional increase in observed CO2
at a given altitude found to 1.9±0.2. Assuming a quiet time scale height of 8-12 km, the estimated
temperature increase required in the lower atmosphere is ∼25-50 K. Their density increase of 1.9±0.2
is slightly larger than found in this thesis. One likely explanation is the difference in methods; most
notably Chaufray et al., 2019 implement an isothermal temperature to derive densities. Although this
147
is not entirely accurate, it is the best estimate.
Using NGIMS data, Liu et al., 2018 present different species’ variability during dust loading events,
similar to Figures 6.2.2a 6.2.3a. As found here, there is a mass dependence in the variability; CO2
varies by less than 50% at 170 km, increasing up to 200% above 210 km. Although our variability
is less than this, the difference could be attributed to the method used to determine quiet and storm
periods. Liu et al., 2018 found thermospheric expansion, as found here, of ∼10 km and attribute the
mass-dependent variability to differing respective scale height. They do not investigate further possible
effects. As mentioned in in Chapter 1, atmospheric expansion is not an isolated effect of the neutral
atmosphere. A recent study by Girazian et al., 2019 is apropos to atmospheric expansion during dust
storm events. Data from the Mars Express (MEX) Mars Advanced Radar for Subsurface and Iono-
sphere Sounding (MARSIS) experiment is used to evaluate altitude change in the ionospheric peak
during the June 2018 global dust storm. At solar zenith angles similar to those sampled by MAVEN
during this period (50-70°), the ionospheric peak altitude rises by 10-15 km based on a nominal peak
of 137 km. The variability in the peak altitude during storm periods more than doubled; this sug-
gests that dynamical processes that couple the lower and upper atmosphere are enhanced during dust
storms. The enhanced inter-hemispheric circulation that permits polar warming presented in Chapter 4
is evidenced by Girazian et al., 2019. MARSIS and MAVEN (Radio Occultation Science Experiment -
ROSE) sampled the dust season concurrently at different latitudes (30°S and 50°N, respectively). At
the storm onset, both instruments observe a rise in ionospheric peak location simultaneously implying
dust storms can affect a large area over a short time period. The dust storm was local to the southern
hemisphere; however, the effects lasted longer in the northern hemisphere as indicated by a more rapid
descent in ionospheric peak measurements by MARSIS. Qin et al., 2019 used MGS radio occultation
profiles during MY27. The mean shift in electron peak from pre-onset to post-onset is from 135.1 km to
140.5 km. Given CO2 dominates in this region, it is unsurprising this is commensurate with neutral shifts.
The lower atmosphere temperature has been estimated, and this is expected to be significantly
warmer than in the thermosphere, as is investigated now. By inspection, the averaged density profiles
are approximately linear in log-space. Therefore isothermal temperatures can be calculated from their
respective scale heights. The pre- and post-onset temperatures for inbound passes are 177 K and 191 K
for CO2 , 187 K and 199 K for Ar, and 176 K and 180 K for N2 , respectively. For outbound passes pre-
and post-onset temperatures are 219 K and 225 K for CO2 , 219 and 223 K for Ar and 217 K and 205 K
148
6.3. DUST STORM DECAY
for N2 . By its very nature, the assumption of an isothermal atmosphere means altitudinal temperature
variations cannot be observed. However, pre-onset temperatures are in 5/6 of the cases lower than
post-onset temperatures. CO2 exhibits the most substantial temperature increase of 14 K during the
inbound passes. During the inbound passes N2 has a temperature increase of 4 K, but a decrease of
12 K during the outbound passes. Again, mass dependence on temperature seems apparent here; thus,
it could be inferred that O temperatures would decrease during the storm. Nevertheless, the increase
in temperatures agrees with previous studies of heating within the atmosphere. Thus, the conclusion
that upper atmosphere temperatures increases are not expected to be as substantial as for the lower
atmosphere due to lack of increased direct heating can be made. We can infer that storm-induced
heating is most intense below the homopause (∼25-50 K) with slight warming in the upper region. This
is confirmed further by Jain et al., 2020 who explored MAVEN/IUVS limb scans. IUVS discovered a
significant increase in peak airglow intensities, symptomatic of an increase in neutral densities. Jain
et al., 2020 end by noting an increase in thermospheric temperature of ∼20 K. The prediction of in-
tense lower atmosphere heating is verified by Guerlet et al., 2019 who used the Atmospheric Chemistry
Suite (ACS) on board TGO and found a considerable increase of 50 K in the 5-45 km range, which is
the prime region of dust heating. On a side note, the above study infers strengthened and dominant
diurnal modes during storm periods, in contrast to dominant semi-diurnal modes during quieter periods.
Similarly, Kass et al., 2019 exploited temperature and density data from MCS (Mars Climate Sounder)
to explore warming/cooling around 25 km during global dust storm events. During quiet years, average
temperatures are ∼180-200 K. During dust storm periods (June 2001 and June 2018), the zonal average
temperatures increase to 220-230 K. The atmospheric warming of 30-50 K is in excellent agreement with
that postulated in our study. On a different note, evidence of the dust storm ’shading’ the surface can
be implied by considering surface temperatures. Using MCS data, Streeter et al., 2019 found an average
global surface warming of 0.9 K, with local cooling on the dayside of up to 16 K. More intense heating
above the surface evinces shading by the dust storm. The 2001 dust storm saw smaller warming/cooling
due to its storm, with only increases/decreases in the surface temperature of 2-3 K (Cantor, 2007).
6.3 Dust Storm Decay
In Chapter 1, the main phases comprising a dust storm were discussed: onset, growth, and decay. The
previous section investigated the onset, and now the decay phase is studied. Withers and Pratt, 2013
149
presents an exposition on dust storms and their decay rates. They focus on the magnitude and timescale
of changes in upper atmosphere densities throughout the regional Noachis dust storm. They focused
on aerobraking accelerometer data, UV stellar occultation data and radio occultation data from MGS.
MGS accelerometer data are presented for 130 km, 140 km, 150 km and 160 km for both inbound
and outbound legs. Data span from Ls ∼185-300°with the storm onset at ∼224°. Their technique is
described now and explained in the context of NGIMS data. Figures 6.3.1a-d and 6.3.2a-d show inter-
polated background total densities at 170 km, 180 km, 190 km, and 200 km for inbound and outbound
legs, respectively. Vertical dashed and dotted lines signify storm onset and storm peak, respectively, as
in Figures 6.2.1a and b. Pre-onset densities are shown by stars in the range Ls=185-189°. During the
growth stage of the storm, taken in this study to be Ls=189-194°, densities are shown by diamonds.
Lastly, post-onset densities are shown by circles. Two exponential functions are fit such that time con-
stants can be derived to get a measure of the decay timescale. The first timescale accounts for changes
in density due to changing sampling location; densities generally decrease in solar longitude here due to
moving towards the nightside. To ensure that only the seasonal trend is captured only data taken from
Ls=185-189° and Ls=220-240° are used; this aims to omit dust storm influences. The fit is shown by
solid orange lines in Figures 6.3.1a-d and 6.3.2a-d and its decay timescale is denoted τs . The second
timescale is a combination of seasonal changes and abatement of the dust storm and is calculated using
all densities post-onset (Ls≥193°) and denoted τf it . This fit is shown by the dashed orange line in
Figures 6.3.1a-d and 6.3.2a-d. Given exponential functions have been used, the decay timescale of the
dust storm can be isolated and derived using 1/τf it = 1/τs + 1/τcor . This formulation is taken from
Withers and Pratt, 2013. The ratio (r) between the average peak storm density and pre-onset density
is computed for each altitude level to quantify the severity of the global storm. The average peak
storm density is calculated from all orbits in the range Ls=193-195°; this is the same range as used for
post-onset densities in Figures 6.2.2a and 6.2.3b. The pre-onset density is taken from the seasonal fit
(solid orange dashed line) at Ls=189°. Results for the Noachis regional dust storm are shown in Table
6.3.1. Timescales found from this current study are shown in Table 6.3.2.
Figure 6.3.3a shows τcor as a function of r. Results from this thesis are shown by circles and results
found by Withers and Pratt, 2013 are shown crosses. The apparent dichotomy indicated by Tables 6.3.1
and 6.3.2 becomes less so when visualised in Figure 6.3.3a. Although r and τcor values are notably
different in both studies, a similar trend is observed across both dust storm events. In both cases, τcor
decreases with increasing r; thus, more severe enhancements decay more rapidly. Severity here means
150
Figure 6.3.1: (a), (b), (c) and (d) show interpolated inbound total densities at 170, 180, 190 and 200
km, respectively. Grey dashed lines show the first detection of the dust storm in the upper atmosphere
(Ls∼189°). Densities prior to this onset are shown by stars. Densities during dust storm growth are
shown by crosses. A peak in the storm is taken at Ls∼193° as shown by grey dotted lines. Decay time
densities are shown by circles. Orange dashed lines show an exponential fit to post-onset densities. Solid
orange lines show an exponential fit to the pre-storm and last 10° Ls densities and act as a seasonal fit.
the ratio between pre- and storm time densities. Further, and more intriguing, perhaps, is the likeness of
the trends. Similar gradients are observed for both dust storms; these are -55°and -70°for results from
this thesis and Withers and Pratt, 2013, respectively. Thus, it can be inferred that although there is not
a single relationship that describes all dust storms, dust storm abatement phases are similar across dust
storms. The most likely behaviour is that all decay phases have similar gradients to those presented
here. Further behaviour that could be implied from the data, and verified with further storm data, is
that decay phases are translated diagonally along a line from the lower left to the upper right. For
example, decay data from another storm may lay along a line between the coordinates (1.6, 100°) and
151
Figure 6.3.2: (a), (b), (c) and (d) show interpolated outbound total densities at 170, 180, 190 and 200
km, respectively. Grey dashed lines show the first detection of the dust storm in the upper atmosphere
(Ls∼189°). Densities prior to this onset are shown by stars. Densities during dust storm growth are
shown by crosses. A peak in the storm is taken at Ls∼193° as shown by grey dotted lines. Decay time
densities are shown by circles. Orange dashed lines show an exponential fit to post-onset densities. Solid
orange lines show an exponential fit to the pre-storm and last 10° Ls densities and act as a seasonal fit.
(2.3, 50°). Future work may elucidate relationships between dust storms. Withers and Pratt, 2013 note
that taking the product of τcor and r produces a narrow range of values (205-222° Ls). This is not true
here with the product producing values of 80-100°. Instead, τcor is now divided by r for both studies.
Figure 6.3.3b shows τcor /r as a function r. This exercise provides a compelling commonality between
the two data sets with τcor /r for both studies taking values of 20-60°. Further studies will elucidate
whether this is true for all dust storms. Decay timescales have been derived for the lower thermosphere
during the 2018 global dust storm. Guzewich et al., 2019 used atmospheric optical depth measurements
from Mastcam on board the Curiosity rover and found decay constant of 43±2 sols, equivalent to 25° Ls.
152
Altitude (km) Direction r(-) τf it (◦ ) τs (◦ ) τcor (◦ )
130 Inbound 2.24 42.93±2.78 80 93±3

130 Outbound 2.60 53.31±4.53 160 80±5
140 Inbound 2.45 39.64±1.72 90 71±2
140 Outbound 2.36 59.65±3.95 187 87±4
150 Inbound 2.44 41.65±1.98 76 93±2
150 Outbound 2.02 73.12±4.98 234 106±5
160 Inbound 2.08 50.94±3.24 100 104±3
160 Outbound 1.88 82.98±6.50 279 118±7
Table 6.3.1: Dust storm decay timescales determined for Noachis dust storm. r is ratio of peak storm
density to pre-storm density. τf it is the decay timescale obtained by a direct exponential fit to the data,
τs is the background decay timescale associated with changes in season and latitude, and τcor is the
corrected decay timescale. All uncertainties are 1σ. Adapted from Withers and Pratt, 2013
Although this is comparable with fitted decay constants in the upper atmosphere, it is marginally lower
than the corrected constants. There is inherent sampling bias; MAVEN sampled on the dayside around
50°S whereas Curiosity is located in Gale Crater (∼5°S), and it is not unexpected for different regions
to recover at different rates. Further, Wolkenberg et al., 2020 compared dust storm decay rates using
Thermal Emission Spectrometer (TES) on board MGS and the Planetary Fourier Spectrometer (PFS)
on board Mars Express to compare global dust storm events in MY25 (2001), MY28 (2007) and MY34
(2018). The decay constants found for the respective years are 58°, 16° and 7°, and as found in both
Withers and Pratt, 2013 and this study, the more severe the storm, the faster its recovery.
Other studies of global dust storm events have provided decay timescales for comparison. Dust
opacity decay timescales found for Mariner 9 during the 1971 global dust storm were around 20° Ls,
whereas lower atmospheric decay timescales found were longer taking values of around 36° Ls (Withers,
2006). Lemmon et al., 2015 found a constant of 43 sols using data from the Mars Exploration Rovers
during the 2007 global dust storm. This constant is similar to the 51 sols calculated by Pollack et al.,
1979 during the 1977 global dust storm. Cantor, 2007 made use data from of the Mars Observer
153
Altitude (km) Direction r(-) τf it (◦ ) τs (◦ ) τcor (◦ )
170 Inbound 1.52 20.88±5.76 29.47 71.61

170 Outbound 1.85 13.23±3.60 18.57 46.07
180 Inbound 1.59 17.29±3.18 23.58 64.84
180 Outbound 1.93 11.54±3.17 16.10 40.74
190 Inbound 1.66 15.37±4.29 20.65 60.09
190 Outbound 1.95 11.11±2.64 15.57 38.84
200 Inbound 1.48 15.38±2.78 19.50 72.89
200 Outbound 1.99 11.48±2.67 16.50 37.75
Table 6.3.2: Dust storm decay timescales determined for June 2018 dust storm. r is ratio of peak
storm density to pre-storm density. τf it is the decay timescale obtained by a direct exponential fit to
the data, τs is the background decay timescale associated with changes in season and latitude, and τcor
is the corrected decay timescale. All uncertainties are 1σ.
Camera (MOC) on board MGS to analyse the 2001 global dust storm event. Using the MOC, daily
global maps were created from which over 5000 dust storms were identified, yet only one was a global
dust storm event. Cantor, 2007 used atmospheric opacity (τ ) to quantify storm conditions and assume
an exponential decay during the abatement phase. Without accounting for seasonal variations in τ ,
decay constants ranged from 30-117 sols. From Table 1.4.2, the equivalent solar longitude timescale is
∼20°-70° Ls. Consequently, the decay phase lasted 60° Ls. These values are remarkably similar to those
in this current study. The results imply that regional storms, such as the Noachis storm, have longer
decay times compared to global dust storms. One possible explanation is that with a global dust storm,
the complete shading of the surface removes the energy source, thus decaying more rapidly. The values
attained in these other studies do fall into the behaviour described in this study. At the time of writing,
this is the only study to quantify the 2018 dust storm decay timescale in this manner; it is unanimously
agreed that the dust storm extended and ended around September 2018. Hence, no direct comparisons
can be made. However, the dust storm’s decline has been hinted at being precipitous (Zurek et al.,
2018). It is unclear as to whether the decay is steeper than the increase or sharper in comparison to
previous dust storms.
154
6.4. DUST STORM EFFECTS ON GRAVITY WAVES
Figure 6.3.3: (a) Storm decay rate (τcor ) as a function of density enhancement (r) from this study
(circles) and Withers and Pratt, 2013 (crosses). (b) Storm decay rate normalised by density enhancement
as a function of density enhancements
6.4 Dust Storm Effects on Gravity Waves
This current study is one of few that studies the effects of dust storms on gravity waves. Figures 6.4.1a
and b show Ar gravity waves (blue lines) captured pre- and post-onset of the 2018 global dust storm
using the same of orbits in Section 6.2. Waves are extracted by the method outlined in Chapter 5.
The local maximum and minimum amplitudes are computed for each wave and then binned in 5 km
intervals; this creates an envelope showing the average wave amplitude throughout each orbital range.
This envelope is shown by the red dashed lines in Figures 6.4.1a and b. In Figure 6.4.1a shows waves
taken pre-onset with average amplitudes of ∼10-20%, as expected from work in Chapter 5. There is no
convincing growth in altitude. Figure 6.4.1b shows waves extracted in the post-onset range. Significant
growth is apparent at all altitudes with an average amplitude of ∼30-40%; this amounts to a factor of
two increase compared to pre-onset waves. These are surprisingly large given the solar zenith angle. To
ensure this behaviour is storm-induced the solar zenith angles of the two sampled regions are computed.
A correlation between solar zenith angle and amplitude was shown in Chapter 5; larger zenith angles
are associated with a larger amplitude. Figure 6.4.1d and e show the periapsis solar zenith angle for
each orbit in the pre-onset and post-onset ranges, respectively. Pre-onset zenith angles range from
155
53-57° and continue to increase to 65-70° during post-onset. From Chapter 5, such a relatively small
increase in solar zenith angle should not induce a factor of two increase in amplitudes, as seen here. For
further validation of increased storm-induced amplitudes, a quiet period is considered that samples the
same solar zenith angles as during the dust storm, but two months earlier (April 2018). This removes
the potential for seasonal effects to influence wave amplitudes. Figures 6.4.1c and f show Ar wave
profiles and the solar zenith angle range during April 2018. Average wave amplitudes are ∼10-15%;
this is in very good agreement with pre-onset amplitudes, thus confirming significant wave growth takes
place during global dust storms. By extension, waves in-situ of regional dust storms most likely grow
significantly. While moderate temperature increases are observed in the upper atmosphere, the posited
dissipation processes in Chapter 5 are outweighed by the processes allowing significant wave growth.
Elrod et al., 2019 noted increased turbulence and wave structures during this dust storm. Creasey et al.,
2006b observed no noticeable change in wave-like perturbations during the Noachis storm. Local dust
storms are more common than global dust storms; however, their locality reduces the probability of
observation. As such, local dust storms can be studied using models. One such study performed by
Spiga et al., 2013 investigated, what they proposed as, rocket dust storms. These localised storms form
and inject dust high into the atmosphere (30-40 km). This creates a dust layer similar to that observed
using Mars Reconnaissance Orbiter (MRO) date (Heavens et al., 2011). Spiga et al., 2013 identified
the emission of mesoscale gravity waves due to rocket dust storms. Large temperature perturbations
were observed.
There have been very few studies investigating the relationship between dust storms and gravity
waves. One possible explanation is an increase in wind speed during a dust storm. Ryan and Henry,
1979 and Ryan and Sharman, 1981 use Viking Lander 1 and 2 data to characterise the response of the
lower atmosphere to major dust storms at Mars. Prior to the dust storm, typical wind speeds at the
Viking 1 site were less than 10 m/s, however, at the onset speeds increased to 17.7 m/s with gusts
peaking at 25.6 m/s. As noted in the text, these are significantly higher than previously recorded at the
site. The Viking 2 landing site experienced slightly lower, but wind speeds notably increased at storm
onset. Initially, wind speeds increased to 13.9 m/s with 21.2 m/s gusts, rising over the next few sols to
17.3 m/s and gusts of 25.7 m/s. Sánchez-Lavega et al., 2019 used ground-based images to understand
the growth of the dust storm. At the onset, the storm expanded at a speed of 17±5 m/s, increasing to
40±5 m/s around 7-8 June. If this expansion is commensurate with surface winds, then a significant
enhancement in topographically-induced gravity wave activity is expected. Das et al., 2011 performed
156
Figure 6.4.1: (a) Ar density perturbations (blue lines) extracted prior to the global dust storm onset
using orbits with Ls=185-189°. (b) Ar density perturbations extracted post global dust storm onset
using orbits with Ls=192-194°. (c) Ar density perturbations extracted at similar solar zenith angles to
those sampled during the storm using orbits taken with Ls=159-161°. (d), (e) and (f) show periapsis
solar zenith angles during the sampled periods.
an Earth-based study investigating gravity waves over the Thar Desert, located at the India/Pakistan
border. They present evidence for a sudden enhancement of warming over a broad region of ∼0.8 K/day
near 3.5 km altitude. This region is known to be a warm spot due to dust storms (Deepshikha et al.,
2006). A significant perturbation is observed in the middle atmospheric temperature profiles, which is
posed to be a potential source of gravity waves. As gravity wave generation requires a source of perturb-
ation, localised heating due to dust storm is a plausible explanation. The results presented above show
significant increases in gravity wave amplitudes during a dust storm, as in agreement with Das et al.,
2011. The geographical location of each profile is considered to understand whether localised heating
has an impact on waves at Mars. In both latitude and longitude, there was no clear trend with increase
157
6.5. SUMMARY
wave activity; amplitudes are well mixed in location and time suggesting any local heating effects which
may occur in the lower atmosphere are not visible in the thermosphere. Das et al., 2011 found specific
wavelengths (18-35 km) became prominent during the dust storm, which is not seen outside this region
or during quiet periods. For the three sampling periods studied here, wavelengths are determined using
the method in Chapter 5. From inspection, the general distributions of wavelengths are indistinguishable
between the sampling period. From this, it can be inferred that dust storms enhance amplitudes but do
not alter the wave spectrum.
As the first observational study of gravity waves during a global dust storm, it was unknown as to
how wave characteristics would react to dusty conditions. The results are therefore new, exciting and
surprising in equal measure. As suggested, increased winds may lead to a growth in wave activity; it
is worth considering the use of a low-level wind/topography product to predict gravity wave activity.
Revised open questions are stated in Chapter 8. It has been stated ad nauseam that gravity waves during
dust storms have not been studied in depth. Nonetheless, comments have been made in conference
proceedings alluding to significant wave variations, e.g., ’The accelerometer data also reveal high-
frequency fluctuations in periapsis density that had sizable amplitude during this period’ from Zurek
et al., 2018 and ’... the structure of the atmosphere exhibits significantly more turbulence after the
onset of the dust event across all longitudes causing high variability in the altitude vs density as much
as a factor of 5 -10 within a scale height’ from Elrod et al., 2019. Our study is consistent with these
results. Future modelling and further observational studies of local storms may elucidate the connection
between wave behaviour and dust.
6.5 Summary
In this chapter, the response of the Martian upper atmosphere to the June 2018 global dust storm has
been studied. NGIMS data are exploited in an attempt to gain a fuller understanding of dust storm
dynamics are Mars. The main findings are highlighted below.
• CO2 , Ar and N2 density data are utilised to understand atmospheric expansion due to lower
atmosphere heating. All species’ densities increase at the onset of the dust storm with a mass
dependence; CO2 , Ar and N2 densities increase by factors of ∼1.7, 1.5 and 1.1, respectively. The
stability of N2 densities compared to other species during a dust storm is a new result and poses
the question as to what may cause this. Atmospheric expansion is 2-7 km, in agreement with
158
6.5. SUMMARY
previous studies (Bougher et al., 1997; Bougher et al., 1999, Wang and Nielsen, 2003, Qin et al.,
2019). Lower atmosphere heating has been predicted to be ∼25-50 K, whereas thermospheric
heating is typically less than 10 K.
• A comprehensive comparison with work by Withers and Pratt, 2013 investigating atmospheric
recovery timescales after dust storms. For the global storm studied here, storm decay times-
cales range from ∼30-70° Ls compared to a range of 71-118° found for the Noachis dust storm.
Nonetheless, similar local trends are observed with more substantial density enhancements having
recovery more quickly. The offset between Noachis and global dust storm values suggest that
storms exhibit similar behaviour, as shown by a common gradient in severity versus decay times-
cales. However, there is not a single relationship between severity and recovery for all storms.
This is the first time decay timescales in the upper atmosphere have been calculated for a global
dust storm using this method. These results allow comparison with previous dust storms.
• The response of upper atmospheric gravity waves to a global dust storm has been studied for
the first time. Pre-storm waves are compared to those at the peak of the dust storm. Wave
amplitudes significantly increase during the dust storm (by a factor of ∼1.5); however, the wave
spectrum remains similar. The cause of growth is currently unknown; however, increased wind
speeds generating larger-scale waves has been proposed.
159
Chapter 7
Results from Trace Gas Orbiter

Aerobraking Campaign
7.1 Introduction
For the first time, Mars’ thermosphere has been measured in-situ concurrently by multiple orbiting
spacecraft. Starting in March 2017, ESA’s Trace Gas Orbiter (TGO) began its year-long aerobraking
campaign. During this period MAVEN continued to take measurements of the thermosphere. Although
not a dedicated multi-spacecraft mission, much can be learned about the thermosphere from the con-
current datasets. Mars Express is still collecting data but at lower altitudes. The large scale temporal
and spatial variations can begin to be decoupled. As shown earlier, MAVEN and TGO sampled a similar
region at the same time but different altitudes. This opportune occurrence is exploited to understand
further density and temperature structures in the upper atmosphere with techniques employed to con-
nect the two sampled regions. The evolution of wave structures is explored using case studies alongside
general characteristics. The latter allows wave growth in altitude to be examined using TGO data in
conjunction with MAVEN. Questions to answer include what are typical densities and temperatures in
this previously unstudied region, and how do values compare to the MCD? We hope to study wave
evolution by utilising both the TGO and MAVEN datasets. How do amplitudes behave with increasing
height? As postulated, can we observationally show that larger wavelengths dominate with altitude?
160
7.2. SEPARATION OF BACKGROUND AND WAVE PROFILES
7.2 Separation of Background and Wave Profiles
The derivation of densities from spacecraft accelerometers was discussed in Chapter 2, and have been
performed from TGO accelerometer measurements by Sean Bruinsma and J-C Marty, CNES. This is a
new, unique dataset that no one has currently worked on. Figures 7.2.1a-c show three typical density
profiles from TGO, with periapsis locations show in Table 7.2.1.
Panel Date and Time Latitude LST Ls
a 2017-12-22 20:57:50 62.8°S 16.7 hr 104°

b 2018-01-25 15:39:55 68.6°S 7.1 hr 120°
c 2018-02-10 01:48:39 46.5°S 2.9 hr 128°
Table 7.2.1: Periapsis locations for profiles shown in Figures 7.2.1a-c
Within each panel, there are four density profiles: two for inbound and two for outbound. Inbound
and outbound profiles are shown in red and blue, respectively. The two profiles for each leg are derived
from TGO density data (solid) and fitted background densities (dotted). Densities are fitted to TGO
profiles using a third-order polynomial to the logarithm of the density, as done for MAVEN/NGIMS
profiles in Chapter 5. Figure 7.2.1a shows an example of inbound and outbound densities differing by
over a factor of two, implying the presence of a strong horizontal density gradient. Figure 7.2.1b exhibits
similar behaviour only above around 110 km. Below this, densities converge to a common profile. The
opposite is seen in Figure 7.2.1c. Inbound and outbound profiles align above around 112 km, below
which densities can differ by a factor of two, again implying horizontal density gradients. The outbound
portion in (a) and inbound portion in (c) show stagnation in density near periapsis. For the former
(latter) case, densities are constant around 4×10−8 kg/m3 (3×10−8 kg/m3 ). This is likely due to the
offset explained in Chapter 2, whereby closest approach and point of maximum density do not coincide.
Perturbations are observed in TGO density profiles as seen in NGIMS profiles, for example, below
110 km in Figure 7.2.1c, as density oscillate around the background profile. These have been inter-
preted as gravity waves and are extracted using the same process described in Chapter 5 by subtracting
the background profile from the density data. Then, relative perturbations are determined by dividing
through by background densities. Figure 7.2.2b shows the extracted perturbations. By inspection, the
161
Figure 7.2.1: TGO (solid) and fitted (dashed) density profiles for pass (a), (b) and (c). Inbound and
outbound legs are signified by red and blue, respectively. I=inbound, O=outbound. Periapsis locations
in the format [date-time, latitude, LST, and Ls] are (a) - [2017-12-22 20:57:50, 62.75◦ S, 16.7 hr and
104◦ ], (b) - [2018-01-25 15:39:55, 68.6◦ S, 7.1 hr and 120◦ ], and (c) - [2018-02-10 01:48:39, 46.5◦ S,
2.9 hr and 128◦ ]
inbound wave has an amplitude of 5% relative to the background. Around 114 km, the amplitude
reaches 10%. The outbound wave exhibits considerably smaller perturbations with amplitudes around
2.5% below 114 km and about 3-5% above.
Wave characteristics can be calculated following different methods starting with an initial wave
such as that shown in Figure 7.2.2b. Using MAVEN data, Terada et al., 2017 calculated the root-mean-
square of the perturbations, and England et al., 2016 considered sections along each orbit and computed
the three most dominant waves using a Lomb-Scargle spectral analysis. Using TGO accelerometer
data, Jesch et al., 2019 took the standard deviation of perturbations as the amplitude. The same
162
Figure 7.2.2: (a) Inbound (red) and outbound (blue) TGO (solid) and fitted (dotted) density profiles
as shown in Figure 7.2.1a. (b) Extracted inbound and outbound waves from the orbit shown in Figure
7.2.1a. (c) and (d) show inbound and outbound wave spectra, respectively, computed from Fourier
transforms of waves shown in (a).
technique used to extract NGIMS waves in Chapter 5 is used here for consistency. Both amplitudes and
wavelengths can thus be computed. As noted in the study mentioned above, caution should be taken
with this technique; the maximum wavelength that can be extracted with confidence is equal to the
altitude range considered. Consequently, the apparent dominant wave may not indeed be the presiding
wave. For nearly all orbits periapsis altitudes are below 110 km, this allows a maximum obtainable
wavelength of up to 10 km. For comparison, scales heights are ∼5 km in this region, as will be shown
in Section 7.4. Figures 7.2.2c and d show the spectra determined by performing a Fourier transform
on the inbound and outbound waves, respectively, and coloured accordingly. This is demonstrated in
wavelength. For the inbound wave, the dominant wavenumber is approximately 0.19 km−1 , equivalent
to a wavelength of ∼6 km. The amplitude of this component is 3.3%. For the outbound leg, the
163
7.3. BACKGROUND DENSITY STRUCTURES
dominant wavenumber is around 0.21 km−1 , equivalent to a wavelength of 4 km. The amplitude of
this component is 3%. It should be emphasised that only the most dominant composite waves have
been highlighted. However, the spectra reveal other similarly dominant waves at 0.25 km−1 and 0.08
km−1 for inbound and outbound legs, respectively. Spectra are determined as above for all orbits and
analysed in Section 7.5. Wavelengths greater than 10 km are likely present but are not captured here.
7.3 Background Density Structures
7.3.1 TGO and MAVEN Densities
In Section 7.2, it was shown that each density profile could be split into two components - background
values and waves. In this section, we present the global structure and seasonal variability of Mars’ back-
ground thermosphere. The global properties of atmospheric waves are examined in Section 6. While
the spacecraft observations provided density measurements at vertical locations which changed from
one orbit to another, we have interpolated these onto a common vertical grid. Thereby, we can more
easily perform comparisons between orbits and analysis of density variations across the parameter space
of solar zenith angle (SZA), longitude, local solar time (LST) and latitude. Red symbols in Figure
7.3.1 show TGO densities at 110 km altitude versus (a) SZA, (b) longitude (c) LST and (d) latitude.
Thermosphere densities at 110 km do not show consistent trends with most of these parameters, having
typical constant values of 5×10−9 -5×10−8 kgm−3 . Assuming a CO2 dominated region, these densities
equate to number densities of around 6×1016 -6×1017 m−3 . These values are within an order of mag-
nitude of those derived from EUV solar occultation data by Thiemann et al., 2018. A slight trend of
densities decreasing towards the south pole can, however, be seen in Figure 7.3.1d.
Figure 7.3.1 also shows densities at 150 km (green symbols) and 190 km (blue) from MAVEN ob-
servations. MAVEN total densities are calculated from CO2 , Ar and N2 data. Densities have been
calculated using measured count rates obtained from MAVEN NGIMS Level 1 (L1) export, versions 9
and 10, revision 1 data files, data that is available publicly in MAVEN NGIMS Level 1b (L1b) files on
the NASA Planetary Data System (PDS). The reader is referred to Stone et al., 2018 and references
therein for discussion into data reduction. Possible seasonal effects have been partially mitigated by
selecting MAVEN densities collected during similar solar longitudes as the data from TGO’s aerobraking
phase. Solar longitude range is taken to be 0-150◦ Ls.
164
Figure 7.3.1: TGO densities at 110 km (red) and MAVEN densities at 150 (km) and 190 (km) in (a)
SZA, (b) longitude, (c) LST and (d) latitude.
The expected trend of solar EUV-heating on the dayside is strikingly more evident when looking at
densities as a function of SZA (Figure 7.3.1a). Both MAVEN altitudes (150 and 190 km) show densities
as being relatively constant on the dayside (0-90◦ ) and decreasing towards the terminator and into the
nightside. Across the SZA range sampled by TGO at 110 km (90-175◦ ) the densities show no clear trend.
As seen from Figure 7.3.1b densities at all altitudes show no trend with longitude, implying that any
systematic density structures caused by surface topography can no longer be seen in the thermosphere.
Figure 7.3.1c shows densities as a function of LST. At the higher altitudes, densities are larger during
daytime hours (10-15 LST). The behaviour observed in the accelerometer dataset at 160 km (Liu et al.,
2019) appears inverse at 110 km where daytime densities are at their lowest around 12-15 LST, increas-
ing towards the nightside. The larger dayside densities at higher altitudes are consistent with in-situ
solar EUV-driven behaviour, solar heating on the dayside expanding the atmosphere. At 110 km, Mars’
thermosphere is less driven by in-situ solar heating and instead more controlled by atmospheric winds.
165
Figure 7.3.1d shows atmospheric densities as a function of latitude. At 110 km, TGO covers the entire
Southern hemisphere from 75◦ S to 0◦ , while MAVEN data covers latitudes extending from 75◦ S-75◦ N.
MAVEN densities at any particular latitude show variability covering 1-2 orders of magnitude, so it is
difficult to extract any latitudinal trends with confidence. Densities at 190 km suggest a weak trend of
decrease towards the south pole.
In the upper thermosphere, it is known that solar heating is a strong driver of density; good agreement
between Viking 1 and predominantly solar EUV-driven circulation model has been observed Bougher et
al., 1990. Solar extreme ultraviolet (EUV) is mainly absorbed in the altitude range 100-200 km (Bougher
et al., 2015). Thiemann et al., 2018 used MAVEN solar occultation data to understand thermospheric
variability. By considering density profiles at the dawn and dusk terminators, Thiemann et al., 2018
obtained a positive correlation between background temperature and EUV flux; thus, this region is solar
EUV-driven. The behaviour at 110 km is different from that of the upper thermosphere. There is a
pronounced drop in density across the terminator at MAVEN altitudes, whereas a small increase is seen
at TGO altitudes. It can be inferred that dynamics may play a significant role in density structure in this
region. One proposed explanation for the higher nightside temperatures is atmospheric downwelling on
the nightside. The convergence of global circulation forces a downward wind, leading to the atmosphere
‘collapsing’, increasing the density. Elrod et al., 2017 used MAVEN/NGIMS data to reveal He density
bulges on the nightside. They conclude that strong local vertical advection is responsible on the night-
side. This is a potential explanation for density behaviour observed here. It can be seen for densities
at 140 km that their behaviour is a combination of drivers seen at 190 km and 110 km. A decrease in
density is observed across the terminator; however, this is not as rapid as seen at 190 km. Clearly, the
driver of density change is dependent on altitude. Additionally, the optical depth at UV wavelengths is
above one in the lower thermosphere so the atmosphere can no longer be heated by sunlight.
Although the above findings are not new, they do provide observational evidence which agrees with
global-scale modelling of the lower thermosphere being dynamically driven. Some of these modelling
studies are discussed now, with comparisons made to our study. Medvedev et al., 2011 used the gravity
wave scheme of Yiǧit et al., 2008 in the GCM described in Hartogh et al., 2005. They found gravity
waves impart their dynamical influence on the lower thermosphere (∼100-130 km). Such revealed effects
include deceleration of the mean zonal wind. They note that temperature enhancements are complex,
with winter polar warming in the upper atmosphere. The introduction of gravity waves can typically alter
166
thermospheric temperatures by around ±15 K. Using simulations from the Max Planck Institute Martian
General Circulation Model (MPI-MGCM) (Medvedev et al., 2013), Yiğit et al., 2018 demonstrated large
temperature fluctuations around 120 km, as induced by gravity waves. By modelling an entire Martian
year, they found the global mean temperature to cool by around 9% with the addition of gravity waves.
These results highlight the importance and influence of dynamic processes, such as propagating gravity
waves, on the lower thermosphere. Our observations are in good agreement with the predictions of the
two aforementioned global-scale modelling studies that the lower thermosphere is dynamically driven.
7.3.2 Comparison with Laboratoire de Météorologie Dynamique Mars Climate Data-

base
In the following, TGO density data are compared to those extracted from the Laboratoire de Météoro-
logie Dynamique Mars Climate Database v5.3 (LMD MCD). The MCD is a database of atmospheric
statistics compiled from state-of-the-art Global Climate Model (GCM) simulations of the Martian at-
mosphere (Forget et al., 1999; Millour et al., 2015). The solar minimum scenario is employed. It
extends from the surface up to 250 km and accounts for physical processes such as the condensation
of the CO2 atmosphere during polar nights, the generation and evolution of water ice clouds and the
distribution of airborne dust (González-Galindo et al., 2015). The MCD is validated using available
measurements across a variety of missions. Atmospheric temperatures are compared to measurements
from MGS/TES (Thermal Emission Spectrometer), MRO/MCS (Mars Climate Sounder) and MGS and
Mars Express radio occultation experiments. Surface pressures and temperatures are compared to TES,
Viking, Pathfinder, Phoenix and Mars Science Laboratory (MSL) measurements (Millour et al., 2015).
Amongst many other useful outputs, density is used as a comparison parameter between observations
and modelling. Densities have been extracted from the MCD and compared to the TGO background
profiles at 110 km (see above section). Figure 7.3.2a-d shows TGO (red circles) and MCD (blue stars)
background densities at 110 km in SZA, longitude, LST and latitude, respectively.
In Figure 7.3.2a, the MCD successfully captures both the order of observed densities and the beha-
viour in SZA; the MCD correctly identifies the atmosphere not to be solar EUV-driven at this altitude.
Orbit-to-orbit variability is visible in the MCD data, though to a lesser degree. This is due to horizontal
changes, not in time. Figure 7.3.2b shows densities in longitude. As shown from both TGO and MAVEN
data, any influence of topography on thermospheric densities is not visible in the data, yet the MCD
shows a density increase of around 50% near 60◦ E when compared to TGO densities. Similarly, a slight
167
Figure 7.3.2: TGO (red circles) and MCD (blue stars) densities interpolated at 110 km shown in (a)
SZA, (b) longitude, (c) LST and (d) latitude. Here, only background profiles are used.
dip can be seen near 250◦ E. These deviations are most likely associated with the Hellas Basin and
Tharsis Region, respectively. Figure 7.3.2c shows densities in LST. Orbit-to-orbit variability is observed
in the midnight-dusk sector with MCD densities typically larger than TGO densities. Moving towards
to the dayside sees well-matched TGO and MCD densities. In the dusk-midnight sector, MCD densities
are several factors lower than observed by TGO. Again, the variability is not captured. The final panel,
Figure 7.3.2d, shows densities in latitude. The general behaviour is captured well with both datasets
exhibiting diminishing densities toward the south polar region. There is a smattering of larger-than-
expected MCD densities around 30◦ S-60◦ S which are most likely associated with the topographical
regions discussed above.
This next exercise stems from that above. Now, for both inbound and outbound legs on each pass
densities have been interpolated to altitudes of 105 km, 110 km, 115 km and 120 km. Densities are
168
taken from the background fitted profiles (Figure 7.2.1). Then, for each inbound and outbound profile,
the ratio of TGO to MCD densities is computed at the stated altitudes. Data have been binned using a
bin size of 0.5 hr × 5◦ . A mean value is then computed for each bin. Figure 7.3.3a-d show these ratios
in LST and latitude at 105 km, 110 km, 115 km and 120 km, respectively. Here, blue (red) shows an
overestimation (underestimation) by the MCD.
Figure 7.3.3: TGO to MCD density ratios in LST and latitude at (a) 105 km, (b) 110 km, (c) 115 km
and (d) 120 km. Data have been binned using a bin size of 0.5 hr × 5◦
Evidently, the lowest altitudes are the least sampled. At 105 km, TGO densities are larger by a factor
two in the dawn sector, fairly comparable towards noon, and lower by around 0.5 on the nightside. We
can infer that day-night variations are larger than predicted by the MCD. As will be shown now, this is
not the case at higher altitudes. Inspection of the next altitude level (110 km - Figure 7.3.3b) displays
more coverage, extending onward from dusk towards midnight. The strong underestimation in the dusk
sector is now weakened with TGO and MCD densities differing by around 50%. In the dusk sector, TGO
169
7.4. TGO TEMPERATURES
densities are generally larger, shown by a density ratio of about 1.5. Figure 7.3.3c shows this behaviour
is duplicated at 115 km with differences in densities between dawn and dusk becoming larger. The
faint underestimation seen around 6-8 LST at 110 km is no longer visible. MCD densities surpass TGO
densities here. This behaviour is repeated at 120 km (Figure 7.3.3d). Again, ratios decrease in the dusk
and noon sectors, implying MCD densities are decreasing at a faster rate with altitude. That is to say,
the scale heights differ, leading to the conclusion that temperatures observed by TGO are cooler than
predicted by the MCD in this region of the atmosphere.
Two main results were presented: TGO densities being a factor two larger than MCD densities in the
dusk sector compared to being smaller in the dawn and noon sectors and MCD densities decreasing at a
faster rate in altitude, leading to more extreme TGO-MCD ratios. The former indicates a deficiency in
local time dependencies in the thermosphere. This disparity appears to weaken with altitude suggesting
the solar EUV-driven nature of the atmosphere is captured, but the lower thermosphere is not. The
latter result can be attributed to TGO gathering data within a cooler atmosphere than expected. Within
the thermosphere, the MCD has been validated against MGS accelerometer data, Viking 1 and 2 data,
and Pathfinder. A full validation document is planned Millour et al., 2019b. MGS sampled altitudes and
solar longitudes similar to those in this study. However, MGS did not sample the noon sector. Thus, one
recommendation is using TGO observations to constrain the MCD in this region further. Similarly, MGS
sampled the dusk sector at latitudes below 50◦ , thus further constraints are needed for this region. As
explained in Chapter 3, discrepancies between the MCD and observational data are expected given the
relatively recent extension of the model into the thermosphere. This work will help to further constrain
the model densities, allowing to further understand the physics required to replicate the thermosphere.
7.4 TGO Temperatures
7.4.1 Deriving TGO Temperatures
In the following, TGO densities are used to derive temperatures within the lower thermosphere. For
reasons explained now, a different method is implemented to compute temperatures, and as such, only
a single temperature is derived per pass. From Chapter 3, it was seen that when integrating densities
downwards to attain temperature profiles, a convergence of profiles was observed after several scale
heights for a wide range of upper boundary conditions. A 100 K difference in upper boundary conditions
at 240 km leads a <10 K difference in temperature profiles at 200 km. And this uncertainty diminishes
170
further will lowering altitude. While this technique is favourable for MAVEN, it is not auspicious for TGO
as the latter spacecraft only sampled ∼15 km which is not large enough for profiles to suitably converge.
If the upper boundary conditions were known exactly, then profiles could be derived. For these reasons, a
new technique is implemented to determine a single temperature is retrieved at periapsis with confidence.
Density structures within an isothermal atmosphere can be described in altitude by ρ = ρ0 exp[−(z −

z0 )/Hs ], where ρ0 is density at a reference altitude z0 . Hs is the density scale height. This is derived in
Chapter 1. In this idealised situation a spacecraft would measure identical density profiles during inbound
and outbound passes. As evidenced in Figure 7.2.1a-c this is not necessarily the case. Zurek et al., 2017
present a modified version of the above equation which allows more complex density structures to be
taken into account. Equation 7.1 shows the along-track density as a function of altitude (z) and time
from periapsis (t) with fitted constants periapsis density (ρ0 ), along-track density gradient at periapsis

∂ρ0
∂t , density scale height at periapsis (Hs ) and along-track scale height gradient at periapsis ∂Hs
∂t .
!
∂ρ0 z

ρ(z, t) = ρ0 + t exp − (7.1)
∂t Hs + ∂H
∂t t
s
Zurek et al., 2017 approximates Equation 7.1 to first order by assuming gradient terms are small given
time variations are slow,
1 ∂ρ0 z 1 ∂Hs
ln ρ(z, t) = ln ρ0 + t− + tz (7.2)
ρ0 ∂t Hs Hs ∂t
By fitting Equation 7.2 to TGO density profiles, Hs can be determined, and subsequently, temper-
atures can be derived using Equation 3.2 where a mean molecular mass of 43.03 amu is assumed.
7.4.2 Background TGO Temperatures
Temperatures are computed for each TGO orbit using the technique described above. Figures 7.4.1a-d
shows periapsis temperatures in altitude, SZA, LST and latitude, respectively. As periapsis altitude
varies throughout TGO’s aerobraking period data points are coloured correspondingly to altitude. Thus,
panel (a) serves as a colour bar. To further compare observations with the MCD, temperatures have
been extracted and shown as a rolling mean by the black dashed line in each panel. Figure 7.4.1a
shows temperature as a function of altitude. Fairly good agreement is observed between TGO and
MCD at the higher altitudes where temperatures are predicted by the MCD to asymptote around 120
K above ∼108 km. TGO temperatures are observed to be ∼10 K cooler than this. This temperature
gap widens at the lowest sampled altitudes (∼102-105 km) as MCD temperatures increase up to ∼130
171
K. In contrast, TGO temperatures decrease on average to ∼80-90 K. This verifies the inferences made
earlier regarding TGO observing cooler temperatures. Figure 7.4.1b shows temperature as a function
of SZA. Here, altitudes variations are well mixed such that any trends in SZA should emerge. Again,
MCD temperatures are on average ∼20 K warmer than observed by TGO, nonetheless the same trend
is seen with fairly invariant temperatures in zenith angle. Figure 7.4.1c shows temperature as a function
of local solar time. The midnight-dawn sector is well captured by the MCD with the rolling mean in
good agreement with the average TGO temperature (∼120 K). The results within the noon sector imply
TGO temperatures are significantly cooler than expected from the MCD, however, on inspection of the
night sector, a slight increase in MCD temperatures is observed; the commonality between these sectors
is the altitude at which temperatures were calculated. Hence, the discrepancies are due to altitude,
not local time dependencies. Finally, Figure 7.4.1d shows temperature in latitude. Closest to the south
pole, lower altitudes were sampled, hence the divergence between MCD and TGO temperatures. MCD
temperatures are fairly invariant in latitude, however slight warming towards the equator is observed in
TGO temperatures.
Jesch et al., 2019 performed a similar exercise and found temperatures of ∼120 K in this region. Pre-
vious studies have also determined temperatures in the lower thermosphere. Gröller et al., 2018 exploited
MAVEN Imaging UV Spectrograph occultation data and required temperatures of ∼70-170 K to explain
deduced densities at 120 km. Stone et al., 2018 derived temperature profiles from MAVEN/NGIMS
data during the DD campaigns. These campaigns probed the thermosphere down to 120 km. At these
altitudes, temperatures generally span 90-150 K. Overall, temperatures derived in our present study are
in broad agreement other studies. Such cool temperatures (below ∼100 K) are not wholly unexpected
when considering energy sources in particular regions of the atmosphere. Energy sources were touched
upon in Section 5.1. Withers, 2006 utilised MGS and ODY aerobraking data to derive background
temperatures at similar altitudes to those in this study. Data from MGS Phase 2 is most comparable,
given its similar seasonal sampling. Typical afternoon temperatures at 120 km found by Withers, 2006
are 100-150 K and during this phase, relatively invariant in latitude. The difference in altitude may
account for the shift in temperatures observed between the two studies. Further, the trend in this study
suggests temperatures increase with altitude, thus converging to those found by Withers, 2006. MGS
morning temperatures are slightly cooler by ∼10-20 K than their afternoon counterparts. This is in
broad agreement with the findings here. Overall, there is good agreement between these studies. As
stated earlier, Vals et al., 2019 explored perturbations in MGS, ODY and MRO density profiles. They
172
Figure 7.4.1: Derived periapsis temperatures from TGO density data shown in (a) altitude, (b) SZA,
(c) LST and (d) latitude. Data are coloured according to altitude. Rolling means from extracted MCD
temperatures are shown by the black dashed lines.
investigated the correlation between amplitudes and temperatures; this wealth of temperature data is
now compared to those from TGO. ODY temperatures are 125-200 K, which are generally higher than
those found here. Seasonal and latitudinal sampling differences most likely account for higher tem-
peratures. Similar temperatures were found during MGS’ Phase 2 aerobraking campaign. Again, the
majority of TGO temperatures are below this range. The final measurements discussed are from ODY.
As stated earlier, ODY sampled a region alike to that by TGO. The temperatures found by Vals et al.,
2019 reflect this; they found a range of ∼125-160 K. This overlaps with the range observed for TGO,
but temperatures remain warmer for MRO. One potential cause is that TGO temperatures are derived
around periapsis, whereas Vals et al., 2019 found the mean temperature from a profile. As temperatures
are known to increase with altitude in the region sampled, the mean value is skewed by higher altitude
temperatures (Stone et al., 2018).
173
7.5. WAVES
7.5 Waves
This section aims to further understand waves in the thermosphere, with particular emphasis on wave
growth, allowing the posed questions to be answered. Following the process outlined in Section 7.2
whereby waves were separated from background structures, wave amplitudes and wavenumbers are
calculated for all passes. Spectra are combined and averaged with wavenumbers being replaced by
wavelengths. Figure 7.5.1a shows averaged amplitudes of wavelengths using a bin size of 2 km. These
are averaged over all altitudes sampled by TGO accelerometers (around 100-120 km). Peak amplitudes
relative to the background are around 6±4%. Also shown is the averaged spectrum derived from waves
extracted from MAVEN data in Chapter 5.
As vertical wavelengths may scale with scale height, a normalised wavelength is introduced; this is
the vertical wavelength divided by the appropriate scale height. For TGO altitudes, we have determined
a mean scale height of 5.4±2.0 km, with the latter value being the standard deviation. This uncertainty
corresponds to a temperature of ∼40 K. Scale heights are determined from the gradient of a linear fit to
log(ρ)-altitude plots, treating inbound and outbound profiles separately (Figure 7.2.1). And for MAVEN
altitudes, the scale height is taken to be 12 km (Zurek et al., 2017). Figure 7.5.1b shows amplitude as a
function of normalised vertical wavelength. This figure begins to elucidate how particular waves within
the thermosphere behave depending on their relative wavelength. For discussion, 7.5.1b is simplified
by taking the ratio of MAVEN amplitudes to TGO amplitudes for each bin. This is shown in Figure
7.5.1c. The horizontal dashed line shows a value of one. These results allow the evolution of waves to
be explored due to each characteristic - amplitude and wavelength - shedding light on how waves de-
velop. As density falls exponentially, amplitudes necessarily grow exponentially to conserve momentum.
Thereby, understanding amplitudes at distinct altitudes allows these theories to be tested and regions
of damping to be identified.
Figure 7.5.1c infers two regimes observable at altitudes studied by TGO and MAVEN. The first is
for normalised wavelengths less than one; that is, measured wavelengths are less than the scale height
at which they were captured. In this regime, the ratio of amplitudes is unity leading to the inference
that waves of these scales are saturated by the time they reach the upper atmosphere. At the lowest
normalised wavelengths (<0.5), ratios are marginally lower than one, indicating damping is present. The
second regime occurs beyond normalised wavelengths of unity. Here, MAVEN amplitudes are larger by
174
7.5. WAVES
Figure 7.5.1: (a) Binned TGO (red) and MAVEN (blue) wave spectrum with width 2 km. Error bars
are one standard deviation. (b) Same as (a), but wavelengths have been normalised by scale height.
(c) Amplitude ratio between MAVEN and TGO waves using normalised data from (b)
up to a factor of two compared to their TGO counterparts. Given MAVEN’s higher altitude orbits, it
can be inferred that longer-wavelength waves, those with a normalised wavelength larger than one, will
continue to grow with altitude. An undamped wave is expected to grow by a factor of around nine,
assuming an average scale height between 120 km and 160 km of 9 km. We can therefore infer that
175
7.5. WAVES
although larger wavelengths continue to grow with altitude, they are still damped but to a lesser degree
than smaller waves. This is the first study to observationally infer these results at Mars.
From Figure 7.5.1a, the most dominant wavelengths are ∼5-15 km. Given the altitude range studied
by TGO is about 15 km, wavelengths of this length can be extracted confidently from the data. Creasey
et al., 2006b used MGS occultation profiles to determine wave properties at 0-30 km altitude. Near the
surface, a bi-modal distribution of vertical wavelengths with peaks at 8-10 km and 13-15 km were found.
The former range is in good agreement with the range found in this current study. Consequently, it is
plausible that dominant wavelengths remain sufficiently unchanged through the lower atmosphere with
amplitudes increasing with altitude; Creasey et al., 2006b found amplitudes typically less than 5%. The
latter range is found to still be relatively dominant at TGO altitude; however, 13-15 km wavelengths
are nearing the point of not being confidently extracted from the profiles via the Fourier transform. In
Chapter 5, we found vertical wavelengths of 10-30 km above 140 km. It is possible that atmospheric
wave filtering is being observed above 100 km as smaller-scale waves are absorbed by background winds
(Fritts and Alexander, 2003). These results imply shorter wavelengths (<10 km) are dominant at TGO
altitudes, whereas larger wavelengths (>10 km) are dominant at MAVEN altitudes (>140 km).
Jesch et al., 2019 found average wave amplitudes of 6-8% with an altitudinal dependent standard
deviation of 1.5-4%. These amplitudes are in a very good agreement with values found in this current
study; this validates our technique. Using MGS accelerometer data, Fritts et al., 2006 extracted amp-
litudes from wave profiles. Fortuitously, during the latter stages of its aerobraking phase, MGS sampled
a similar region to TGO. MGS sampled nightside southern latitudes during northern spring/summer.
At 105 km, amplitudes are around 5-10% and gradually increase to about 15% at 120 km. Evidently,
values are very repeatable under similar atmospheric conditions. They did not differentiate between
wavelengths. Similarly, Creasey et al., 2006a extracted perturbations from MGS aerobraking data.
During similar seasonal conditions as TGO experienced, Creasey et al., 2006a found amplitudes in the
Northern hemisphere rarely reach 5% which is comparable to those found in this present study. During
the same phase, MGS sampled southern latitudes also. Here, amplitudes are larger than above the
equator and hence larger than TGO amplitudes. MGS sampled higher altitudes than TGO, and this
difference may account for the wave growth, as justified above. Vals et al., 2019 determined the RMS
perturbation amplitudes from MGS, Odyssey (ODY) and Mars Reconnaissance Orbiter (MRO) density
data. ODY sampled Mars in approximately the opposite half of the Martian year (Ls=260-310◦ ). Amp-
176
7.5. WAVES
litudes are 5-20% with an average value above 10%. Seasonal effects are expected to vary amplitudes.
MGS values are comparable to those found by Creasey et al., 2006a, expected. MRO sampled the
comparable region to TGO; measurements were taken in the southern hemisphere at Ls=30-100◦ . RMS
amplitudes are typically 5-10% which is in very good agreement with those found in this current study.
The two underlying findings from our amplitudes are that they are in agreement with previous
spacecraft data and there is clear wave growth in the thermosphere. Studies mentioned in this section
have shown slight amplitude growth with altitude within their specific altitude range. Our technique for
extracting waves could be applied to smaller altitude sections, as in Siddle et al., 2019, and amplitudes
calculated for each segment. It would not be unexpected to see wave growth as in previous studies. A
reduction in altitude range limits the extent of extractable wavelengths; thus, dominant waves are likely
to be overlooked. Further, Medvedev et al., 2016 used the MPI-MGCM to interpret MAVEN Imaging
Ultraviolet Spectrograph (IUVS) observations. By comparing density profiles to zonal mean densities,
relative deviations in altitude were determined, and it was demonstrated that density perturbations
could grow up to a factor of around two as they propagate from the lower to the upper thermosphere.
This is in good agreement with results found in our study where wavelengths typically larger than the
background scale height grow by up to a factor three between TGO and MAVEN periapsis altitudes
(around 100-140 km). Shorter wavelengths appear saturated at these altitudes. Overall, the finding that
waves unquestionably grow with altitude is not unexpected, it has been shown using multiple spacecraft
data for the first time, increasing confidence in the results.
A case study is now presented that shows the possibility that the same gravity has been observed in
both the TGO and MAVEN dataset. Figure 7.5.2a shows an outbound TGO wave (red line) sampled
on 28 Jan 2018 at 16:53. Less than an hour later, MAVEN sampled a similar region of the atmosphere
during its inbound leg. Geographical locations of each spacecraft are shown in Figure 7.5.2. The
remarkable similarity is observed in characteristics between the waves as quantified by Figures 7.5.2b
and c, which show the wave spectra of the sampled MAVEN and TGO waves, respectively. Dominant
amplitudes are ∼10-20% with wavelengths 5-10 km. Although these are typical wavelengths for TGO,
these are short for MAVEN, thus increasing the confidence that the same wave has been sampled. There
are less than ten occasions where TGO and MAVEN orbits reside is such similar locations. And of these
few occasions, not all exhibit similar wave characteristics across both spacecrafts’ datasets.
177
7.6. COMBINING MAVEN AND TGO DATA FOR CONTINUOUS VERTICAL PROFILES
Figure 7.5.2: (a) Extracted inbound MAVEN wave (blue) from 28 Jan 2018 and outbound TGO wave
(red) also from 28 Jan 2018. (b) and (c) are wave spectra of the extracted waves in (a) for MAVEN and
TGO, respectively. (c) show the trajectory of each spacecraft’s pass in longitude and latitude. Colours
are consistent in all figures.
7.6 Combining MAVEN and TGO Data for Continuous Vertical Profiles
Earlier, TGO and MAVEN waves were compared with one case study presented, which posited the
possibility that the same wave had been observed in both datasets. This was achieved due to the
proximity of both orbits. This fortuitous closeness is exploited to hydrostatically connect the datasets
such that density and temperature behaviour in the ‘unsampled’ region, typically in the range 120-140
km, can be derived.
178
7.6.1 Comparing TGO and MAVEN Densities
A short calibration exercise is undertaken to ensure that TGO and MAVEN densities are comparable.
By this, it is meant that an analysis is conducted to test whether both spacecraft would measure the
same densities at the same location and remove the possibility of any systematic bias between the
instruments. MAVEN did not undertake any DD campaigns during TGO’s aerobraking phase; thus,
similar altitudes were not sampled. Consequently, the requirement for data concurrency is lifted, and
all DD campaigns are considered. Further, to ensure an altitudinal overlap between the datasets, TGO
densities are no longer cut-off at 120 km; they now extend up to 125 km. Background densities within
both datasets are interpolated at 125 km. A total along-track density is constructed for MAVEN using
CO2 , Ar and N2 . At this altitude CO2 is dominant, so the omission of other species does not pose
an issue. For each dataset, densities are then interpolated further in solar zenith angle on a fixed grid
with spacing 1°. Where possible, MAVEN densities are divided by TGO densities for each interpol-
ated solar zenith angle. Figure 7.6.1 shows the ratio of MAVEN to TGO densities as a function of solar
zenith angle. A least-squares fit is shown by the dashed line. A ratio of unity is signified by a dotted line.
Figure 7.6.1: Ratio between TGO and MAVEN densities at 125 km altitude as a function of solar
zenith angle. A linear best line is shown with by the blue dashed line. For ease of comparison, a grey
dotted line shows a ratio of one.
179
The underlying result is that MAVEN densities are typically larger than respective TGO densities.
There is a great deal of variability in the ratio for the majority of sampled solar zenith angles. As
mentioned earlier in the chapter and in Chapter 2, data are not gathered concurrently at similar locations
leading to further variability potentially caused by the following factors: atmospheric tides, seasonal
variations, and solar activity. The latter two being addressed in Chapter 4. MAVEN sampled during a
time of comparatively higher solar activity. Nonetheless, the majority of MAVEN densities are within a
factor of two of TGO densities. A good agreement is observed between the datasets due to no clear
offset, and as such, they can be ‘synthesised’ for analysis in the next section.
7.6.2 Hydrostatically Connecting TGO and MAVEN Density Profiles
From Chapter 2 it was shown that MAVEN concurrently sampled a similar region as TGO during the
tail-end of the latter’s aerobraking phase. During this period it is possible to hydrostatically connect
these two regions, allowing densities and temperatures within the unsampled altitudes to determined.
This has implications on modelling efforts where a continual profile is required, such as within diffu-
sion models. In the following, the process of connecting individual TGO and MAVEN profile is described.
Profiles are selected where similar TGO and MAVEN latitudes and longitudes are sampled and
periapses nominally within three hours of each other. In practice, consecutive TGO and MAVEN orbits
occur within one hour. The two regions’ densities cannot be connected by a simple function such as
a polynomial as this would create a nonphysical atmosphere, whereby the temperature structure would
not successfully reproduce the density profile. For the densities to be connected, a physical temperature
profile is initially required. As alluded to in Chapter 3 and explained earlier in this chapter, there would
be large uncertainty associated with temperature profiles derived from TGO data; thus a rolling mean
of 1 km is performed on TGO densities. This removes sharp variations in density which are present.
The gradient of the log-density is calculated every 1 km using a window of 2 km. This is identical to
computing the upper boundary condition but applied over multiple smaller ranges. A mean molecular
mass of 43.03 amu is used, taking into account a CO2 dominated atmosphere. NGIMS temperature
profiles are derived via downward integration (Section 3.2). The two derived temperature profiles are
now connected hydrostatically to create a continuous profile, and this is achieved by assuming the
missing temperature profile takes the form of a fourth-order polynomial shown by Equation 7.3.
180
T (z) = az 4 + bz 3 + cz 2 + dz + e (7.3)
Coefficients are determined using calculable boundary conditions, as shown in Equation 7.4. The first
boundary condition is the bottom temperature (T (z1 )) of the missing profile must match the temperature
at the highest altitude in the TGO region (z1 ). Similarly, the upper temperature (T (z2 ) in the missing
region must match the temperature at the lowest MAVEN altitude (z2 ). Thirdly, the temperature
gradient (T 0 (z1 )) at z1 must match that of the bottom temperature gradient in the missing region.
Fourthly, the temperature gradient (T 0 (z2 )) at z2 must match that of the top temperature gradient in
the missing region. The final boundary condition is an assumption that there is a local minimum or
maximum in the missing temperature profile at an altitude of zm located between z1 and z2 . By solving
Equation 7.4, a hydrostatic temperature profile can be determined.
     
 z14 z13 z12 z1 1 a  T (z1 ) 
     
 z4 z23 z22 z2   b   T (z2 ) 
1    
 2
     
(7.4)
    0
·  c  = T (z1 )
 3
 4z1 3z12 2z1 1 0

     
0
     
 4z 3 3z22 2z2 1 0 d T (z2 )
   
 2 
     
3
4zm 2
3zm 2zm 1 0 e T 0 (z3 )
zm is the sixth unknown. However, it is constrained between z1 and z2 . Values of zm are iterated
through in increments of 0.1 km. Each profile is plausible concerning temperature; however, a further
condition is required to be met to determine the actual value of zm . The final condition is the density at
the highest altitude in the missing region must match the density of the lowest altitude in the MAVEN
profile. A new density profile is created in the missing region using Equation 7.5,
dz

ρ(z) = ρ(z − dz) exp − (7.5)
H(z)
where the scale height H(z) is defined as,
1 m(z)g(z) 1 dT (z)
= + (7.6)
H(z) kT (z) T (z − dz) dz
where dz is 0.1 km and ρ(z) and T (z) are the missing region’s density temperature profile. The ini-
tial ρ(z − dz) and T (z − dz) values are the top TGO values. All other symbols have been defined
previously. For each zm , the temperature profile is input into Equation 7.5 and a density profile is
181
derived. The ratio between the density at the highest altitude in the missing region and the density of
the lowest altitude in the MAVEN profile is calculated and subtracted from unity. This is defined as the
offset. The zm value that derives the density profile with the lowest offset is taken forward as the best fit.
Figures 7.6.2a and b show connected density and temperature profiles, respectively. MAVEN (TGO)
data are shown by blue (red) lines. Data are taken on 29 Jan 2018 in the southern hemisphere, as shown
in Figure 7.6.2c. For these orbits MAVEN probes down to just below 140 km and TGO samples just
above 120 km. As these are typical altitudes, there is a 20 km range which is not sampled. The well-
matched density profile confirms that a simple interpolation between regions is not suitable, but rather
a more complicated temperature structure is required. The necessity for the temperature profile to have
a maximum or minimum does not introduce any dramatic differences compared to derived profiles.
Figure 7.6.2c shows the coverage of MAVEN and TGO in longitude and latitude. Longitude is util-
ised as geographically similar profiles are required. Results are not expected to differ dramatically if local
time comparisons are made. This section aims to show how this technique works and how applicable it
could be to future datasets. The connecting density and temperature profiles should not be assumed
to be entirely accurate as the boundary conditions create a volatile profile. If the TGO profile were
truncated below 120 km, a different temperature profile would be observed. Similarly, if the MAVEN
profile were cut-off below ∼160 km, the resultant temperature profile would be significantly different
from what has been derived.
Given time constraints, the following proposed exercise has not been undertaken, but would nonethe-
less be intriguing. As mentioned, the Martian community is fortunate to have upwards of ten spacecraft
gather atmospheric data at the Red Planet. The conflation of density datasets may reveal some inter-
esting trends which have been missed due to lack of sampling. By luck, MGS may have studied similar
regions to MAVEN, allowing seasonal and even yearly trends to be explored using profiles extending
upwards of 100 km.
182
7.7. SUMMARY
Figure 7.6.2: (a) Connecting TGO and MAVEN density datasets. TGO and MAVEN densities are
shown by red markers and blue lines, respectively. The connecting density profile derived from the
connecting temperature profile in (b) is shown by the black dashed line. (b) Derived TGO and NGIMS
temperature profiles. Connecting temperature profile (black dashed line) is fit according to boundary
conditions specified in the text. It takes the form of a fourth-order polynomial. (c) Latitude and local
solar time of used TGO (red) and MAVEN (blue) data.
7.7 Summary
This chapter has focused on concurrent density measurements from MAVEN and TGO. Much of this
chapter is an expansion of results previously presented in this thesis. The key findings are outlined below.
• Concurrent TGO and MAVEN data have been utilised to understand the main drivers of density
structures within the thermosphere. It is apparent that at TGO altitudes (<120 km) the atmo-
sphere is predominantly driven by dynamics, demonstrated by invariant densities in solar zenith
angle. Above this region, the atmosphere is primarily solar-driven as implied by an apparent
183
7.7. SUMMARY
decrease in density with increasing solar zenith angle in MAVEN data (at 150 km and 190 km).
• TGO and MCD densities have been compared in altitude, local solar time and latitude. The MCD
is unable to capture the local solar time dependence in density, as demonstrated by larger than
expected TGO densities in the dusk sector. By comparing densities in altitude and computing
temperatures, it is evident that the MCD predicts warmer temperatures than observed by 10-20
K.
• Waves have been extracted from TGO densities profiles following the same procedure outlined in
Chapter 5. Typical amplitudes and wavelengths are ∼5-10% and 5-15 km. Using results from
Chapter 5, TGO and MAVEN spectra have been compared by scaling wavelengths by respective
scale heights. Waves with wavelengths smaller than their scale heights are saturated within the
thermosphere as indicated by similar TGO and MAVEN amplitudes. For longer waves, MAVEN
amplitudes increase by a factor of two, thus demonstrated wave growth as expected. Damping is
evident as growth is not as rapid as expected.
• Lastly, it has been shown that TGO and MAVEN profiles can be hydrostatically connected. Com-
plete density and temperature profiles are now available from ∼100 km up to over 200 km. These
can be utilised for use within diffusion models. Additionally, this technique has the potential to
connect future TGO and MAVEN datasets.
184
Chapter 8
Conclusions and Future Work
8.1 Summary and Conclusions
In this study, Mars’ upper atmosphere has been studied in depth using in-situ density data from NGIMS
on board MAVEN and accelerometer data from TGO. Mars’ atmosphere is appreciably dynamic, both
temporally and spatially.
Chapter 1 introduced the relevant background knowledge required for this thesis. The history of
Martian observation and exploration has been detailed. The planetary and orbital properties of Earth
and Mars have been compared, and the consequences of any dissimilarities have been outlined. The
underlying physical principles that determine the observed density and temperature structures have
been explained regarding spacecraft data. One particular Martian GCM, the LMD-MCD, has been used
throughout this study. The capabilities of the model, and thus reasons for its selection, are highlighted.
A summary of Martian GCM development has also been presented for completeness. Observations
of gravity waves in the Martian thermosphere were discussed in anticipation of later work, with their
generation in the lower atmosphere and evolution up to the thermosphere being described. Lastly, our
current understanding of dust storms, from their origin and formation to their trajectory and decay, has
been described.
Chapter 2 introduced the MAVEN and TGO missions and datasets used in this thesis. The operating
principle, collection, and reduction of data from the Neutral Gas and Ion Mass Spectrometer on board
MAVEN have been discussed in detail. NGIMS is capable of measuring individual species’ densities,
allowing the composition of Mars’ upper atmosphere to be studied in detail. The three primary species
185
8.1. SUMMARY AND CONCLUSIONS
used in this study are CO2 , Ar and N2 . CO2 is used for its dominance and Ar for its inertness, thus allows
the dynamics of the atmosphere to be studied without considering chemistry. Both MAVEN and TGO
have on board accelerometers from which densities can be retrieved from raw measurements. The oper-
ation of accelerometers and subsequent density retrieval has been discussed in the context of MAVEN.
MAVEN accelerometers densities have been compared directly to NGIMS data, with densities typically
differing by ∼20% near periapsis. During TGO’s aerobraking phase MAVEN continued to gather data.
Fortunately, during the early months of 2018, TGO and MAVEN sampled similar regions concurrently.
The potential science achievable with these overlapping events is discussed and undertaken in Chapter 7.
Chapter 3 introduced the data analysis techniques needed throughout this study. In particular, the
derivation of temperature profiles by integrating density data downwards has been discussed. Both the
advantages and disadvantages of the presented technique were addressed. The Mars Climate Database
was used to test this technique by extracting along-track densities, deriving temperature profiles and com-
paring them with obtained along-track temperature profiles. The most substantial variations between
derived and extracted temperature profiles occurred near the terminator. This is posited due to the
derivation technique relying on vertical density variations, however spacecraft travel quasi-horizontally
near periapsis, thus sampling horizontal rather than vertical variations. This manifests as a divergence
between temperature profiles, with derived temperatures ∼20 K warmer than extracted temperatures.
A key aim of this chapter was to assess whether there is a systematic difference between NGIMS and
MCD densities. The local time dependence is not captured successfully, with NGIMS CO2 densities
larger at the subsolar point up to a factor of five. This discrepancy reduces towards the terminator.
Beyond a solar zenith angle of ∼90° , MCD CO2 densities surpass NGIMS up to a factor of ten at the
antisolar point. Similar behaviour is seen for Ar and N2 species however MCD densities begin to exceed
NGIMS densities at zenith angles of ∼30°. In light of these findings, further open questions that need
to be addressed are the following. What is explicitly causing this discrepancy? Can refined gravity
schemes account for inconsistencies? What further information is required to constrain model densities
further? An offset between timings of the closest approach and maximum density are highlighted in
the TGO density dataset. It has been suggested that strong latitudinal density gradients are responsible.
Chapter 4 explored density and temperature variations over three timescales: diurnally, monthly
and seasonally. The first question to answer was what are typical dayside and nightside temperatures
at the Red Planet and how do they compare with modelling efforts? Diurnal temperatures variation has
186
been shown in solar zenith angle using background profiles. Dayside temperatures are typically ∼250
K, reducing to ∼150 K towards the nightside. Significant variations are seen across the terminator
owing to the atmosphere no longer being directly heated by solar radiation. The MCD most successfully
reproduces observed temperatures below ∼150 km on the dayside. Above this, the MCD is too cool by
at least 20 K. The MCD is consistently warmer on the nightside by ∼20 K than observed in the data.
The inclusion of refined gravity wave thermal effects should account for some of this. The The effect of
the 28-day solar rotation on the thermosphere has been identified in density data, further highlighting
the strong influence of the Sun on the atmosphere. However, this effect is not continually observed in
the data. Other, more influential factors may cause solar rotation effects to be diminished. Seasonal
variations in density and temperature were investigated by identifying periods where MAVEN sampled
regions multiple times. This would, for the first time, allow the effect of season to be quantified. Each
region spanned ∼2° latitude and ∼0.5 hr local time. The criteria led to 27 periods being identified, each
of which contained ∼30 orbits. Wave activity was removed by using background profiles. For a fixed
altitude, density is enhanced by nearly a factor of 100 during the summer months as the atmosphere
expands due to intensified heating in the lower atmosphere, especially during the dust storm season.
Likewise, this was performed using temperature profiles. A trend as above was not visible. Throughout
a Martian year, temperature appears to vary by no more than 100 K. Still, a more representative upper
limit is ∼60 K. The Mars Global Ionosphere-Thermosphere model accurately reproduces these results.
In view of work carried in this chapter, there are two main questions going forward. The first is a revision
of seasonal temperature trends. Can models reproduce this unexpected result, or is there an issue with
the method used? The second is born from the discussion of polar warming. Enhanced circulation by
dust near perihelion has been used to explain warming around 120 km, can the same mechanism explain
observed southern warming? If so, how far does this effect extend upwards? If not, what could cause
such warming
Chapter 5 investigated wave activity within the upper atmosphere using Ar NGIMS data. Perturb-
ations were extracted from density and temperature profiles and interpreted as vertically propagating
gravity waves. Amplitudes are typically 10-20%, and wavelengths vary from 10-30 km. These values are
compared to and agree with results derived from other Martian datasets. Waves amplitudes are correl-
ated with temperature with the largest amplitudes arising in the coolest regions. Amplitude continues
to grow with altitude on the nightside, yet appear saturated by ∼120 km on the dayside. The deposition
of wave momentum and energy have been discussed, and typical values compared to previous results.
187
Any evidence of the origin of presented gravity waves appear erased by the time they reach the studied
altitudes, given there is no correlation between activity and topography. Results have been introduced
which indicate that waves rarely appear persistent across consecutive orbits suggesting they can survive
several hours. One consequence of such persistence is the potential for CO2 cloud formation due to
lasting cold pockets of air. Other factors, such as nucleation rates and density, need to be considered
for further analysis. The posed open questions have been successfully answered, as expected. One
additional question which is discussed in Further Work, is can complete gravity wave structures (i.e.
vertical and horizontal wavelengths) be computed from purely inbound and outbound trajectory data?
Chapter 6 examined the response of the upper atmosphere to the 2018 global dust storm. Dust
storm onset is visible around Ls=187°consistent with previous studies. This is slightly later than first
detected by rovers (Ls≈185°) most likely caused by sampling bias either spatially and/or altitudinally.
For a fixed altitude, densities increased by ∼10-80%, with the most substantial increases associated with
more massive species (i.e. CO2 ). Combined with Elrod et al., 2019, it can be inferred that species lighter
than ∼22 amu will have density decrease during a global dust storm event for a given altitude. Upper
atmosphere heating is minimal (O(1) K), whereas the lower atmosphere is expected to be warm by
25-50 K. For the first time, gravity waves have been analysed during a dust storm. Average amplitudes
are up to 50% larger during storm periods than during immediately preceding intervals. Storm-waves
were compared to waves in a comparable region to account for changes in sampled regions and were
found to be significantly larger. One potential explanation is localised dust heating, causing more sig-
nificant perturbations than typically observed. Another factor may be enhanced winds, which induce
larger topographically-induced gravity waves. The relationship between heating and gravity waves has
been seen on Earth. Following the technique employed in previous studies, the decay of the June 2018
dust storm has been explored, and it appears that each dust storm exhibits similar abatement behaviour;
the severity of the dust storm is anti-correlated with the time needed to recover. However, each storm
has a unique recovery relationship with severity. Decay timescales are similar to the 2001 and 2007
dust storm; normally around 60°Ls, equivalent to several months. Modelling efforts of dust storms
have been discussed and compared to results in this thesis. The response of the upper atmosphere has
been compared to the lower atmosphere with more severe heating is observed in the latter. Two new
results in this chapter require new open questions to be posed. The first is the interesting finding that
N2 densities appear relatively stable during global dust storms. Why is this? Can a diffusion model
explain this surprising finding? A GCM is most certainly required to complete this work. The second
188
8.2. FUTURE WORK
question delves into dust storm gravity waves. What causes a significant increase in wave amplitude?
Are waves generated with larger amplitudes, or does the dust storm allow substantial growth? It is
worth considering the use of a low-level wind/topography product to predict gravity wave activity
Chapter 7 exploited concurrent MAVEN and TGO data. Similar geographic locations were sampled
with TGO sampling up to ∼120 km and MAVEN down to ∼140 km (120 km during DD campaigns).
Initially, densities were interpolated to 125 km and further interpolated in solar zenith angle. Overall,
there was good agreement between the two datasets, thus confirming they can be ’synthesised’ for
future use. This allows profiles from TGO and MAVEN to be hydrostatically connecting to understand
density and temperature structures in the ‘unsampled’ region, typically in the range of 120-140 km.
By comparing background densities at 110 km (TGO), 150 km (MAVEN) and 190 km (MAVEN), it
can be inferred that atmosphere structure at higher altitudes (above at least 150 km) is solar-driven,
as indicated by evident day-night disparities in density. TGO densities imply that the lower portion
of the thermosphere is more driven by dynamics. This new dataset enabled further comparisons with
the MCD. Two main conclusions have been drawn from this. The first is a deficiency of the MCD to
successfully capture the density trend in local solar time, as found for NGIMS. The second inference is
the MCD predicting cooler temperatures than observed by TGO; this is implied by the different rates at
which densities decrease by with increasing altitude and explicit calculations of scale heights. The work
in Chapter 5 has been extended with the addition of extracted TGO waves. By comparing TGO and
MAVEN wave amplitudes, it was found that shorter-wavelength waves (typically less than their scale
height) are saturated within this region. In contrast, longer waves continue to grow by over a factor of
two. For the first time, the same wave has possibly been observed across both datasets. As found with
MCD-NGIMS comparisons, TGO-MCD comparisons highlight deficiencies with the MCD in reproducing
observations. The most obvious question to ask is why. What physics is missing from the MCD to
prevent replication of observations? One key factor is most likely the inclusion of accurate gravity wave
schemes, which allow warming/cooling to be included. Will the addition of MY34 atmospheric and solar
conditions in the MCD account for some of this difference?
8.2 Future Work
In the following, potential future work in outlined and concludes with pertinent questions currently
unanswered in this thesis.
189
8.2. FUTURE WORK
Extension of MAVEN Mission
The MAVEN mission has been granted an extension beyond its initial mission and will act as a commu-
nication relay for current and future landers and rovers, including Mars 2020 (Beswick et al., 2020). As
such, the new orbit is more circular with a higher periapsis (∼200 km) than during its previous missions
(R Yelle 2019, personal communication). NGIMS is still capable of retrieving in-situ density data at
this altitude; therefore, there is potential to extend results found in this thesis. As touched upon in
Chapter 3, MAVEN sampled the Red Planet primarily during solar minimum and moderate conditions.
If MAVEN continues to sample the atmosphere beyond 2025, the possibility of sampling more frequent
and extreme coronal mass ejections will likely increase, thus furthering our understanding of their impact
on atmospheric loss. By an extended mission, the interannual variability in density and temperature
can be considered. It would be beneficial to seasonal variations as well as the variations during the
solar cycle. Additional observational will, needless to say, allow validation of all models described in
this study. Moreover, further wave profiles will inform our understanding of wave behaviour in currently
missing locations, such as at the smallest and largest solar zenith angles during aphelion. The latter
location is expected to produce the largest observed gravity waves. With an increased mission length,
the probability of another global dust storm event rises, more so for a regional dust storm. The same
analyses would be performed in Chapter 6. This additional potential study could confirm several spec-
ulations posited. The first is the behaviour of different species; does O still decrease during a global
dust storm? Does the atmosphere expand at the same rate? It is hoped that similar decay behaviour is
observed, thus allowing predictions of storm decay times to be made.
Gravity Wave Validation
This thesis has presented gravity wave characteristics determined within the upper thermosphere. Given
the reliance of observations required to validate the gravity wave scheme, it would be informative to
perform a short literature review comparing observational and model wave characteristics. Do known
amplitudes compare favourably with model values in the upper thermosphere? Are wavelengths suffi-
ciently long/short in schemes to match observations? Refining wave inputs will improve model outputs.
Dust Storms
Dust storms have been investigated with emphasis on the response of individual species and gravity waves
to such grand events. The former has been explored in previous studies; however, an in-depth analysis of
190
8.2. FUTURE WORK
diffusion during a dust storm has not been performed. Diffusion coefficients can be determined by com-
paring observed and model density profiles for measured species. This will further our understanding of
atmospheric responses to dust storms. Also, gravity wave schemes can be tested under dust storm con-
ditions. Do they accurately predict wave activity? If not, which parameters need to tuning to rectify this?
The rarity of global dust storm events made studying the 2018 event a new priority for this thesis.
While comparisons have been drawn between global dust storms both in the upper and lower atmosphere
results are derived using independent methods outlined in various studies. A more rigorous overview
would benefit from implementing the same technique on each dataset to study each storm. Further,
as asserted by Cantor, 2007, over 5000 local dust storms were identified in MOC global maps. A
worthwhile, yet potentially interminable task, is mining through lower and upper atmosphere data and,
where possible, performing the same analyses on each dataset. Decay rates would be studied such that
the effects of severity, location, and altitude can be explored. Cantor et al., 2019 identified 93 regional
storms within the Mars 2020 landing site within the last ∼6 years. It is probable that, if possible to
study these storms, they would reveal interesting traits. This could inform dust storm models. Are the
decay rates and postulations presented in Chapter 6 repeatable? Can we get an early indicator of the
duration of the global dust storm? And do global and regional dust storms only differ in size and length?
Can results from both be combined into one dataset?
Mars Climate Database Version 6.0
A new version (v6.0) of the Mars Climate Database was due for release in October 2019 Millour et
al., 2019a. The first benefit of this updated model is the best estimate of conditions during MY34,
which is currently unavailable. This is the best representation of daily atmospheric dust loading and
daily solar EUV input. This scenario spans May 2017-March 2019, which includes ∼80% of TGO’s
aerobraking period. The implementation of this scenario should improve the comparisons. Additionally,
the dust cycle has been updated. Purely vertically propagating gravity waves initiated in the MCD were
utilised in Chapter 5 with the examination of spacecraft trajectory on measured characteristics. v6.0
will include horizontally propagating gravity waves. And with the addition of this feature, this study can
be expanded. Can horizontal and vertical wave components be determined with knowledge of apparent
wavelengths and spacecraft trajectory? From this, can the propagation of each wave be calculated and
potentially reproduced with a gravity wave scheme?
191
8.2. FUTURE WORK
Utilisation of Previous and Upcoming Missions
Throughout this study, where possible, direct comparisons have been made with previous results, e.g.
similarities between MAVEN and MGS gravity waves. A proposed extension of this study is to conflate
dataset, as suggested in Chapter 7, to understand trends further whether this be temperature, density
or gravity wave behaviour. Previous studies have combined the results of MGS, ODY and MRO (e.g.
Keating et al., 2003; Bougher et al., 2006; Withers, 2006). These can be further extended with MAVEN
data expanding the sampled region in all parameters - latitude, LST, altitude and season amongst others
- such that initially entangled trends can be decoupled. Further, future aerobraking missions will only
extensively add to the datasets already available. Both similar and new analyses can be performed to
fill in gaps in our understanding, such as simply investigating regions currently unsampled. One such
mission is Mars Orbiter Mission 2 (launch date unknown; Haider et al., 2018). Can a global overview
of density and temperature structure be assembled using data from all thermosphere spacecraft? How
does this picture compare to model outputs? Can occultation data from TGO be combined with in-situ
MAVEN data to create further hydrostatically connected temperature and density profiles?
As the above shows, there are still myriad topics for the next generation to undertake. Now, more
than ever is a fascinating time to be studying the Red Planet. Techniques presented in this thesis do
not apply only to Mars, but all studied planetary atmospheres.
192
Appendices
193
Appendix A
Visualisation of Regions Sampled

Multiple Times by MAVEN
Figure A.0.1 shows the regions in latitude and local time which have been sampled twice (or more) by
MAVEN as listed in Tables 4.6.1-4.6.3 in Chapter 4. Boxes show the minimum and maximum local
time and latitude values. Each box contains two colours which relate to the two Ls values Mars was at
whilst MAVEN sampled the region.
Figure A.0.1: Graphical representation of coverage overlap detailed in Tables 4.6.1,4.6.2 and 4.6.3.
Boxes show the latitude and local time coverage of each overlap scenario. The two colours in each show
the two seasons the overlap periods occurred in.
194
Appendix B
Deriving Lower Atmosphere Heating

Caused by a Dust Storm
This section outlines the derivation for an estimation of dust storm warming in the lower atmosphere as
shown in Chapter 7 and Chaufray et al., 2019.
Assume two density distributions - one prior to the dust storm denoted by subscript 1 and one
post-onset denoted by subscript 2. These are given by:
∆z ∆z

n1 ∝ exp − n2 ∝ exp −
hHi1 hHi2
where n1 is the density above the heated region ∆ z prior to the storm, hHi1 is the average scale
height prior to the onset within the heated region. Likewise, n2 is the density above the heated region
post-onset, hHi2 is the average scale height after the storm onset.
Then,

∆z
n2 exp hHi1
=
n1 ∆z
exp hHi2
which can be simplified to,
n2 ∆z ∆z

= exp −
n1 hHi1 hHi2
195
Taking the natural logarithm of both sides leads to,
n2 ∆z ∆z

ln = −
n1 hHi1 hHi2
hHi1
Multiplying each term by ∆z results in,
hHi1 n2 hHi1

ln =1−
∆z n1 hHi2
Rearranging the above equation leads to,
hHi1 n2 hHi1

1− ln =
∆z n1 hHi2
And finally, reciprocating the above equation results in,
hHi2 1
= hHi1

hHi1 1− ln n2
∆Z n1
196
Appendix C
Software Used
I am indebted to the below communities who have developed wondrous Python packages that have
made this PhD possible.
• Python - programming language (https://www.python.org/)
• Matplotlib - a Python library for plotting (https://matplotlib.org/)
• Pandas - a Python library for data science (https://pandas.pydata.org/)
• NumPy - a Python library for array manipulation (https://docs.scipy.org/doc/numpy/)
• SciPy - a Python library for scientific computing and technical computing (https://docs.
scipy.org/doc/scipy/reference/)
• seaborn - a Python library for statistical data visualization (https://seaborn.pydata.org/

index.html)
197
Appendix D
Data Used
The various datasets used in this thesis can be found in the following locations:
• MAVEN Accelerometer Data Level 3 - https://pds-atmospheres.nmsu.edu/data_and_services/

atmospheres_data/MAVEN/maven_acc.html
• MAVEN Neutral Gas and Ion Mass Spectrometer Data Level 3 - https://pds-atmospheres.
nmsu.edu/data_and_services/atmospheres_data/MAVEN/ngims.html
– With special thanks to Roger Yelle and Shane Stone from LPL, University of Arizona for
providing initial data and helping with the interpretation of results.
• TGO Accelerometer Data - derived from raw accelerometers counts.
– With further thanks to Sean Bruinsma and Jean-Charles Marty from CNES, Toulouse for
deriving density data and helping with the interpretation of results.
198
Bibliography
Aboudan, A., G. Colombatti, C. Bettanini, F. Ferri, S. Lewis, B. Van Hove, O. Karatekin and S. Debei
(2018). ‘ExoMars 2016 Schiaparelli Module Trajectory and Atmospheric Profiles Reconstruction’. In:
Space Science Reviews 214.5, p. 97.
Allison, M. and R. Schmunk (2018). Technical Notes on Mars Solar Time as Adopted by the Mars24
Sunclock.
Ando, H., T. Imamura and T. Tsuda (2012). ‘Vertical wavenumber spectra of gravity waves in the
Martian atmosphere obtained from Mars Global Surveyor radio occultation data’. In: Journal of the
Atmospheric Sciences 69.9, pp. 2906–2912.
Bagnold, R. A. (1941). The physics of blown sand and desert dunes. Courier Corporation.
Barnes, J. R. (1990). ‘Possible effects of breaking gravity waves on the circulation of the middle atmo-
sphere of Mars’. In: Journal of Geophysical Research 95, pp. 1401–1421.
Barnes, J. (1999). ‘Initiation and Spread of Martian Dust Storms’. In: The Fifth International Conference
on Mars.
Bell, J. M., S. W. Bougher and J. R. Murphy (2007). ‘Vertical dust mixing and the interannual variations
in the Mars thermosphere’. In: Journal of Geophysical Research: Planets 112.E12.
Benna, M., S. W. Bougher, Y. Lee, K. J. Roeten, E. Yiğit, P. R. Mahaffy and B. M. Jakosky (2019).
‘Global circulation of Mars’ upper atmosphere’. In: Science 366.6471, 1363 LP - 1366.
Benna, M. and M. Elrod (2018). NGIMS Planetary Data System software interface specification (MAVEN-
NGIMS-SIS-0001). Tech. rep.
Beswick, R., S. Demcak, B. Young, S. E. McCandless, J. R. Carpenter, R. Burns and B. Jakosky (2020).
‘MAVEN Orbital Trajectory Analysis: Design and Implementation of Lander Relay Support’. In: AIAA
Scitech 2020 Forum. AIAA SciTech Forum. American Institute of Aeronautics and Astronautics.
Bougher, S. W. (1995). ‘Comparative thermospheres: Venus and Mars’. In: Advances in Space Research
15.4, pp. 21–45.
199
BIBLIOGRAPHY
Bougher, S. W., J. M. Bell, J. R. Murphy, M. A. Lopez-Valverde and P. G. Withers (2006). ‘Polar

warming in the Mars thermosphere: Seasonal variations owing to changing insolation and dust
distributions’. In: Geophysical Research Letters 33.2.
Bougher, S. W. and R. E. Dickinson (1988). ‘Mars mesosphere and thermosphere: 1. Global mean heat
budget and thermal structure’. In: Journal of Geophysical Research: Space Physics 93.A7, pp. 7325–
7337.
Bougher, S. W., R. E. Dickinson, E. C. Ridley and R. G. Roble (1988a). ‘Venus mesosphere and ther-
mosphere’. In: Icarus 73.3, pp. 545–573.
— (1988b). ‘Venus mesosphere and thermosphere: III. Three-dimensional general circulation with coupled
dynamics and composition’. In: Icarus 73.3, pp. 545–573.
Bougher, S. W., R. E. Dickinson, R. G. Roble and E. C. Ridley (1988c). ‘Mars thermospheric general
circulation model: Calculations for the arrival of Phobos at Mars’. In: Geophysical Research Letters
15.13, pp. 1511–1514.
Bougher, S. W., S. Engel, R. G. Roble and B. Foster (1999). ‘Comparative terrestrial planet thermo-
spheres: 2. Solar cycle variation of global structure and winds at equinox’. In: Journal of Geophysical
Research: Planets 104.E7, pp. 16591–16611.
— (2000). ‘Comparative terrestrial planet thermospheres: 3. Solar cycle variation of global structure
and winds at solstices’. In: Journal of Geophysical Research: Planets 105.E7, pp. 17669–17692.
Bougher, S. W., D. M. Hunten and R. G. Roble (1994). ‘CO2 cooling in terrestrial planet thermospheres’.
In: Journal of Geophysical Research 99, pp. 14609–14622.
Bougher, S. W., J. Murphy and R. M. Haberle (1997). ‘Dust storm impacts on the Mars upper atmo-
sphere’. In: Advances in Space Research 19.8, pp. 1255–1260.
Bougher, S. W., D. Pawlowski, J. M. Bell, S. Nelli, T. McDunn, J. R. Murphy, M. Chizek and A. Ridley
(2015). ‘Mars Global Ionosphere-Thermosphere Model: Solar cycle, seasonal, and diurnal variations
of the Mars upper atmosphere’. In: Journal of Geophysical Research: Planets 120.2, pp. 311–342.
Bougher, S. W. and R. G. Roble (1991). ‘Comparative terrestrial planet thermospheres: 1. Solar cycle
variation of global mean temperatures’. In: Journal of Geophysical Research: Space Physics 96.A7,
pp. 11045–11055.
Bougher, S. W., R. G. Roble, E. C. Ridley and R. E. Dickinson (1990). ‘The Mars thermosphere: 2.
General circulation with coupled dynamics and composition’. In: Journal of Geophysical Research:
Solid Earth 95.B9, pp. 14811–14827.
200
BIBLIOGRAPHY
Bougher, S. W., K. J. Roeten, K. Olsen, P. R. Mahaffy, M. Benna, M. Elrod, S. K. Jain, N. M. Schneider,

J. Deighan, E. Thiemann, F. G. Eparvier, A. Stiepen and B. M. Jakosky (2017). ‘The structure and
variability of Mars dayside thermosphere from MAVEN NGIMS and IUVS measurements: Seasonal
and solar activity trends in scale heights and temperatures’. In: Journal of Geophysical Research:
Space Physics 122.1, pp. 1296–1313.
Bougher, S., S. Engel, D. P. Hinson and J. M. Forbes (2001). ‘Mars Global Surveyor radio science
electron density profiles : Neutral atmosphere implications’. In: Geophysical Research Letters 28.16,
pp. 3091–3094.
Bruinsma, S., D. Tamagnan and R. Biancale (2004). ‘Atmospheric densities derived from CHAMP/STAR
accelerometer observations’. In: Planetary and Space Science 52.4, pp. 297–312.
Cantor, B. A. (2007). ‘MOC observations of the 2001 Mars planet-encircling dust storm’. In: Icarus
186.1, pp. 60–96.
Cantor, B. A., N. B. Pickett, M. C. Malin, S. W. Lee, M. J. Wolff and M. A. Caplinger (2019). ‘Martian
dust storm activity near the Mars 2020 candidate landing sites: MRO-MARCI observations from
Mars years 28–34’. In: Icarus 321, pp. 161–170.
Chaufray, J.-Y., M. Chaffin, J. Deighan, S. Jain, N. Schneider, M. Mayyasi and B. Jakosky (2019).
‘Effect of the 2018 Martian global dust storm on the CO2 density in the lower nightside thermosphere
observed from MAVEN/IUVS Lyman-alpha absorption’. In: Geophysical Research Letters.
Clette, F., L. Svalgaard, J. M. Vaquero and E. W. Cliver (2014). ‘Revisiting the sunspot number’. In:
Space Science Reviews 186.1-4, pp. 35–103.
Creasey, J. E., J. M. Forbes and G. M. Keating (2006a). ‘Density variability at scales typical of gravity
waves observed in Mars’ thermosphere by the MGS accelerometer’. In: Geophysical Research Letters
33.22.
Creasey, J. E., J. M. Forbes and D. P. Hinson (2006b). ‘Global and seasonal distribution of gravity
wave activity in Mars’ lower atmosphere derived from MGS radio occultation data’. In: Geophysical
Research Letters 33.1.
Das, S. K., A. Taori and A. Jayaraman (2011). ‘On the role of dust storms in triggering atmospheric grav-
ity waves observed in the middle atmosphere.’ In: Annales Geophysicae (09927689) 29.9, pp. 1647–
1654.
Davies, D. W. (1979). ‘Effects of dust on the heating of Mars’ surface and atmosphere’. In: Journal of
Geophysical Research: Solid Earth 84.B14, pp. 8289–8293.
201
BIBLIOGRAPHY
Deepshikha, S., S. K. Satheesh and J. Srinivasan (2006). ‘Dust aerosols over India and adjacent con-
tinents retrieved using METEOSAT infrared radiance Part I: sources and regional distribution’. In:
Annales Geophysicae. Vol.∼24. 1, pp. 37–61.
Dickinson, R. E., E. C. Ridley and R. G. Roble (1984). ‘Thermospheric general circulation with coupled
dynamics and composition’. In: Journal of the Atmospheric Sciences 41.2, pp. 205–219.
Eliassen, A. and E. Palm (1961). ‘On the transfer of energy in stationary mountain waves’. In: Geofysiske
Publikasjoner 22(3), pp. 1–23.
Elrod, M. K., S. W. Bougher, K. Roeten, R. Sharrar and J. Murphy (2019). ‘Structural and compositional
changes in the upper atmosphere related to the PEDE-2018 dust event on Mars as observed by
MAVEN NGIMS’. In: Geophysical Research Letters.
Elrod, M. K., S. Bougher, J. Bell, P. R. Mahaffy, M. Benna, S. Stone, R. Yelle and B. Jakosky (2017).
‘He bulge revealed: He and CO2 diurnal and seasonal variations in the upper atmosphere of Mars as
detected by MAVEN NGIMS’. In: Journal of Geophysical Research: Space Physics 122.2, pp. 2564–
2573.
England, S. L., G. Liu, E. Yiğit, P. R. Mahaffy, M. Elrod, M. Benna, H. Nakagawa, N. Terada and
B. Jakosky (2016). ‘MAVEN NGIMS Observations of Atmospheric Gravity Waves in the Martian
thermosphere’. In: Journal of Geophysical Research: Space Physics 122.2, pp. 2310–2335.
Evans, J. S., M. H. Stevens, J. D. Lumpe, N. M. Schneider, A. I. F. Stewart, J. Deighan, S. K. Jain, M. S.
Chaffin, M. Crismani and A. Stiepen (2015). ‘Retrieval of CO2 and N2 in the Martian thermosphere
using dayglow observations by IUVS on MAVEN’. In: Geophysical Research Letters 42.21, pp. 9040–
9049.
Forbes, J. M., S. Bruinsma and F. G. Lemoine (2006). ‘Solar rotation effects on the thermospheres of
Mars and Earth’. In: Science 312.5778, pp. 1366–1368.
Forget, F., F. Hourdin, R. Fournier, C. Hourdin, O. Talagrand, M. Collins, S. R. Lewis, P. L. Read and
J.-P. Huot (1999). ‘Improved general circulation models of the Martian atmosphere from the surface
to above 80 km’. In: Journal of Geophysical Research 104, pp. 24155–24176.
Forget, F., F. Montmessin, J.-L. Bertaux, F. González-Galindo, S. Lebonnois, E. Quémerais, A. Reberac,
E. Dimarellis and M. A. López-Valverde (2009). ‘Density and temperatures of the upper Martian
atmosphere measured by stellar occultations with Mars Express SPICAM’. In: Journal of Geophysical
Research: Planets 114.E1.
Fox, J. L. (2004). ‘Advances in the aeronomy of Venus and Mars’. In: Advances in Space Research 33.2,
pp. 132–139.
202
BIBLIOGRAPHY
Fritts, D. C. (1989). ‘A review of gravity wave saturation processes, effects, and variability in the middle
atmosphere’. In: Pure and Applied Geophysics 130.2, pp. 343–371.
Fritts, D. C. and M. J. Alexander (2003). ‘Gravity wave dynamics and effects in the middle atmosphere’.
In: Reviews of Geophysics 41.1.
Fritts, D. C., L. Wang and R. H. Tolson (2006). ‘Mean and gravity wave structures and variability
in the Mars upper atmosphere inferred from Mars Global Surveyor and Mars Odyssey aerobraking
densities’. In: Journal of Geophysical Research: Space Physics 111.A12.
Fulchignoni, M., F. Ferri, F. Angrilli, A. J. Ball, A. Bar-Nun, M. A. Barucci, C. Bettanini, G. Bianchini,
W. Borucki and G. Colombatti (2005). ‘In situ measurements of the physical characteristics of Titan’s
environment’. In: Nature 438.7069, p. 785.
Garcia, R. F., P. Drossart, G. Piccioni, M. López-Valverde and G. Occhipinti (2009). ‘Gravity waves in
the upper atmosphere of Venus revealed by CO2 nonlocal thermodynamic equilibrium emissions’.
In: Journal of Geophysical Research: Planets 114.E5.
Gardner, L. C. and R. W. Schunk (2011). ‘Large-scale gravity wave characteristics simulated with a
high-resolution global thermosphere-ionosphere model’. In: Journal of Geophysical Research: Space
Physics 116.A6.
Girazian, Z., Z. Luppen, D. D. Morgan, F. Chu, L. Montabone, E. M. B. Thiemann, D. A. Gurnett,
J. Halekas, A. J. Kopf and F. Němec (2019). ‘Variations in the Ionospheric Peak Altitude at Mars
in Response to Dust Storms: 13 Years of Observations from the Mars Express Radar Sounder’. In:
Journal of Geophysical Research: Planets.
Girazian, Z., P. Mahaffy, M. Benna, M. Elrod, D. Drob and S. Bougher (2017). ‘A Global Empirical
Model Of The Thermosphere Of Mars Based On In Situ Mass Spectrometer Data’. In: The Mars
Atmosphere: Modelling and Observation, p. 3312.
González-Galindo, F., S. W. Bougher, M. A. López-Valverde, F. Forget and J. Murphy (2010). ‘Thermal
and wind structure of the Martian thermosphere as given by two General Circulation Models’. In:
Planetary and Space Science 58.14-15, pp. 1832–1849.
González-Galindo, F., M. A. López-Valverde, F. Forget, M. García-Comas, E. Millour and L. Montabone
(2015). ‘Variability of the Martian thermosphere during eight Martian years as simulated by a ground-
to-exosphere global circulation model’. In: Journal of Geophysical Research (Planets) 120, pp. 2020–
2035.
Greeley, R. (2002). ‘Saltation impact as a means for raising dust on Mars’. In: Planetary and Space
Science 50.2, pp. 151–155.
203
BIBLIOGRAPHY
Gröller, H., F. Montmessin, R. V. Yelle, F. Lefèvre, F. Forget, N. M. Schneider, T. T. Koskinen, J.

Deighan and S. K. Jain (2018). ‘MAVEN/IUVS stellar occultation measurements of Mars atmo-
spheric structure and composition’. In: Journal of Geophysical Research: Planets 123.6, pp. 1449–
1483.
Guerlet, S., R. Young, N. Ignatiev, F. Forget, E. Millour, T. Fouchet, L. Montabone, A. Shakun, A.
Grigoriev, A. Trokhimovskiy, F. Montmessin and O. Korablev (2019). ‘Investigation of the 2018
Global Dust Event from ACS-TIRVIM on board ExoMars/TGO’. In: EPSC-DPS Joint Meeting
2019.
Guo, J., W. Wan, J. M. Forbes, E. Sutton, R. S. Nerem, T. N. Woods, S. Bruinsma and L. Liu (2007).
‘Effects of solar variability on thermosphere density from CHAMP accelerometer data’. In: Journal
of Geophysical Research: Space Physics 112.A10.
Gurwell, M. A., E. A. Bergin, G. J. Melnick and V. Tolls (2005). ‘Mars surface and atmospheric tem-
perature during the 2001 global dust storm’. In: Icarus 175.1, pp. 23–31.
Guzewich, S. D., M. Lemmon, C. L. Smith, G. Martínez, Á. Vicente-Retortillo, E. Newman C, M.
Baker, C. Campbell, B. Cooper, J. Gómez-Elvira, A. M. Harri, D. Hassler, F. J. Martin-Torres, T.
McConnochie, J. E. Moores, H. Kahanpää, A. Khayat, I. Richardson M, D. Smith M, R. Sullivan,
M. Torre Juarez, A. R. Vasavada, D. Viúdez-Moreiras, C. Zeitlin and M. Maria-Paz Zorzano (2019).
‘Mars Science Laboratory Observations of the 2018/Mars Year 34 Global Dust Storm’. In: Geophysical
Research Letters 46.1, pp. 71–79.
Haberle, R. M. (2015). ‘Solar System/Sun, Atmospheres, Evolution of Atmospheres | Planetary Atmo-
spheres: Mars’. In: Encyclopedia of Atmospheric Sciences (Second Edition). Ed. by G. R. North, J.
Pyle and F. B. T. Zhang. Oxford: Academic Press, pp. 168–177.
Haberle, R. M., M. M. Joshi, J. R. Murphy, J. R. Barnes, J. T. Schofield, G. Wilson, M. Lopez-Valverde,
J. L. Hollingsworth, A. F. C. Bridger and J. Schaeffer (1999). ‘General circulation model simula-
tions of the Mars Pathfinder atmospheric structure investigation/meteorology data’. In: Journal of
Geophysical Research: Planets 104.E4, pp. 8957–8974.
Haberle, R. M., C. B. Leovy and J. B. Pollack (1982). ‘Some effects of global dust storms on the
atmospheric circulation of Mars’. In: Icarus 50.2-3, pp. 322–367.
Haider, S. A., A. Bhardwaj, M. Shanmugam, S. K. Goyal, V. Sheel, J. Pabari and D. Prasad Karanam
(2018). ‘Indian Mars and Venus Missions: Science and Exploration’. In: 42nd COSPAR Scientific
Assembly. Vol.∼42, pp. 1–10.
204
BIBLIOGRAPHY
Hall, L. A. and H. E. Hinteregger (1970). ‘Solar radiation in the extreme ultraviolet and its variation
with solar rotation’. In: Journal of Geophysical Research 75.34, pp. 6959–6965.
Hartmann, W. K., M. Malin, A. McEwen, M. Carr, L. Soderblom, P. Thomas, E. Danielson, P. James
and J. Veverka (1999). ‘Evidence for recent volcanism on Mars from crater counts’. In: Nature
397.6720, p. 586.
Hartogh, P., A. S. Medvedev, T. Kuroda, R. Saito, G. Villanueva, A. G. Feofilov, A. A. Kutepov and
U. Berger (2005). ‘Description and climatology of a new general circulation model of the Martian
atmosphere’. In: Journal of Geophysical Research: Planets 110.E11.
Heavens, N. G., M. I. Richardson, A. Kleinböhl, D. M. Kass, D. J. McCleese, W. Abdou, J. L. Benson,
J. T. Schofield, J. H. Shirley and P. M. Wolkenberg (2011). ‘The vertical distribution of dust in the
Martian atmosphere during northern spring and summer: Observations by the Mars Climate Sounder
and analysis of zonal average vertical dust profiles’. In: Journal of Geophysical Research: Planets
116.E4.
Hedin, A. E., H. B. Niemann, W. T. Kasprzak and A. Seiff (1983). ‘Global empirical model of the Venus
thermosphere’. In: Journal of Geophysical Research: Space Physics 88.A1, pp. 73–83.
Hinson, D. P. and J. M. Jenkins (1995). ‘Magellan Radio Occultation Measurements of Atmospheric
Waves on Venus’. In: Icarus 114.2, pp. 310–327.
Hinson, D. P. and J. A. Magalhães (1993). ‘Inertio-Gravity Waves in the Atmosphere of Neptune’. In:
Icarus 105.1, pp. 142–161.
Hinson, D. P. and G. Tyler (1983). ‘Internal gravity waves in Titan’s atmosphere observed by Voyager
radio occultation’. In: Icarus 54.2, pp. 337–352.
Imamura, T., A. Watanabe and Y. Maejima (2016). ‘Convective generation and vertical propagation of
fast gravity waves on Mars: One- and two-dimensional modeling’. In: Icarus 267, pp. 51–63.
Jain, S. K., S. W. Bougher, J. Deighan, N. M. Schneider, F. González-Galindo, A. I. F. Stewart, R.
Sharrar, D. Kass, J. Murphy and D. Pawlowski (2020). ‘Martian thermospheric warming associated
with the Planet Encircling Dust Event of 2018’. In: Geophysical Research Letters, e60064.
Jakosky, B. M., J. M. Grebowsky, J. G. Luhmann, J. Connerney, F. Eparvier, R. Ergun, J. Halekas,
D. Larson, P. Mahaffy, J. McFadden, D. L. Mitchell, N. Schneider, R. Zurek, S. Bougher, D. Brain,
Y. J. Ma, C. Mazelle, L. Andersson, D. Andrews, D. Baird, D. Baker, J. M. Bell, M. Benna, M.
Chaffin, P. Chamberlin, Y.-Y. Chaufray, J. Clarke, G. Collinson, M. Combi, F. Crary, T. Cravens,
M. Crismani, S. Curry, D. Curtis, J. Deighan, G. Delory, R. Dewey, G. DiBraccio, C. Dong, Y. Dong,
P. Dunn, M. Elrod, S. England, A. Eriksson, J. Espley, S. Evans, X. Fang, M. Fillingim, K. Fortier,
205
BIBLIOGRAPHY
C. M. Fowler, J. Fox, H. Gröller, S. Guzewich, T. Hara, Y. Harada, G. Holsclaw, S. K. Jain, R. Jolitz,

F. Leblanc, C. O. Lee, Y. Lee, F. Lefevre, R. Lillis, R. Livi, D. Lo, M. Mayyasi, W. McClintock, T.
McEnulty, R. Modolo, F. Montmessin, M. Morooka, A. Nagy, K. Olsen, W. Peterson, A. Rahmati,
S. Ruhunusiri, C. T. Russell, S. Sakai, J.-A. Sauvaud, K. Seki, M. Steckiewicz, M. Stevens, A. I. F.
Stewart, A. Stiepen, S. Stone, V. Tenishev, E. Thiemann, R. Tolson, D. Toublanc, M. Vogt, T.
Weber, P. Withers, T. Woods and R. Yelle (2015a). ‘MAVEN observations of the response of Mars
to an interplanetary coronal mass ejection’. In: Science 350.6261.
Jakosky, B. M., R. P. Lin, J. M. Grebowsky, J. G. Luhmann, D. F. Mitchell, G. Beutelschies, T. Priser, M.
Acuna, L. Andersson, D. Baird, D. Baker, R. Bartlett, M. Benna, S. Bougher, D. Brain, D. Carson,
S. Cauffman, P. Chamberlin, J.-Y. Chaufray, O. Cheatom, J. Clarke, J. Connerney, T. Cravens, D.
Curtis, G. Delory, S. Demcak, A. DeWolfe, F. Eparvier, R. Ergun, A. Eriksson, J. Espley, X. Fang,
D. Folta, J. Fox, C. Gomez-Rosa, S. Habenicht, J. Halekas, G. Holsclaw, M. Houghton, R. Howard,
M. Jarosz, N. Jedrich, M. Johnson, W. Kasprzak, M. Kelley, T. King, M. Lankton, D. Larson, F.
Leblanc, F. Lefevre, R. Lillis, P. Mahaffy, C. Mazelle, W. McClintock, J. McFadden, D. L. Mitchell,
F. Montmessin, J. Morrissey, W. Peterson, W. Possel, J.-A. Sauvaud, N. Schneider, W. Sidney,
S. Sparacino, A. I. F. Stewart, R. Tolson, D. Toublanc, C. Waters, T. Woods, R. Yelle and R.
Zurek (2015b). ‘The Mars Atmosphere and Volatile Evolution (MAVEN) Mission’. In: Space Science
Reviews 195.1, pp. 3–48.
Jakosky, B. M., M. Slipski, M. Benna, P. Mahaffy, M. Elrod, R. Yelle, S. Stone and N. Alsaeed (2017).
‘Mars’ atmospheric history derived from upper-atmosphere measurements of 38Ar/36Ar’. In: Science
355.6332, 1408 LP - 1410.
Jakosky, B. M. and T. Z. Martin (1987). ‘Mars: North-polar atmospheric warming during dust storms’.
In: Icarus 72.3, pp. 528–534.
Jesch, D., A. S. Medvedev, F. Castellini, E. Yiğit and P. Hartogh (2019). ‘Density Fluctuations in the
Lower Thermosphere of Mars Retrieved From the ExoMars Trace Gas Orbiter (TGO) Aerobraking’.
In: Atmosphere 10.10, p. 620.
Kass, D. M., J. T. Schofield, A. Kleinböhl, D. J. McCleese, N. G. Heavens, J. H. Shirley and L. J.
Steele (2019). ‘Mars Climate Sounder observation of Mars’ 2018 global dust storm’. In: Geophysical
Research Letters.
Keating, G. M., S. W. Bougher, R. W. Zurek, R. H. Tolson, G. J. Cancro, S. N. Noll, J. S. Parker,
T. J. Schellenberg, R. W. Shane, B. L. Wilkerson, J. R. Murphy, J. L. Hollingsworth, R. M. Haberle,
M. Joshi, J. C. Pearl, B. J. Conrath, M. D. Smith, R. T. Clancy, R. C. Blanchard, R. G. Wilmoth,
206
BIBLIOGRAPHY
D. F. Rault, T. Z. Martin, D. T. Lyons, P. B. Esposito, M. D. Johnston, C. W. Whetzel, C. G. Justus

and J. M. Babicke (1998). ‘The Structure of the Upper Atmosphere of Mars: In Situ Accelerometer
Measurements from Mars Global Surveyor’. In: Science 279.5357, pp. 1672–1676.
Keating, G. M., M. Theriot, R. Tolson, S. Bougher, F. Forget and J. Forbes (2003). ‘Brief review on the
results obtained with the MGS and Mars Odyssey 2001 Accelerometer Experiments’. In: Proceedings,
International workshop: Mars atmosphere modeling and observations, Granada, Spain, pp. 13–15.
Korablev, O., F. Montmessin, A. Trokhimovskiy, A. A. Fedorova, A. V. Shakun, A. V. Grigoriev, B. E.
Moshkin, N. I. Ignatiev, F. Forget and F. Lefèvre (2018). ‘The atmospheric chemistry suite (ACS)
of three spectrometers for the ExoMars 2016 Trace Gas Orbiter’. In: Space Science Reviews 214.1,
p. 7.
Korablev, O. et al. (2019). ‘No detection of methane on Mars from early ExoMars Trace Gas Orbiter
observations’. In: Nature 568.7753, pp. 517–520.
Kuroda, T., A. S. Medvedev, E. Yiğit and P. Hartogh (2015). ‘A global view of gravity waves in
the Martian atmosphere inferred from a high-resolution general circulation model’. In: Geophysical
Research Letters 42.21, pp. 9213–9222.
Lee, C. O., T. Hara, J. S. Halekas, E. Thiemann, P. Chamberlin, F. Eparvier, R. J. Lillis, D. E. Larson,
P. A. Dunn and J. R. Espley (2017). ‘MAVEN observations of the solar cycle 24 space weather
conditions at Mars’. In: Journal of Geophysical Research: Space Physics 122.3, pp. 2768–2794.
Leighton, R. B., B. C. Murray, R. P. Sharp, J. D. Allen and R. K. Sloan (1965). ‘Mariner IV Photography
of Mars: Initial Results’. In: Science 149.3684, pp. 627–630.
Lemmon, M. T., M. J. Wolff, J. F. Bell, M. D. Smith, B. A. Cantor and P. H. Smith (2015). ‘Dust
aerosol, clouds, and the atmospheric optical depth record over 5 Mars years of the Mars Exploration
Rover mission’. In: Icarus 251, pp. 96–111.
Leovy, C. B., G. A. Briggs, A. T. Young, B. A. Smith, J. B. Pollack, E. N. Shipley and R. L. Wildey
(1972). ‘The martian atmosphere: Mariner 9 television experiment progress report’. In: Icarus 17.2,
pp. 373–393.
Leovy, C. and Y. Mintz (1969). ‘Numerical Simulation of the Atmospheric Circulation and Climate of
Mars’. In: Journal of the Atmospheric Sciences 26.6, pp. 1167–1190.
Lewis, S. R. (2003). ‘Modelling the martian atmosphere’. In: Astronomy & Geophysics 44.4, pp. 6–4.
Lewis, S. R., M. Collins, P. L. Read, F. Forget, F. Hourdin, R. Fournier, C. Hourdin, O. Talagrand
and J.-P. Huot (1999). ‘A climate database for Mars’. In: Journal of Geophysical Research: Planets
104.E10, pp. 24177–24194.
207
BIBLIOGRAPHY
Lindzen, R. S. (1981). ‘Turbulence and stress owing to gravity wave and tidal breakdown’. In: Journal
of Geophysical Research: Oceans 86.C10, pp. 9707–9714.
Liu, G., S. L. England, R. J. Lillis, P. Withers, P. R. Mahaffy, D. E. Rowland, M. Elrod, M. Benna,
D. M. Kass and D. Janches (2018). ‘Thermospheric expansion associated with dust increase in the
lower atmosphere on Mars observed by MAVEN/NGIMS’. In: Geophysical Research Letters 45.7,
pp. 2901–2910.
Liu, G., S. England, R. J. Lillis, P. R. Mahaffy, M. Elrod, M. Benna and B. Jakosky (2017). ‘Longitudinal
structures in Mars’ upper atmosphere as observed by MAVEN/NGIMS’. In: Journal of Geophysical
Research: Space Physics 122.1, pp. 1258–1268.
Liu, J., S. Jin and Y. Li (2019). ‘Seasonal Variations and Global Wave Distributions in the Mars
Thermosphere from MAVEN and Multi-Satellites Accelerometer-derived Mass Densities’. In: Journal
of Geophysical Research: Space Physics 124.11.
Lorenz, R. D., L. A. Young and F. Ferri (2014). ‘Gravity waves in Titan’s lower stratosphere from
Huygens probe in situ temperature measurements’. In: Icarus 227, pp. 49–55.
Lyons, D. T., J. G. Beerer, P. Esposito, M. D. Johnston and W. H. Willcockson (1999). ‘Mars Global
Surveyor: Aerobraking Mission Overview’. In: Journal of Spacecraft and Rockets 36.3, pp. 307–313.
Magalhães, J. A., J. T. Schofield and A. Seiff (1999). ‘Results of the Mars Pathfinder atmospheric
structure investigation’. In: Journal of Geophysical Research: Planets 104.E4, pp. 8943–8955.
Mahaffy, P. R., M. Benna, M. Elrod, R. V. Yelle, S. W. Bougher, S. W. Stone and B. M. Jakosky
(2015a). ‘Structure and composition of the neutral upper atmosphere of Mars from the MAVEN
NGIMS investigation’. In: Geophysical Research Letters 42.21, pp. 8951–8957.
Mahaffy, P. R., M. Benna, T. King, D. N. Harpold, R. Arvey, M. Barciniak, M. Bendt, D. Carrigan, T.
Errigo, V. Holmes, C. S. Johnson, J. Kellogg, P. Kimvilakani, M. Lefavor, J. Hengemihle, F. Jaeger,
E. Lyness, J. Maurer, A. Melak, F. Noreiga, M. Noriega, K. Patel, B. Prats, E. Raaen, F. Tan, E.
Weidner, C. Gundersen, S. Battel, B. P. Block, K. Arnett, R. Miller, C. Cooper, C. Edmonson and
J. T. Nolan (2015b). ‘The Neutral Gas and Ion Mass Spectrometer on the Mars Atmosphere and
Volatile Evolution Mission’. In: Space Science Reviews 195.1, pp. 49–73.
Mahaffy, P. R., C. R. Webster, S. K. Atreya, H. Franz, M. Wong, P. G. Conrad, D. Harpold, J. J. Jones,
L. A. Leshin, H. Manning, T. Owen, R. O. Pepin, S. Squyres and M. Trainer (2013). ‘Abundance
and Isotopic Composition of Gases in the Martian Atmosphere from the Curiosity Rover’. In: Science
341.6143. Ed. by O. Kemppinen et al., pp. 263–266.
208
BIBLIOGRAPHY
Malin, M. C., M. H. Carr, G. E. Danielson, M. E. Davies, W. K. Hartmann, A. P. Ingersoll, P. B. James,

H. Masursky, A. S. McEwen and L. A. Soderblom (1998). ‘Early views of the martian surface from
the Mars Orbiter Camera of Mars Global Surveyor’. In: Science 279.5357, pp. 1681–1685.
Malin, M. C. and K. S. Edgett (2000). ‘Evidence for Recent Groundwater Seepage and Surface Runoff
on Mars’. In: Science 288.5475, 2330 LP - 2335.
Martin, T. Z. (1979). ‘Global thermal behavior of Mars in dusty conditions, 1, The 15-# m band
measurements’. In: Journal of Geophysical Research 84, B6.
Mayyasi, M., C. Narvaez, M. Benna, M. Elrod and P. Mahaffy (2019). ‘Ion-Neutral Coupling in the
Upper Atmosphere of Mars: A Dominant Driver of Topside Ionospheric Structure’. In: Journal of
Geophysical Research: Space Physics 124.5, pp. 3786–3798.
McDunn, T. (2012). ‘On the Structure and Dynamics of the Martian Middle Atmosphere’. PhD thesis.
Medvedev, A. S., F. González-Galindo, E. Yiğit, A. G. Feofilov, F. Forget and P. Hartogh (2015). ‘Cooling
of the Martian thermosphere by CO2 radiation and gravity waves: An intercomparison study with
two general circulation models’. In: Journal of Geophysical Research: Planets 120.5, pp. 913–927.
Medvedev, A. S. and P. Hartogh (2007). ‘Winter polar warmings and the meridional transport on Mars
simulated with a general circulation model’. In: Icarus 186.1, pp. 97–110.
Medvedev, A. S., H. Nakagawa, C. Mockel, E. Yiğit, T. Kuroda, P. Hartogh, K. Terada, N. Terada, K.
Seki and N. M. Schneider (2016). ‘Comparison of the Martian thermospheric density and temperature
from IUVS/MAVEN data and general circulation modeling’. In: Geophysical Research Letters 43.7,
pp. 3095–3104.
Medvedev, A. S. and E. Yiğit (2012). ‘Thermal effects of internal gravity waves in the Martian upper
atmosphere’. In: Geophysical Research Letters 39.5.
Medvedev, A. S., E. Yiğit and P. Hartogh (2011). ‘Estimates of gravity wave drag on Mars: Indication
of a possible lower thermospheric wind reversal’. In: Icarus 211.1, pp. 909–912.
Medvedev, A. S., E. Yiğit, T. Kuroda and P. Hartogh (2013). ‘General circulation modeling of the
Martian upper atmosphere during global dust storms’. In: Journal of Geophysical Research: Planets
118.10, pp. 2234–2246.
Medvedev, S. A. and E. Yiğit (2019). ‘Gravity Waves in Planetary Atmospheres: Their Effects and
Parameterization in Global Circulation Models’. In: Atmosphere 10.9.
Mendillo, M., C. Narvaez, J. Trovato, P. Withers, M. Mayyasi, D. Morgan, A. Kopf, D. Gurnett, F.
Němec and B. Campbell (2018). ‘Mars Initial Reference Ionosphere (MIRI) Model: Updates and
209
BIBLIOGRAPHY
Validations Using MAVEN, MEX, and MRO Data Sets’. In: Journal of Geophysical Research: Space
Physics 123.7, pp. 5674–5683.
Michaels, T. I., A. Colaprete and S. C. R. Rafkin (2006). ‘Significant vertical water transport by
mountain-induced circulations on Mars’. In: Geophysical Research Letters 33.16.
Millour, E., F. Forget, A. Spiga, M. Vals, L. Zakharov V. Montabone, F. Lefèvre, F. Montmessin, J.-Y.
Chaufray, M. A. López-Valverde, F. González-Galindo, S. R. Lewis, P. L. Read, M.-C. Desjean and
F. Cipriani (2018). ‘The Mars Climate Database (Version 5.3)’. In: Scientific Workshop: “From Mars
Express to ExoMars”.
— (2019a). ‘The Mars Climate Database (version 6)’. In: European Planetary Science Congress 2019.
Millour, E., F. Forget, A. Spiga, M. Vals, V. Zakharov, L. Montabone, F. Lefevre, F. Montmessin, J.-Y.
Chauffray and M. A. Lopez-Valverde (2019b). ‘The Latest Mars Climate Database (Version 6.0)’.
In: LPI Contributions 2089.
Millour, E., F. Forget, A. Spiga, T. Navarro, J.-B. Madeleine, L. Montabone, A. Pottier, F. Lefèvre,
F. Montmessin and J.-Y. Chaufray (2015). ‘The Mars Climate Database (MCD version 5.2)’. In:
European Planetary Science Congress 2015. Vol.∼10. EPSC.
Mitrofanov, I., A. Malakhov, B. Bakhtin, D. Golovin, A. Kozyrev, M. Litvak, M. Mokrousov, A. Sanin,
V. Tretyakov, A. Vostrukhin, A. Anikin, L. M. Zelenyi, J. Semkova, S. Malchev, B. Tomov, Y.
Matviichuk, P. Dimitrov, R. Koleva, T. Dachev, K. Krastev, V. Shvetsov, G. Timoshenko, Y.
Bobrovnitsky, T. Tomilina, V. Benghin and V. Shurshakov (2018). ‘Fine Resolution Epithermal
Neutron Detector (FREND) Onboard the ExoMars Trace Gas Orbiter’. In: Space Science Reviews
214.5, p. 86.
Miyoshi, Y., J. M. Forbes and Y. Moudden (2011). ‘A new perspective on gravity waves in the Martian
atmosphere: Sources and features’. In: Journal of Geophysical Research: Planets 116.E9.
Moriyama, S. (1975). ‘Effects of Dust on Radiation Transfer in the Martian Atmosphere (II)’. In: Journal
of the Meteorological Society of Japan. Ser. II 53.3, pp. 214–221.
Moudden, Y. and J. M. Forbes (2011). ‘First detection of wave interactions in the middle atmosphere
of Mars’. In: Geophysical Research Letters 38.4.
Mueller-Wodarg, I. C. F., T. T. Koskinen, L. Moore, J. Serigano, R. V. Yelle, S. Hörst, J. H. Waite
and M. Mendillo (2019). ‘Atmospheric waves and their possible effect on the thermal structure of
Saturn’s thermosphere’. In: Geophysical Research Letters 46.5, pp. 2372–2380.
Mueller-Wodarg, I. C. F., D. F. Strobel, J. I. Moses, J. H. Waite, J. Crovisier, R. V. Yelle, S. W. Bougher
and R. G. Roble (2008). ‘Neutral Atmospheres’. In: Space Science Reviews 139.1-4, pp. 191–234.
210
BIBLIOGRAPHY
Mueller-Wodarg, I. C. F., R. V. Yelle, N. Borggren and J. H. Waite (2006). ‘Waves and horizontal
structures in Titan’s thermosphere’. In: Journal of Geophysical Research: Space Physics 111.A12.
Mueller-Wodarg, I. C. F., S. Bruinsma, J.-C. Marty and H. Svedhem (2016). ‘In situ observations of
waves in Venus’s polar lower thermosphere with Venus Express?aerobraking’. In: Nature Physics 12,
767 EP -.
Nier, A. O. and M. B. McElroy (1977). ‘Composition and structure of Mars’ upper atmosphere: Results
from the neutral mass spectrometers on Viking 1 and 2’. In: Journal of Geophysical Research 82.28,
pp. 4341–4349.
Parish, H. F., G. Schubert, M. P. Hickey and R. L. Walterscheid (2009). ‘Propagation of tropospheric
gravity waves into the upper atmosphere of Mars’. In: Icarus 203.1, pp. 28–37.
Park, J., H. Lühr, C. Lee, Y. H. Kim, G. Jee and J. Kim (2014). ‘A climatology of medium-scale gravity
wave activity in the midlatitude/low-latitude daytime upper thermosphere as observed by CHAMP’.
In: Journal of Geophysical Research: Space Physics 119.3, pp. 2187–2196.
Pätzold, M., B. Häusler, M. K. Bird, S. Tellmann, R. Mattei, S. W. Asmar, V. Dehant, W. Eidel, T.
Imamura, R. A. Simpson and G. L. Tyler (2007). ‘The structure of Venus’ middle atmosphere and
ionosphere’. In: Nature 450, pp. 657–660.
Pollack, J. B., D. S. Colburn, F. M. Flasar, R. Kahn, C. E. Carlston and D. Pidek (1979). ‘Properties and
effects of dust particles suspended in the Martian atmosphere’. In: Journal of Geophysical Research:
Solid Earth 84.B6, pp. 2929–2945.
Pollack, J. B., D. Colburn, R. Kahn, J. Hunter, W. Van Camp, C. E. Carlston and M. R. Wolf (1977).
‘Properties of aerosols in the Martian atmosphere, as inferred from Viking Lander imaging data’. In:
Journal of Geophysical Research 82.28, pp. 4479–4496.
Pollack, J. B., M. E. Ockert-Bell and M. K. Shepard (1995). ‘Viking Lander image analysis of Martian
atmospheric dust’. In: Journal of Geophysical Research: Planets 100.E3, pp. 5235–5250.
Qin, J. F., H. Zou, Y. G. Ye, Z. F. Yin, J. S. Wang and E. Nielsen (2019). ‘Effects of local dust storms on
the upper atmosphere of Mars: observations and simulations’. In: Journal of Geophysical Research:
Planets.
Rafkin, S. C. R. (2009). ‘A positive radiative-dynamic feedback mechanism for the maintenance and
growth of Martian dust storms’. In: Journal of Geophysical Research: Planets 114.E1.
Rafkin, S. C. R., R. M. Haberle and T. I. Michaels (2001). ‘The Mars Regional Atmospheric Modeling
System: Model Description and Selected Simulations’. In: Icarus 151.2, pp. 228–256.
211
BIBLIOGRAPHY
Ridley, A. J., Y. Deng and G. Toth (2006). ‘The global ionosphere–thermosphere model’. In: Journal of
Atmospheric and Solar-Terrestrial Physics 68.8, pp. 839–864.
Ryan, J. A. and R. M. Henry (1979). ‘Mars atmospheric phenomena during major dust storms, as
measured at surface’. In: Journal of Geophysical Research: Solid Earth 84.B6, pp. 2821–2829.
Ryan, J. A. and R. D. Sharman (1981). ‘Two major dust storms, one Mars year apart: Comparison from
Viking data’. In: Journal of Geophysical Research: Oceans 86.C4, pp. 3247–3254.
Sánchez-Lavega, A., T. del Río-Gaztelurrutia, J. Hernández-Bernal and M. Delcroix (2019). ‘The on-
set and growth of the 2018 Martian Global Dust Storm’. In: Geophysical Research Letters 46.11,
pp. 6101–6108.
Schunk, R. and A. Nagy (2009). Ionospheres: Physics, Plasma Physics, and Chemistry. 2nd∼ed. Cam-
bridge: Cambridge University Press.
Scuka, D. (2017). Aerobraking: Back to the Future.
Seiff, A. and D. B. Kirk (1977). ‘Structure of the atmosphere of Mars in summer at mid-latitudes’. In:
Journal of Geophysical Research 82.28, pp. 4364–4378.
Shirley, J. H. (2015). ‘Solar System dynamics and global-scale dust storms on Mars’. In: Icarus 251,
pp. 128–144.
Siddle, A. G., I. C. F. Mueller-Wodarg, S. Bruinsma and J.-C. Marty (2020). ‘Density structures in
the Martian lower thermosphere as inferred by Trace Gas Orbiter accelerometer measurements’. In:
Icarus, p. 114109.
Siddle, A. G., I. C. F. Mueller-Wodarg, S. W. Stone and R. V. Yelle (2019). ‘Global characteristics of
gravity waves in the upper atmosphere of Mars as measured by MAVEN/NGIMS’. In: Icarus 333,
pp. 12–21.
Snowden, D., R. V. Yelle, J. Cui, J. E. Wahlund, N. J. T. Edberg and K. Ågren (2013). ‘The thermal
structure of titan’s upper atmosphere, I: Temperature profiles from Cassini INMS observations’. In:
Icarus 226.1, pp. 552–582.
Spiga, A., J. Faure, J.-B. Madeleine, A. Määttänen and F. Forget (2013). ‘Rocket dust storms and
detached dust layers in the Martian atmosphere’. In: Journal of Geophysical Research: Planets 118.4,
pp. 746–767.
Squyres, S. W., J. P. Grotzinger, R. E. Arvidson, J. F. Bell, W. Calvin, P. R. Christensen, B. C. Clark,
J. A. Crisp, W. H. Farrand, K. E. Herkenhoff, J. R. Johnson, G. Klingelhöfer, A. H. Knoll, S. M.
McLennan, H. Y. McSween, R. V. Morris, J. W. Rice, R. Rieder and L. A. Soderblom (2004). ‘In Situ
212
BIBLIOGRAPHY
Evidence for an Ancient Aqueous Environment at Meridiani Planum, Mars’. In: Science 306.5702,
1709 LP - 1714.
Stone, S. W., R. V. Yelle, M. Benna, M. K. Elrod and P. R. Mahaffy (2018). ‘Thermal Structure of
the Martian Upper Atmosphere from MAVEN NGIMS’. In: Journal of Geophysical Research: Planets
123.11, pp. 2842–28867.
— (2019). ‘Transport of Water to the Martian Upper Atmosphere amid Regional and Global Dust
Storms’. In: LPI Contributions 2089, p. 6363.
Streeter, P. M., S. R. Lewis, M. R. Patel, J. A. Holmes and D. M. Kass (2019). ‘Surface warming during
the 2018/Mars Year 34 Global Dust Storm’. In: Geophysical Research Letters.
Tellmann, S., M. Patzold, B. Hausler, D. P. Hinson and G. L. Tyler (2006). ‘Observations of the Mar-
tian neutral atmosphere with the Radio Science Experiment MaRS on Mars Express’. In: European
Planetary Science Congress 2006, p. 320.
Terada, N., F. Leblanc, H. Nakagawa, A. S. Medvedev, E. Yiğit, T. Kuroda, T. Hara, S. L. England,
H. Fujiwara, K. Terada, K. Seki, P. R. Mahaffy, M. Elrod, M. Benna, J. Grebowsky and B. M.
Jakosky (2017). ‘Global distribution and parameter dependences of gravity wave activity in the
Martian upper thermosphere derived from MAVEN/NGIMS observations’. In: Journal of Geophysical
@TheMarsSociety (2016). A nice graphic showing all of the robotic mission to #Mars through 2015,
successful & failed, as well as a few planned missions.
Thiemann, E. M. B., F. G. Eparvier, L. A. Andersson, C. M. Fowler, W. K. Peterson, P. R. Mahaffy,
S. L. England, D. E. Larson, D. Y. Lo, N. M. Schneider, J. I. Deighan, W. E. McClintock and B. M.
Jakosky (2015). ‘Neutral density response to solar flares at Mars’. In: Geophysical Research Letters
42.21, pp. 8986–8992.
Thiemann, E. M. B., F. G. Eparvier, S. W. Bougher, M. Dominique, L. Andersson, Z. Girazian, M. D.
Pilinski, B. Templeman and B. M. Jakosky (2018). ‘Mars Thermospheric Variability Revealed by
MAVEN EUVM Solar Occultations: Structure at Aphelion and Perihelion and Response to EUV
Forcing’. In: Journal of Geophysical Research: Planets 123.9, pp. 2248–2269.
Thomas, N., G. Cremonese, R. Ziethe, M. Gerber, M. Brändli, G. Bruno, M. Erismann, L. Gambicorti,
T. Gerber and K. Ghose (2017). ‘The colour and stereo surface imaging system (CaSSIS) for the
ExoMars trace gas orbiter’. In: Space Science Reviews 212.3-4, pp. 1897–1944.
213
BIBLIOGRAPHY
Tolson, R. H., G. M. Keating, G. J. Cancro, J. S. Parker, S. N. Noll and B. L. Wilkerson (1999).

‘Application of accelerometer data to Mars Global Surveyor aerobraking operations’. In: Journal of
Spacecraft and Rockets 36.3, pp. 323–329.
Tolson, R. H., G. M. Keating, R. W. Zurek, S. W. Bougher, C. J. Justus and D. C. Fritts (2007).
‘Application of Acclerometer Data to Atmospheric Modeling During Mars Aerobraking Operations’.
In: Journal of Spacecraft and Rockets 44.6, pp. 1172–1179.
Tolson, R. H., E. Bemis, K. Zaleski, G. Keating, J. D. Shidner, S. Brown, A. Brickler, M. Scher, P.
Thomas and S. Hough (2008). ‘Atmospheric Modeling Using Accelerometer Data During Mars Re-
connaissance Orbiter Aerobraking Operations’. In: Journal of Spacecraft and Rockets 45.3, pp. 511–
518.
Tomasko, M. G., L. R. Doose, M. Lemmon, P. H. Smith and E. Wegryn (1999). ‘Properties of dust in
the Martian atmosphere from the Imager on Mars Pathfinder’. In: Journal of Geophysical Research:
Planets 104.E4, pp. 8987–9007.
Toon, O. B., J. B. Pollack and C. Sagan (1977). ‘Physical properties of the particles composing the
Martian dust storm of 1971–1972’. In: Icarus 30.4, pp. 663–696.
Vago, J. L., F. Westall, A. J. Coates, R. Jaumann, O. Korablev, V. Ciarletti, I. Mitrofanov, J.-L. Josset,
M. C. De Sanctis and J.-P. Bibring (2017). ‘Habitability on early Mars and the search for biosignatures
with the ExoMars Rover’. In: Astrobiology 17.6-7, pp. 471–510.
Vallis, G. K. (2006). Atmospheric and Oceanic Fluid Dynamics. Cambridge, U.K.: Cambridge University
Press, p. 745.
Vals, M., A. Spiga, F. Forget, E. Millour, L. Montabone and F. Lott (2019). ‘Study of gravity waves
distribution and propagation in the thermosphere of Mars based on MGS, ODY, MRO and MAVEN
density measurements’. In: Planetary and Space Science 178, p. 104708.
Vandaele, A. C., F. Daerden, R. Drummond, E. Neefs, J.-J. Lopez-Moreno, J. Rodriguez Gomez, M. R.
Patel, G. Bellucci and N. Team (2011). ‘NOMAD, a spectrometer suite for nadir and solar occultation
observations on the ExoMars Trace Gas Orbiter’. In: Space Science Reviews 214.
Vandaele, A. C., O. Korablev, F. Daerden, S. Aoki, I. R. Thomas, F. Altieri, M. López-Valverde, G.
Villanueva, G. Liuzzi and M. D. Smith (2019). ‘Martian dust storm impact on atmospheric H 2 O
and D/H observed by ExoMars Trace Gas Orbiter’. In: Nature 568.7753, p. 521.
Walterscheid, R. L. (1981). ‘Dynamical cooling induced by dissipating internal gravity waves’. In: Geo-
physical Research Letters 8.12, pp. 1235–1238.
214
BIBLIOGRAPHY
Wang, H. and M. I. Richardson (2015). ‘The origin, evolution, and trajectory of large dust storms on
Mars during Mars years 24–30 (1999–2011)’. In: Icarus 251, pp. 112–127.
Wang, J.-S. and E. Nielsen (2003). ‘Behavior of the Martian dayside electron density peak during global
dust storms’. In: Planetary and Space Science 51.4, pp. 329–338.
Withers, P. and R. Pratt (2013). ‘An observational study of the response of the upper atmosphere of
Mars to lower atmospheric dust storms’. In: Icarus 225.1, pp. 378–389.
Withers, P. (2006). ‘Mars Global Surveyor and Mars Odyssey Accelerometer observations of the Martian
upper atmosphere during aerobraking’. In: Geophysical Research Letters 33.2.
Withers, P. and M. Mendillo (2005). ‘Response of peak electron densities in the martian ionosphere
to day-to-day changes in solar flux due to solar rotation’. In: Planetary and Space Science 53.14,
pp. 1401–1418.
Wolff, M. J., M. D. Smith, R. T. Clancy, N. Spanovich, B. A. Whitney, M. T. Lemmon, J. L. Bandfield,
D. Banfield, A. Ghosh, G. Landis, P. R. Christensen, J. F. Bell III and S. W. Squyres (2006).
‘Constraints on dust aerosols from the Mars Exploration Rovers using MGS overflights and Mini-
TES’. In: Journal of Geophysical Research: Planets 111.E12.
Wolkenberg, P., M. Giuranna, M. D. Smith and D. Grassi (2020). ‘Similarities and differences of global
dust storms in MY 25, 28 and 34’. In: Journal of Geophysical Research: Planets 124.
Yelle, R. V., A. Mahieux, S. Morrison, V. Vuitton and S. M. Hörst (2014). ‘Perturbation of the Mars
atmosphere by the near-collision with Comet C/2013 A1 (Siding Spring)’. In: Icarus 237, pp. 202–
210.
Yiǧit, E., A. D. Aylward and A. S. Medvedev (2008). ‘Parameterization of the effects of vertically
propagating gravity waves for thermosphere general circulation models: Sensitivity study’. In: Journal
of Geophysical Research: Atmospheres 113.D19.
Yiǧit, E., S. L. England, G. Liu, A. S. Medvedev, P. R. Mahaffy, T. Kuroda and B. M. Jakosky (2015a).
‘High-altitude gravity waves in the Martian thermosphere observed by MAVEN/NGIMS and modeled
by a gravity wave scheme’. In: Geophysical Research Letters 42.21, pp. 8993–9000.
Yiğit, E. and A. S. Medvedev (2009). ‘Heating and cooling of the thermosphere by internal gravity
waves’. In: Geophysical Research Letters 36.14.
— (2017). ‘Influence of parameterized small-scale gravity waves on the migrating diurnal tide in Earth’s
thermosphere’. In: Journal of Geophysical Research: Space Physics 122.4, pp. 4846–4864.
— (2019). ‘Obscure waves in planetary atmospheres’. In: Physics Today 72, pp. 40–46.
215
BIBLIOGRAPHY
Yiǧit, E. and A. S. Medvedev (2010). ‘Internal gravity waves in the thermosphere during low and high
solar activity: Simulation study’. In: Journal of Geophysical Research: Space Physics 115.A8.
Yiğit, E., A. S. Medvedev and P. Hartogh (2018). ‘Influence of gravity waves on the climatology of
high-altitude Martian carbon dioxide ice clouds’. In: Annales Geophysicae 36.6.
Yiǧit, E., A. S. Medvedev and P. Hartogh (2015b). ‘Gravity waves and high-altitude CO2 ice cloud
formation in the Martian atmosphere’. In: Geophysical Research Letters 42.11, pp. 4294–4300.
Young, L. A., R. V. Yelle, R. Young, A. Seiff and D. B. Kirk (2005). ‘Gravity waves in Jupiter’s
stratosphere, as measured by the Galileo ASI experiment’. In: Icarus 173.1, pp. 185–199.
Young, L., R. Yelle, R. Young, A. Seiff and D. Kirk (1997). ‘Gravity Waves in Jupiter’s Thermosphere’.
In: Science 276.5309, 108 LP - 111.
Zurek, R. W. (1978). ‘Solar heating of the Martian dusty atmosphere’. In: Icarus 35.2, pp. 196–208.
Zurek, R. W., S. W. Bougher, R. Tolson, D. Baird, S. Demcak, D. M. Kass, R. Lugo and B. M. Jakosky
(2018). ‘The 2018 Planet-Encircling Dust Storm: Effects on the Mars Upper Atmosphere as seen in
MAVEN Accelerometer Data’. In: AGU Fall Meeting Abstracts. Vol.∼2018, P43J–3864.
Zurek, R. W. and L. J. Martin (1993). ‘Interannual variability of planet-encircling dust storms on Mars’.
In: Journal of Geophysical Research: Planets 98.E2, pp. 3247–3259.
Zurek, R. W., R. A. Tolson, S. W. Bougher, R. A. Lugo, D. T. Baird, J. M. Bell and B. M. Jakosky
(2017). ‘Mars Thermosphere as seen in MAVEN Accelerometer Data’. In: Journal of Geophysical
Zurek, R. W., R. H. Tolson, D. Baird, M. Z. Johnson and S. W. Bougher (2015). ‘Application of MAVEN
Accelerometer and Attitude Control Data to Mars Atmospheric Characterization’. In: Space Science
Reviews 195.1, pp. 303–317.
Zurek, R., J. Barnes, R. Haberle, J. Pollack, J. Tillman and C. Leovy (1992). ‘Dynamics of the atmo-
sphere of Mars’. In: Mars. Ed. by H. H. Kieffer, B. M. Jakosky, C. W. Snyder and M. S. Matthews,
pp. 835–933.
216

Siddle A 2021 PHD Thesis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Siddle A 2021 PHD Thesis

Uploaded by

Copyright:

Available Formats

Aeronomy of Mars: An Observational Study of the

Dynamic Martian Upper Atmosphere

Department of Physics, Imperial College London

Submitted in partial fulfilment of the requirements for the degree of

Alex Siddle - December 2020

1.8.1 Dust Storm Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2 Instrumentation and Data 50

3 Data Analysis and Model Comparison 67

4 Background Density and Temperature Analysis 83

4.3 Investigating Diurnal Temperature Variations . . . . . . . . . . . . . . . . . . . . . . . 85

5 Gravity Waves in the Martian Upper Atmosphere 111

7 Results from Trace Gas Orbiter Aerobraking Campaign 160

7.3.1 TGO and MAVEN Densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

8 Conclusions and Future Work 185

A Visualisation of Regions Sampled Multiple Times by MAVEN 194

B Deriving Lower Atmosphere Heating Caused by a Dust Storm 195

C Software Used 197

D Data Used 198

1.3.1 Infographic of previous missions to Mars . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.2.1 MAVEN’s coverage in altitude, local time and latitude . . . . . . . . . . . . . . . . . . 53

3.2.1 Diagram showing convergence of temperature profiles after applied offsets . . . . . . . . 69

4.2.1 First Deep-Dip derived temperature profiles . . . . . . . . . . . . . . . . . . . . . . . . 85

7.2.1 Example TGO density profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

A.0.1Latitude-local time overlap period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

1.4.1 Comparison of Mars and Earth’s planetary properties . . . . . . . . . . . . . . . . . . . 22

2.2.1 MAVEN Deep Dip Ephemeris . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.5.1 Extracted periods observed in density due to solar rotation effects . . . . . . . . . . . . 96

5.2.1 Characteristics of extracted example waves . . . . . . . . . . . . . . . . . . . . . . . . . 116

7.2.1 Periapsis locations for profiles shown in Figures 7.2.1a-c . . . . . . . . . . . . . . . . . . 161

amu Atomic Mass Unit = 1.66×10−27 kg.

AU Astronomical Unit = 1.496×108 km.

CME Coronal Mass Ejection.

EDM Entry, Descent and Landing Demonstrator Module.

ESA European Space Agency.

EUV Extreme Ultraviolet.

GCM General (Global) Circulation Model.

IMU Inertial Mass Unit.

IUVS Imaging Ultraviolet Spectrograph.

LMD-MCD Laboratoire de Météorologie Dynamique Mars Climate Database.

LST Local Solar Time.

M-GITM Mars Global Ionosphere-Thermosphere Model.

M-S Distance Mars-Sun Distance.

MARSIS Mars Advanced Radar for Subsurface and Ionosphere Sounding.

MAVEN Mars Atmosphere and Volatile Evolution Mission.

MEX Mars Express.

MGS Mars Global Surveyor.

MRAMS Mars Regional Atmospheric Modeling System.

MRO Mars Reconnaissance Orbiter.

NASA National Aeronautics and Space Administration.

NGIMS Neutral Gas and Ion Mass Spectrometer.

ODY Mars Odyssey.

ROSCOSMOS Roscosmos State Corporation for Space Activities.

TGO Trace Gas Orbiter.

UAMS Upper Atmospheric Mass Spectrometer.

less in the order each topic appears.

1.2 Motivation for Studying Mars

1.3 Previous Observations of Mars

1.3.1 Pre-Spacecraft Observations

1.3.2 Previous Spacecraft Observations

1.4 Mars Planetary Properties

Property Mars Earth

Equatorial radius (km) 3394 6378

Month No. Ls Range (◦ ) Sol range Duration (sol) Details